Electric Machines and Drives PDF

ELECTRIC MACHINES
AND DRIVES
Edited by Miroslav Chomat
Electric Machines and Drives
Edited by Miroslav Chomat
Published by InTech
Janeza Trdine 9, 51000 Rijeka, Croatia
Copyright © 2011 InTech

All chapters are Open Access articles distributed under the Creative Commons
Non Commercial Share Alike Attribution 3.0 license, which permits to copy,
distribute, transmit, and adapt the work in any medium, so long as the original
work is properly cited. After this work has been published by InTech, authors
have the right to republish it, in whole or part, in any publication of which they
are the author, and to make other personal use of the work. Any republication,
referencing or personal use of the work must explicitly identify the original source.
Statements and opinions expressed in the chapters are these of the individual contributors
and not necessarily those of the editors or publisher. No responsibility is accepted
for the accuracy of information contained in the published articles. The publisher
assumes no responsibility for any damage or injury to persons or property arising out
of the use of any materials, instructions, methods or ideas contained in the book.
Publishing Process Manager Katarina Lovrecic

Technical Editor Teodora Smiljanic
Cover Designer Martina Sirotic
Image Copyright demarcomedia, 2010. Used under license from Shutterstock.com
First published February, 2011

Printed in India
A free online edition of this book is available at www.intechopen.com

Additional hard copies can be obtained from [email protected]
Electric Machines and Drives, Edited by Miroslav Chomat

p. cm.
ISBN 978-953-307-548-8
free online editions of InTech
Books and Journals can be found at
www.intechopen.com
Contents
Preface IX
Chapter 1 Premium Efficiency Motors 1

M. Benhaddadi, G. Olivier, R. Ibtiouen, J. Yelle, and J-F Tremblay
Chapter 2 Electric Motor Performance Improvement Using

Auxiliary Windings and Capacitance Injection 25
Nicolae D.V
Chapter 3 Magnetic Reluctance Method for Dynamical

Modeling of Squirrel Cage Induction Machines 41
Jalal Nazarzadeh and Vahid Naeini
Chapter 4 Minimization of Losses in Converter-Fed

Induction Motors – Optimal Flux Solution 61
Waldiberto de Lima Pires, Hugo Gustavo Gomez Mello,
Sebastião Lauro Nau and Alexandre Postól Sobrinho
Chapter 5 Sensorless Vector Control of Induction Motor Drive

- A Model Based Approach 77
Jogendra Singh Thongam and Rachid Beguenane
Chapter 6 Feedback Linearization of Speed-Sensorless

Induction Motor Control with Torque Compensation 97
Cristiane Cauduro Gastaldini, Rodrigo Zelir Azzolin,
Rodrigo Padilha Vieira and Hilton Abílio Gründling
Chapter 7 From Dynamic Modeling to Experimentation

of Induction Motor Powered by Doubly-Fed
Induction Generator by Passivity-Based Control 113
M. Becherif, A. Bensadeq, E. Mendes,
A. Henni, P. Lefley and M.Y Ayad
Chapter 8 A RMRAC Parameter Identification

Algorithm Applied to Induction Machines 145
Rodrigo Z. Azzolin, Cristiane C. Gastaldini,
Rodrigo P. Vieira and Hilton A. Gründling
VI Contents
Chapter 9 Swarm Intelligence Based Controller for Electric

Machines and Hybrid Electric Vehicles Applications 161
Omar Hegazy, Amr Amin, and Joeri Van Mierlo
Chapter 10 Operation of Active Front-End Rectifier

in Electric Drive under Unbalanced Voltage Supply 195
Miroslav Chomat
Chapter 11 Space Vector PWM-DTC Strategy

for Single-Phase Induction Motor Control 217
Ademir Nied, José de Oliveira, Rafael de Farias Campos,
Seleme Isaac Seleme Jr. and Luiz Carlos de Souza Marques
Chapter 12 The Space Vector Modulation PWM Control

Methods Applied on Four Leg Inverters 233
Kouzou A, Mahmoudi M.O and Boucherit M.S
Preface
This book focuses on a very important and diverse field of electric machines and
drives. The history of the electric machine, which is the keystone of electromechanical
energy conversion, dates back to the beginning of the nineteenth century. The names
of famous scientists, such as Michael Faraday, Joseph Henry or Nikola Tesla, are associ-
ated with the invention of the rotating electric machine. Electric drives have quickly
become an integral part of our everyday lives and we can hardly imagine our civiliza-
tion without them. Electric drives play a vital part in industry, transportation as well as
in modern households. If we counted the number of electric drives around every one of
us today, we would certainly be surprised how big the number is.
Since the invention of the first electric machine, novel principles and designs have been
appearing and the properties and parameters of electric machines have been steadily
improving. The advent of power electronics and modern control circuitry at the end
of the twentieth century caused a revolution in the field of electric drives. Nowadays,
when modern technologies are available and advanced materials and techniques com-
monly utilized, formerly inconceivable results can be achieved in the field of modern
electric drives.
The twelve chapters of the book written by renowned authors, both academics and
practitioners, cover a large part of the field of electric machines and drives. Various
types of electric machines, including three-phase and single-phase induction ma-
chines or doubly fed machines, are addressed. Most of the chapters focus on modern
control methods of induction-machine drives, such as vector and direct torque control.
Among others, the book addresses sensorless control techniques, modulation strate-
gies, parameter identification, artificial intelligence, operation under harsh or failure
conditions, and modelling of electric or magnetic quantities in electric machines. Sev-
eral chapters give an insight into the problem of minimizing losses in electric ma-
chines and increasing the overall energy efficiency of electric drives, which is currently
viewed as a priority.
I would like to express my gratitude to all the authors for their contributions, in which
they shared their valuable experience and knowledge with the readers. It was their im-
mense involvement that enabled the publication of this book. I would also like to thank
the InTech staff for their great effort and support in preparation of the book. I hope it
X Preface
will benefit the field of electric machines and drives, provide the readers with a new
point of view on this interesting branch of electrical engineering and possibly initiate
many inventions and innovations in the future.
Miroslav Chomat
Institute of Thermomechanics AS CR, v.v.i.
Czech Republic
1
Premium Efficiency Motors

M. Benhaddadi1, G. Olivier1, R. Ibtiouen2, J. Yelle3, and J-F Tremblay3
1École
Polytechnique de Montréal, dépt. de génie électrique, C.P. 6079
Succursale Centre-ville, Montréal, Québec, H3C 3A7
2École Nationale Polytechnique d’Alger, dépt. de génie électrique, Avenue Pasteur
BP 182, El Harrach, 16200 Alger

3Cégep du Vieux Montréal, dépt. Technologie de génie électrique, 255 Ontario-Est
Montréal, Québec, Canada H2X 1X6

1,3Canada
2Algérie
1. Introduction
Despite its considerable potential for energy savings, energy efficiency is still far from
realizing this potential. This is particularly true in the electrical sector (IEA, 2010). Why?
There is no probably just one single answer to this question. A consequential response
requires major multiform research and an analytical effort. No doubt that analysis of the
interaction between energy efficiency policies and energy efficiency performance of
economies accounts for a significant part of the effort.
In the future sustainable energy mix, a key role will be reserved for electricity, as GHG
emissions reduction in this sector has to be drastically reduced. In this option, obvious
conclusion is that large market penetration Premium motors needs a complex approach
with a combination of financial incentives and mandatory legal actions, as industry doesn’t
invest according to least life cycle costs (DOE, 2010).
This present work illustrates the induced enormous energy saving potential, permitted by
using high-efficiency motors. Furthermore, the most important barriers to larger high-
efficiency motors utilization are identified, and some incentives recommendations are given
to overcome identified impediments.
In the present work, experimental comparison of the performance characteristics of 3 hp
Premium efficiency motors from three different manufacturers has been presented. The
motors were tested according to Standard IEEE 112-B.
2. Energy, climate change and electricity

According to last report Intergovernmental Panel on Climate Change IPCC report (IPCC,
2007), the observed increase in global average temperatures since the mid-20th century is
very likely due to the observed increase in anthropogenic greenhouse gas concentrations.
Moreover, there is no doubt that discernible human influences now extend to other aspects
of climate, including ocean warming, continental-average temperatures, temperature
extremes and wind patterns. Stabilizing atmospheric carbon dioxide concentrations at twice
2 Electric Machines and Drives
the level of pre-industrial times is likely to require emissions reductions up to 90 % below

current levels by 2100. Clearly, reductions of this magnitude can be achieved only by taking
action globally and across all sectors of the economy. The electricity sector will undoubtedly
need to assume a major share of the weight, according to its contribution to overall
emissions estimated to be more 10 000 Mt (million tone) CO2eq per year.
As can be seen in fig.1, the electricity generation is dominantly produced from fossil fuels
(coal, oil, and gas), and today’s situation is the same as forty years ago (DOE, 2010). In the
last XXI world energy congress, it is highlighted that electricity generation will still depend
on fossil sources. In the meantime, according to (IEA, 2010), industry accounts for more 40 %
of the world 20 000 TWh (terawatt hours, or so called billion kilowatt hours) electricity
consumption, weighting more 4 000 Mt CO2eq per year. Within the industrial sector, motor
driven systems account for approximately 60% to 65% of the electricity consumed by North
American (RNC 2004, DOE 2010) and European Union industries. Implementing high
efficiency motor driven systems, or improving existing ones just by 1 to 2 %, could save up
to 100-200 TWh of electricity per year. This would significantly reduce the need for new
power plants. It would also reduce the production of greenhouse gases by more 100 million
CO2eq per year and push down the total environmental cost of electricity generation.
The worldwide electric motors above 1 hp can be estimated to be nowadays more 300
million units, with the annual sales of 34 million pieces. Typically, one-third of the electrical
energy use in the commercial sector and two-thirds of the industrial sector feed the electrical
motors (DOE, 2010). Moreover, the low voltage squirrel cage induction motor constitutes the
industry workhorse. In particular industrial sector such as the Canadian petroleum and
paper industry, the share of the energy used by electrical motors can reach 90% (RNC 2004).
Since induction motors are the largest electrical energy user, even small efficiency
improvements will result in very large energy savings and contribute to reduce greenhouse
gas emissions GHG. Furthermore, the declining resources combined environmental global
warming concerns and with increasing energy prices make energy efficiency an imperative
objective.
Fig. 1. Electricity generation by fuel

Premium Efficiency Motors 3
3. Motor losses segregation and efficiency

The impact of a motor in terms of energy and economical costs depends on its performance
during its lifetime. The motor performances are characterized by the efficiency with which it
converts electrical energy into mechanical energy.
In Standard IEEE 112-B the losses are segregated and the efficiency is estimated by the
following formula:
ΔPstr = Pin – Pout – (ΔPel1 + ΔPel2 + ΔPcore + ΔPmech) (1)

Where the electric input power, Pin, is measured with a power analyser and the output
power, Pout, with a torque meter. The overall precision of efficiency assessment mainly
depends on the torque estimation, and with the improved accuracy of recent power
analysers and torque meters, this method can be considered accurate and reliable.
Motor efficiency is defined as a ratio motor mechanical output power and electrical input
power. Hence in order to have a motor perform better, it is important to reduce its losses.
The major motor losses are resistive losses in the stator and the rotor windings, and
magnetic losses (hysteresis and eddy current losses) in the cores. Other losses include
mechanical (bearing friction and ventilation), and stray load losses. High efficiency motor
losses relative distribution is not so different at low efficiency one’s; it’s more dependent on
the power. Their general distribution is illustrated in fig.2.
Fig. 2. Induction motor losses distribution

There are many ways to improve electric motor efficiency; the majority of them make the
motor larger in diameter or overall sizes and, of course, more expensive.
• Winding stator (∆Pel1) and rotor (∆Pel2) losses are due to currents flowing through the
stator windings and rotor bars. These losses can be reduced by decreasing the
conductor current density in the stator windings, in the rotor bars and in the end rings.
Using larger conductors lowers stator resistance, while the use of copper instead of
aluminum reduces rotor losses (Parasiliti et al. 2002). Another way of decreasing stator
losses is by reducing the number of turns. Unfortunately, this increases the starting
current and maximum torque, as worsen the power factor.
• Magnetic losses ∆Pm occurring in the stator and rotor laminations are caused by the
hysteresis and eddy current phenomena. These losses can be decreased by using better
grade magnetic steel, thinner laminations and by lowering the flux density (i.e larger
magnetic cores). The better grade of laminations steel are still relatively very expensive.
Cheaper manufacturing methods other than stamping are expected to become available
in the near future.
• Mechanical losses ∆Pmec are due to bearing friction and cooling fan air resistance.
Improving the fan efficiency, the air flow and using low friction bearings result in a
more efficient design. As these losses are relatively small, the efficiency gain is small
too, but every improvement is welcome.
• Stray load losses ∆Pst are due to leakage fluxes induced by load current, non-uniform
current distribution, mechanical air-gap imperfection…These losses can be reduced by
design optimisation and manufacturing method improvements.
As can be deducted, one of the most established methods of increasing motor efficiency is to
use higher quality materials, inexorably increasing the motor cost, as most high performance
materials are expensive materials. In a recurrent manner, the same problem of increased cost
holds true for better construction techniques, such as smaller air gaps, copper rather than
aluminum in the rotor construction, higher conductor slot fill, and segmented core stator
construction. The resulting increase in motor cost is evaluated to be between 15 % and 30 %.
4. Testing standards
In North America, the prevailing testing method is based on direct efficiency measurement
method, as described in the Institute of Electrical and Electronics Engineers (IEEE) “Standard
Test Procedure for Polyphase Induction Motors and Generators” IEEE 112-B and in its Canadian
CSA 390 adaptation. The standard first introduced in 1984 and updated in 2004, requires the
measurement of the mechanical power output and the electric input, and provide a value for
the motor losses, where the additional stray load losses are extrapolated from their total by
the following formula (1). So, the efficiency is extrapolated by:
Pout Pout
η= = (2)
Pin Pout + ΔPel 1 + ΔPel 2 + ΔPm +ΔPmec + ΔPst
In Europe, the prevailing testing method is based on an indirect efficiency measurement as
defined in IEC 34-2 standard “Rotating electrical machines – Part 2: Methods for
determining losses and efficiency of rotating electrical machinery from tests”. The standard
first introduced in 1972 and updated in 1997, attribute a fixed value, equal to o.5 % of input
power to the additional stray load losses.
These standards differ mainly by the method used to take into account the additional load
losses (Aoulkadi & Binder, 2008, Boglietti et al 2004, Nagorny et al. 2004, Elmeida et al.
2002…). Many papers have been published and some authors have illustrated, that IEC 34 –
2 has drawback with a noticeable influence on the testing of high efficiency motors, as the
efficiency of this motor type is overestimated, particularly in the small motor size cases.
Ultimately, standard IEC 34 – 2 was found to be unrealistic with its 0.5 % Pin value for stray
losses (Aoulkadi & Binder, 2008, Renier et al. 1999, Boglietti et al 2004…). That is why, in
2007, IEC published a revised standard for efficiency classification no. 60034-2-1 which
includes a test procedure largely comparable to IEEE 112-B or CSA C390. Newly
harmonized standards for energy efficiency testing IEC 60034-2-1 can contribute to lowering
barriers in global trade for energy efficient motor systems.
5. Minimum energy performance standard MEPS and efficiency motor

classification
There are many different worldwide definitions for energy efficient motors, as until these
last years, there was no consensus on what really represents an energy efficient motor.
Technical barriers include non harmonized testing standards and efficiency classification. In
reality, the key mandatory instrument is minimum energy performance standards (MEPS).
5.1 MEPS in North America

On October 1992, US Congress voted law, Energy Policy Act EPAct, which mandates strict
energy efficiency standards for electrical appliances and equipment, including electric
motors. Motor MEPS were for the first time introduced in 1992 when all partners were
finally persuaded that voluntary measures are too slow, and no significant market
transformation towards more efficient motors was possible otherwise.
EPAct requires that the general purpose electric motors meet the higher nominal efficiency
requirements defined in the table of National manufacturer association NEMA Standard,
and the implementation of the motor MEPS went into effect in 1997.
The Canadian Standard association developed a Canadian standard in 1993, and updated it
in 1998. CAN/CSA C-390 set the requirement for minimum efficiency for new motors made
or sold in Canada at the same value as the NEMA energy-efficient level.
The Energy Policy Act EPAct-92 motors covered are:
• General purpose
• Definite or special purpose in a general purpose application
• Continuous duty
• 2, 4 & 6 Pole
• 1-200 HP
• 230/460/, 60 Hz
The Canadian standard was furthermore extended to 575 V and IEC motors, and included
75 % full load to reach maximum efficiency.
Some motors were not covered:
• Definite or special purpose in a non-general purpose application
• Slower speeds
• Inverter duty
• Multi-speed
• Totally enclosed air over TEAO, and totally enclosed not ventilated TENV
As a result of the mandatory standard that was endorsed as part of the EPAct-92, North
America had a motor standard foundation that leads the new century world.
In 2002, NEMA and Consortium for energy efficiency CEE established a voluntary NEMA
Premium level of efficiency, and the manufacturers began the next step in evolution with
the implementation on voluntary basis MEPS NEMA Premium efficiency motors. NEMA
premium efficiency standards (CEE 2007) have remained voluntary for a long period of 10
years. In spite of this, NEMA premium motors have been progressively gaining market
share, as the overall benefits of Premium motors is incommensurable (more reliable, last
longer, have longer warranties, run more quietly and cooler and produce less waste heat
than their less-efficient counterparts). The trend is particularly well depicted in the work,
illustrated in fig. 3.
The evolution of MEPS based on NEMA Premium is now moving from voluntary basis to
legislated regulation, as the law implementation is awaited for December 2010. So, 1 –200
HP general purpose motors already covered by EPAct will change from NEMA MG-1, Table
12-11 Energy Efficient (Annex 1) to Table 12-12 NEMA Premium efficiencies levels (Annex
2), except for fire pump motors which remain at EPAct-92 level.
Moreover, the proposal expands the scope of enclosed 1-200 Hp motors, as several motor
types not previously covered by EPAct-92 must meet EPAct efficiency levels. The added
motors are:
• U-frame
• Design C
• Footless
• Close-coupled pump
• Vertical solid shaft normal thrust
• 8 poles
• All low voltages ≤ 600 V not previously covered, including IEC Metric frame motors
from 90 frames and up.
• 201 to 500 HP low voltage 2-8 pole general purpose motors, where Design B represent
something like ¾ of total range.
It’s appropriate to highlight that this additional coverage means that MEPS is extended to
more 90 %of the motors in the 1 to 200 Hp range, despite the fact that some manufacturers
expressed technical skepticism about meeting premium efficiency levels design C and 900
rpm motors.
Fig. 3. Efficiency trend 2001-2006 and efficiency expected 2008-2013

On the other hand, EISA does not apply to motors exported outside North America, as well
as EISA doesn’t contain any requirement to replace electric motors in use. In addition, the
law applies only to motors manufactured after December 19, 2010 and motors in inventory
on that date can be sold or used without any additional constraints, but some customers
may not probably want them.
Some Motor configurations are not covered by EISA law. These are:
• Design D with high slip
• Inverter duty with optimized windings for adjustable speed drives ASD use
• Customized OEM mounting
• Intermittent duty
• Integral with gearing or brake where motor cannot be used separately
• Submersible motors
• Single Phase motors
• DC motors
• Two-digit frames (48-56)
• Multi-speed
• Medium voltage motors
• Repaired motors
• TENV and TEAO enclosures
It’s relevant to notice that regulation concerning single phase and three-phase small (42/48
and 56 frame) motors is ready, as Department of energy (DOE 2010) published in 2010 its
final rule giving efficiency limits that must be meet by small general purpose electric
motors, starting from 2015. Nevertheless, NEMA doesn’t support regulation of this product,
as many questions are still not solved while small induction motor is continually losing
market part to switched reluctance and permanent magnet types.
5.2 New IEC standard

In 2007, IEC published a revised standard for efficiency testing no. 60034-2-1 which includes
a test procedure largely comparable to IEEE 112-B or CSA C390, and in September 2008 a
new standard no. 60034-30 for efficiency classification of electric motors.
New standard for efficiency classification is applicable to single speed, three-phase
induction motors with the following parameters:
• Power from 0.75 to 375 kW
• Voltage under 1000 V
• 50 and 60 Hz frequency
• 2, 4 and 6 poles
• Duty S1 or S3
• All IP1x to 6x and IC0x to 4x
• Networked
• All types of fixing, shaft extension, accessories
Some Motor configurations are not covered by IEC 60034-30. These are:
• Motors with reinforced isolation specifically designed for variable speed drives
applications
• Motors which are fully integrated in a machine and cannot be tested separately
Newly 2008 harmonized standard for energy efficiency class 60034-30-2008 follow
International Electrotechnical Commission (IEC) protocol and defines four induction motor
efficiency classes:
• Super Premium efficiency level IE4

• Premium efficiency level IE3
• High efficiency level IE2
• Standard efficiency level IE1
It’s noticeable that, nowadays, Premium efficiency IE3 is the most efficient motor. Super
Premium efficiency IE4 is a future new generation motor. It is awaited that in average, the
losses reduction of IE4 should be 15 % compared to IE3. So, IE4 is not a standard in fact, but
just a level.
5.3 MEPS in EU and BRIC countries

In Europe, the European committee of manufacturers of electrical machines and power
electronics CEMEP has classified 2 & 4 pole 1-90 kW motors into three levels:
• High (EFF1),
• Improved (EFF2), and
• Standard efficiency (EFF3).
The CEMEP classification has induced substantial EFF3 motors reduction and EFF2 market
share promotion. Nevertheless, EFF1 market part is still modest. Meanwhile, European
Union is considering prohibiting the sale of motors that don’t meet EFF2 criteria in the near
2011 future.
The new EU MEPS system is based on the latest IEC standard, and it represents a significant
step towards worldwide harmonization of efficiency regulations.
The challenges that EU countries facing are:
• From June 16, 2011, motors shall not be less efficient than IE1 (i.e. EFF2) and EFF3
motors will be banned
• From January 1, 2015, motors with rated output of 7.5 to 375 kW shall not be less
efficient than IE2
• From January 1, 2017, motors with rated output of 7.5 to 375 kW shall not be less
efficient than IE3
• From January 1, 2017, all motors with rated 0.75 to 375 kW shall not be less efficient
than IE2 if equipped with adjustable speed drive (ASD).
BRIC countries (Brazil-Russia-India-China) motors and motor driven equipment are still
relatively less efficient. For example, the efficiency of over 80 % Chinese motors is 2-5 %
lower than international advanced ones. Chinese scientists consider that if efficiency of
motor systems could be raised to the North American level, then 150 TWh of electricity
would be saved each year. But China is making progress, by formulating a number of
policies, laws and regulations on energy conservation. Beginning this 2010 year, the MEPS
efficiency of newly added motors should reach international first-rate level Class 2, which is
equivalent to IE1, or EFF2.
At the present time, 10 countries with a half of global electricity demand have motor MEPS
endorsement. In the next two years, 15 new countries targeted for a next round until 2012 to
reach around 80% of global electricity demand.
As earlier mentioned, the International Electrotechnical Commission (IEC), a worldwide
organization for standardization comprising all national standards committees, defined in
IEC 60034-30 (2009) four efficiency classes for single-speed cage-induction motors, and
specified test procedures. It’s important to notice that the so-called IE4 Super Premium
Efficiency products are not commercially available yet, while lower efficiency motors in use
now Eff3 disappears in the new classification. This new classification will be probably soon
adopted worldwide in place of regional or local classification, as illustrated in table 1. This
new standard defines efficiency classes and their containing minimum values (conditions).
Efficiency Class IEC USA/Canada CEMEP China

Super Premium
IE4 - - -
efficiency
Premium efficiency IE3 NEMA Premium - -
High efficiency IE2 EPAct EFF1 Class 1
Standard efficiency IE1 - EFF2 Class 2
Below standard
- - EFF3 Class 2
efficiency
Table 1. International motor efficiency classification
6. Life cycle cost Premium motors

An electric motor is somewhat cheap to buy, but expensive to run. For example, a 3 hp
Premium efficiency motor functioning 6 000 hours per year consumes about 1000 $ of
electricity at $0.07/kWh. The purchase price for such a motor is about 500 $ and over the
motor’s 15-year life, the acquisition price represents only 3 % of the lifetime costs, while the
cost of electricity accounts for 97 %. Finally, a 2 % increase in Premium motor efficiency over
EFF1 translates in energy savings over that time nearly twice the cost difference. In addition,
with a larger motor, the saving potential will be larger, and therefore payback periods
would be shortened. For the 100 Hp motor, the acquisition price represents only 1 % of the
lifetime costs, while the cost of electricity accounts for 99 %!!! Fig. 4 depicts typical lifetime
cycle cost motor in the conservative case (Benhaddadi & Olivier, 2010a).
The average life cycle of the small power motors is of the order of 15 years, i.e. the
equivalent of the average car range. The fundamental difference is in the fact that during
this period, the cost of the electricity will represent 97 % of the cost of useful life cycle of the
electric motor, while for the car motor, it represents only 10 %. Moreover, the car’s internal
combustion motors can rarely overcome 50 % efficiency, with an enormous negative impact
to be paid in environmental pollution. We can deduct from this fact that the improvement of
1 % of the electrical motor efficiency will have the same impact as the reduction of the 10 %
gas consumption car.
Moreover, the Canadian electricity costs are presently up to two times cheaper than elsewhere
and high electricity prices reduce payback period. In addition, some Canadian utility
companies and public agencies like Hydro Québec in Québec offer rebate programs to
encourage customers to upgrade their standard motors to Premium efficiency (Benhaddadi &
Olivier, 2010b). For motors from 1 to 75 hp, this program allows 600 $/hp to the customer and
150 $/hp to the distributor for each saved hp. Unfortunately, as a consequence of the lack of
energy saving importance, the purchase of a new motor, as well as the rewinding of defective
standard-efficiency motors, the choice of the motor is often driven by short term investment
considerations, not on the cost of the electricity which can be saved.
The first law for energy efficient motors is the Energy Policy Act (EPAct) which mandates
strict energy efficiency standards for electrical appliances and equipment. This law was first
adopted in USA and became effective in Canada with the adoption of Standard CAN/CSA-
C390-98. Today more than 75 % of the motors sold in North America are Premium efficiency
and EPAct machines. This clearly indicates the positive effect of the energy law. In light of
the above, and taking into consideration the very slow market transformation with just
voluntary and incentive measures, there’s no doubt that Premium motors will monopolize
the dominant part of the market in the near 2013 future. This is well illustrated in fig. 3.
There is no doubt that the appropriate legislation is the best way of achieving that goal
(Benhaddadi & Olivier, 2009b). Only the latest energy efficient motor technologies should be
manufactured and used. In general terms, North America is not on the leading edge for
energy saving and conservation. Motor efficiency is an exception that should be at least
maintained by EISA law implementation.
Fig. 4. Lifetime motor cost
7. Barriers to high efficiency motors market penetration

Despite the colossal energy saving potential and financial incentives programs, many
companies are still reticent to invest in energy-efficiency motors. The reasons why the well-
known potential for energy saving energy is not exploited have been investigated, and the
authors have identified several reasons why this potential is not yet fully exploited. The
Grand paradox is that cost effective measures are not taken because of several illogical
barriers. So, the most important barriers to high efficiency motors promotion are
(Benhaddadi & Olivier, 2008a):
• The energy costs are relatively so small that energy efficiency improvement isn’t taken
into consideration,
• Lower priority of energy savings importance, when other factors such as availability
service, reliability, and first costs are of premium importance,
• Industry reluctance to change what is, a priori, a good functioning system,
• Doubt about success of energy efficiency programs, or the discount rates used to justify
energy efficiency programs are too low,
• Downtime replacement cost look like peanuts, but shutdown time to install new
equipment is expensive and many companies don’t accept this inconvenient,
• Reduced budget often makes reducing energy consumption as « poor parent »,
inducing a lack of encouragement to make a decision,
• Implementing making-decision responsibility is often shared with many internal
conflicting pressures and divergence and ultimate choice don’t always belong to
electrician engineers, who are energy savings conscious,
• Distributors regularly represent two or more motor manufacturers and they can
advantage products from the manufacturer that offers the highest discount rather than
high-efficiency ones,
• Usual predisposition to use stocked old motors rather than purchase high efficiency
ones,
• It is not economically pragmatic to change a motor until it fails,
• Penchant to have the failed motor repaired rather than replaced by high efficiency ones,
• Degradation efficiency of repaired motor cannot be simply illustrated,
• Annual running hours are not sufficiently high to induce satisfactory payback.
8. Incentive policies to overcome barriers

Experience derived from many energy saving initiatives around the world showed that the
most successful programs are based on a combination of technical and promotional
information, educational tools and financial incentives. If technicians and engineers would
be trained in system design integration and least lifecycle cost as a goal, no doubt that the
problem of inefficient industrial equipment should be solved (Benhaddadi & Olivier, 2008a).
Consequently, to overcome the identified obstacles need a combination of the following
measures:
• Premium priority: For the companies, the energy saving status has to arrive at
legislative endorsement, like is the case for safety and quality insurance,
• Incentive programs: to reinforce energy savings promotion politics, much higher
discount rates should be used to evaluate the cost-effectiveness of energy efficiency
policies, programs or measures,
• Highlighted information and diffusion: this information must be of practical value, and
sufficiently demonstrative with real pilot projects,
• Environmental concern: It’s necessary to reinforce ecological policy criteria and support
environmental friendly companies. This follows the principle that the saved energy is
the most environmentally friendly one. A particularly promising concept is the
emissions trading scheme, which could be enable companies to claim emissions credits
for investments that reduce energy consumption,
• Legislation: to legislate against recalcitrant and to impose to market the gradual
approach of the « carrot and the stick », where the carrot represents the incentives and
the stick stands for refractory.
The authors strongly believed in the need to enhance policy measures aimed at reducing the
demand for energy and the resultant environmental impact. We therefore welcome the
increased interest in energy conservation in Canada. But, the accumulated experience clearly
show that putting in place incentives and voluntary measures for the energy efficiency of
electric motors is not sufficient, as it is prerequisite to implement mandatory measures for
larger market penetration (fig. 5). The carbon savings from this measure have the potential
to make a significant contribution to emissions target reduction.
Fig. 5. Incentives, voluntary and mandatory measures impact
9. Experimental setup and results

Figure 6 depicts the experimental laboratory set-up. The motors are fed by three single
phase autotransformers and direct torque control, DTC drives. The motors are mechanically
loaded by dc machine connected through a precision torque-speed transducer. A
motor/harmonic power meter is used for measuring real and reactive power, currents,
voltages and power factor. The motors are mechanically loaded by dc machine connected
through a high precision torque-speed transducer.
The measurements were taken in similar conditions and each motor was loaded until
thermal equilibrium was reached, while each of the two benches can be used to determine
the efficiency.
The two benches are sufficiently flexible and require minimum adjustments when different
Premium motors are tested (Benhaddadi et al. 2010b). Several 3 hp Premium motors from
different 3 manufacturers were tested. Figures 7 and 8 show the stators and rotors of the
three different motors.
The measurements were taken in similar conditions and each motor was loaded until
thermal equilibrium was reached.
For each of the motors B & C, Fig. 9 shows the variation of the efficiency versus the
mechanical torque in the case of direct 60 Hz feeding, without drive. The results for motor A
are not provided, as they are the same as for the motor B. One can deduct that when the
motors are feed from the rated 230 V grid, the difference between the measured efficiency at
rated load and the nameplate value is less a 0.2 %,. But, most electric motors are designed to
function at 50 % to 100 % of rated value. Fig. 9 also shows that maximum Premium motor
Fig. 6. Measurement setup with instrumentation

efficiency is near 75 % of rated load, and tends to decrease substantially below about 50 %
load. Moreover, experienced overloaded motors don’t significantly lose efficiency, as they
are designed with a 1.15 factor service. It’s also relevant to notice that the two motors give
the same efficiency value just for the rated regime, as this efficiency difference is significant
(1.5 %) when the motor is under loaded (Benhaddadi et al. 2010b).
Next step is to analyze the feeding voltage impact. The voltages choice are made taking into
account practical considerations: 230V, and 200.
• The first one is the motor nameplate indication, which is in accordance with an
industrial available 240 V feeding voltage,
• The second one is the real full-time laboratory available voltage. These two voltages
were obtained with three one-phase transformers.
The experimental results obtained in the grid feeding case (without drive) and illustrated in
fig. 10 show that feeding voltage has an important impact on efficiency value. With rated
230 V voltage and load values, efficiency is 89.3 %, i.e. 0.2 % less than nameplate value.
When the feeding voltage is decreased to 208 V and 200 V values, the efficiency decrease up
to 2 %. So, additional losses occur when a 230 V motor is operated at or below 208 volts. The
motor show lower full-load efficiency, slips more, and produces less torque.
In another hand, the ASD (adjustable speed drive) deployment to control motor can
substantially reduce energy consumption. The ASD advantages and energy consumption
reduction are nowadays well documented (Benhaddadi & Olivier, 2007). But, at the present
Fig. 7. Stator Premium motors
Fig. 8. Rotor Premium motors

time, many companies use ASD to feed their motors, whenever they don’t need speed
regulation. In the energy saving point of view, we must be careful with this making-
decision, as ASD introduces supplementary losses in the motor and the drive. As can be
seen in fig. 11 and 12, the drive introduces a noticeable reduction of the motor system
efficiency. This decrease reaches approximately 4 % in the rated regime, withdrawing totally
energy savings induced by Premium efficiency motor use. Further investigations to
correctly understand the extents of the losses introduced by the controller for all frequencies
are under consideration.
The other problem is that if Premium motors are misapplied, they may not achieve
predicted energy savings and may result in diminished performance efficiency. For
centrifugal pumps, an increase in operating speed will increase the required power by the
third power of the speed ratio. For example, by substituting Premium 1760 rpm motor to
EPAct 1740 rpm one, a 20 rpm increase in the speed induce 3.5 % increase in the load, as
(1760 ÷ 1740)3 = (1.014)3 = 1.035. So, when replacing a standard efficiency motor, one must be
careful, as a Premium motor with lower or equal full-load speed must be selected to avoid
the energy increase that may negate the predicted energy savings resulting from a higher
efficiency.
It’s important to notice that to date, there is no agreement that allows the determination of
ASD system efficiency at any given frequency. The ideal situation is to obtain a family of
efficiency curves for diverse torques and frequencies, including overrated values, as
experimentally illustrated in fig. 13.
Fig. 9. Manufacturing technology impact Fig. 10. Viltage feeding impact

Fig. 11. Drive impact Fig. 12. Drive impact

But, to generalize results, there are two difficulties: the first one can be illustrated by the
results presented in fig.14, where we can see that the same 3 Hp motors issued from two
different manufacturers can have the same efficiency for 60 Hz frequency feeding voltage,
but different efficiency for another 30 Hz frequency (fig. 14). For each of these two motors B
& C, Fig. 14 shows the variation of the efficiency versus the mechanical torque. In the 60 Hz
case, the difference between the measured efficiency at rated load is negligible, while it
reaches 2 % in the 30 Hz feeding frequency.
The second difficulty is about 50-60 Hz feeding frequency dilemma. As earlier illustrated, 50
Hz frequency gives better efficiency beginning from 2/3 load, while 60 Hz is better for low
loads. It’s noticeable that the same results were obtained for Motor B.
Considering that for the same frequency, two Premium motors issued from two different
manufacturers can show significantly different efficiency in low frequencies, one must be
careful in generalizing the conclusions of this research. Moreover, the same motor tested for
different frequencies can yield to different losses repartitions. So, before claiming that the
obtained results are, or are not in agreement with findings of other authors, several other
Premium efficiency motors from different constructors should be tested. The mentioned
work is under consideration.
10. Conclusions
In the future sustainable energy mix, a key role will be reserved for electricity, as GHG
emissions reduction in this sector has to be drastically reduced. In this option, obvious
conclusion is that large market penetration Premium motors needs a complex approach
with a combination of financial incentives and mandatory legal actions, as industry doesn’t
invest according to least life cycle costs.
The US Energy Policy Act and the Canadian Energy Efficient Act, along with the
implementation of NEMA Premium efficiency levels, have lead to North American
leadership on motor efficiency implementation. In general terms, North America is not on
the leading edge for energy saving and conservation. Motor efficiency is an exception that
should be at least maintained. Next step is to get Tax incentives to promote early retirement
of older inefficient pre-EPAct motors by replacing instead of repairing
Experimental comparison of the performance characteristics of 3 hp Premium efficiency
induction motors has been presented. The motors were tested according to Standard IEEE
112-B. In the rated frequency and voltage case, the experimental results are in good
agreement with nameplate manufacturer’s information. Particularly, a comparison of the
rated operating point shows that, the discrepancy is approximately 0.2 %.
However, in low voltage/frequency applications, the use of a variable speed drive
introduces extra losses and the overall efficiency can be noticeably reduced. The
experimental results show that feeding voltage has an important impact on efficiency value,
while efficiency at low frequencies depends on a certain level at manufacturer technology.
From a global energy saving point of view, the ASD application to Premium efficiency
motors should be promoted just when adjustable speed is needed.
Fig. 13. Frequency impact Fig. 12. Frequency & manufacturing

technology impact
ODP TEFC
HP 2 Pole 4 Pole 6 Pole 8 Pole 2 Pole 4 Pole 6 Pole 8 Pole
1.0 0.0 82.5 80.0 74.0 75.5 82.5 80 74
1.5 82.5 84.0 84.0 75.5 82.5 84 85.5 77
2.0 84.0 84.0 85.5 85.5 84 84 86.5 82.5
3.0 84.0 86.5 86.5 86.5 85.5 87.5 87.5 84
5.0 85.5 87.5 87.5 87.5 87.5 87.5 87.5 85.5
7.5 87.5 88.5 88.5 88.5 88.5 89.5 89.5 85.5
10.0 88.5 89.5 90.2 89.5 89.5 89.5 89.5 88.5
15.0 89.5 91.0 90.2 89.5 90.2 91 90.2 88.5
20.0 90.2 91.0 91.0 90.2 90.2 91 90.2 89.5
25.0 91.0 91.7 91.7 90.2 91 92.4 91.7 89.5
30.0 91.0 92.4 92.4 91.0 91 92.4 91.7 91
40.0 91.7 93.0 93.0 91.0 91.7 93 93 91
50.0 92.4 93.0 93.0 91.7 92.4 93 93 91.7
60.0 93.0 93.6 93.6 92.4 93 93.6 93.6 91.7
75.0 93.0 94.1 93.6 93.6 93 94.1 93.6 93
100.0 93.0 94.1 94.1 93.6 93.6 94.5 94.1 93
125.0 93.6 94.5 94.1 93.6 94.5 94.5 94.1 93.6
150.0 93.6 95.0 94.5 93.6 94.5 95 95 93.6
200.0 94.5 95.0 94.5 93.6 95 95 95 94.1
250.0 94.5 95.4 95.4 94.5 95.4 95 95 94.5
300.0 95.0 95.4 95.4 94.5 95.4 95.4 95 0
350.0 95.0 95.4 95.4 94.5 95.4 95.4 95 0
400.0 95.4 95.4 0.0 0.0 95.4 95.4 0 0
450.0 95.8 95.8 0.0 0.0 95.4 95.4 0 0
500.0 95.8 95.8 0.0 0.0 95.4 95.8 0 0
Annex 1. NEMA MG-1 Table 12-11 Full-Load Efficiencies of Energy Efficient Motors (EPAct)
ODP TEFC
HP 2 Pole 4 Pole 6 Pole 2 Pole 4 Pole 6 Pole
1 77 85.5 82.5 77 85.5 82.5
1.5 84 86.5 86.5 84 86.5 87.5
2 85.5 86.5 87.5 85.5 86.5 88.5
3 85.5 89.5 88.5 86.5 89.5 89.5
5 86.5 89.5 89.5 88.5 89.5 89.5
7.5 88.5 91 90.2 89.5 91.7 91
10 89.5 91.7 91.7 90.2 91.7 91
15 90.2 93 91.7 91 92.4 91.7
20 91 93 92.4 91 93 91.7
25 91.7 93.6 93 91.7 93.6 93
30 91.7 94.1 93.6 91.7 93.6 93
40 92.4 94.1 94.1 92.4 94.1 94.1
50 93 94.5 94.1 93 94.5 94.1
60 93.6 95 94.5 93.6 95 94.5
75 93.6 95 94.5 93.6 95.4 94.5
100 93.6 95.4 95 94.1 95.4 95
125 94.1 95.4 95 95 95.4 95
150 94.1 95.8 95.4 95 95.8 95.8
200 95 95.8 95.4 95.4 96.2 95.8
250 95 95.8 95.4 95.8 96.2 95.8
300 95.4 95.8 95.4 95.8 96.2 95.8
350 95.4 95.8 95.4 95.8 96.2 95.8
400 95.8 95.8 95.8 95.8 96.2 95.8
450 95.8 96.2 96.2 95.8 96.2 95.8
500 95.8 96.2 96.2 95.8 96.2 95.8
Annex 2. NEMA MG-1 Table 12-12 Full-Load Efficiencies for 60 Hz NEMA Premium
Efficient Electric Motors Rated 600 Volts or less
Rating Poles Rating Rating Poles

Rating (kW) (hp) 2 4 6 (kW) (hp) 2 4 6
0.75 1 72.1 72.1 70.0 0.75 1 77.0 78.0 73.0
1.1 1.5 75.0 75.0 72.9 1.1 1.5 78.5 79.0 75.0
1.5 2 77.2 77.2 75.2 1.5 2 81.0 81.5 77.0
2.2 3 79.7 79.7 77.7 2.2 3 81.5 83.0 78.5
3 4 81.5 81.5 79.7 3 4 84.5 85.0 83.5
4 5.5 83.1 83.1 81.4 4 5.5 86.0 87.0 85.0
5.5 7.5 84.7 84.7 83.1 5.5 7.5 87.5 87.5 86.0
7.5 10 86.0 86.0 84.7 7.5 10 87.5 88.5 89.0
11 15 87.6 87.6 86.4 11 15 88.5 89.5 89.5
15 20 88.7 88.7 87.7 15 20 89.5 90.5 90.2
18.5 25 89.3 89.3 88.6 18.5 25 89.5 91.0 91.0
22 30 89.9 89.9 89.2 22 30 90.2 91.7 91.7
30 40 90.7 90.7 90.2 30 40 91.5 92.4 91.7
37 50 91.2 91.2 90.8 37 50 91.7 93.0 91.7
45 60 91.7 91.7 91.4 45 60 92.4 93.0 92.1
55 75 92.1 92.1 91.9 55 75 93.0 93.2 93.0
75 75 92.1 92.1 91.9 75 75 93.0 93.2 93.0
90 125 92.7 92.7 92.6 90 125 93.0 93.5 94.1
110 150 93.0 93.0 92.9 110 150 94.1 94.5 94.1
132 175 93.3 93.3 93.3 150 200 94.1 94.5 94.1
160 215 93.5 93.5 93.5 185 250 94.1 94.5 94.1
200 270 93.8 93.8 93.8 200 270 94.1 94.5 94.1
260 350 94.0 94.0 94.0 260 350 94.1 94.5 94.1
300 400 94.0 94.0 94.0 300 400 94.1 94.5 94.1
335 450 94.0 94.0 94.0 335 450 94.1 94.5 94.1
375 500 94.0 94.0 94.0 375 500 94.1 94.5 94.1
Table 3. Standard Efficiency IE1 50Hz Table 4. Standard Efficiency IE1 60Hz
Annex 3. CEI efficiencies

Rating Poles Rating Rating Poles

Rating (kW) (hp) 2 4 6 (kW) (hp) 2 4 6
0.75 1 77.4 79.6 75.9 0.75 1 75.5 82.5 80.0
1.1 1.5 79.6 81.4 78.1 1.1 1.5 82.5 84.0 85.5
1.5 2 81.3 82.8 79.8 1.5 2 84.0 84.0 86.5
2.2 3 83.2 84.3 81.8 2.2 3 85.5 87.5 87.5
3 4 84.6 85.5 83.3 3.7 5 87.5 87.5 87.5
4 5.5 85.8 86.6 84.6 5.5 7.5 88.5 89.5 89.5
5.5 7.5 87.0 87.7 86.0 7.5 10 89.5 89.5 89.5
7.5 10 88.1 88.7 87.2 11 15 90.2 91.0 90.2
11 15 89.4 898.0 88.7 15 20 90.2 91.0 90.2
15 20 90.3 90.6 89.7 18.5 25 91.0 92.4 91.7
18.5 25 90.9 91.2 90.4 22 30 91.0 92.4 91.7
22 30 91.3 91.6 90.9 30 40 91.7 93.0 93.0
30 40 92.0 92.3 91.7 37 50 92.4 93.0 93.0
37 50 92.5 92.7 92.2 45 60 93.0 93.6 93.6
45 60 92.9 93.1 92.7 55 75 93.0 94.1 93.6
55 75 93.2 93.5 93.1 75 75 93.6 94.5 94.1
75 75 93.8 94.0 93.7 90 125 94.5 94.5 94.1
90 125 94.1 94.2 94.0 110 150 94.5 95.0 95.0
110 150 94.3 94.5 94.3 150 200 95.0 95.0 95.0
132 175 94.6 94.7 94.6 185 250 95.4 95.4 95.0
160 215 94.8 94.9 94.8 200 270 95.4 95.4 95.0
200 270 95.0 95.1 95.0 260 350 95.4 95.4 95.0
260 350 95.0 95.1 95.0 300 400 95.4 95.4 95.0
300 400 95.0 95.1 95.0 335 450 95.4 95.4 95.0
335 450 95.0 95.1 95.0 375 500 95.4 95.4 95.0
375 500 95.0 95.1 95.0
Table 5. High Efficiency IE2 50Hz Table 6. High Efficiency IE2 60Hz (EPACT)

( )
Rating Rating Poles
(kW) (hp) 2 4 6
Rating Poles
0,75 1 77,0 85,5 82,5
Rating (kW) (hp) 2 4 6
0.75 1 80.7 82.5 78.9 1,1 1,5 84,0 86,5 87,5
1.1 1.5 82.7 84.1 81.0 1,5 2 85,5 86,5 88,5
1.5 2 84.2 85.3 82.5 2,2 3 86,5 89,5 89,5
2.2 3 85.9 86.7 84.3 3,7 5 88,5 89,5 89,5
3 4 87.1 87.7 85.6 5,5 7,5 89,5 91,7 91,0
4 5.5 88.7 88.6 86.8 7,5 10 90,2 91,7 91,0
5.5 7.5 89.2 89.6 88.0 11 15 91,0 92,4 91,7
7.5 10 90.1 90.4 89.1 15 20 91,0 93,0 91,7
11 15 91.2 91.4 90.3
18,5 25 91,7 93,6 93,0
15 20 91.9 92.1 91.2
22 30 91,7 93,6 93,0
18.5 25 92.4 92.6 91.7
22 30 92.7 93.0 92.2 30 40 92,4 94,1 94,1
30 40 93.3 93.6 92.9 37 50 93,0 94,5 94,1
37 50 93.7 93.9 93.3 45 60 93,6 95,0 94,5
45 60 94.0 94.2 93.7 55 75 93,6 95,4 94,5
55 75 94.3 94.6 94.1 75 75 94,1 95,4 95,0
75 75 94.7 95.0 94.6 90 125 95,0 95,4 95,0
90 125 95.0 95.2 94.9 110 150 95,0 95,8 95,8
110 150 95.2 95.4 95.1 150 200 95,4 96,2 95,8
132 175 95.4 95.6 95.4
185 250 95,8 96,2 95,8
160 215 95.6 95.8 95.6
200 270 95,8 96,2 95,8
200 270 95.8 96.0 95.8
260 350 95.8 96.0 95.8 260 350 95,8 96,2 95,8
300 400 95.8 96.0 95.8 300 400 95,8 96,2 95,8
335 450 95.8 96.0 95.8 335 450 95,8 96,2 95,8
375 500 95.8 96.0 95.8 375 500 95,8 96,2 95,8
Table 7. Premium Efficiency IE3 50Hz Table 8. Premium Efficiency IE3 60Hz
(NEMA Premium)
Annex 4: Acronyms
• ASD Adjustable speed drives
• ACEEE American Council for and Energy Efficient Economy
• CEE Consortium for Energy Efficiency
• DOE Department of Energy
• EISA Energy Independence & Security Act
• EPAct Energy Policy act
• EU European Union
• GHG Greenhouse gas emissions
• IPCC International Panel on climate change
• OEM Original equipment manufacturer
• MG 1 Motor Generator
• MEPS Minimum energy performance standard MEPS
• MDM Motor Decisions Matter (MDM
• NEMA National Electrical Manufacturers Association
11. References
Aoulkadi, M., Binder, A.: “Evaluation of different measurement methods to determine stray
load losses in induction machines,”. IEEE Trans. On Industrial Electronics, vol. 2
No 1, 2008
Benhaddadi M., Olivier G.: (2010a) “La promotion de l’économie d’énergie électrique passe
par son juste prix,”. 79e Congrès de l’AFAS, Montréal, Canada, 2010
Benhaddadi M., Olivier G., and Yelle J.: (2010b) “Premium efficiency motors effectiveness,”.
IEEE International symposium on power electronics, electrical drives, automation
and motion SPEEDAM 2010, Pisa, Italy, 2010
Benhaddadi M., Olivier G., Labrosse D., Tétrault P.: (2009a) “Premium efficiency motors and
energy saving potential,”. IEEE International electric machines and drives
conference, IEEE_IEMDC, Miami, USA, 2009
Benhaddadi M., Olivier G.: (2009b) “L’économie d’énergie : une affaire de législation,”.
Communication présentée au 78e Congrès de l’ACFAS, Ottawa, Canada, 2009
Benhaddadi M., Olivier G.: (2008a) “Barriers and incentive policies to high-efficiency motors
and drives market penetration,”. IEEE International symposium on power
electronics, electrical drives, automation and motion SPEEDAM 2008, Ischia, Italy,
Benhaddadi M., Olivier G.: (2008b) “Le génie électrique à la rescousse des économies
d’énergie dans l’industrie,”. Communication présentée au 76e Congrès de l’ACFAS,
Québec, Canada, 2008
Benhaddadi M., Olivier G.: (2008c) “Dilemmes énergétiques,”. Presses de l’Université du
Québec, Québec, 2008, 216p.
Benhaddadi, M., Olivier G.: (2007) “Energy savings by means of generalization adjustable
speed drive Utilization,”. IEEE Canadian Conference on Electrical and Computer
Engineering, Vancouver, 2007.
Benhaddadi M., Olivier G.: (2004) “Including Kyoto in electrical engineering curriculum,”.
IEEE Canadian Conference on Electrical and Computer Engineering, Niagara Falls,
2004
Boglietti, A. Cavagnino, A. Lazzari, M. Pastorelli, , M.: (2004) “International standards for
the induction motor efficiency evaluation: a critical analysis o A. f the stray-load
loss determination,”. IEEE Trans. On Industry Appl., vol.40, No 5, 2004
Bonnett, A.H., Yung, C.: (2008) “Increased efficiency versus increased reliability: A
comparison of pre-EPAct, EPAct, and premium-efficiency motors,”. IEEE Industry
application Magazine, vol.2, 2008
BP (2010) British Petroleum statistical review of world energy (2010), June 2010
CEE (2007) Energy-efficiency incentive programs: Premium-efficiency motors and
adjustable speed drives in the US and Canada, (2007) prepared by consortium for
energy efficiency, may 2007, www.cee.org
DOE (2010) Energy Information Administration, Official energy statistics from the US
government,
DOE/EIA http://www.eia.doe.gov
Elmeida, A.I. Ferreira, F.FJ. Busch, J.F. Angers P. (2002) “Comparative analysis of IEEE-112 B
and IEC 34-2 efficiency testing standards using stray load losses in low-voltage
three phase cage induction motors,” IEEE Industry applications, March-April 2002,
pp. 608-614
Energy Policy Act EPACT and motor testing understanding, IEEE-112 Method B,
Evans, B.D., Crissman, J., Gobert, G.: (2008) “Test results for energy savings,”. IEEE Industry
application Magazine, vol.2, 2008
Finlay, W.R. Veerkamp B., Gehring D., and Hanna: P. (2009)“Improving motor efficiency
levels globally,”. IEEE Industry application Magazine, vol.15, 2009
IEA (2010) International Energy Agency, OECD/IEA http://oecd.org./
IEA (2009) The experience with energy efficiency policies and programs in IEA countries:
learning from the critics, (2009) International energy agency information paper,
august 2009
IPCC (2007) Contribution of working group I to the fourth assessment report of the
intergovernmental panel on climate change, summary for policymakers,
http://www.ipcc.ch
Nagorny, A., Wallace, A., Von Jouanne, A.: “Stray load loss efficiency connections,”. IEEE
Industry application Magazine, vol.10, issue 3, may-june 2004
NRC (2004) Energy-efficient motor systems assessment guide, Canadian industry program
for energy conservation (2004), CIPEC, Natural Resources Canada, 2004
Parasiliti, F., Villani, M., Paris, C., Walti, O. Songini, G. Novello, A. Rossi T.: (2004) “Three-
phase induction motor efficiency improvements with die-cast cooper rotor cage
and premium steel,”. IEEE International symposium on power electronics,
electrical drives, automation and motion SPPEDAM 2004, Capri, Italy, 2004
RNC (2004) Ressources Naturelles Canada: (2004) “Guide d’évaluation du rendement des
systèmes moteurs éconergétiques,” RNC, 2004
Renier, B. Hameyer, K., Belmans, R.: (1999) “Comparison standards for determining
efficiency of three phase induction motors,”. IEEE Trans. On Energy Conversion,
vol.14, No 3, 1999
Rooks, J.A., Wallace: (2004) “Energy efficiency of VSDs,” IEEE Industry applications
Magazine, vol.10, issue 3, 2004
UE (2003) Commission Européenne, Direction générale énergie et transport: (2003) “Motor
challenge programme,”. UE, Bruxelles, 2003
2
Electric Motor Performance Improvement Using

Auxiliary Windings and Capacitance Injection
Nicolae D.V
Tshwane University of Technology
South Africa
1. Introduction
Generally, some electric machines such as induction machines and synchronous reluctance
motors require reactive power for operation. While the reactive power required by a
synchronous machine can be taken from the power source or supplied by the machine itself
by adjustment of the field current, the power factor of an induction machine is always
lagging and set by external quantities (i.e., the load and terminal voltage). Poor power
factor adversely affects the distribution system and a cost penalty is frequently levied for
excessive VAr consumption.
Power factor is typically improved by installation of capacitor banks parallel to the motor. If
the capacitor bank is fixed (i.e. that it can compensate power factor only for a fixed load),
when the load is variable, then the compensation is lost. Some authors (El-Sharkawi et al,
1984, Fuchs and Hanna, 2002) introduced the capacitors using thyristor/triac controllers; by
adjusting the firing angle, the capacitance introduced in parallel with the motor becomes
variable and thus compensating the power factor for any load. Other works (Suciu et al,
2000.) consider the induction motor as an RL load and power factor is improved by inserting
a variable capacitor (through a bridge converter) which is adjusted for unity according with
the load. For the above methods, the capacitive injection is directly into the supply. Another
method conceived for slip ring induction motor was to inject capacitive reactive power
direct into the rotor circuit (Reinert and Parsley, 1995; Suciu, et al. 2002).
The injection of reactive power can be done through auxiliary windings magnetically
coupled with the main windings (E. Muljadi et al. 1989; Tamrakan and Malik, 1999;
Medarametla et al. 1992; Umans, and H. L. Hess, 1983; Jimoh and Nicolae, 2006, 2007). This
compensating method has also been applied with good results not only for induction
motors but also for a synchronous reluctance motor (Ogunjuyigbe et al. 2010).
2. Method description
2.1 Physical solution
The method described in this chapter makes use of two three-phase stator windings. One
set, the main winding (star or delta), is connected directly to the source. The other set of
windings - auxiliary, is only magnetically coupled to the main winding. All windings have
the same shape and pitch, but may have different turn numbers and wire sizes; usually
smaller in order to be accommodated in the slots together with the stator. The windings are
arranged in slots such that there is no phase shift between the two windings. Figure 1 shows
a possible arrangement of the windings for a four pole induction machine.
-C +A +A
-C +A
-C -Z +X +X -B
-Z +X
+B -Z -B
-Y
+B +Y -Y -B
+Y -Y
+B +C
+Y +Z
-A +C
-X +Z
-A -X +Z +C
-A -X -X -A
+C +Z -X -A
+Z -X
+C -A
+Z +Y
+C +B
-Y +Y
-B -Y +Y +B
-Y -Z
-B +B
+X -Z
-B +X +X -Z
-C
+A -C
+A +A -C
Fig. 1. High Power factor induction machines-windings arrangement
2.2 Auxiliary windings connections

As mention above, the main winding can have delta or star connection. Figure 2 shows the
main winding connected in star and the auxiliary windings connected in generic (a), star (b)
and delta (c) to the capacitor bank via a static switch.
Figure 3 shows a simpler way to inject capacitive reactive power. In this method, the
auxiliary windings are in “single –phase connection” with the apparent advantage of using
only one capacitor and static switch.
2.3 Variable Capacitors

In order to achieve a compensation for various loading of the machine, the compensating
capacitor should be able to be varied. This capability is obtained through connecting a fixed
capacitor via a static switch. The static switch can be achieved using thyristors or IGBTs in
bidirectional configuration.
Electric Motor Performance Improvement Using Auxiliary Windings and Capacitance Injection 27
Ia
Va *
* Vx
Ib
Vb * ● Rotor
* Vy
Ic
Vc *
* Vz
Static Switch
(a)
●Va ●Vb ● Vc
●
● ●
Cz Va Cx
●
Vz Vx
● ● ● ●
● Vx ● Vy ● Vz Vc● Vy
●Vb
● ●
Cx Cy Cz
● Cy
(b) (c)
Fig. 2. Auxiliary windings: a) generic connection; b) star connection; c) delta connection
Ia
Va
Vx
Ib
Vb ● Rotor
Vy
Ic
Vc
Vz
Va Iax
Static Switch
Fig. 3. Auxiliary windings: “single –phase connection”

2.3.1 Thyristor-based variable capacitor

Figure 4 shows the use of thyristor to accomplish a variable capacitor. The inductor Lr is
introduced to reduce – limit the surge current; it is relatively small and does not affect the
overall capacitive behaviour.
Th2
C Lr
Th1
Fig. 4. Variable capacitor using bidirectional thyristor
The equivalent capacitance depends on the delay angle. Due to the phase angle control, the
device introduces harmonic currents.
2.3.2 IGBT-based variable capacitor

The above drawback can be address using IGBTs in bidirectional configuration (Figure 5)
and a switching frequency higher then operational frequency (50 Hz).
Fig. 5. IGBT in bidirectional topology

Figure 6 shows a configuration to achieve a variable capacitor using two bidirectional static
switches. The main capacitor C1 is introduced in the auxiliary winding circuit, via a
bidirectional switch Sw1, for a period of time depending on the duty cycle (δ) of the
switching frequency; in this time the bidirectional switch Sw2 is OFF. When Sw1 is OFF, the
capacitor is discharged. The reactor Lr limits the capacitive surge current without affecting
the capacitive behaviour.
●
Ix
Sw1
Lr
●
Vx
C2 C1 Sw2
● ●
Fig. 6. Variable capacitor using two IGBTs in bidirectional topology
The capacitor C2, much smaller than C1 is connected to mitigate the voltage spikes during
switching off the main capacitor. Thus, the equivalent capacitor can be written as:
Ceq =δ×C1 +C 2 (1)
2.3.3 Variable capacitor H-topology

Figure 7 shows a single-phase H topology to achieve a variable capacitor. This
configuration using H-bridge bidirectional topology obtains a higher equivalent capacitance
for the same fixed one as reference. In this configuration, the reactor Lr has the same
purpose of limiting the surge capacitive current, while C2 also of small value mitigates the
voltage spikes. The equivalent capacitance could be express as:
C1
Ceq =C 2 + (2)
( 2δ-1)
2
It can be notice that the equivalent capacitance could increase significant when the duty
cycle approaches 50 %. In practice, the switches are not ideal and there is no “infinite
increase” of the equivalent capacitance.
Vx
Ix
● ●
C2
●
Sw1 Sw2
Lr
● ●
C1
Sw3 Sw4
●
Fig. 7. Variable capacitor using H-bridge bidirectional topology
Ix
Iy
Iz C
Fig. 8. Variable capacitor using three-phase H-bridge topology

Another solution to achieve a variable capacitance, or rather to generate a capacitive current

was proposed using a three-phase H topology as PWM inverter (E. Muljadi, et al 1989;
Tamrakan and Malik, 1999) as presented in Figure 8. The converter injects capacitive
reactive power into auxiliary windings and thus improving the power factor of the motor.
3. Mathematical model
The machine is treated as having two three-phase windings and the voltages equations
system can be written as:
d
[Vabcs ]=[R 1 ][I abc ]+ [λ abc ] (3)
dt
d
0=[R 2 ][I xyz ]+ [λ xyz ]+Vc xyz (2)
dt
d
0=[R r ][I abcr ]+ [λ abcr ] (5)
dt
where
T
Vabc = ⎣⎡Va Vb Vc ⎦⎤ (6)
T T
I abcs = ⎡⎣I a Ib I c ⎤⎦ ; λ abc = ⎡⎣λ a λb λ c ⎤⎦ (7)
⎡ra 0 0⎤ ⎡rx 0 0⎤
⎢ ⎥ ⎢ ⎥
[R 1 ]= ⎢ 0 rb 0 ⎥ [R 2 ]= ⎢ 0 ry 0⎥ (8)
⎢⎣ 0 0 rc ⎥⎦ ⎢0 0 rz ⎥⎦
⎣
Note that indices “1” refer to the main winding and “2” to the auxiliary winding.
⎡ λ abcs ⎤ ⎡ L abc L abcsxyz L abcsr ⎤ ⎡I abcs ⎤

⎢ ⎥ ⎢ ⎥⎢ ⎥
⎢ λ xyz ⎥ = ⎢L xyzabcs L xyz L xyzabcr ⎥ ⎢ I xyz ⎥ (9)
⎢λ ⎥ ⎢ L L abcrxyz
⎥
L abcr ⎦⎥ ⎢⎣I abcr ⎥⎦
⎣ abcr ⎦ ⎣⎢ abcrs
The inductances in eq. (9) are time dependent, and this make the equation difficult and time
consuming to solve. In order to obtain constant parameters, the voltage equations (3-5) are
then transformed to the rotor reference frame. To achieve this, the equations are multiplied
with an appropriate transformation matrix K(θ) to obtain:
d
[K(θ)][Vabcs ]=[R 1 ][K(θ)][I abcs ]+ [K(θ)][λ abcs ] (10)
dt
d
0=[R 2 ][K(θ)][I xyzs ]+ [K(θ)][λ xyzs ]+⎡⎣Kθ ⎤⎦⎡⎣ Vcxyzs ⎤⎦ (11)
dt
d (12)
0=[R r ][K(θ r )][I abcr ]+ [K(θ r )][λ abcr ]
dt
When these equations are expanded, after substantial matrix manipulations, it resolves to:
d
⎡ Vqdo1 ⎤ = ⎡⎣R s1 ⎤⎦ ⎡I qdo1 ⎤ + ⎡ λ qd01 ⎤ +ϖ ⎡ λ qdo1 ⎤ (13)
⎣ ⎦ ⎣ ⎦ dt ⎣ ⎦ ⎣ ⎦
d
0= ⎡⎣R 2 ⎤⎦ ⎡⎣I qdo2 ⎤⎦ + ⎡ λ qd02 ⎤ +ϖ ⎡ λ qdo2 ⎤ + ⎡ Vcqd02 ⎤ (14)
dt ⎣ ⎦ ⎣ ⎦ ⎣ ⎦
d
0= ⎡⎣R r ⎤⎦ ⎡⎣I qdor ⎤⎦ + ⎡ λ qd0r ⎤ +ϖ r ⎡ λ qdo1 ⎤ (15)
dt ⎣ ⎦ ⎣ ⎦
where
⎡ 0 1 0⎤ ⎡ 0 1 0⎤
⎢ ⎥ ⎢ ⎥
ϖ = ω ⎢ −1 0 0 ⎥ ; ϖ r = (ω − ωr ) ⎢ −1 0 0 ⎥ (16)
⎢⎣ 0 0 0 ⎥⎦ ⎢⎣ 0 0 0 ⎥⎦
and
∫
1
⎡ Vcqdo2 ⎤ = ⎡I qd02 ⎤ dt+ϖ ⎡ Vcqd02 ⎤ (17)
⎣ ⎦ C ⎣ ⎦ ⎣ ⎦
Neglecting the ‘0’ sequence since we initially are assuming a balanced system, the
expression for the stator and rotor flux linkages, resolves into the matrix:
⎡ λ q1 ⎤ ⎡ L ls1 +L m 0 L lm +L m 0 Lm 0 ⎤ ⎡ I q1 ⎤
⎢ ⎥ ⎢ ⎥⎢ ⎥
⎢ λ d1 ⎥ ⎢ 0 L ls1 +L m 0 L lm +L m 0 L m ⎥ ⎢ I d1 ⎥
⎢ λ ⎥ ⎢ L +L 0 L ls2 +L m 0 Lm 0 ⎥ ⎢ I q2 ⎥
⎢ q2 ⎥ = ⎢ lm m ⎥⎢ ⎥ (18)
⎢ λ d2 ⎥ ⎢ 0 L lm +L m 0 L ls2 +L m 0 L m ⎥ ⎢ I d2 ⎥
⎢ ⎥ ⎢ ⎢ ⎥
⎢ λ qr ⎥ ⎢ L m 0 Lm 0 L lsr +L m 0 ⎥ ⎢ I qr ⎥
⎥
⎢ λ ⎥ ⎣⎢ 0 Lm 0 Lm 0 L lsr +L m ⎦⎥ ⎢ I ⎥
⎣ dr ⎦ ⎣ dr ⎦
4. Equivalent model
4.1 Symmetrical loaded auxiliary windings
When each phase of the auxiliary windings is loaded with equal capacitors (C), eventually
star connected, the entire circuit is having a symmetrical behaviour and the equivalent
circuit is very simple as shown in Figure 9.
R1 Ll1
I1
Llm Llr
R2 Ll2
V1
Rr/s
Lm
C
Fig. 9. Equivalent circuit for symmetrical loading auxiliary winding

This equivalent circuit has two branches, each having separate resistance (R1 – main and R2 -
auxiliary) and leakage reactance (Ll1 – main and Ll2 – auxiliary) together with a common
mutual leakage inductance Llm, which occurs due to the fact that the two set of windings
occupy the same slots and therefore mutually coupled by their leakage flux. The mutual
inductance that occurs between main winding and rotor is represented by Lm.
Other parameters of the equivalent circuit are: main winding resistance R1, auxiliary
winding resistance R2, main winding leakage reactance L1, auxiliary winding leakage
inductance L2; mutual leakage inductance Llm, rotor leakage inductance Llr; and the rotor
resistance Rr and “s” is the slip.
In this analysis there is no need to refer the auxiliary winding quantities to the main
winding, because the two sets of windings are wound for the same number of turns with a
transformation ratio of one.
The above equivalent circuit helps us to determine the input impedance seen from the
supply and the condition for unity power factor.
Zin =R 1 +jX l1 +
( R 2 R 3 -X 2 X 3 ) +j ( X 2 R 3 +X 3R 3 ) (19)
( R 2 +R 3 ) +j ( X 2 +X 3 )
Where: R1 and jXl1 are the components of the main winding (per-phase impedance), X2 is the
equivalent reactance of the auxiliary winding including the compensating capacitor, R2 is
the resistance (per-phase) of the auxiliary winding, R3 and X3 are the equivalent resistance
and reactance of the paralleling the rotor branch with magnetizing branch:
1
X2 = -X l2 (20)
ωC
{
X 3 =Im jX lm + ⎡⎣( R r /s ) +jX lr ⎤⎦ //( jX m ) } (21)
{
R3 =Re ⎡⎣( R r /s ) +jX lr ⎤⎦ //( jXm ) } (22)
The condition for unity power factor is:
Im {Z in } = 0 (23)
Which, after some mathematical manipulation can be written as:
αX 22 +βX 2 +γ=0 (24)
With:
α=X 3 +X l1 (25)
β=-R 2 R 3 + ( R 1 +R 2 ) R 3 +2X 3 X l1 (26)
γ=-X 3 ⎡⎣R 2 R 3 - ( R 2 +R 3 ) R 2 ⎤⎦ +X l1 ⎡( R 2 +R 3 ) +X 23 ⎤
2
(27)
⎣ ⎦
The equation (24) together with relations (25) to (27) produces two solutions, which means
for a given slip (s) there are two values for the capacitor satisfying the condition for unity
power factor. The practical/recommended value is the high X2 or small C connected to the
auxiliary winding in order to have a small current through it.
4.2 Asymmetrical loaded auxiliary windings.

If the auxiliary windings are connected as “single-phase configuration”, then the system has
an asymmetrical behaviour. This connection is obtained by connecting in series the three
auxiliary windings and thus we can write: Ix = Iy = Iz = IX. Using this condition of current in
the expansion that results to equations (3) – (18) it can be written: Id2 = Iq2 = 0, thus the
resulting expression from this can be represented in a d-q-0 equivalent circuit of Figure 10
shown below.
Id1 R1 Ll1 ωλq1

Llm (ω-ωr)λqr Llr
R2 Ll2
Vd1
Lm Rr
Vd2=0
(a)
Iq1 R1 Ll1 ωλd1

Llm (ω-ωr)λdr Llr
R2 Ll2
Vq1
Lm Rr
Vq2=0
(b)
I01 R1 Ll1 Ll2 R2
V01 V02
Llm
Ceq
(c)
Fig. 10. The “d-q-0” equivalent circuit: a) “d” – equivalent circuit; b) “q” – equivalent circuit;
c) “0” – equivalent circuit
The particularity of connecting the auxiliary windings in a single phase winding creates an
asymmetrical situation which brings about the relevance of the zero sequence.
The power factor of the machine could be defined by the argument of Za, which is Va/Ia.
And this is expressed as:
Va Vq1 +V01
Za = = (28)
Ia Iq1 +I01
(Z01 +Zlm )
V01 (C)=(-jωC+Z02 )×I02 × (29)
Zlm
As can be observed from the equations (28) and (29), the power factor of the machine depends
on C and thus it can be brought to unity by means of adjusting the equivalent capacitance.
It should be also noticed that the asymmetrical behaviour of the auxiliary windings
connected as “single-phase configuration” has got a significant draw-back namely creating
torque ripple. This disadvantage should not be overseen by the simplicity of the physical
solution. Given this, further in this, we will consider only the symmetrical loading of the
auxiliary winding.
5. Concept validation
For the validation of this concept of improving performances of the induction motor
injecting capacitive reactive power via auxiliary windings, a standard 4 kW, 380V, 50Hz, 4
pole, 36 slots with frame size DZ113A induction motor have been used. The stator was
rewound with two full-pitches, single layer windings (see Figure 1).The main windings –
delta connected have same number of turns like in the original motor but the size of the wire
has been reduced to accommodate the auxiliary windings. The auxiliary windings were
connected in delta in order to reduce current through them. For this study, the auxiliary
windings have the same number of turns as main windings. The rotor is unchanged. Table 1
shows the parameters of the modified motor under test obtained using the IEEE test
procedure. The mutual leakage inductance Xlm is very small compared to the other
reactance’s (E. Muljadi et al, 1989).
Description of data Values

Main Winding Rated Voltage 380 V
Auxiliary Winding Rated Voltage 380V
Number of poles 4
Magnetizing Reactance (Xm) 37.86 Ω
Main winding phase resistance (R1) 4.33 Ω
Auxiliary winding phase resistance (R2) 18.1 Ω
Main winding leakage reactance (Xl1) 6.97 Ω
Auxiliary winding leakage reactance(Xl2) 6.97 Ω
Rotor resistance (Rr) 1.35 Ω
Rotor leakage reactance (Xlr) 1.97 Ω
Full load main winding current 8.6 A
Table 1. Specific parameters of the induction motor under test
Firstly, the motor was tested without compensation; the results are presented in Table 2.
IL(A) TL(Nm) P(W) S(VA) PF Slip RPM

5.43 1.83 678.6 3568.7 0.17 0.0026 1496
5.56 4.73 1734 3655.1 0.43 0.0060 1491
6.02 5.49 1992 3957.5 0.49 0.0086 1487
6.88 7.53 2709 4522.9 0.61 0.0153 1477
7.74 9.89 3540 5088.2 0.70 0.0213 1468
8.61 12.1 4179 5653.6 0.74 0.056 1416
Table 2. Experimental data for the motor under test
5.1 Simulation results

Based on Matlab platform, a simulation model has been built (Figure 11). The parameters
from Table 1 have been used for Matlab model after delta-star transformation.
The rotor resistor was simulated using a variable resistor depending on slip. The model was
run firstly to show the capability of adjusting the power factor via the static switch.
Continuous
0 t
pow ergui
Clock
R1 Xl1 Xl2 R2
i
+
-
Auxiliary
Main
Ia
Xr C
+
v
-
Xm 0
0.056
Va
s Control
g
Ideal Switch 1
In1
Conn1
m
2
V
I
Active & Reactive

Power
PQ
Conn2
Rr/s
Lr
Q P (Variable Resistor)
Fig. 11. Basic simulation model
5.1.1 Thyristor-based variable capacitor

One of the first solutions to introduce a variable capacitor was to use an “ac-ideal” switch
based on two thyristors anti-parallel connected. The test was done using a 100 µF and a
delay angle of 450 for a slip of 0.0026 (no load). This results in a reduction of reactive power
drawn by the motor from 1193 VAr to 46 VAr. The main drawback of this switching solution
is the strong distorting currents introduced as could be seen in Figure 12.
As could be noticed, both currents main and auxiliary are very much distorted. Obviously,
the power factor compensation is accompanied with degrading of all other performances of
the motor.
300
va (t)
200
ia (t)*10
Relative amplitude (V, A)

100
-100
-200
ix (t)*10
-300
0.145 0.15 0.155 0.16 0.165 0.17 0.175 0.18
Time (sec)
Fig. 12. Electric parameters (va, ia and ix) for compensated model using thyristor-based
variable capacitor
5.1.2 IGBT-based variable capacitor

The other solution tested to obtaining variable capacitor was that of Figure 6. The values for
the two capacitors were chosen as: C1 = 100 µF and C2 = 1 µF. The validation simulation was
done for a slip of s = 0.0026; the switching frequency was 1 kHz and then the duty cycle
manually adjusted. Figure 13 shows the main and auxiliary currents toward supply voltage
for a duty cycle of 40 % (Fig.13. a) and 80% (Fig.13. a).
400
300
300 va
va 200
200
ia*10 100
100 ia*10
0 0
-100 -100
ix*10 ix*10
-200 -200
-300 b
a -300
-400
0.145 0.15 0.155 0.16 0.165 0.17 0.175 0.18 0.145 0.15 0.155 0.16 0.165 0.17 0.175 0.18
Time (sec) Time (sec)
Fig. 13. Main and auxiliary currents for s = 0.0026, 1 kHz and: a) δ = 40 %; b) δ = 80 %.
As can be noticed a duty cycle of 40 % does adjust the shift between the main current and
voltage marginally increasing the power factor from 0.15 to 0.32 lagging. When the duty
cycle was increased to 80%, then the main current arrived in phase with the voltage. It can
also be noticed a relative high ripple especially for low duty cycle.
Figure 14 shows the same simulation conditions, but now the switching frequency was
raised to 4 kHz. As can be noticed the ripple has been reduced significant.
The other methods of producing variable capacitive reactive power, meaning the H-bridge
topologies require a more complex control system and are not addressed in this study.
400 400
va
300 300
va
200 200

ia*10
100 100 ia*10
0 0
-100 -100
ix*10 ix*10
-200 -200
-300 -300
-400 -400
0.145 0.15 0.155 0.16 0.165 0.17 0.175 0.18 0.145 0.15 0.155 0.16 0.165 0.17 0.175 0.18
Time (sec) Time (sec)
Fig. 14. Main and auxiliary currents for s = 0.0026, 4 kHz and: a) δ = 40 %; b) δ = 80 %.
5.1.3 Capacitance versus slip for unity power factor

Now, the model was run for each value of the slip (speed) given in Table 2. The result of
these simulations is the capacitance producing unity power factor (see Table 3).
s CY Ia Pa Qa Sa
PF
(n) (µF) (A) (W) (VAr) (VA)
0.0026 0 5.49 184 1193 1207 0.152
(1496) 81 1.65 362 9 363 0.999
0.006 0 5.65 366 1187 1242 0.295
(1491) 81 2.57 565 11 565 0.999
0.0086 0 5.86 505 1186 1289 0.392
(1487) 81.6 3.27 718 15 719 0.999
0.0153 0 6.71 855 1200 1473 0.58
(1477) 87 5.21 1143 1 1143 0.999
0.0213 0 7.68 1160 1229 1689 0.686
(1468) 90 6.82 1499 17 1499 0.999
0.056 0 8.31 1330 1252 1826 0.728
(1491) 90.6 7.66 1684 27 1686 0.999
Table 3. Simulation results: capacitance versus slip for unity power factor
It is interesting to observe that the capacitance producing unity power factor does not vary
so much as presented in other studies (E. Muljadi et al. 1989; Tamrakan and Malik, 1999)
concerning similar method of injecting capacitive reactive power.
Moreover, it could be found a fix value of 75 µF which maintain a high power factor
irrespective of the load/slip. Table 4 shows the simulation results for the fixed capacitor and
variable load/slip.
Figure 15 shows the variation of the power factor versus slip and Figure 16 shows the
variation of the supply current versus slip for compensated with a fix 75 µF capacitor per
phase and uncompensated machine.
s CY Ia Pa Qa Sa
PF
(µF) (A) (W) (VAr) (VA)
0.0026 75 1.57 330 101 345 0.956
0.006 75 2.47 531 111 542 0.978
0.0086 75 3.16 683 123 694 0.984
0.0153 75 4.91 1066 170 1079 0.987
0.0213 75 6.44 1396 230 1415 0.987
0.056 75 7.29 1580 272 1603 0.986
Table 3. Simulation results: capacitance versus slip for unity power factor
1.0
● ● ● ●
●
●
C o m p e n s a t e d m ac h i n e
0.9
0.8
0.7
● Unc omp ens at ed machi n e
Pow er F act or
0.6 ●
0.5
0.4 ●
0.3 ●
0.2
●
0.1
0 0.01 0.02 0.03 0.04 0.05 0.06
Slip
Fig. 15. Power Factor versus slip
U n c o m p e ns a t e d m a c h i n e ●
8
●
●
S u p pl y c ur r e n t I a ( A )
7
●
●
C om p e ns a t e d m ac h i n e
6 ●
●
●
5 ●
●
3
2
●
1
0 0.01 0.02 0.03 0.04 0.05 0.06
Slip
Fig. 16. Supply current versus slip
5.2 Experimental validation

After rewinding the machine under test with both sets of windings delta connected, three
identical capacitors were connected to the auxiliary. Then the “compensated” motor was
loaded gradually to get the same speeds as for the “uncompensated” motor.
The testing started with a capacitance of 25 µF as found during the simulation: CΔ = CY / 3.
This did not produce the results found via the simulation. Then a value of 28 µF was found to
give a relative even power factor of approximate 0.95 in the entire loading range (see Table 4).
Figure 17 shows the phase voltage (van) and current (ia) for the uncompensated (Fig. 17.a)
and compensated (Fig. 17.b) with slip of 0.0026. It can be observed, the power factor
increased from 0.17 to 0.937 which represent 450% improvement for no load. For full load
the improvement in power factor is 29%.
IL(A) TL(Nm) P(W) S(VA) PF Slip RPM

1.71 2.81 1058 1129 0.937 0.0026 1496
2.71 4.73 1694 1789 0.947 0.0060 1491
3.49 5.83 2191 2304 0.951 0.0086 1487
5.46 7.63 2709 3604 0.955 0.0153 1477
6.92 9.29 3442 4567 0.952 0.0213 1468
7.85 13.8 4937 5181 0.953 0.056 1416
Table 2. Experimental data for the motor under test
. . . . . . . .
v v
. . . . . . . . . . . . . . . .
. . . . . . a . . . . . . . . a .. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..
i
. . . . . . . . . . . . . . . .
. . . . . . . .
. . . . . . . . .
.
.
.
.
.
.
.
.
.
.
. a .. .
.
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..
. . . . . . . . . . . . . . . .
i
. . . . . . . . . . . . . . . .
. a
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. .
1) [THS720P].CH1 100 V 2 mS
. . . . . . .
1) [THS720P].CH1 .
100 V 5 mS. . . . . .
2) [THS720P].CH2 5 A 2 mS 2) [THS720P].CH2 1 A 5 mS
(a) (b)
Fig. 17. Voltage and current for: a) uncompensated motor; b) compensated motor
6. Comments
This study proved that direct injecting capacitance reactance through auxiliary winding
does improve the power factor of the induction motor; it also increases the ratio torque over
current. What is also very interesting is this method achieves a “flat” variation of power
factor with respect of load variation. This is a very important improvement given the fact
that majority of induction motors do not work at constant full load where the classic design
produces maximum performances.
However, this method introduces extra copper losses reducing the overall efficiency and
increasing the operation temperature.
This study did not intended to elucidate the full effect of the method upon the torque
especially the existence of the influence produced by the current flowing through auxiliary
windings. The only aspect clearly noticed was the torque ripple introduced by the “single-
phase” auxiliary winding connection.
Further more, the same method was applied to a synchronous reluctance motor
(Ogunjuyigbe et al, 2010). The same type of stator winded similarly as above was used with
a salient milled rotor obtained from the corresponding squirrel cage induction motor. The
experimental results show an improvement of the power factor from 0.41-0.69 to 0.93-0.97
for the entire loading range.
The economic benefits are related with the savings on demand especially for places where a
large number of three-phase induction motors are used under variable loading.
7. References
M. A El-Sharkawi, S. S Venkata, T. J Williams, and N. G Butler, “An adaptive Power Factor
Controller for Three-Phase Induction Generators”, Paper 84 SM 672-2 presented at
the IEEE/PES Summer Meeting, Seattle, Washington, July 15-20, 1984.
Fuchs, E.F. Hanna, W.J.; “Measured efficiency improvements of induction motors with
thyristor/triac controllers”, , IEEE Transaction on Energy Conversion, Volume 17,
Issue 4, Dec. 2002 pp. 437 – 444
Suciu, C.; Dafinca, L.; Kansara, M.; Margineanu, I.; “Switched capacitor fuzzy control for power
factor correction in inductive circuits”, IEEE 31st Annual Power Electronics Specialists
Conference, 2000. Vol. 2, pp. 773 - 777
C. Suciu, M. Kansara, P. Holmes and W. Szabo, “Performance Enhancement of an Induction
Motor by Secondary Impedance Control, IEEE Trans. On Energy Conversion, Vol. 17,
No. 2, June 2002
J. Reinert, M.J. Parsley, ‘‘Controlling the speed of 8×1 induction motor by resonating the
rotor &wit,” in IEEE Transactions on Industry Applications, Vol. 31, No. 4,
July/August 1995, pp. 887-891.
E. Muljadi, T.A. Lipo, D.W. Novotny, "Power Factor Enhancement of Induction Machines by
Means of Solid State Excitation," IEEE Trans. on Power Electronics, Vol. 4, No. 4, pp.
409418, Oct. 1989.
I. M Tamrakan and O.P Malik, “Power Factor Correction of Induction motors Using PWM
Inverter Fed Auxiliary Stator Winding”, IEEE Transaction on Energy Conversion, Vol.
14, No.3, Sept, 1999, pp. 426-432
J. B. Medarametla, M. D. Cos, and Baghzouz, “Calculations and measurement of unity plus
three-phase induction motor,” IEEE Transactions on Energy Conversion, vol. 7, no. 4,
pp. 732-738, 1992.
S. D. Umans, and H. L. Hess, “Modelling and analysis of a the Wanlass three-phase
induction motor configuration,” IEEE Transaction on Power Apparatus and Systems,
vol. 102, no. 9, pp. 2912-2916, 1983.
R. Spée and A. Wanllace, “Comparative Evaluation Of Power-Factor Improvement
Techniques For Squirrel cage Induction Motors”, Industry Applications Society
Annual Meeting, 1990.
A.A. Jimoh and D.V. Nicolae, “Performance Analysis of a Three-Phase Induction Motor with
Capacitance Injection”, OPTIM’06, 10th International Conference on Optimization of
Electrical and Electronic Equipments, Brasov, Romania, May 17-20, 2006
D.V. Nicolae and A.A. Jimoh, “Three-Phase Induction Motor with Power Electronic
Controlled Single-Phase Auxiliary Stator Winding”, PESC’07, The 38th IEEE Power
Electronics Specialists Conference, Orlando, USA, June 17-21, 2007
A.S.O. Ogunjuyigbe, A.A. Jimoh, D.V. Nicolae and E.S. Obe, “Analysis of Synchronous
Reluctance Machine with Magnetically Coupled Three Phase Windings and Reactive
Power compensation”, IET Electric Power Applications, 2010, Vol. 4, Iss. 4, pp 291-303
3
Magnetic Reluctance Method for Dynamical

Modeling of Squirrel Cage Induction Machines
Jalal Nazarzadeh and Vahid Naeini
Faculty of Engineering
Shahed University, Tehran
Iran
1. Introduction
Nowadays, induction machines play important role in electromechanical energy conversion in
industry. These machines are often operated in critical conditions where can cause unexpected
failures and outages. Generally, stator and bearing faults, broken rotor bar and end-rings,
air-gap irregularities are some of the major faults in an induction machine (Al-Shahrani, 2005;
Sprooten, 2007) which may be situated the induction machines in out of service (Siddique
et al., 2005). Fourier analysis for stator currents (Bellini et al., 2001; Benbouzid, 2000; Jung
et al., 2006), torque and rotor speed, acoustic noise and temperature analysis (Siddique et
al., 2005) are some classical techniques which introduced for identification and diagnosis of
induction machines faults. Additionally, other heuristic methods were proposed to monitor
of the induction machines for fault detection. For instance, neural network modelling were
applied to monitor an induction machine for fault detection (Su & Chong, 2007). Also, space
vector of rotor magnetic field (Mirafzal & Demerdash, 2004) based on artificial intelligent
approaches and pendulous oscillation of the rotor magnetic field were proposed. Recently,
a new technique based on the analysis of three-phase stator current envelopes was presented
(Mirafzalet & Demerdash, 2008). In all monitoring and fault detection techniques, we need
to tune up the monitoring systems based on response of induction machines for proper
operations. However, experimental set up for testing any arbitrary fault conditions are not
practical. Thus, an accuracy dynamic and steady state models of induction machines are very
important for this propose.
Also, for dynamical modelling of induction machines, space harmonic distribution, core
saturation and loss are often neglected in abc quantitative and two-axis methods (Krause et
al., 1995). Thus, these approaches do not have an efficient accuracy for modelling of induction
machines in asymmetrical and non-linear conditions. For considering distribution rotor bars,
coupled magnetic circuit method (Muñoz & Lipo, 1999), abc quantitative based on rotor bar
currents (Alemi & Nazarzadeh, 1996) can be utilized. Furthermore, winding function method
may be used to include the stator winding distribution effect in the air gap flux (Luos et al.,
1995). However, in all mentioned methods, the core saturation, stator and rotor teeth effects
and distributions of the rotor and stator windings can not be investigated, simultaneously.
Also, Finite Element Method (FEM) is a professional technique for analysis of any
electromagnetic systems, which needs to magnetic and geometry details of the systems (Faiz
et al., 2002). This method is very accurate and flexible, but due to complexity, the dynamical
modelling of an induction machine is quite complicated. Contrary to FEM, the Magnetic
Equivalent Circuit Method (MECM) can apply to analysis of the electro magnetics problems
with lower complexity. Magnetic saturation, space harmonics in stator and rotor teeth, stator
windings and distributed structure of a squirrel-cage rotor can be considered by MECM for
modelling and analysis of any induction machines (Jeong et al., 2003; Ostovic, 1989). In this
approach, the non-linear reluctances of flux paths use to configure magnetic equivalent circuit.
This method has less complexity than FEM for dynamical modelling of induction machines.
Therefore, developing an exact details model of induction machine for analysis of transient,
sensitivity and fault diagnosis in the asymmetrical conditions are very essential.
The present chapter introduces methodology of MECM for modelling and analysis of
asymmetrical non-linear systems in transient and steady state conditions. MECM is very
suitable method for finding a generalized accurate dynamical model of squirrel cage induction
machines with asymmetrical conditions. For evaluation of the method, several simulations
in linear and non-linear conditions are made. Also, some simulations results for induction
machines with broken bar faults and core saturation conditions are included to illustrate
capability of the method in asymmetrical conditions.
2. Electric and magnetic based model of squirrel cage induction machines

For detailed modelling of any electromagnetic systems, we have to find a correlation between
electric and magnetic variables of the system. Generally, a set of non-linear differential
equations presents dynamical model of a electromagnetic system that by using numerical
analysis, transient response of the electrical variables can be obtained. In addition, non-linear
algebraic equations illustrate non-linear relations between electrical and magnetic variables.
MECM provides an augmented model of the electromagnetic systems, in which we can
achieve all variables of the systems in transient and steady state, simultaneously. Also, the
main advantages of MECM for modelling of induction machines are; simple algorithm for
including distribution winding, stator and rotor teeth effects and magnetic core saturation
phenomena. Global non-linear model of squirrel cage induction machines can be offered
in algebraical (magnetic) and differential (electric) equations which will be presented in the
following sections.
3. Magnetic equivalent circuit of induction machines

Fig. 1 shows a part of rotor and stator structures for a typical squirrel cage induction
machine that magnetic circuit elements are presented for rotor and stator teeth and yoke.
Numbers of rotor and stator teeth are considered by h and l, respectively. Also in this figure,
magnetic mutual permeances of rotor and stator teeth in air gap are shown. Due to non-linear
characteristics of flux and magnetic current in iron core, permeances of the rotor and stator in
the magnetic cores are illustrated as non-linear elements. Gi,j is linear permeance of flux path
between i th stator and jth rotor teeth in the air gap.
3.1 Magnetic node equations

Due to the fact that magnetic permeances of each stator and rotor teeth make several magnetic
loops in the air gap, we apply node magnetic potential equations to the each air gap nodes for
Magnetic Reluctance Method for Dynamical Modeling of Squirrel Cage Induction Machines 43
Rotor Axis
Jt
Stator Axis
Fig. 1. Magnetic equivalent circuit of induction machine
simplicity in algebraic magnetic equations. For instance, sum of magnetic fluxes in i th air gap
node (stator tooth) have to be zero; thus we can write:
s
u i − u si+1 G sσ + u si − u si−1 G sσ + (u si − ur1 ) Gi1 + · · · + (u si − url ) Gil = φist (1)
where u si , uri and φist are magnetic potential of stator, rotor and flux of i th stator tooth and
G sσ and Gi,j are permeances of stator neighbour slots and mutual permance between i th stator
with jth rotor teeth, respectively.
Similar Eq. (1), continuity principle in the teeth fluxes at i th rotor node yields as:
r
u i − uri+1 Grσ + uri − uri−1 Grσ + (uri − u1s ) Gi1 + · · · + (uri − u sh ) Gih = − φirt (2)
In Eq. (2), φist and Grσ are flux of i th rotor tooth and permeance of rotor slot, respectively.
Consequently, node potential equations for equivalent circuit of Fig. 1 along the air gap can
be written as:
Ass Ust + Asr Urt = Ψst (3)
Ars Uss + Arr Urt = − Ψrt (4)
where Ass ∈ R h×h âĂć, Arr ∈ R l ×l and Asr ∈ R h×l are air gap permeance coefficients matrices
that we can written as:
⎡ l
⎤
⎢ 2G + ∑ G1,j − G sσ ··· − G sσ
sσ 0 0 0 ⎥
⎢ j =1 ⎥
⎢ ⎥
⎢ l ⎥
⎢ − G sσ 2G sσ + ∑ G2,j − G sσ ··· 0 0 0 ⎥
⎢ ⎥
Ass = ⎢ j =1 ⎥ (5)
⎢ .. .. .. .. .. .. .. ⎥
⎢ ⎥
⎢ . . . . . . . ⎥
⎢ ⎥
⎣ l ⎦
− G sσ 0 0 ··· 0 − G sσ 2G sσ + ∑ Gh,j
j =1
⎡ ⎤
h
⎢ 2G + ∑ Gi,1 − Grσ ··· − Grσ
rσ 0 0 0 ⎥
⎢ i =1 ⎥
⎢ h ⎥
⎢ − Grσ 2Grσ + ∑ Gi,2 − Grσ ··· ⎥
⎢ 0 0 0 ⎥
Arr = ⎢
⎢ i =1 ⎥
⎥ (6)
⎢ .. .. .. .. .. .. .. ⎥
⎢ . . . . . . . ⎥
⎢ ⎥
⎣ h ⎦
− Grσ 0 0 ··· 0 − Grσ 2Grσ + ∑ Gi,l
i =1
⎡ ⎤
G1,1 G1,2 ··· G1,l
⎢ G2,1 G2,2 ··· G2,l ⎥
⎢ ⎥
Asr = Ars
T
=⎢ . .. .. .. ⎥ (7)
⎣ .. . . . ⎦
Gh,1 Gh,2 ··· Gh,l
Also, Ψst ∈ R h×1 , Ψrt ∈ R l ×1, Ust ∈ R h×1 and Urt ∈ R l ×1 are stator and rotor teeth fluxes
vectors, stator and rotor magnetic scalar potentials vectors, respectively. These vectors can be
presented as:

T
Ψst = φ1st φ2st · · · φhst (8)

rt T
Ψrt = φ1 φ2rt · · · φlrt (9)

st T
Ust = u1 u2 · · · u h
st st (10)

rt T
Urt = u1 urt 2 · · · urt
l (11)
In Eq. (5), G sσ is linear permeances between each successive stator slots with constant
geometric permeability. This permeance can be obtained from (see Fig. 1):
Asσ
G sσ = μ0 (12)
L sσ
Similarly, for Grσ we can write:
Arσ
Grσ = μ0 (13)
Lrσ
where A , L , A , L and μ0 are cross section and length of stator and rotor slot opening
sσ sσ rσ rσ
and air permeability, respectively. Furthermore, mutual permeance between i th and jth slots
of the rotor and stator teeth depends to rotor mechanical angle (γi (t)), momentarily. Fig.2-a
(a) Geometric of the stator and rotor teeth
(b) Approximation function
Fig. 2. A typical air gap permeance between stator and rotor teeth
shows a typical geometry of the stator and rotor teeth. In this case, stator and rotor mutual
permeance can be approximated by:
⎧
⎪ Gmax 0 < γ < γt 2π − γt < γ < 2π
⎪
⎪
⎪
⎪ γ−γ
⎪
⎪ 1 + cosπ γ −γt
⎪
⎪ Gmax
⎨
t t
γt < γ < γt
Gi,j (γ ) = 2 (14)
⎪
⎪ γ −2π + γ
1 + cosπ γ −γ t
⎪
⎪
⎪
⎪
t t
2π − γt < γ < 2π − γ
⎪ Gmax
⎪ 2
⎪
⎩
0 γt < γ < 2π − γt
where γt and γ t are two mechanical angles which are depended to the stator and rotor teeth
geometry. For instance, in Fig. 2-a; γt and γ t can be obtained as:
wst + wrt + oss + osr

γt = (15)
Dag
|wst − wrt |
γt = (16)
Dag
where wst , wrt , oss and osr are dimensions of stator and rotor which are shown in Fig. 2-a and
Dag is :
Dsi + Dro
Dag = (17)
2
The maximum value of air gap permeance Gmax can be written as
l × min [ wst , wrt ]

Gmax = μ0 (18)
δ
where l and δ are lengths of machine and air gap.
3.2 Magnetic mesh equations

Eqs. (3) and (4) present l + h node equations which are shown relation between magnetic
potentials of teeth nodes with stator and rotor teeth fluxes. Attention to Fig. 1, two neighbour
stator teeth and yoke paths make a simple mesh in each stator slots, that sum of the magnetic
potentials in these mesh have to be zero. For instance, magnetic mesh equation in i th and
(i + 1)th stator teeth and yoke can be written as:
sy sy
i − u i −1 − i −1 φi −1 + i φi + i φi = Fi
u st st st st st st S
(19)
Similarly, mesh equation for i th rotor tooth and yoke with (i + 1)th rotor tooth can be obtained
as:
ry ry
i − u i −1 + i −1 φi −1 − i φi + i φi = Fi
urt rt rt rt rt rt r
(20)
Thus, magnetic mesh equations for all rotor and stator meshes can be expressed as:
Aust Ust + AΨsy Ψsy + A ϕst Ψst = F s (21)
Aurt Urt + Aφry Ψry + Aφrt Ψrt = Fr (22)

In Eqs. (21) and (22), Aφst ∈ R h×h and Aφrt ∈ R l ×l
are diagonal coefficients matrices of the
stator and rotor teeth reluctance that we can find as:
⎡ ⎤
B1st 0 0 ··· 0 − Bhst
⎢ − B1st B2st ⎥
⎢ 0 st · · · 0 0 ⎥
⎢ 0 − B2 st B3 ··· 0 0 ⎥
⎢ ⎥
⎢ . . . . . . ⎥
Aφst = ⎢ .. .. .. .. .. .. ⎥ (23)
⎢ ⎥
⎢ ⎥
⎢ 0 ··· 0 · · · Bhst−1 0 ⎥
⎣ ⎦

0 ··· 0 · · · − Bhst−1 Bhst
⎡ ⎤
− B1rt 0 0 ··· 0 Blrt
⎢ B1rt − B2rt 0 ··· 0 0 ⎥
⎢ ⎥
⎢ 0 B2rt − B3rt ··· 0 0 ⎥
⎢ ⎥
⎢ .. .. .. .. .. .. ⎥
Aφrt = ⎢ . ⎥ (24)
⎢ . . . . . ⎥
⎢ ⎥
⎢ 0 ··· 0 ··· − Blrt−1 0 ⎥
⎣ ⎦

0 ··· 0 ··· Blrt−1 − Blrt
also Aφsy ∈ R h×h and Aφry ∈ R l ×l are stator and rotor yoke reluctance coefficients matrices
which can be written as

sy sy sy
Aφsy = diag ( B1 ), ( B2 ), · · · , ( Bh ) (25)

ry ry ry
Aφry = diag ( B1 ), ( B2 ), · · · , ( Bl ) (26)
which Ψsy ∈ R h×1 , Ψry ∈ R l ×1 are stator and rotor yoke fluxes vectors and F s ∈ R h×1 and
Fr ∈ R l ×1 are stator and rotor ampere-turn vectors, respectively. These vectors are considered
as:

sy sy sy T
Ψsy = φ1 φ2 ··· φh (27)

ry ry ry T
Ψry = φ1 φ2 ··· φl (28)

T
Fs = F1s F2s ··· Fhs (29)

T
Fr = F1r F2r ··· Flr (30)
Also, Aust ∈ R h×h and Aurt ∈ R l ×l are constant matrices which are given by:
⎡ ⎤
1 0 0 ··· 0 −1
⎢ −1 1 0 ··· 0 0 ⎥
⎢ ⎥
⎢ 0 −1 1 ··· 0 0 ⎥
⎢ ⎥
Aust = Aurt = ⎢ .. .. .. .. .. .. ⎥ (31)
⎢ . . . . . . ⎥
⎢ ⎥
⎣ 0 0 0 ··· 1 0 ⎦
0 0 0 ··· −1 1
3.3 Teeth and yoke flux relations

For completing of the magnetic model of an induction machine, we need to find a relation
between teeth and yoke fluxes in mesh equations (Eqs. (21) and (22)). Attention to Fig. 1,
the relations between stator teeth and yoke fluxes can be obtained by applying magnetic flux
continuity principle in the yoke nodes. Thus, we have:
sy sy
φi = φist + φi+1 (32)
Similarly, relation between the rotor fluxes are:

ry ry
φj+1 = φrt
j + φj (33)
Eqs. (32) and (33) can be presented in matrix form as:
Ψst = Asyt Ψsy (34)
Ψrt = Aryt Ψry (35)

where Asyt ∈ R h×h and Aryt ∈ R l ×l can be written as
⎡ ⎤
−1 1 0 ··· 0 0
⎢ 0 −1 1 ··· 0 0 ⎥
⎢ ⎥
⎢ 0 0 −1 ··· 0 0 ⎥
⎢ ⎥
Aryt = −Asyt = ⎢ .. .. .. .. .. .. ⎥ (36)
⎢ . . . . . . ⎥
⎢ ⎥
⎣ 0 0 0 ··· −1 1 ⎦
1 0 0 ··· 0 −1
In a squirrel cage induction machine, a stator winding is not concentrated in single slot, but it
is distributed along air gap for harmonics reduction, full utilization of core and reduction of
mechanical stress to the winding. Thus, flux of stator coil equals the sum of fluxes of stator
teeth in the coil pitch. If three phase flux vector of the stator windings denotes as Ψ3φ , we can
find a matrix relation between stator teeth and windings fluxes as:
Ψ3φ = M c f Ψst (37)
where M c f ∈ R 3×h is a connected matrix which can be obtained based on the connection
diagram of stator windings.
4. Core saturation characteristic

Generally, magnetic cores of any electrical machines have non-linear characteristic curve
( B − H ), thus the elements of reluctance matrices in Eqs. (23) and (24) are depended to their
fluxes. For inserting the non-linearity characteristic of the magnetic core to mesh equations,
a non-linear permeability dependent to the core field density are used. For this purpose, a
non-linear permeability is defined as:
∂B
μ ( B ) = μ0 μr ( B ) = (38)
∂H
So, the reluctances of flux path in the stator and rotor cores can be written as:
L ki k = sy, st, ry, rt
Bik = (39)
μ ( Bik ) Aki i = 1, 2, . . .
where L ki and Aki are i th length and cross section of flux path in the stator and rotor cores,
respectively. Tangent, exponential and piecewise linear functions may be used to approximate
the saturation curve (Chen et al., 2005). In this chapter, the magnetic core permeability is
2.5

2.0

1.5
B Tesla

1.0

0.5

0.0

0 5000 10 000 15 000 20 000 25 000 30 000
A
H
m
Fig. 3. Core saturation curve
considered as:
μ (B) = b × e a× B
2
(40)
where a and b are two constants which can be chosen for the best fitting of the saturation curve
( B − H ) with this relation. By integrating Eq. (38) and combining with Eq. (40), the magnetic
field intensity H can be written as
√ √
1 πEr f a×B
H (B) = dB = √ (41)
μ (B) 2b a
where Er f is error function which is defined as: (Gautschi, 1964):

2
Er f ( B ) = e B dB (42)
For instance, Fig. 3 shows a saturation curve ( B − H ) for a silicon steel core that we can find
two constants in Eq. (40) for the best fitting of the saturation curve. In this case, the two
constants can be found as:
a = −0.8
(43)
b = 1000
5. Excitation vectors of reluctance network

For determining relations between three phase currents with ampere-turn in the stator
magnetic circuits, we give a contour between (i − 1)th and i th stator teeth, yoke and slot in
Fig. 1. Also, we suppose that combining of three phase conductors are placed in the stator
slots. Therefore, by applying Ampere’s law to this contour, (i )th ampere-turn of the stator
magnetic circuits ( Fis ) in Eq. (21) can be expressed as:
Fis = wia i a + wbi ib + wci ic (44)

Fig. 4. Single layer distributed winding with 24 slot
where wia , wbi and wci are conductor numbers of phase a, b and c in the i th stator slot,
respectively. Values and signs of these parameters will depend to connections and directions
of windings in the i th slot. For instance, Fig. 4 shows a single layer three phase winding with
two poles and 24 slots, which can generate a rotating magnetic field in the air gap. In this
case, relation between ampere-turn of the stator magnetic circuits and stator currents can be
determined as: ⎡ s ⎤ ⎡ ⎤
F1 −wa 0 0
⎢ F s ⎥ ⎢ −wa 0 0 ⎥
⎢ 2 ⎥ ⎢ ⎥
⎢ Fs ⎥ ⎢ 0 0 wc ⎥ ⎡ ⎤
⎢ 3 ⎥ ⎢ ⎥ ia
⎢ . ⎥ ⎢ . . . ⎥
⎢ . ⎥=⎢ . .. .. ⎥ ⎣ ⎦
⎢ . ⎥ ⎢ . ⎥ ib (45)
⎢ s ⎥ ⎢ ⎥ i
⎢ F22 ⎥ ⎢ 0 w b 0 ⎥ c
⎢ s ⎥ ⎢ ⎥
⎣ F23 ⎦ ⎣ − w a 0 0 ⎦
s
F24 −wa 0 0
Thus, the stator ampere-turn vector (Fs ) in Eq. (21) may be presented as
Fs = Wc is (46)
where W c ∈ R h×3 , is a connection matrix that shows number of conductors in all of the stator
slots. Also, is is a vector of three phase currents that we can arrange as:

T
is = ia ib ic (47)
In single layer winding, each slot can carry only one phase current, thus rows of the connection
matrix in Eq. (46) have only one non-zero element. These non-zero elements can be
determined by attention to stator winding topology.
In other respect, currents of rotor bars in squirrel cage induction machines are equal to
current of conductors in the rotor slots. Therefore, similar to above mentioned method; the
ampere-turn in rotor slots (Fr ) in Eq. (22) can be determined by:
Fr = i b (48)
where ib , the rotor bars currents vector is defined as

T
ib = i1b i2b ··· ilb (49)
Substituting Eqs. (34), (35), (46) and (48) into (21) and (22) and combining with Eqs. (3) and
(4), the magnetic algebraic equations of squirrel cage induction machines in matrix form can
be augmented as
⎡ ⎤⎡ ⎤ ⎡ ⎤
Aφsy + Aφst Asyt 0 Aust 0 Ψsy Wc is
⎢ Aφry + Aφrt Aryt Aurt ⎥ ⎢ ⎥ ⎢ ib ⎥
⎢ 0 0 ⎥⎢ Ψry ⎥=⎢ ⎥ (50)
⎣ − Asyt 0 Ass Asr ⎦⎣ Ust ⎦ ⎣ 0 ⎦
0 Aryt Ars Arr Urt 0
Due to core saturation characteristic is a non-linear curve, the matrix Eq. (50) is non-linear and
some coefficient matrices depends to the core fluxes density. Thus, ordinary methods cannot
be employed for solving Eq. (50). Furthermore, rotor and stator currents are depended to
stator three phase source voltages and rotor speed in differential equation forms. Therefore,
for detailed analysis of squirrel cage induction machine, it is necessary to solve an electric,
mechanic and magnetic algebraic differential equations, simultaneously
6. Electrical voltage equations of squirrel cage induction machines

Generally, a set of differential equations in an electrical machine is used to describe rates of
the electrical and mechanical variables. These equations establish the relationship between
fluxes, currents and the three phase source voltage variables. In the next section, rotor and
stator voltage relations are be derived.
6.1 Stator voltage equations

For an induction machine, we can write electrical differential equations in stator windings as:
d
Vs = Rs is + Λs (51)
dt
where Vs and Rs are voltage and stator resistances matrix, respectively which are defined by

T
Vs = v a vb vc (52)

Rs = diag r a rb rc (53)
Moreover, Λs is linkage flux vector and equals to the product of turn number of stator
windings and the phase fluxes. Thus, we can write:

T
Λs = λa λb λc (54)
Λs = wc Ψ3ϕ (55)
By substituting Eqs. (34) and (37) into (55), we obtain:
Λs = wc M c f Asyt Ψsy (56)

Fig. 5. Structure of squirrel cage rotor
6.2 Rotor voltage equations

Fig. 5 shows a structure of squirrel cage rotor in an induction machine. The rotor topology has
l + 1 mesh, thus we can write l + 1 independent differential equations to describe electrical
dynamic of rotor variables. Suppose, the rotor rings segments have a symmetrical structure,
thus number of l independent equations will be enough for modelling of the rotor dynamics.
According to Fig. 1, j th tooth of rotor is surrounded by jth and ( j + 1)th rotor bars. Thus, based
on currents directions of bars and ring segments which are shown in the Fig. 4; the voltage
equation in this loop can be given by:
d rt
φ = −2irj+1 rrj+1 − i bj r bj + i bj+1 r bj+1 (57)
dt j
This relation can be rearranged in the matrix form and the electrical differential equations for
the rotor cage can be obtained as
d
Ψrt = −2Jr Rr ir − Aryt Rb ib (58)
dt
where the vector ir ∈ R l ×1 ,Rb ∈ R l ×l and Rr ∈ R l ×l denote ring segment currents, rotor bar
and ring segment resistance matrices, receptively. These vectors are given by:

T
ir = i1r i2r ··· irl (59)

Rr = diag r1r r2r ··· rrl (60)

Rb = diag r1b r2b ··· rlb (61)
Also, constant matrix Jr ∈ R l ×l is defined as

⎡ ⎤
0 1 0 ··· 0 0
⎢ 0 0 1 ··· 0 0 ⎥
⎢ ⎥
⎢ 0 0 0 ··· 0 0 ⎥
⎢ ⎥
Jr = ⎢ .. .. .. .. .. .. ⎥ (62)
⎢ . ⎥
⎢ . . . . . ⎥
⎣ 0 0 0 ··· 0 1 ⎦
1 0 0 ··· 0 0
Moreover, there are two current vectors in left hand side of Eq. (58) which have to find a
relation between them. For this propose, Kirchhoff’s Current Law (KCL ) can be used to
determined a relation between rotor bars and ring segments currents. By applying KCL to
jth node (bar) in the rotor cage (Fig. 4), we can write:
i bj = irj+1 − irj (63)
Therefore, for l nodes in the rotor cage; Eq. (63) can be expressed in matrix form as:
ib = Aryt ir (64)
By substituting Eqs. (64) into (58), we can get:
d
Ψrt = −2Jr Rr + Aryt Rb Aryt ir (65)
dt
Up to this stage, two separate sets of non-linear equations are derived to established of
algebraical (magnetic) and dynamicalal (electric) models of a squirrel cage induction machine.
By combining of Eqs. (50), (51) and (65), the total algebria-differential equations of the system
can be augmented as:
⎡ ⎤⎡ ⎤ ⎡ ⎤
wc M c f Asyt p 0 0 0 Rs 0 Ψsy Vs
⎢ 0 A p 0 0 0 2Jr Rr − Aryt Rb Aryt⎥⎢ Ψry ⎥ ⎢ 0 ⎥
⎢ ryt ⎥⎢ ⎥ ⎢ ⎥
⎢Aφsy + Aφst Asyt 0 A 0 −Wc 0 ⎥⎢ Ust ⎥ ⎢ 0 ⎥
⎢ ust ⎥⎢ ⎥=⎢ ⎥ (66)
⎢ 0 A + A A 0 Aurt 0 Aryt ⎥⎢ Urt ⎥ ⎢ 0 ⎥
⎢ φry φrt ryt ⎥⎢ ⎥ ⎢ ⎥
⎣ −Asyt 0 Ass Asr 0 0 ⎦⎣ is ⎦ ⎣ 0 ⎦
0 − Aryt Ars Arr 0 0 ir 0
where p denotes time derivative operator ( dt

d
).
6.3 Mechanical differential equations

Some of the matrix coefficients in Eq.(66) are depended to mechanical instantaneous angle
γ. In variable speed conditions, the mechanical variables of the system can be determined
by solving differential equations of the rotor angel and speed. Generally, mechanical torque
balance equation can be expressed as:
dω
J = Tem − Tl (67)
dt
Quantity Symbol Value

Power P 1.1kw
Voltage V 220V
Frequency f 50Hz
Number of pole p 2
Stator resistance rs 5Ω
Rotor bar resistance rb 20μΩ
Rotor ring sector resistance re 1.1μΩ
Inertia moment j 0.02kgm2
Number of turns per slot 68
Number of rotor slots 18
Table 1. Parameters of squirrel cage induction machine
in which
dγ
=ω (68)
dt
also, ω, J, Tl and Tem are the rotor angular speed, total inertia on the shaft, load torque and
electromagnetic torque, respectively. The electromagnetic torque is depended to mmf along
air gap and derivative of air gap permeances with respect to rotor angel (γ ) (Ostovic, 1989).
For an induction machines, we can express as:
h l 2 dG (γ )
i,j
Tem = ∑∑ i − uj
u st rt
dγ
(69)
i =1 j =1
By substuting Eqs.(7), (10) and (11) into (69), air gap electromagnetic torque can be obtained
as: d
Tem = Ust T
− Urt
T
Asr (Ust − Urt ) (70)
dγ
Therefore, the mechanical non-linear differential equations (67) to (70), with the magnetic
non-linear differential-algebraic equations (66), describe the generalized non-linear dynamic
model of squirrel cage induction machine. Because of non-linearity of the model, the
advantage numerical solution methods must be used. In the next section, some simulation
results are presented for demonstration of capability validation of the new model. After that,
asymmetrical situations of squirrel cage induction machine such as broken rotor bar fault with
saturation effects are analysed and evaluated.
7. Simulations and experiment of results

Table 1 shows the electrical parameters of the squirrel cage induction machine which is used
to obtain numerical simulation and experimental results. Air gap, rotor and stator slots and
other main dimensions of the induction machines are presented in Fig. (5) and windings
connection diagram are assumed similar to Fig. (4). In the next section, some different
dynamical conditions are implemented to obtain numerical results of the non-linear model
of induction machines.
Fig. 6. Basic construction and main diameters and stator and rotor slot shapes
7.1 Simulation in healthy condition

For simulation study, the healthy conditions of a squirrel cage induction machine with
saturable iron core are given. At first, free starting-up is considered and at t = 0.3sec, a
mechanical load (Tl = 10Nm) is applied to the rotor shaft. The results of the simulations
are shown in Figs. 7 and 8. Speed-torque transient acceleration of the system is presented in
Fig. 7-a, that rotor speed reaches to steady state value at about t = 0.2sec. Also, transient
electromagnetic torque are shown in Fig. 7-a where low order harmonics are appeared in the
electromagnetic torque on the transient duration. This is a common behaviour in the induction
machines, which can be obtained from classical method such as two axes theory. But the high
order harmonics in transient and steady state are also appeared in the results. At t = 0.3sec
load torque changes form zero to Tl = 10N.m, electromagnetic torque increase by decreasing
of the rotor speed. Fig. 7-b shows dynamic of three phase stator currents which decrease
by increasing the rotor speed. Other dynamic performance for flux and rotor bar currents
are shown in Fig. 8. The results show that, high order harmonics are observed in the rotor
and stator teeth fluxes and currents. The rotor teeth and slots are moved opposite the stator
teeth and slots, thus radial fluxes will have variable permeances in the air gap. This is caused
a slot harmonics appears in the the all variables of the system. Fig 8-a illustrates that, slot
harmonics deform the stator teeth flux in the transient and steady state conditions. Because of
the machine is symmetric, all variables of machine such as the stator currents and stator teeth
flux are balanced.
7.2 Simulation in faulty condition

In this simulation , analysis of the induction machine with a broken bar is done to determined
the steady state performances of the machine. Fig. 9-a shows the steady state current of the
stator winding with linear and non-linear core characteristics. The rotor speed and torque
are 307.7rad/sec and 10N.m, respectively. Similar to the last conditions, slot harmonics are
appeared on the stator currents in the both causes (linear and non-linear magnetic core).
However, stator current has low order harmonics with the non-linear iron core. Stator teeth
fluxes in two cases are approximately equal. Also, Fig. 9-c presents that the bar currents near
to the broken bar are strongly changed. But attention to Fig 9-d, this situation is not appeared
in induction machine with non-linear magnetic core. Therefore, disturbances of rotor broken
bar in the saturated machines cannot effectively transfer to the stator currents.
70.
300
60.
250
Ωr Radsec
50.
T em N.m
200 40.
150 30.
20.
100
10.
50 0.
0 10.
0.0 0.1 0.2 0.3 0.4 0.5 0.6
Time Sec
(a) Torque-speed dynamic
40 40.
20 20.
is A
0 0.
20 20.
40 40.
0.0 0.1 0.2 0.3 0.4 0.5 0.6
Time Sec
(b) Stator currents
4 4.
AG mweb
2 2.
0 0.
2 2.
4 4.
0.0 0.1 0.2 0.3 0.4 0.5 0.6
Time Sec
(c) Air gap flux
Fig. 7. Starting-up transient of induction machine
8. Conclusion
In this chapter, a generalized non-linear dynamical model of the squirrel cage induction
machines was presented. In this modelling, magnetic saturation effects in iron core, space
harmonic distributed of fluxes in the rotor and stator teeth, the stator windings and rotor bar
distribution were considered. Some simulations results showed that, presented model has
high accuracy and efficiency for asymmetrical analysis such as broken bar conditions. The
effects of broken bar will not clearly appear when the machine core is saturated.
0.6 0.6
0.4 0.4
st mweb
0.2 0.2
0.0 0.
0.2 0.2
0.4 0.4
0.6 0.6
0.0 0.1 0.2 0.3 0.4 0.5 0.6
Time Sec
(a) Stator teeth Fluxes
4 4.
2 2.
i b KA
0 0.
2 2.
4 4.
0.0 0.1 0.2 0.3 0.4 0.5 0.6
Time Sec
(b) Rotor bar currents
300 40.
200
20.
v as Volts
100
ias A
0 0.
100
20.
200
300 40.
0.0 0.1 0.2 0.3 0.4 0.5 0.6
Time Sec
(c) Stator voltage and current
1.0 0.6
0.4
0.5
st mweb
as web
0.2
0.0 0.
0.2
0.5
0.4
1.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6
Time Sec
(d) Tooth and phase fluxes
Fig. 8. Starting-up transient of induction machine

20 20.
ia Nonlinear core A

15.
ia Linear core A
10 10.
5.
0 0.
5.
10 10.
15.
20 20.
0.00 0.02 0.04 0.06 0.08 0.10
Time Sec
(a) Stator current
Φst NonLinear coremweb

0.6 0.6
Φst Linear coremweb
0.4 0.4
0.2 0.2
0.0 0.
0.2 0.2
0.4 0.4
0.6
0.00 0.02 0.04 0.06 0.08 0.10
Time Sec
(b) Stator flux
0.6
0.4
0.4
0.2
0.2
Time Sec,
Time Sec,
0.0 0.0
0.2
0.2
0.4
0.4
0.6
0.5 1.0 1.5 2.0 0.5 1.0 1.5 2.0
mweb Φrt mweb Φrt
(c) Rotor fluxes with linear core (d) Rotor fluxes with non-linear core
Fig. 9. Steady state fluxes and currents of faulty induction machine with a broken bar for
linear and non-linear core
9. References
Alemi, P. & Nazarzadeh, J. (2006). The Induction machines modeling based on bars current
for rotor fault analysis, 21th International Power System Conference, PSC 2006, 477-483,
Niroo Research Institute, 13-15 Nov. 2006, Tehran, Iran
Al-Shahrani, A. S. (2005). Influence of adjustable speed drive on induction motor fault detection
using stator current monitoring. Ph.D. thesis, Oregon State University
Bellini, A.; Filippetti, F.; Franceschini, G.; Tassoni, C. & Kliman G. B. (2001). Quantitative
evaluation of induction motor broken bars by means of electrical signature analysis.
IEEE Transactions on Industry Applications, Vol. 37, No. 5, (Sep./Oct. 2001), 1248-1255
Benbouzid, M. E. H. (2000). A review of induction motors signature analysis as a medium for
faults detection.IEEE Transactions on Industry Electronics, Vol. 47, No. 5, (Oct. 2000),
984-993
Chen, S. D.; Lin, R. L & Cheng, C. K.(2005). Magnetizing inrush model of transformers based
on structure parameters.IEEE Transactions on Industry Electronics, Vol. 20, No. 3, (July
2005), 1947-1954
Faiz, J.; Sharifian, M. B. B.; Feyzi, M. R.; & Shaarbafi, K. (2002). A complete lumped equivalent
circuit of three-phase squirrel-cage induction motors using two-dimensional
finite-elements technique. IEEE Transactions on Energy Conversion, Vol. 17, No. 3,
(Spet. 2002), 363-367
Gautschi W. (1964). Error function and Fresnel integrals, Handbook of Mathematical Functions, NBS
Appl. Math. Series, Vol. 55, U.S. Government Printing Office, Washington, D.C.
Jeong, J. H.; Lee E. W. & Cho, H. K. (2003). Analysis of transient state of the squirrel cage
induction motor by using magnetic equivalent circuit method, Sixth International
Conference on Electrical Machines and Systems, Vol. 2, 720-723, 2003
Jung, J. H.; Lee, J. J. & Kwon, B. H. (2006). Online diagnosis of induction motors using MCSA.
IEEE Transactions on Industry Electronics, Vol. 53, No. 6, (Dec 2006), 1842-1852
Krause, P. C.; Wasynczuk, O. & Sudhoff, S. D. (2002). Analysis of Electric Machinery and Drive
Systems (2nd Edition), Wiley-IEEE Press, ISBN 047114326X, New York.
Luos, X.; Liao, Y.; Toliyaty, H.; El-Antably, A. & Lipos, T. A. (1995). Multiple coupled circuit
modeling of induction machines. IEEE Transactions on Industry Applications, Vol. 31,
No.2, (Mar./Apr. 1995), 311-318
Mirafzal, B.& Demerdash, N. A. O. (2004). Induction machine broken-bar fault diagnosis
using the rotor magnetic field space-vector orientation. IEEE Transactions on Industry
Applications, Vol. 40, No. 2, (Mar./Apr. 2004), 534-542
Mirafzal, B.& Demerdash, N. A. O. (2008). Induction machine broken bar and stator
short-circuit fault diagnostics based on three-phase stator current envelopes. IEEE
Transactions on Industry Applications, Vol. 55, No. 3, (March 2008), 1310-1318
Muñoz, A. R. & Lipo, T. A. (1999). Complex vector model of the squirrel-cage induction
machine including instantaneous rotor bar currents. IEEE Transactions on Industry
Applications, Vol. 35, No. 6, (Nov./Dec. 1999), 1332-1340
Ostovic, V. (1989). A novel method for evaluation of transient states in saturated electric
machine. IEEE Transactions on Industry Applications, Vol. 25, No. 1, (Feb. 1989), 96-1000
Ostovic, V. (1989). Dynamics of Saturated Machines, Springer-Verlag, ISBN 0387970797, New
York
Siddique, A.; Yadava, G. S. & Singh, B. (2005). A review of stator fault monitoring techniques
of induction motors. IEEE Transactions on Energy Conversion, Vol. 20, No. 1, (March
2005), 106-114
Sprooten, J. (2007). Finite element and electrical circuit modeling of faulty induction machines study
of internal effects and fault detection techniques. Ph.D. thesis, Department of Bio, Electro
and Mechanical Systems (BEAMS), University Libre de Bruxelles
Su, H. & Chong, K. T. (2007). Induction machine condition monitoring using neural network
modelling. IEEE Transactions on Industry Applications, Vol. 54, No. 1, (Feb. 2007),
241-249
4
Minimization of Losses in Converter-

Fed Induction Motors – Optimal Flux Solution
Waldiberto de Lima Pires, Hugo Gustavo Gomez Mello,
Sebastião Lauro Nau and Alexandre Postól Sobrinho
WEG Equipamentos Eletricos S.A. – Motores
Research and Development of Product Department
Av. Pref. Waldemar Grubba, 3000 – malote 41
Jaraguá do Sul, SC - 89256-900
Brazil
1. Introduction
When a TEFC induction motor fed by static frequency converter drives a load which
demands constant torque throughout the operation range, the low speeds are thermally
critical, because the motor losses (heat sources) do not vary much as a function of the speed,
but the ventilation efficiency decreases as the operation speed falls down, since the fan is
installed on the very motor shaft. In such cases, when operating at low speeds, the
temperature rise often exceeds the limits of the motor thermal class due to the lack of
cooling. In order to prevent this problem the industry has traditionally adopted one of the
following solutions: independent ventilation (a small auxiliary motor is used to exclusively
drive the fan that provides the main motor cooling) or oversizing (the motor used in the
application provides a higher torque than the rated load demand). However, neither one
nor the other of these two options are attractive, as both, besides increasing the space
required for the installation, still increase the motor price [1].
Frequency converters usually apply to the motor a constant voltage/frequency ratio
throughout the operation range, so that no loss control is provided. But the study on the
motor losses composition and its relation with voltage, frequency, magnetic flux and
current, allied with the study on the influence of the ventilation on the temperature rise of
the motor, has led to an optimal voltage/frequency ratio, which minimizes the total motor
losses at each speed [2]. This way, by implementing the automatic control of the
voltage/frequency ratio in the converter, the motor loss minimization can be automatically
obtained throughout the frequency range, so that the motor temperature rise is kept within
the thermal class limits even at low speeds with reduced ventilation.
2. Determination of losses
The fast growth of the number of industrial applications using static frequency converters
recently observed in variable speed drives has encouraged the meticulous study of losses in
magnetic materials under PWM supply by several researchers [3, 4, 5, 6, 7, 8]. They have
shown that such losses depend on a number of control parameters, such as the modulation
index, the number of levels and pulses of the frequency converter and the duration of PWM
signal pulses. On the other hand, Boglietti et alli have concluded that the flux waveforms
resulting from PWM supply differ from those resulting from sinusoidal supply just except
for a small ripple, which depends on the switching frequency, and that above approximately
5 kHz the iron losses can be considered independent of this parameter [9]. Such studies
represent the first step towards the understanding of the losses behavior in electric motors
under PWM supply, that involves a higher degree of complexity and is not restricted
exclusively to the magnetic materials issue, but includes also additional losses in the
conductors and due to the cooling system and depends, besides the control parameters, on
some machine design parameters, such as the flux density, the lamination geometry and the
connection of windings [3, 10, 11, 12], as well as on other variables inherent to the
manufacturing process [13].
For the purposes of this study, however, the analysis of the motor losses can be simplified,
so that it is enough to separate the total motor losses in three key components:
P = Pfe + Pj + Pmec (1)

where:
Pfe – Iron losses, which depend on the flux density (or magnetic induction), the frequency
and the quality of the magnetic material.
Pj – Losses by Joule effect, which depend on the currents flowing through the stator
windings and the rotor bars.
Pmec – Mechanical losses due to cooling system (fan coupled to the shaft) and friction,
which depend on the speed.
The iron losses are classically considered as being composed of two portions: Hysteresis losses
(pH) and induced eddy current (Foucault) losses (pF). For a lamination sample tested in
Epstein Frame with sinusoidal supply, the hysteresis losses are directly proportional to the
frequency (f) and to the square of the magnetic induction (B2), while the eddy current losses
are proportional to the square of both the frequency (f2) and the magnetic induction (B2),
especially for induction values above one Tesla (1 T). However, in the induction motor the iron
losses present a much more induction-dependent behavior than the quadratic ratio obtained
with normalized samples of the magnetic material in Epstein Frame tests. In low-voltage three-
phase induction motors manufactured with fully processed steel laminations tested under
different saturation levels, iron losses presented a dependance on the induction close to B4 for
inductions above 1,2 T (usual value for industrial motors), as presented in Fig. 1.
The fundamental theory of the electric machines shows that the torque provided by the
induction motor is directly proportional to the product between the magnetic flux and the
electric current [14, 15]. Then in order to keep a constant torque, if the flux increases the
current can decrease (and vice-versa). As the Joule losses are directly proportional to the
square of the current, these losses can be considered as inversely proportional to the square
of the magnetic flux. From the Faraday-Lenz law of induction, one can easily demonstrate
that the magnetic flux in the motor is directly proportional to the ratio between the
electromotive force (E) and the frequency. Considering the steady-state model of equivalent
circuit of the induction motor per-phase (Fig. 2), it can be noted that at the base frequency
the voltage drop in the primary impedance has little significance, so that the flux can be
considered as proportional to the V1/f (voltage/frequency) ratio.
Minimization of Losses in Converter-Fed Induction Motors – Optimal Flux Solution 63
3,25
Motor 30kW IV poles 400V 50Hz
3,00
2,75
2,50
pfe (p.u.)
2,25
pfe = (E/f)4
2,00
1,75
1,50
1,25
1,00
1,00 1,05 1,10 1,15 1,20 1,25 1,30
E/f (p.u.)
Fig. 1. Iron losses x Magnetic induction for an industrial three-phase induction motor
Fig. 2. Steady state equivalent circuit of the induction motor per-phase

For low operating frequencies, however, in which the input voltage is reduced, the voltage
drop in the primary resistance becomes important and can be no longer despised. By
neglecting the influence of the primary reactance, electromotive force E is given by
(2)
The voltage drop at the stator branch (ΔV) then depends directly on the stator current (I1).
As Fig. 2 shows, the motor current can be decomposed into two components: one
concerning magnetization and the other concerning torque production.
(3)
Taking rated voltage as the base, E/f ratio per unit can be written as:
(4)
where:
ΔVn - Voltage drop per unit with rated frequency and load.
f – Motor operating frequency per unit, considering rated frequency fn as the base.
In (4), the square root results in a correction factor, which is function of the motor current
and whose terms are explained in the following paragraphs.
It should be taken into account that, as the frequency (and consequently, the rotation) is
reduced, the mechanical losses decrease in a nearly cubic proportion to it (f3). The
mechanical losses do not affect the iron losses, but they act as an additional load to the
motor, therefore they must be considered as a torque to be added to the rated torque
available on the shaft. Its reduction implies current reduction and so reduction of Joule
losses in the conductors (Pj).
Thus it is possible to rewrite the induction motor total losses per unit p, for operation with
variable voltage and frequency, as follows:
(5)
where:
pin – Total Joule losses with the motor operating at rated conditions of load, voltage and
frequency.
pHn – Total hysteresis losses with the motor operating at rated conditions of load, voltage
and frequency.
pFn – Total eddy current losses with the motor operating at rated conditions of load, voltage
and frequency.
Motors manufactured with low loss magnetic core (fully processed silicon steel) operating at
rated conditions typically present values of 80%, 12% and 8%, for parameters pin, pHn and
pFn, respectively. The remaining parameters of (5) will be opportunely explained ahead.
The term of (5) in brackets refers to the motor Joule losses and depends on the total motor
current. The second and third terms refer to the motor iron losses for hysteresis and eddy
currents, respectively. The magnetic induction was conveniently replaced by the E/f ratio.
There is no explicit term for the motor total mechanical losses in the equation, because they
are embedded in the first term, in accordance with what was mentioned before, by means of
the factor km defined below:
(6)
where pmn is the mechanical losses at rated speed referred to the rated output power Pn.
The aim of this study is to minimize the motor losses, in order to reduce its temperature rise,
so that the need of both the torque reduction (oversizing) and the use of independent
ventilation can be prevented. In (5), this is considered by means of the derating factor kT,
which will be addressed later on this paper. It should be noted that the torque affects only
the current-dependent losses, not influencing the iron losses. Therefore, km and kT are
torque correction factors required to compensate for the effects of the speed variation, which
influences the portion of losses related to the load current.
Ki0n is the no-load current factor, defined by (7).
(7)
where:
Io – No-load current under rated voltage and frequency.
In – Full-load current under rated voltage and frequency.
Due to the non-linearity of the magnetization curve of the laminations, the E/f ratio increase
causes the no-load current to increase according to (8). This peculiar behavior of the no-load
current was observed experimentally (Fig. 3), and is taken into account in (5) by means of
the factor ki0.
(8)
Fig. 3. Magnetizing current x E/f

3. Minimization of losses
The analysis of (5) shows that the induction motor global losses depend on both the
operation frequency and the induction (or magnetic flux). Then the values of V/f that
minimize the motor global losses change with the operation frequency, so that it is necessary
to find the minimum losses at each frequency, with different values of V/f. Fig. 4 shows the
total losses calculated as function of the frequency for various values of the V/f ratio.
Fig. 4. Total losses x frequency curve for several V/f ratios at rated torque
Fig. 5 derives from the family of curves above and represents the V/f ratios theoretically
obtained, which minimize the total losses of the motor at each operation frequency.
Fig. 5. V/f x frequency curve for minimization of total losses

4. Influence of ventilation reduction

Thermal calculations implemented for a number of motors of distinct frame sizes and power
ratings considering speed variation, combined with experiments and tests performed with
several motors at rated load, varying separately the fan speed from zero to base speed, led
to the conclusion that TEFC three-phase motors of a wide output range present a similar
thermal behavior. Fig. 6 represents the temperature rise per unit of low-voltage 4-pole cage
induction motors manufactured with die cast iron frame as a function of the fan speed per
unit.
Fig. 6. Temperature rise x fan speed at rated load

In this approach, for each desired value of frequency and for each value of V/f ratio
according to Fig. 5, the total losses p are calculated according to (5). Fig. 6 determines the
influence of the ventilation reduction on the motor temperature rise (ΔTmotor). So in order
for the required motor temperature rise to be assured, it is necessary to calculate a new
value of p, henceforth referred to as p’, according to (9).
(9)
where:
p’ – total motor losses for the required temperature rise, considering the ventilation
reduction.
pn - total motor losses at rated conditions.
If, for instance, the maximum required temperature rise is the limit of the insulation system
thermal class, then ΔT = ΔTclass. If, otherwise, the maximum required temperature rise is
ΔTn, then ΔT=1.
Once known the total losses p’ that will cause the required temperature rise in the motor with
reduced ventilation, then it is possible to calculate the convenient derating factor using (10):
(10)
where kHVF is the harmonic voltage factor as defined by NEMA [16]. It was placed in the
equation originally conceived as (5) for the influence of the PWM supply voltage harmonics
to be also considered on the motor temperature rise. For most of the modern static
frequency converters kHVF is 0.95.
As a consequence of the cooling reduction, kT is usually lower than 1. However, if the
minimized total losses are such that reduce the motor temperature rise even with poor
ventilation, kT can be higher than 1. Similarly, if kT is calculated for an insulation class
temperature rise, it will be normally higher than 1 if the rated motor temperature rise is
much below the insulation class temperature rise limit.
Fig. 7 presents an example of loss reduction achieved with the proposed method. A three-
phase, 30 kW, 4-pole induction motor was tested at constant rated torque within the
frequency range from 0.1 to 1.0 (p.u.). The results are presented for three different situations:
calculation with constant flux, calculation with optimal flux and testing with optimal flux.
The loss reduction obtained with the proposed technique, as shown in Fig. 7, results in a
better thermal performance of the motors operating with optimal flux. Comparing the motor
temperature rises when operating at constant flux condition to those when operating at
optimal flux condition, it is remarkable a behavior similar to that outlined in Fig. 8, as can be
checked in the experimental results presented in section VI.
Fig. 7. Total losses (p.u.) x frequency (p.u.)

ΔTn: temperature of thermal stabilization of the motor fed by sinusoidal supply (mains)
ΔT0: temperature of thermal stabilization of the motor fed by converter supply
fn: operation base frequency of the motor
Fig. 8. Temperature rise x operation frequency, sketch of the motor performance under
different flux conditions
5. Validation and implementation of the optimal flux curve

To validate the proposed technique, temperature rise tests with speed/ventilation variation
were accomplished for a wide range of industrial motor ratings. This way, it was possible to
compare the thermal performances of converter-fed motors when under constant flux (rated
losses) and optimal flux (minimum losses) conditions.
Before the implementation of the automatic function for optimal flux selection by the
converter, drives with suitably modified softwares were used, so that specific flux values
could be manually adjusted. This allowed the practical correction of the curve obtained by
means of mathematical calculations and the finding of the actual optimal flux curve that was
implemented in the converters used in the tests, into which were incorporated the automatic
function for the optimal flux setting.
As the sensorless vector control enables the magnetic flux of the motor to be directly altered,
this was the control type employed in all tests with converter. The switching frequency used
in all tests with converter was 2.5 kHz.
A. NEMA High Efficiency (NHE) Motors
The motors to be tested were selected considering the worst horsepower/frame size ratios
criteria. The following machines were used in the tests, all of them 4-pole (predominant
polarity in low-voltage industrial applications) and all of them with class F insulation: 5 hp
(NEMA 184T); 20 hp (NEMA 256T); 50 hp (NEMA 326T) and 150 hp (NEMA 444T).
Occasionally, for investigation of specific issues, tests were also realized with motors that
are not related above.
It should be noted that the loss minimization technique was conceived and developed
specifically for low operation frequencies (below 0.5 p.u.). Thus, at base (rated) frequency, in
this case 60 Hz, the flux value for the minimum losses condition is the rated flux. The tests
were conducted with full load being continuously applied until the motor thermal
stabilization.
B. NEMA Premium Efficiency (NPE) Motors
Considering that NEMA Premium Efficiency motors tend to present lower temperature rises
than NEMA High Efficiency motors, few NPE motors were tested just for the effectiveness
of the proposed loss minimization technique to be corroborated with such motors. Despite
that tests were conducted with 2- and 4-pole motors of very distinct horsepower rates, in
order to extend the solution also for machines of other polarities. The following motors were
tested: 5 hp (NEMA 184T) and 150 hp (NEMA 444T).
The tests were conducted in a way similar to the followed for NHE motors: continuous
application of full load until the thermal stabilization of the motor.
C. EFF1 (IE3) Motors
The several motors tested were chosen according to the criterion of the most critical frame
size/horsepower ratios. They were all 4-pole machines, since this polarity is typical in low-
voltage applications involving speed variation. Motors with base frequencies of both 50 and
60 Hz were tested, for a wider verification of the proposed solution, as listed below:
- fbase = 60 Hz: 3 hp (IEC 90L), 12.5 hp (IEC 132M), 50 hp (IEC 200L), 75 hp (IEC 225SM) and
150 hp (IEC 280 SM).
- fbase = 50 Hz: 2hp (IEC 90L), 10 hp (IEC 132M), 40 hp (IEC 200L), 75 hp (IEC 225SM) and
150 hp (IEC 280 SM).
Regarding the temperature rise tests, it should be noted that at 50 or 60 Hz the motors were
always fed by the mains (AC power line), that is, sinusoidal supply was used, while all the
other tests were accomplished with the motors fed by converter in optimal flux condition for
loss minimization. Due to the fact that EFF1 (IE3) motors usually present lower efficiency
levels than NHE and NPE motors, they tend to operate a little bit warmer than the latter.
Because of that, unlike the tests performed with NHE and NPE motors, sometimes derating
factors were applied in the tests with EFF1 (IE3) motors at low frequencies. However, in all
cases the tests were performed following S1 duty as well.
6. Experimental results
The results contained in Tables I and II evidence the effectiveness of the proposed loss
minimization technique for NEMA High Efficiency and NEMA Premium Efficiency motors:
all the motors tested with optimal flux got the thermal stabilization at a lower temperature
than with constant flux, at all frequencies analysed. Furthermore, Tables I and II show that,
when operating at minimum loss condition, NHE and NPE motors can provide the rated
torque continually throughout the operation range, even at low frequencies.
Tables I and II present some lacking results of low frequency (typically 5 Hz or below) tests.
In these cases, the respective temperature rise test at full load could not be concluded, either
because the motor was already running too hot, forcing the interruption of the test before
thermal stabilization was reached, or because the dynamometer could no longer apply the
required test load after a period, due to its functioning principle (braking/loading caused by
the Foucault effect) and given the increased slip of the motor when it is heated. Temperature
rise tests that could not be concluded properly with constant flux, but could be normally
performed in the optimal flux condition, are themselves evidences of the thermal
performance improvement of the motor, provided by the loss minimization technique.
Table I. Temperature rise tests results (K) 4-pole nhe motors – s1 duty
*Tests that have not been performed

Table II. Temperature rise tests results (K) npe motors – s1 duty
The analysis of the data presented in the tables above by means of graphs makes the
interpretation of results easier. Fig. 9 evidences the advantages of using the optimal flux
technique with a NHE, 20 hp, 4-pole motor.
Some results of temperature rise tests realized with EFF1 (IE3) motors under optimal flux
condition are graphically presented in Figs. 10 to 13.
20 hp - 4 poles - NEMA High Efficiency

DT (K)
Constant Torque (100% Tr)
120
104.7 thermal class F limit
100
80.6 79.8 thermal class B limit

80
64.7 62.4
60
48.1
40
20
0
freq (Hz)
5 10 15 60
optimal flux constant flux 60 Hz sinusoidal
Fig. 9. Examples of results of temperature rise tests accomplished with NHE motor
DT (K) EFF1 motors (fbase = 60 Hz)

Optimal Flux - Constant Torque - 10 Hz
120.0
106.0
99.0 DT 105
100.0
91.7 89.7
86.7
80.4 82.0
80.0 73.5 72.9
DT 80
61.8 59.9
60.0
40.0
30.5
20.0
0.0 motor
3 hp [90L] 50 hp [200L] 75 hp [225 SM] 150 hp [280 SM]
rated load 80% rated load 60 Hz sinusoidal supply
Fig. 10. Results of temperature rise tests conducted with EFF1 (fbase = 60 Hz) motors running
at 10 Hz
DT (K)
EFF1 motors (fbase = 50 Hz)
120.0
DT 105
100.0 94.0
89.5
84.9
79.6
80.0 76.5 76.9 DT 80
73.8
65.3
62.5
59.2 57.4
60.0
80% load!
47.9
40.0
20.0
0.0 motor
2 hp [90L] 10 hp [132M] 75 hp [225 SM] 150 hp [280 SM]
rated load 81,8% rated load 50 Hz sinusoidal supply
at 10 Hz

DT (K)
120.0
98.6 DT 105
100.0
77.8 78.8 78.5 76.2

80.0 72.9
DT 80
70.8
64.5
61.8 59.9
60.0
40.0
30.5
20.0
0.0 motor
3 hp [90L] 50 hp [200L] 75 hp [225 SM] 150 hp [280 SM]
87,4% rated load 76,7% rated load 60 Hz sinusoidal supply
at 5 Hz

DT (K)
120.0
109.3
DT 105
100.0
85.8
80.0
80.0 74.5 DT 80
66.0 67.6
63.2 62.5 64.8 65.3
57.4
full load!
85% rated load!

60.0
47.9
40.0
20.0
0.0 motor
2 hp [90L] 10 hp [132M] 75 hp [225 SM] 150 hp [280 SM]
89,1% rated load 77,4% rated load 50 Hz sinusoidal supply
at 5 Hz
7. Conclusions
The study on the behavior of 3-phase induction low voltage motors’ losses as a function of
parameters such as load, operation frequency and magnetic induction (flux density),
brought about the conception of an equation for magnetic flux that minimizes the total
motor losses in each operation frequency. This is very interesting, given that variable speed
drives usually involve operation throughout a speed range. Frequency reduction causes the
motor iron losses to decrease, thus allowing the magnetic induction (or flux) to be increased
at low operation speeds, enabling the motor to provide torque with lower current. The flux
increase results from an increment in voltage/frequency ratio, accomplished in a growingly
more accentuated manner the lower the operation frequency.
The mathematical equation that has been created was implemented in commercial static
frequency converters and the study validation was carried out by means of the analysis of
specific motor lines and the execution of tests with a number of frequency converters and
induction motors of distinct sizes, using the optimal flux ratios theoretically determined.
The tests led to the definition of the actual optimal flux curve as well as new derating curves
valid for motors working under optimal flux condition, which evidence the advantages
provided by the proposed technique (Fig. 14).
According to Fig. 14, it is possible to note that if temperature rise of thermal class F (105ºC)
is permitted on the motor windings, the loss minimization technique prevents the need of
torque reduction for operation of the motor at low frequencies (until 0.1 p.u.). If temperature
rise of thermal class B (80ºC) is required, the loss minimization technique allows for the
application of a smoother torque derating than the needed for the motor running with
constant flux and rated losses.
DERATING CURVES
1.10
1.00
Torque reduction (p.u.)
0.90
0.80
0.70
0.60
0.50
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6
Frequency (p.u.)
Optimal flux (105 K) Optimal flux (80 K) Constant flux (105 K) Constant flux (80 K)
Fig. 14. The new derating (torque reduction) curves proposed evidence the advantages
provided by the optimal flux solution
The implementation of the optimal flux curve into commercial frequency converters
resulted in the development of a solution (converter + motor) to optimize/improve the
variable speed drives with constant torque loads.
The next step of this development (ongoing) is improving the Optimal Flux Solution, so that
it can be advantageously used with any type of load driven by converter-fed induction
motors.
8. References
[1] PCT/BR2004/000074, “Static frequency converter with automatic function for optimizing
magnetic flux and minimizing losses in electric induction motors”.
[2] S. L. Nau and A. P. Sobrinho, “Optimal voltage/frequency curve for inverter-fed motor”.
Proceedings of the 3rd International Conference on Energy Efficiency in Motor
Driven Systems (EEMODS), Treviso, Italy, 2002.
[3] R. Kaczmarek; M. Amar and F. Protat, “Iron loss under PWM voltage supply on Epstein
frame and in induction motor core”. IEEE Transactions on Magnetics, Vol. 32, No.1,
January 1996.
[4] A. Boglietti; P. Ferraris; M. Lazzari and F. Profumo, “Iron losses in magnetic materials
with six-step and PWM inverter supply”. IEEE Transactions on Magnetics, Vol. 27,
No. 6, November 1991.
[5] L. T. Mthombeni and P. Pillay, “Core losses in motor laminations exposed to high-
frequency or nonsinusoidal excitation”. IEEE Transactions on Industry
Applications, Vol. 40, No. 5, September-October 2004.
[6] M. S. Lancarotte; C. Goldemberg and A. A. Penteado Jr, “Estimation of FeSi core losses
under PWM or DC bias ripple voltage excitations”. IEEE Transactions on Energy
Conversion, Vol. 20, No. 2, June 2005.
[7] A. J. Moses and N. Tutkun, “Investigation of power loss in wound toroidal cores under
PWM excitation”. IEEE Transactions on Magnetics, Vol. 33, No. 5, September 1997.
[8] A. Boglietti; M. Chiampi; M. Repetto; O. Bottauscio and D. Chiarabaglio, “Loss
separation analysis in ferromagnetic sheets under PWM supply”. IEEE
Transactions on Magnetics, Vol. 34, No. 4, July 1998.
[9] A. Boglietti; P. Ferraris; M. Lazzari and M. Pastorelli, “About the possibility of defining a
standard method for iron loss measurement in soft magnetic materials with
inverter supply”. IEEE Transactions on Industry Applications, Vo. 33, No. 5,
September-October 1997.
[10] C. Cester; A. Kedous-Lebouc and B. Cornut, “Iron loss under practical working
conditions of a PWM powered induction motor”. IEEE Transactions on Magnetics,
Vol. 33, No. 5, September 1997.
[11] A. Ruderman and R. Welch, “Electrical machine PWM loss evaluation basics”.
Proceedings of the 4th International Conference on Energy Efficiency in Motor
Driven Systems (EEMODS), Heidelberg, Germany, 2005.
[12] M. Sokola; V. Vuckovic and E. Levi, “Measurement of iron losses in PWM inverter fed
induction machines”. Proceedings of the 30th Universities Power Engineering
Conference (UPEC), London, UK, 1995.
[13] A. C. Smith and K. Edey, “Influence of manufacturing processes on iron losses”.
Proceedings of the 7th International Conference on Electrical Machines and Drives -
Conference Publication No. 412. September 1995.
[14] K. Kostenko e L. Piotrovski, Máquinas Eléctricas – Volume II – Máquinas de Corrente
Alternada (tradução de original russo), Ed. Lopes da Silva, Porto, Portugal,
1979.
[15] P. C. Krause, Analysis of Electric Machinery, McGraw Hill, New York, USA, 1996.
[16] NEMA Standard MG1-2003, Part 30 – Application considerations for constant
speed motors used on a sinusoidal bus with harmonic content and general purpose
motors used with adjustable-voltage or adjustable-frequency controls or both.
5
Sensorless Vector Control of Induction Motor

Drive - A Model Based Approach
Jogendra Singh Thongam1 and Rachid Beguenane2
1Departmentof Renewable Energy Systems,
STAS Inc., Chicoutimi, QC
2Department of ECE, Royal Military College,
Kingston, ON
Canada
1. Introduction
Induction machines are more rugged, compact, cheap and reliable in comparison to other
machines used in similar applications. Vector controlled induction motor drive outperforms
the dc motor drive because of higher transient current capability, increased speed range and
lower rotor inertia.
Sensors widely used in electric drives degrade the reliability of the system especially in
hostile environments and require special attention to electrical noise. Moreover, it is difficult
to mount sensors in certain applications in addition to extra expenses involved. Therefore, a
lot of researches are underway to develop accurate speed estimation techniques. With
sensorless vector control we have a decoupled control structure similar to that of a
separately excited dc motor retaining the inherent ruggedness of the induction motor at the
same time. Speed sensorless control technique first appeared in (Abbondante & Brennen,
1975). The commonly used methods for speed estimation are Model Reference Adaptive
System (MRAS) (Schauder, 1992; Tajima & Hori, 1993; Peng & Fukao, 1994; Choy et al.,
1996), Neural Networks (Simoes & Bose, 1995; Fodor et al., 1995; Ben-Brahim & Kudor, 1995;
Kim et al., 2001; Toqeer & Bayindir, 2003; Haghgoeian et al., 2005), Extended Kalman Filter
(EKF) (Kim et al.,1994, Comnac et al., 2001; Ma & Gui, 2002; Du et al., 1995; Thongam &
Thoudam, 2004) and Nonlinear Observer (Bodson et al., 1995; Liu et al., 2001, Pappano et al.,
1998).
The aim of this chapter is to provide with a brief overview of high performance sensorless
induction motor drive. There exist two approaches to speed estimation for sensorless control
of induction machine: non model based approach and model based approach. The non
model based technique tracks a machine anisotropy: either saturation for flux estimation or
rotor slotting for rotor position estimation, whereas the model based technique rely mostly
on back emf voltage associated with fundamental component excitation of the machine.
Model-based observers are considered very well adapted for state estimation and allow, in
most cases, a stability proof and a methodology to tune observer gains. Among the
observers the reduced order observers are more frequently implemented than the full order
ones as they don’t require heavy computations. Two sensorless vector control strategies
using machine model-based estimation are presented in this chapter.
2. Speed estimation
Rotor speed has been considered as a constant by many researchers in speed estimation
problem (Schauder, 1992; Tajima & Hori, 1993; Peng & Fukao, 1994; Kim et al., 1994;
Comnac et al., 2001; Ma & Gui, 2002; Du et al., 1995; Minami et al., 1991; Veleyez-Reyes &
Verghese., 1992; Veleyez-Reyes et al., 1989). The idea is that the speed changes slowly
compared to electrical variables. Adopting such an approach allowed speed estimation
without requiring the knowledge of mechanical parameters of the drive system such as load
torque, inertia etc. In (Schauder, 1992; Tajima & Hori, 1993; Peng & Fukao, 1994) speed was
estimated using model reference adaptive system considering it as an unknown constant
parameter. In (Kim et al., 1994; Comnac et al., 2001; Ma & Gui, 2002; Du et al., 1995) the
speed was considered as an unknown constant state of the machine and extended kalman
filter (EKF) was used to estimate it. Recursive least square estimation method was used in
(Minami et al., 1991; Veleyez-Reyes & Verghese, 1992; Veleyez-Reyes et al., 1989) for speed
estimation considering speed as an unknown constant parameter and found out the value of
estimated speed that best fits the measured and calculated data in the dynamic equations of
the motor.
In this section we present a sensorless vector control strategy using machine model-based
speed estimation (Thongam & Ouhrouche, 2007). The proposed method does not require
taking derivative of the measured signals unlike that of (Peng & Fukao, 1994; Minami et al.,
1991; Veleyez-Reyes & Verghese, 1992; Veleyez-Reyes et al., 1989). The method is also
simpler to implement than implementing EKF. In this method the model of the motor used
for estimation is derived by introducing a new variable which is a function of rotor flux and
speed assuming that rotor speed varies slowly in comparison to electrical states and hence
its derivative can be conveniently equated to zero in the machine model used for estimation.
The sensorless method presented here in this chapter is based on observing this new
variable. A reduced order observer is implemented for estimating the new variable using
which the rotor speed is estimated.
2.1 Induction machine model

The induction motor model in stationary stator reference frame α − β may be written in
vector matrix form as
dψ r
= A11ψ r + A12 is (1)
dt
di s
= A21ψ r + A22 is + A23v s (2)
dt
Lm
where A11 = − ( Rr / Lr ) I + ω J , A12 = ( Lm Rr / Lr ) I , A21 =
σ Ls Lr
{( Rr / Lr ) I − ω J } ,
{ (
A22 = − Rs / (σ Ls ) + Rr L2m / σ Ls L2r )} I , A23 = 1 / (σ Ls ) I ,
⎡1 0 ⎤
I =⎢ ⎥,
⎣0 1 ⎦
⎡ 0 −1⎤
J =⎢
⎣1 0 ⎦
⎥,
T T T
ψ r = ⎡⎣ψ rα ψ r β ⎤⎦ is the rotor flux, is = ⎡⎣isα isβ ⎤⎦ is the stator current, v s = ⎡⎣vsα vs β ⎤⎦ is
the stator voltage and σ = 1 − L2m /( Ls Lr ) is the leakage coefficient.

Sensorless Vector Control of Induction Motor Drive - A Model Based Approach 79
Now, we introduce a new quantity into the motor model which when introduced will make
the right hand side of conventional motor model given by equations (1) and (2) independent
of the unknowns – the rotor flux and speed. Let’s define the new quantity as
Z = − A11ψ r (3)
A new motor model is obtained after introducing the new quantity as given below:
dψ r
= A12 is + A14 Z (4)
dt
di s
= A22 is + A23v s + A24 Z (5)
dt
dZ
= A32 is + A34 Z (6)
dt
{ } ( )
where A14 = − I , A24 = Lm / (σ Ls Lr ) I , A32 = Lm Rr2 / L2r I − ω ( Lm Rr / Lr ) J and A34 = A11 .
2.2 Observer structure and speed estimation

The proposed speed estimation algorithm is based on observing the newly defined quantity
which is a function of rotor flux and speed. Equation (5) and (6) are used for constructing a
Gopinath’s reduced order observer (Gopinath, 1971) for estimating the newly defined
quantity. The observer is as given below
dZˆ ⎛ di diˆ ⎞
= A32 is + A34 Zˆ + G ⎜ s − s ⎟ (7)
dt ⎜ dt ⎟⎠
⎝ dt
⎡g − g2 ⎤ diˆs
where G = ⎢ 1 ⎥ is the observer gain. Using equation (5) for the observer equation
⎣ g2 g1 ⎦ dt
becomes
dZˆ ⎛ di ⎞
= A32 is + A34 Zˆ + G ⎜ s − A22 is − A23vs − A24 Zˆ ⎟ (8)
dt ⎝ dt ⎠
The observer poles can be placed at the desired locations in the stable region of the
complex plane by properly choosing the values of the elements of the G matrix. In order to
avoid taking derivative of the stator current in the algorithm we introduce another new
quantity
D = Zˆ − Gis (9)
Finally, the observer is of the following form:
d
F = ( A32 + A34G − GA22 − GA24G ) is − GA23v s + ( A34 − GA24 ) D ` (10)
dt
Zˆ = D + Gis (11)
The block diagram of the Z observer is shown in Fig. 1.
A34 − GA24
+
vs _ + Ẑ
GA23 D
+ +
is
A32 + A34G − GA22 − GA24G
Fig. 1. Block diagram of Z observer

Assuming no parameter variation and no speed error, the equation for error dynamics is
given by
d d
dt
Z=
dt
( )
Z − Zˆ = ( A34 − A24G ) Z (12)
Eigenvalues of ( A34 − A24G ) are the observer poles which are as given below:
⎛R Lm ⎞ ⎛ Lm ⎞
Pobs1,2 = − ⎜⎜ r + g1 ⎟⎟ ± j ⎜⎜ ω − g 2 ⎟⎟ (13)
⎝ Lr σ Ls Lr ⎠ ⎝ σ Ls Lr ⎠
The desired observer dynamics can be imposed by proper selection of observer gain G.
Next, let’s see how the rotor speed is computed. It can be seen that the observed quantity is
a function of rotor flux and speed. Performing matrix multiplication of ψ r T J with equation
(3) we have
(
Zαψ r β − Z βψ rα = ψ r2α + ψ r2β ω ) (14)
This is a simple equation which does not involve derivative or integration. To use it directly
for speed computation we need to know the rotor flux; and as for Zα and Z β we can use the
estimated values. The required flux is obtained from the reference. Rearranging the above
equation we have the equation used for rotor speed computation as given by
Zˆαψ r*β − Zˆ βψ r*α

ωˆ = 2 2
(15)
ψ r*α + ψ r*β
The coefficient matrices A32 and A34 in the observer equation are updated with the estimated
values of rotor speed.
It is to be noted here that the model of the motor used in implementing the observer
algorithm has been developed assuming that the derivative of the rotor speed is zero. It is
valid to make such an assumption since the dynamics of rotor speed is much slower than
that of electrical states. Moreover, such an assumption allows estimation without requiring
the knowledge of mechanical quantities of the drive such as load torque, inertia etc.

Simulation is carried out in order to validate the speed estimation algorithm presented. The
block diagram of the sensorless indirect vector controlled induction motor drive
incorporating the proposed speed estimator is shown in Fig. 2. The results of simulation are
shown in Fig. 3 - Fig. 5.
Vdc _
* *
isa +
*+ Tem *
isd +
ω dq
_ 2h
_ *
INDIRECT
isq *
isb +
ω̂ FOC
ψ r* ρ _
INVERTER
2h
*
isc +
abc
_ 2h
ψ r*
vs isa isb isc
SPEED
ESTIMATOR is IM
Fig. 2. Sensorless indirect VC induction motor drive

Initially, the drive is run at no load. It is accelerated from rest to 150 rad/s at 0.15 sec. and
then, the speed is reversed at 2.5 sec. The speed is reversed again at 5.5 sec. The speed of the
motor (ω ) , estimated speed (ωˆ ) and reference speed (ω * ) are shown in Fig. 3 (a). Fig. 3 (b)
shows speed estimation error (ω − ωˆ ) . The newly defined quantity ( Z ) and its estimated
value ( Zˆ ) are shown in Fig. 3 (c) and its estimation error ( Z − Zˆ ) is shown in Fig. 3 (d).
The estimation algorithm and the drive response are then verified under loading and
unloading conditions. The unloaded drive is started at 0.15 sec and full load is applied at 1
sec; then load is completely removed at 2 s. Later, after speed reversal, full load is applied at
4 sec and the load is completely removed at 5 sec. Fig. 4 shows the speed estimation result
and response of the sensorless drive system.
Then, the sensorless induction motor drive is run under fully loaded condition at various
operating speeds. The drive is started at full load at 0.15 s to 150 rad/s and the speed is
reduced in steps in order to observe the response of the loaded drive at various speeds. Fig.
5 shows the estimation results and response of the loaded drive.
200 15
Reference speed
10
Speed estimation error

Actual speed
100
Speed [ rad/s ]
Estimated speed 5
[ rad/s ]
0 0
-5
-100
-10
-200 -15
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Time [ s ] Time [ s ]
( a) ( b)
120 3
100 2
Z estimation error
80 Actual Z
1
Estimated Z
60 0
Z
40
-1
20
-2
0
-3
0 1 2 3 4 5 6 0 1 2 3 4 5 6
(c) ( d)
Fig. 3. Acceleration and speed reversal at no load; (a) reference, actual and estimated speeds;
(b) speed estimation error; (c) actual Z and estimated Z and (d) Z estimation error
200 15
Reference speed
Actual speed 10
100
Speed [ rad/s ]
Estimated speed
5
[ rad/s ]
0 0
-5
-100
-10
-200 -15
0 1 2 3 4 5 6 0 1 2 3 4 5 6
(a) (b)
120 3
100 2
Z estimation error
80 1
Actual Z
60
0
Z
Estimated Z
40
-1
20
-2
0
-3
0 1 2 3 4 5 6 0 1 2 3 4 5 6
(c) (d)
Fig. 4. Application and removal of load; (a) reference, actual and estimated speeds; (b) speed
estimation error; (c) actual Z and estimated Z and (d) Z estimation error
180 8
Reference speed
150 Actual speed

Estimated speed 4
120
Speed [rad/s]
[rad/s]
90 0
60
-4
30
0
-8
0 1 2 3 4 5 6 0 1 2 3 4 5 6
(a) ( b)
100 4
80 Actual Z
2
Z estimation error
Estimated Z
60
0
Z
40
20 -2
0
-4
0 1 2 3 4 5 6 0 1 2 3 4 5 6
(c) ( d)
Fig. 5. Operation at full load at various speeds; (a) reference, actual and estimated speeds;
(b) speed estimation error; (c) actual Z and estimated Z and (d) Z estimation error
2.4 Improvement in speed estimation

It is observed that the estimation algorithm presented above gives good estimation accuracy
under both dynamic and steady state conditions. However, it is found that the estimation
accuracy decreases with decrease in speed. This is because of the fact that the estimation
algorithm uses the command flux for speed estimation and not the actual rotor flux which
however is little bit different from the command value. With the reduction in speed Z
decreases and therefore the flux error is more prominently visible in the estimated speed. The
problem is overcome in this section by using a rotor flux observer based on the voltage model
of the machine along with the observer of the newly defined quantity. This allows accurate
speed estimation in various operating ranges. The speed is computed using (15) after replacing
the command flux by the estimated one. Further, due to the obvious advantages of dc current
regulators over ac current regulators as regards its robustness, and load and operating point
independence (Rowan & Kerkman, 1986) the control system uses dc current regulators. The
rotor flux estimator and the control scheme are presented in the following subsections.
2.4.1 Flux estimation

The voltage model of induction motor is given by
ψr =
Lr
Lm
{∫ (vs − Rs is )dt − σ Ls is } (16)
The rotor flux can be estimated using (16). However, the integration in (16) produces a
problem of dc off-set and drift component in low speed region. Therefore, a first order low
pass filter (LPF) is used instead of integration. The phase error in the low speed region
produced due to LPF is approximately compensated by adding low pass filtered reference
flux with the same time constant as above, and producing the estimated rotor flux (Ohtani et
al., 1992). The estimator equation is given as
Lr ⎧ τ τs ⎫ * 1 (17)
ψˆ r = ⎨(v s − Rs is ) − σ Ls is ⎬ + Ψr
Lm ⎩ 1+τ s 1+τ s ⎭ 1+τ s
where τ is the LPF time constant. The command rotor flux Ψ r* in (17) is obtained as follows:
⎡Ψ r*α ⎤ ⎡Ψ * cos ρ* ⎤
Ψ *r = ⎢ ⎥=⎢ r ⎥ (18)
⎢Ψ r*β ⎥ ⎢⎣Ψ r* sin ρ* ⎥⎦
⎣ ⎦
where Ψ r* = Lmi*sd and ρ* , the command rotor flux angle, is as given by
ρ* = ∫ ω*e dt (19)
ω*e , the command rotor flux speed, is computed as given below
ω*e = ω*sl + ωˆ (20)
The command slip speed ω*sl is given by
*
Rr iqs
ωsl* = *
(21)
Lr ids
2.4.2 Speed estimation

The equation (15) after modification is used for speed computation. In place of reference
flux, estimated flux is used for speed computation as given below
Zˆαψˆ r β − Zˆ βψˆ rα
ωˆ = (22)
2 2
ψˆ rα + ψˆ r β
2.4.3 Simulation results

Simulation is carried out in order to verify the accuracy of the estimation algorithm and to see
the response of the sensorless drive system. The block diagram of sensorless vector controlled
induction motor drive incorporating the flux and speed estimators is shown in Fig. 6.
First, the sensorless drive is run at no load at various speeds to verify the performance of the
observer under no load condition. The drive is started at no-load and is run at various
speeds by increasing it in steps to 10 rad/s, 50 rad/s, 100 rad/s and 150 rad/s at 0.3 sec, 1.5
sec, 3 sec and 4.5 sec respectively. The speed of the motor (ω ) , estimated speed (ωˆ ) ,
reference speed (ω * ) and speed estimation error (ω − ωˆ ) are shown in Fig. 7 (a). Fig. 7 (b)
shows the actual Z, estimated Z and Z estimation error ( Z − Zˆ ).
+ −
Vdc
v*sa
ω* +
*
isq + v*sq
−
v*sb
ψ r*+ * −
isd v*sd dq abc INVERTER
+
v*sc
ω̂ − −
ψˆ r *
isq
isq isd
dq abc
is ρ*
FLUX VECTOR
GENERATION Ψ r*
is
ψˆ r ROTOR
ROTOR
FLUX
SPEED
ESTIMATOR
ESTIMATOR vs
IM
Fig. 6. Sensorless vector controlled induction motor drive

Then, the performance of the estimator and drive response are verified on loading and
unloading. The drive is started at no-load at 0.25 s to a speed of 150 rad/s and full load is
applied at 1 sec and then the load is removed completely at 2 sec. Later, after speed reversal,
full load is applied at 4 s and the load is completely removed at 5 sec. The response of the
drive on application and removal of load is shown in Fig. 8.
200 100
Reference speed Actual Z
Actual speed 80 Estimated Z
150 Estimated speed
Speed [ rad/s ]
60
100
Z
40
50 20
0
0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
2
6
4
Z estimation error
2
[ rad/s ]
0 0
-2
-4
-6
-2
0 1 2 3 4 5 6 0 1 2 3 4 5 6
( a) ( b)
Fig. 7. No load operation at various speeds; (a) reference, actual, estimated speeds, and
speed estimation error; (b) actual Z, estimated Z, and Z estimation error
200 120
Reference speed
Actual speed 100
100 Estimated speed
80
Speed [ rad/s ]
60
0
Z
40
-100 20 Actual Z
Estimated Z
0
-200
0 1 2 3 4 5 6 0 1 2 3 4 5 6
8 3
2
Z estimation error
1
[ rad/s ]
0 0
-1
-4
-2
-8 -3
0 1 2 3 4 5 6 0 1 2 3 4 5 6
( a) (b)
Fig. 8. Application and removal of load; (a) reference, actual, estimated speeds, and speed
estimation error; (b) actual Z, estimated Z, and Z estimation error.
180 100
Reference speed
Actual Z
Actual speed 80 Estimated Z
Estimated speed
Speed [ rad/s ]
120
60
Z
40
60
20
0
0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
10 2
5
Z estimation error
0
[ rad/s ]
-2
-5
-10 -4
0 1 2 3 4 5 6 0 1 2 3 4 5 6
( a) ( b)
Fig. 9. Full load operation at various speeds; (a) reference, actual, estimated speeds, and
speed estimation error; (b) actual Z, estimated Z, and Z estimation error.
The estimator performance and drive response are then verified under fully loaded
condition of the drive at various operating speeds. The fully loaded machine is accelerated
to 150 rad/s at 0.3 s, and then the speed is reduced in steps to 100 rad/s, 50 rad/s and 10
rad/s at 1.5 sec, 3 sec and 4.5 sec respectively. Fig. 9 shows the estimation results and
response of the drive during the operation.
Good speed estimation accuracy was obtained under both dynamic and steady state
conditions under various operating conditions and response of the VC induction motor
drive incorporating the estimation algorithms was found to be good. The speed estimation
algorithm presented in this section depends upon the knowledge of the rotor flux, whereas,
the rotor flux estimator is independent of rotor speed and requiring only the measurable
stator terminal quantities, the stator voltage and current.
3. Flux and speed estimation

Induction machines do not allow rotor flux to be easily measured. The current model and
the voltage model are the traditional solutions, and their benefits and drawbacks are well
known. Various observers for flux estimation were analyzed in the work by Verghese and
Sanders (Verghese & Sanders, 1988) and Jansen and Lorenz (Jansen & Lorenz, 1994). Over
the years several other have been presented, many of which include speed estimation
(Tajima & Hori, 1993; Kim et al., 1994; Ohtani et al., 1992; Kubota et al., 1993; Sathiakumar,
2000; Yan et al., 2000).
Tajima & Hori (Tajima & Hori, 1993) proposed MRAS (Schauder, 1992) with novel pole
allocation method for speed estimation while rotor flux estimation was done using
Gopinath’s observer. Extended Kalman Filter was used in (Kim et al., 1994) for estimating
the rotor flux and speed using a full order model of the motor assuming that rotor speed is a
constant. Ohtani et al (Ohtani et al., 1992) used the voltage model for flux estimation
overcoming the problem associated with integrator and low pass filter while speed was
obtained using a frequency controller. A speed adaptive flux observer was proposed in
(Kubota et al., 1993) for estimating rotor flux and speed. Gopinath style reduced order
observer was used in (Sathiakumar, 2000) for estimating the rotor flux while the speed was
computed using an equation derived from the motor model. Yan et al (Yan et al., 2000)
proposed a flux and speed estimator based on the sliding-mode control methodology.
In this section, we present a new flux estimation algorithm for speed sensorless rotor flux
oriented controlled induction motor drive (Thongam & Ouhrouche, 2006). The proposed
method is based on observing the variable Z introduced in Section 2 which when
introduced makes the right hand side of the conventional motor model independent of
rotor flux and speed. Rotor flux estimation is achieved using an equation obtained after
introduction of the newly defined quantity into the Blaschke equation or commonly
known as the current model; while, speed is computed using a simple equation obtained
using the new quantity Z.
3.1 Estimation of rotor flux and speed

The speed computation equation (22) obtained in section 2.4.2 requires the knowledge of
rotor flux and Z. Here, we present a joint rotor flux and speed estimation algorithm. The
block diagram of the proposed rotor flux and speed estimation algorithm is shown in
Fig. 10.
vs Ẑ
Z
Estimation ω̂
(Equations Rotor Flux Speed
is 10 & 11) Estimation Computation
(Equations ψ̂ r (Equation 22)
28 & 29)
Ψ *r
Fig. 10. Rotor flux and speed estimator

Rotor flux may be obtained directly using equation (4) which is obtained after introducing
the newly defined quantity Z into the Blaschke equation as
ψˆ r = ∫ ( A12is + A14 Z ) dt (23)
However, rotor flux computation by pure integration suffers from dc offset and drift
problems. To overcome the above problems a low pass filter is used instead of pure
integrator and the phase error due to low pass filtering is approximately compensated by
adding low pass filtered reference flux with the same time constant as used above (Ohtani et
al., 1992). The equation of the proposed rotor flux estimator is given below
τ 1
ψˆ r = ( A12is + A14 Z ) + Ψ r* (24)
1+τ s 1+τ s
where τ is the LPF time constant. The command rotor flux Ψ r* is obtained as follows
⎡Ψ * ⎤ ⎡cos ρ* ⎤ ⎡cos ρ* ⎤
rα
Ψ *r = ⎢ ⎥ = Ψ r* ⎢ ⎥ = Lmi*sd ⎢ ⎥ (25)
⎢Ψ r*β ⎥ ⎢
⎣ sin ρ *⎥
⎦ ⎢⎣ sin ρ* ⎥⎦
⎣ ⎦
The command rotor flux angle ρ* is obtained by integrating the command rotor flux speed
as given by
ρ* = ∫ ω*e dt = ∫ ( ω*sl + ωˆ )dt (26)
The command slip speed ω*sl is given by
*
Rr iqs
ωsl* = *
(27)
Lr ids
We know that the equation of the back emf is given by:
Lm dψ r Lm
e= = ( A12is + A14 Z ) (28)
Lr dt Lr
is
A12
+
τ
+ 1+τ s
Z
A14
Ψ *r 1 +
+ ψˆ r
1+τ s
Fig. 11. Rotor Flux Estimator
G
e − axis
τ G
1+τ s
{( Lr / Lm ) e }
1 G*
Ψr
1+τ s
ζ
G
Ψ *r
G G
ζ ψ̂ r ψ r − axis
1 G*
Ψr
1+τs
Fig. 12. Obtaining estimated rotor flux

Now, equation (24) may also be written as
τ ⎛ Lr ⎞ 1
ψˆ r = ⎜ e ⎟⎟ + Ψ r* (29)
1 + τ s ⎜⎝ Lm ⎠ 1 + τ s
Block diagram of the rotor flux estimator is shown in Fig. 11. Fig.12 explains how estimated
flux is obtained using equation (29).

Simulation is carried out in order to validate the performance of the proposed flux and
speed estimation algorithm. The proposed rotor flux and speed estimation algorithm is
incorporated into a vector controlled induction motor drive. The block diagram of the
sensorless vector controlled induction motor drive incorporating the proposed estimator is
shown in Fig. 13. The sensorless drive system is run under various operating conditions.
First, acceleration and speed reversal at no load is performed. A speed command of 150
rad/s at 0.5 s is given to the drive system which was initially at rest, and then the speed is
reversed at 3 s. The response of the drive is shown in Fig. 14. Fig. 14 (a) shows reference
( ω* ), actual ( ω ), estimated ( ω̂ ) speed, and speed estimation error ( ω − ωˆ ). The module of
the actual ( | Ψ r | ), estimated ( | Ψˆ r | ) rotor flux, and rotor flux estimation error ( | Ψ r | − | Ψˆ r | )
are shown in Fig. 14 (b). Fig. 14 (c) and (d) shows respectively the locus of the actual and
estimated rotor fluxes.
The drive is then run at various speeds under no load condition. It is accelerated from rest to
10 rad/s at 0.5 s, then accelerated further to 50 rad/s, 100 rad/s and 150 rad/s at 1.5 s, 3 s
and 4.5 s respectively. Fig. 15 shows the estimation of rotor flux and speed, and the response
of the sensorless drive system.
Then, the drive is subjected to a slow change in reference speed profile (trapezoidal), the
results of which are shown in Fig. 16.
+ Vdc −
ω* +
*
isq +
v*sq v*sa
v*sb
−
−
ψ r* +
*
isd +
v*sd dq abc INVERTER
ω̂ v*sc
− −
* isq isd is ρ*
ψˆ r isq dq abc
FLUX VECTOR
GENERATION Ψ r*
is
ROTOR FLUX
&
SPEED
ESTIMATOR
vs IM
Fig. 13. Sensorless vector controlled induction motor drive
Further, the performance of the estimator is verified under loaded conditions at various
operating speeds. The fully loaded drive is accelerated to 150 rad/s at 0.5 s and then
decelerated in steps to 100 rad/s, 50 rad/s and 10 rad/s at 1.5 s, 3 s and 4.5 s respectively.
Fig. 17 shows the estimation results and response of the loaded drive system.
Then, we test the performance of the estimator on loading and unloading. The drive at rest
is accelerated at no-load to 150 rad/s at 0.5 s and full load is applied at 1 s; we then remove
the load completely at 2 s. Later, after speed reversal, full load is applied at 4 s, then, the
load is removed completely at 5 s. Fig. 18 shows the estimation results and the response of
the sensorless drive.
200 0.4
Reference speed
S p e e d [ ra d /s ]
F lu x [ W b ]
Actual speed
0 0.2 Actual flux
Estimated speed Estimated flux
-200 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
S p e e d e s tim a tio n e rro r
F lu x e s tim a tio n e rro r

10 0.2
[ ra d /s ]
[ Wb ]
0 0
-10 -0.2
0 1 2 3 4 5 6 0 1 2 3 4 5 6
(a) (b)
E s tim a te d ψ r β [ W b ]
A c tu a l ψ r β [ W b ]
0.2 0.2
0 0
-0.2 -0.2
-0.4 -0.4
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4
Actual ψr [ Wb ] Estimated ψr [ Wb ]
α α
(c) (d)
Fig. 14. Acceleration and speed reversal of the sensorless drive at no-load
0.4
150 Reference speed
S p e e d [ ra d /s ]
F lu x [ W b ]
100 Actual speed

Estimated speed 0.2 Actual flux
50 Estimated flux
0
0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
S p e e d e s tim a tio n e rro r
F lu x e s tim a tio n e rro r
10 0.2
[ ra d /s ]
[ Wb ]
0 0
-10 -0.2
0 1 2 3 4 5 6 0 1 2 3 4 5 6
(a) (b)
E s tim a te d ψ r β [ W b ]
0.2 0.2
0 0
-0.2 -0.2
-0.4 -0.4
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4
α α
(c) (d)
Fig. 15. No-load operation of the sensorless drive with step increase in speeds
200 0.4
S peed [ rad/s ] Reference speed
Actual speed
F lux [ W b ]
Estimated speed
0 0.2 Actual flux
Estimated flux
-200 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
10 0.2
S peed es tim ation error
F lux es tim ation error

0 0
[ rad/s ]
[ Wb ]
-10 -0.2
0 1 2 3 4 5 6 0 1 2 3 4 5 6
(a) (b)
E s tim ated ψ rβ [ W b ]
A c tual ψ rβ [ W b ]
0.2 0.2
0 0
-0.2 -0.2
-0.4 -0.4
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 -0.4 0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4
α α
(c) (d)
Fig. 16. No-load operation of the sensorless drive with trapezoidal reference speed
200 0.4
Reference speed
S p e e d [ ra d / s ]
F lu x [ W b ]
Actual speed
100 Estimated speed 0.2 Actual flux
Estimated flux
0
0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
S p e e d e s t im a tio n e rro r
F lu x e s t im a tio n e rro r
10 0.2
[ ra d /s ]
[ Wb ]
0 0
-10 -0.2
0 1 2 3 4 5 6 0 1 2 3 4 5 6
(a) (b)
E s tim a te d l ψ r β [ W b ]
0.2 0.2
0 0
-0.2 -0.2
-0.4 -0.4
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4
Actual ψr [ Wb ] Estimatedl ψr [ Wb ]
α α
(c) (d)
Fig. 17. Operation of the sensorless drive at full load at various speeds
200 0.4
Reference speed
S peed [ rad/s ]
Actual speed
F lux [ W b ]
Estimated speed Actual flux
0 0.2
Estimated flux
-200 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
S peed es tim ation error
F lux es tim ation error

10 0.2
[ rad/s ]
[ Wb ]
0 0
-10 -0.2
0 1 2 3 4 5 6 0 1 2 3 4 5 6
(a) (b)
E s tim ated ψ r β [ W b ]
A c tual ψ r β [ W b ]
0.2 0.2
0 0
-0.2 -0.2
-0.4 -0.4
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4
α α
(c) (d)
Fig. 18. Drive response on application and removal of load
4. Conclusion and future works

In this chapter we have presented some methods of sensorless vector control of induction
motor drive using machine model-based estimation. Sensorless vector control is an active
research area and the treatment of the whole model based sensorless vector control will
demand a book by itself.
First, a speed estimation algorithm in vector controlled induction motor drive has been
presented. The proposed method is based on observing a newly defined quantity which is a
function of rotor flux and speed. The algorithm uses command flux for speed computation.
The problem of decrease in estimation accuracy with the decrease in speed was overcome
using a flux observer based on voltage model of the machine along with the observer of the
newly defined quantity, and satisfactory results were obtained.
Then, a joint rotor flux and speed estimation algorithm has been presented. The proposed
method is based on a modified Blaschke equation and on observing the newly defined
quantity mentioned above. Good estimation accuracy was obtained and the response of the
sensorless vector controlled drive was found to be satisfactory.
The mathematical model of the motor used for implementing the estimation algorithm was
derived with the assumption that the rotor speed dynamics is much slower than that of
electrical states. Therefore, increase in estimation accuracy of the proposed algorithms will
be observed with the increase in the size of the machine used.
The machine model developed in this chapter may be used in future for machine parameter
estimation. The newly defined quantity presented in this chapter contains rotor resistance
information as well, in addition to that of rotor flux and speed. Therefore, future research
efforts may be made towards developing rotor resistance estimation algorithm using the
new machine model. Further, in the proposed algorithms rotor flux was necessary for speed
estimation. Future research efforts may also be made towards developing a speed
estimation algorithm for which the knowledge of rotor flux is not necessary.
5. References
Abbondante, A. & Brennen, M. B. (1975). Variable speed induction motor drives use
electronic slip calculator based on motor voltages and currents. IEEE Trans. Ind.
Appl, Vol. 1A-11, No. 5, Sept/Oct, pp. 483-488.
Ben-Brahim, L. & Kudor, T. (1995). Implementation of an induction motor speed estimator
using neural networks. Proceedings of International Power Electronics Conference, IPEC
1995, Yokohama, April, pp. 52-58.
Bodson, M.; Chiasson, J. & Novotnak, R. T. (1995). Nonlinear Speed Observer for High
Performance Induction Motor Control. IEEE Trans. Ind. Elec, Vol. 42, No. 4, Aug.
pp. 337-343.
Choy, I.; Kwon, S. H.; Lim J. & Hong, S. W. (1996). Robust Speed Estimation for Tacholess
Induction Motor Drives. IEEE Electronics Letters, Vol. 32, No. 19, pp. 1836-1838.
Comnac V.; Cernat M.; Cotorogea, M. & Draghici, I. (2001). Sensorless Direct Torque and
Stator Flux Control of Induction Machines Using an Extended Kalman Filter",
Proceedings of IEEE Int. Conf. on Control Appl, Mexico, Sept. 5-7, pp. 674-679.
Du T.; Vas, P. & Stronach, F. (1995). Design and Application of Extended Observers for Joint
State and Parameter Estimation in High Performance AC Drives. IEE Proc. Elec.
Power Appl., Vol. 142, No. 2, pp. 71-78.
Fodor, D. ; Ionescu, F. ; Floricau, D. ; Six, J.P. ; Delarue, P. ; Diana, D. & Griva, G. (1995).
Neural Networks Applied for Induction Motor Speed Sensorless Estimation.
Proceedings of the IEEE International Symposium on Industrial Electronics, ISIE’ 95, July
10-14, Athens, pp. 181-186.
Gopinath, B. (1971). On the Control of Linear Multiple Input-Output Systems. Bell System
Technical Journal, Vol. 50, No. 3, March, pp. 1063-1081.
Haghgoeian, F.; Ouhrouche, M. & Thongam, J. S. (2005). MRAS-based speed estimation for
an induction motor sensorless drive using neural networks. WSEAS Transactions on
Systems, Vol. 4, No. 12, December, pp. 2346-2352.
Jansen, P. L. & Lorenz, R. D. (1994). A physically insightful approach to the design and
accuracy assessment of flux observers for field oriented induction machine drives.
IEEE Trans. Ind. App., Vol. 30, No. l, Jan. /Feb., pp. 101-110.
Kim, S. H.; Park, T. S.; Yoo, J. Y. & Park, G. T. (2001). Speed-sensorless vector control of an
induction motor using neural network speed estimation. IEEE Trans. Ind. Elec, Vol.
48, No. 3, June, pp. 609-614.
Kim, Y. R.; Sul S. K. & and Park, M. H. (1994). Speed sensorless vector control of induction
motor using extended Kalman filter. IEEE Trans. Ind. Appl., Vol. 30, No. 5,
Sept/Oct, pp. 1225-1233.
Kubota, H.; Matsuse K. & Nakano, T. (1993). DSP-based speed adaptative flux observer of
induction motor. IEEE Trans. Ind. Appl, Vol. 29, No. 2, March/April, pp. 344-348.
Liu, J. J.; Kung, I. C. & Chao, H. C. (2001). Speed estimation of induction motor using a non-
linear identification technique. Proc. Natl. Sci. Counc. ROC (A), Vol. 25, No. 2, pp.
107-114.
Ma, X. & Gui, Y. (2002). Extended Kalman filter for speed sensor-less DTC based on DSP.
Proc. of the 4th World Cong. on Intelligent Control and Automation, Shanghai, China,
June 10-14, pp. 119-122.
Minami, K.; Veley-Reyez, M.; Elten, D.; Verghese, G. C. & Filbert, D. (1991). Multi-stage
speed and parameter estimation for induction machines. Proceedings of the IEEE
Power Electronics Specialists Conf., Boston, USA, pp. 596-604.
Ohtani, T.; Takada, N. & and Tanaka, K. (1992). Vector control of induction motor without
shaft encoder. IEEE Trans. Ind. Appl, Vol. 28, No. 1, Jan/Feb, pp. 157-164.
Pappano, V.; Lyshevski, S. E. & Friedland, B. (1998). Identification of induction motor
parameters. Proceedings of the 37th IEEE Conf. on Decision and Control, Tampa,
Florida, USA, December 16-18, pp. 989-994.
Peng, F. Z. & Fukao, T. (1994). Robust speed identification for speed sensorless vector
control of induction motors. IEEE Trans. Ind. Appl, Vol. 30, No. 5, Sept/Oct., pp.
1234-1240.
Rowan, T. M. & Kerkman, R. J. (1986). A new synchronous current regulator and an analysis
of current-regulated PWM inverters. IEEE Trans. Ind. Appl, Vol. IA-22, No. 4,
July/Aug., pp. 678-690.
Schauder, C. (1992). Adaptive speed identification for vector control of induction motors
without rotational transducers. IEEE Trans. Ind. Appl, Vol. 28, No. 5, Sept./Oct., pp.
1054-1061.
Sathiakumar, S. (2000). Dynamic flux observer for induction motor speed control.
Proceedings of Australian Universities Power Engineering Conf. AUPEC 2000, Brisbane,
Australia, 24-27 Sept., pp. 108-113.
Simoes, M. G. & Bose, B. K. (1995). Neural network based estimation of feedback signals for
a vector controlled induction motor drive. IEEE Trans. Ind. Appl., Vol. 31,
May/June, pp. 620-629.
Tajima, H. & Hori, Y. (1993). Speed sensorless field-orientation control of the induction
machine. IEEE Trans. Ind. Appl., Vol. 29, No. 1, pp. 175-180.
Thongam, J. S. & Thoudam, V. P. S. (2004). Stator flux based speed estimation of induction
motor drive using EKF. IETE Journal of Research, India, Vol. 50, No. 3. May-June, pp
191-197.
Thongam, J. S. & Ouhrouche, M. (2006). Flux estimation for speed sensorless rotor flux
oriented controlled induction motor drive. WSEAS Transactions on Systems, Vol. 5,
No. 1, Jan., pp. 63-69.
Thongam, J. S. & Ouhrouche, M. (2007). A novel speed estimation algorithm in indirect
vector controlled induction motor drive. International Journal of Power and Energy
Systems, Vol. 27, No. 3, 2007, pp. 294-298.
Toqeer, R. S. & Bayindir, N. S. (2003). Speed estimation of an induction motor using Elman
neural network. Neuro Computing, Volume 55, Issues 3-4, October, pp. 727- 730.
Velez-Reyes, M.; Minami, K. & Verghese, G. C. (1989). Recursive speed and parameter
estimation for induction machines", IEEE/IAS Ann. Meet. Conf. Rec., San Diego, pp.
607-611.
Veleyez-Reyes, M. & Verghese, G. C. (1992). Decomposed algorithms for speed and
parameter estimation in induction machines. IFAC Symposium on Nonlinear Control
System Design, Bordeaux, France, pp. 77-82.
Verghese, G. C. & Sanders, S. R. (1988). Observers for flux estimation in induction machines.
IEEE Trans. Ind. Elec, Vol. 35, No. 1, Feb., pp. 85-94.
Yan, Z.; Jin C. & Utkin, V. I. (2000). Sensorless sliding-mode control of induction motors.
IEEE Trans. Ind. Elec, Vol. 47, No. 6, Dec., pp. 1286-1297.
6
Feedback Linearization of Speed-Sensorless

Induction Motor Control with Torque
Compensation
Cristiane Cauduro Gastaldini1 , Rodrigo Zelir Azzolin2 ,
Rodrigo Padilha Vieira3 and Hilton Abílio Gründling4
1,2,3,4Federal University of Santa Maria
2 Federal University of Rio Grande
3 Federal University of Pampa
Brazil
1. Introduction
This chapter addresses the problem of controlling a three-phase Induction Motor (IM) without
mechanical sensor (i.e. speed, position or torque measurements). The elimination of the
mechanical sensor is an important advent in the field of low and medium IM servomechanism;
such as belt conveyors, cranes, electric vehicles, pumps, fans, etc. The absence of this sensor
(speed, position or torque) reduces cost and size, and increases reliability of the overall
system. Furthermore, these sensors are often difficult to install in certain applications and
are susceptible to electromagnetic interference. In fact, sensorless servomechanism may
substitute a measure value by an estimated one without deteriorating the drive dynamic
performance especially under uncertain load torque.
Many approaches for IM sensorless servomechanism have been proposed in the literature
is related to vector-controlled methodologies. One of the proposed nonlinear control
methodologies is based on Feedback Linearization Control (FLC), as first introduced by
(Marino et al., 1990). FLC provides rotor speed regulation, rotor flux amplitude decoupling
and torque compensation. Although the strategy presented by (Marino et al., 1990) was not
a sensorless control strategy, fundamental principles of FLC follow servomechanism design
of sensorless control strategies, such as (Gastaldini & Grundling, 2009; Marino et al., 2004;
Montanari et al., 2007; 2006).
The purpose of this chapter is to present the development of two FLC control strategies in the
presence of torque disturbance or load variation, especially under low rotor speed conditions.
Both control strategies are easily implemented in fixed point DSP, such as TMS320F2812 used
on real time experiments and can be easily reproduced in the industry. Furthermore, an
analysis comparing the implementation and the limitation of both strategies is presented. In
order to implement the control law, these algorithms made use of only two-phase IM stator
currents measurement. The values of rotor speed and load torque states used in the control
algorithm are estimated using a Model Reference Adaptive System (MRAS) (Peng & Fukao,
1994) and a Kalman filter (Cardoso & Gründling, 2009), respectively.
This chapter is organized as follows: Section 2 presents the fifth-order IM mathematical
model. Section 3 introduces the feedback linearization modelling of IM control. A simplified
FLC control strategy is described in Section 4. The proposed methods for speed and torque
estimation, MRAS and Kalman filter algorithms, respectively, are developed in Sections 5 and
6. State variable filter is used to obtain derivative signals necessary for implementation of the
control algorithm, and this is presented in section 7. Digital implementation in fixed point
DSP TMS320F2812 and real time experimental results are given in Section 8. Finally, Section 9
presents the conclusions.
2. Induction motor mathematical model

A three-phase N pole pair induction motor is expressed in an equivalent two-phase model in
an arbitrary rotating reference frame (q-d), according to (Krause, 1986) and (Leonhard, 1996)
according to the fifth-order model, as
d
Iqs = − a12 Iqs − ω s Ids + a13 a11 λqr − a13 Nωλdr + a14 Vqs (1)
dt
d
I = − a12 Ids + ω s Iqs + a13 a11 λdr + a13 Nωλqr + a14 Vds (2)
dt ds
d
λqr = − a11 λqr − (ω s − Nω ) λdr + a11 L m Iqs (3)
dt
d
λ = − a11 λdr + (ω s − Nω ) λqr + a11 L m Ids (4)
dt dr
d B T
ω = μ · λdr Iqs − λqr Ids − ω − L (5)
dt J J

Te = μ · J · λdr Iqs − λqr Ids (6)

In equations (1)-(6): Is = Iqs , Ids , λ r = λqr , λdr and Vs = Vqs , Vds denote stator current,
rotor flux and stator voltage vectors, where subscripts d and q stand for vector components in
(q-d) reference frame; ω is the rotor speed, the load torque TL , electric torque Te and ω s is the
stationary speed, θ0 is the angular position of the (q-d) reference frame with respect to a fixed
stator reference frame (α-β) , where physical variables are defined. Transformed variables
related to three-phase (RST) system are given by
xαβ = K · xRST (7)

Let
xqd = e jθ0 xαβ (8)
⎡ ⎤
−
1
−
1
cos θ0 − sin θ0 2⎢ 1 √2 √2 ⎥
with e =
jθ and K = ⎣ 3 ⎦
0 .
sin θ0 cos θ0 3 3
0 − −
2 2
xqd and xαβ stand for two-dimensional voltage flux and stator current vector, respectively on
(q-d) and (α-β)reference frame.
The relations between mechanical and electrical parameters in the above equations are

Δ Δ Rr Δ L s Lr R s L2m Δ Lm Δ Lr Δ NL m
a0 = L s Lr − L m , a11 =
2 ,a = + a , a13 = ,a = and μ = ;
Lr 12 a0 L s a0 11 a0 14 a0 JLr
Feedback Linearization of Speed-Sensorless Induction Motor Control with Torque Compensation 99
where Rs , Rr , L s and Lr are the stator/rotor resistances and inductances, L m is the magnetizing
inductance, J is the rotor inertia, B is the viscous coefficient and N is the number of pole pairs.
In the control design, the viscous coefficient of (5) is considered to be approximately zero, i.e.
B ≈ 0.
3. Feedback Linearization Control

The feedback Linearization Control (FLC) general specifications are two outputs - rotor speed
and rotor flux modulus, as
T T
Δ
y1 = ω λ2qr + λ2dr = ω |λ r | (9)
which is controlled by two-dimensional stator voltage vector Vs , on the basis of measured
variables vector y2 = Is . The development concept of this control strategy is completely
described in (Marino et al., 1990) and it will be omitted here. Following the concept of indirect
field orientation developed by Blaschke, (Krause, 1986) and (Leonhard, 1996), the purpose of
FLC control is to align rotor flux vector with the d-axis reference frame, i.e.
λdr = |λ r | λqr = 0 (10)

The condition expressed in (10) guarantees the exact decoupling of flux dynamics of (1)-(4)
from the speed dynamics. Once rotor flux is not directly measured, only asymptotic field
orientation is possible, according to (Marino et al., 1990) and (Peresada & Tonielli, 2000), then
lim λdr = |λ r | lim λqr = 0 (11)

t→ ∞ t→ ∞
T
It is defined y ∗1 = ωre f λr∗ , where ωre f and λr∗ are reference trajectories of rotor speed
and rotor flux. The speed tracking, flux regulation control problem under speed sensorless
conditions is formulated considering IM model (1)-(5) under the following conditions:
(a) Stator currents are measurable;
(b) Motor parameters are known and considered constant;
(c) Load torque is estimated and it is applied after motor flux excitation;
(d) Initial conditions of IM state variables are known;
(e) λr∗ is the flux constant reference value and estimated speed ω and reference speed ωre f are
the smooth reference bounded speed signals.
FLC equations are developed considering the fifth-order IM model under the assumption that
estimated speed tracks real speed, and therefore it is acceptable to replace measured speed
with estimated speed ( i.e. ω k ≈ ω). In addition, the torque value is estimated using a Kalman
filter. Fig. 1 presents the block diagram of FLC Control.
3.1 Flux controller

From the decoupling properties of field oriented transformation (10), the control objective of
the flux controller is to generate a flux vector aligned with the d-axis to guarantee induction
motor magnetization.
Then, substituting (10) in (4)

∗ d 1
ids = a11 λr∗ + |λ r | (12)
dt a11 L m
I ds
-
+ PI
PI vds* Indution
uds vqs* qd
d 3f Motor
d I ds q rst
λr dt
dt
λr SVF Current u IRST
Flux ds - vds* VRST
SVF Control Ids +
Control rst
i ds* ab
µk
ω
ωref I qs i Sαβ vSαβ
-
µk
ω ew iq + PI
PI
- d u qs Speed MRAS
+ PI
PI I qs
dt Estimation
SVF Current u qs - vqs*
iq + µ
ω
Speed - Control Iqs
Control +
d i qs* Kalman
wref µk
ω Filter
dt
SVF T µk
L ω TL
wref
µk
ω q I qs Te Tm w
+
+ ò Ktn +
-
1/J
s + B/J
ωs TL
Fig. 1. Feedback Linearization Control proposed
The rotor flux |λr | is estimated by a model derived from the induction
motor mathematical
model, (3) and (4), that makes use of measured stator currents Iqs , Ids and estimated speed
variables.
ω
d
λr = − a11 λr − j (ω s − Nω ) λr + a11 L m Is (13)
dt
a Lm ∗
where the stationary speed is ω s = N ω + 11 ∗ iqs . The digital implementation of the flux
λ
controller is made using Euler discretization and the derivative rotor flux signal is obtained
by a state variable filter (SVF).
3.2 Speed controller

The speed control algorithm uses the same strategy adopted for the flux subsystem and it is
computed from (5), as

1 L
T d
iq = + ωre f (14)
μλr∗ J dt
To compensate for speed error between estimated speed and reference speed, (i.e.
eω = ω
− ωre f ), a proportional integral compensation is proposed, as follows

k i_iq
iq = k p_iq + eω (15)
s
These gains values (k p_iq , k i_iq ) are determined considering an induction motor mechanical
model. The reference quadrature component stator speed current is derived from (14)-(15), as
∗
iqs = iq − iq (16)
In DSP implementation, the speed controller is discretized using the Euler method and the
rotor speed derivative (14) is computed by a SVF.
3.3 Currents controller

From (1) and (2), the currents controller is obtained, as

1 ∗ ∗ ∗ d
u qs = a12 iqs + ω s ids + a13 λr N ωre f + eω + Iqs (17)
a14 dt
and

1 ∗ ∗ d
u ds = a12 ids + ω s iqs + a11 a13 λr + I (18)
a14 dt ds
where proportional integral gains of the current error

k
u qs = k pv + iv i qs (19)
s
and

k iv
u ds = k pv + i ds (20)
s
in which
iqs = Iqs − iqs∗ and ∗ .
ids = Ids − ids
These gains (k pv , k iv ) are determined considering a simplified induction motor electrical
model, which is obtained by load and locked rotor test. Hence, current controllers are
expressed as
v∗qs = u qs − u qs (21)
v∗ds = u ds − uds (22)

In DSP, currents controller are digitally implemented using discretized equation (17)-(22)
based on the Euler method, and the stator current derivative is obtained by SVF using stator
currents measures.
4. Simplified feedback linearization control

In order to reduce the number of computation requirements, a simplified feedback
linearization control scheme is proposed. In this control scheme, one part of the current
controller (6)-(7) is suppressed and only a proportional integral controller is used. This
modification minimizes the influence of parameters variation in the control system.
Fig. 2 presents the block diagram of the Simplified FLC proposed.
The currents controller of simplified FLC are defined as

k iv
v∗qs = k pv + i qs (23)
s

k iv
v∗ds = k pv + i ds (24)
s
vds* Indution
d vqs* qd 3f Motor
λr I ds
dt q rst
λr Flux i ds* - vds*
SVF + PI
PI IRST VRST
Control
rst
ωref ab
µk
ω ew iq i Sαβ vSαβ
-
+ PI
PI
I qs Speed MRAS
Speed iq - i qs* - vqs* Estimation
+ + PI
PI
Control µ
d ω
wref
dt
Kalman
SVF TL Filter
wref
µk
ω TL
µk
ω q I qs Te Tm w
+
+ ò Ktn +
-
1/J
s + B/J
ωs TL
Fig. 2. Proposed Simplified Feedback Linearization Control

Flux and Speed Controller are computed exactly as in the previous scheme, as (12) and
(14)-(16).
5. Speed estimation - MRAS algorithm

A squirrel-cage three-phase induction motor model expressed in a stationary frame can be
modelled using complex stator and rotor voltage as in (Peng & Fukao, 1994)
d d
vs = Rs is + Ls is + Lm ir (25)
dt dt
for squirrel-cage IM v r = 0
d d
0 = Rr ir − jNωLr ir − jNωL m is + Lr ir + Lm is (26)
dt dt
The voltage and the current space vectors are given as x = xα + jx β , x ∈ {v s , is , ir }, relative
to the transformed variables present in (7). The induction motor magnetizing current is
expressed by
Lr
im = ir + is (27)
Lm
Two independent observers are derived to estimate the components of the
counter-electromotive vectors.

L2 L2 1 1
em = m im = m ωim − im + is
(28)
Lr Lr Tr Tr
d
em = v s − Rs is − σL s is (29)
dt
Lm
where σ = 1 − . The instantaneous reactive power maintains the magnetizing current,
L s Lr
and its value is defined by cross product of the counter-electromotive and stator current vector
q m = is ⊗ em (30)
Substituting (28) and (29) for em in (30) and noting that is ⊗ is = 0, which gives

d
q m = is ⊗ v s − σL s is (31)
dt
and

L2m 1
m =
q (im is ) ω + (im ⊗ is ) (32)
Lr Tr
Then, q m is the reference model of reactive power and q m is the adjustable model. The
estimated speed is produced by the proportional integral adaptation mechanism error of both
models, and an MRAS system can be drawn as in Fig.3
This algorithm is customary for speed estimation and simple to implement in fixed point DSP,
such as in (Gastaldini & Grundling, 2009; Orlowska-Kowalska & Dybkowski, 2010; Vieira
et al., 2009).
The SVF blocks are state variable filters and are explained in greater detail in Section 7. These
filters compute derivative signals and are applied in voltage signals to avoid addition noise
and phase delay among the vectors as was proposed by (Martins et al., 2006).
6. Load torque estimation - Kalman filter

The reduced mechanical IM system can be represented by the following equations
⎡ ⎤ ⎡ ⎤
Bn 1 1
d ω − − ω
=⎣ J J ⎦ + ⎣ J ⎦ Te (33)
dt TL 0 0
TL
0

ω
y= 1 0 (34)
TL
The Kalman Filter could be used to provide the value of torque load or disturbances - TL .
Since (15)-(16) is nonlinear, the Kalman filter linearizes the model at the actual operating
Reference Model
i Sαβ
SVF qm
æ d ö q m = i Sαβ Ä e m
e m = vSαβ - ç R s i Sαβ + sL s i Sαβ ÷ +
-
vSαβ è dt ø
SVF $
q m
Adaptative Model
d µk Ä i - 1 i + 1 i æµ 1 1 ö
im = w m m Sαβ
e$ m = L m ç wk Ä im - im + i Sαβ ÷
dt TR TR è TR TR ø
$ = i Ä e$ m
q m Sαβ
‘
µk
w µ
w Dq m
Kalman Filter PI
Fig. 3. Reactive Power MRAS Speed Estimation
point (Aström & Wittenmark, 1997). In addition, this filter takes into account the signal
noise, which could be generated
⎡ as pulse width⎤ modulation
⎡ drivers.
⎤ Assuming the definitions
T Bn 1 1
− − ⎦
xk = ω k T L , Am = ⎣ J J , Bm = ⎣ J ⎦, Cm = 1 0 and yk = ω.
0 0 0
Then, the recursive equation for the discrete time Kalman Filter (De Campos et al., 2000) is
described by
−1
K(k) = P (k)Cm
T
Cm P (k)Cm
T
+R (35)
where K(k) is the Kalman gain. The covariance matrix P (k) is given by
P (k + 1) = (I − Am ts ) (P (k) − K(k)Cm P (k)) (I − Am ts ) T + (Bm ts ) Q (Bm ts ) T (36)

Therefore, the estimated torque T L is one observed state of the Kalman filter
x k ( k + 1) = ( I − A m t s )
xk (k) + B m ts u (k) + (I − Am ts ) K(k) (ω
− Cm
xk (k)) (37)
T
giving ω≈ω xk (k) = ω
k and k TL .
The matrices R and Q are defined according to noise elements of predicted state variables,
taking into account the measurement noise covariance R and the plant noise covariance Q.
7. State variable filter

The state variable filter (SVF) is used to mathematically evaluate differentiation signals. This
filter is necessary in the implementation of FLC and MRAS algorithms. The transfer function
of SVF is of second order as it is necessary to obtain the first order derivative.
ω sv f
Gsv f = 2 (38)
s + ω sv f
where ω sv f is the filter bandwidth defined at around 5 to 10 times the input frequency signal
u sv f .
The discretized transfer function, using the Euler method, can be performed in state-space as
xsvf (k + 1) = Asvf xsvf (k) + Bsvf u sv f (k) (39)

1 1 0 x1
where Asvf = , Bsvf = and xsvf = .
− ω sv
2
f 1 − 2ω sv f ω 2
svf x2
The state variables x1 and x2 represent the input filtered signal and input derivative signal.
Sensorless control schemes were implemented in DSP based platform using TMS 320F2812.
Experimental results were carried out on a motor with specifications: 1.5cv, 380V, 2.56A, 60
Hz, Rs = 3.24Ω, Rr = 4.96Ω, Lr = 404.8mH, L s = 402.4mH, L m = 388.5mH, N = 2 and
nominal speed of 188 rad/s.
The experimental analyses are carried out with the following operational sequence:
1) The motor is excited (during 10 s to 12 s) using a smooth flux reference trajectory.
2) Starting from zero initial value, the rotor speed reference grows linearly until it reaches the
reference value. Thus, the reference rotor speed value is kept constant.
3) During stand-state, a step constant load torque is applied.
In order to generate load variation for torque disturbance analyses, the DC motor is connected
to an induction motor driving-shaft. Then, the load shaft varies in accordance with DC motor
field voltage and inserting a resistance on its armature. Fig. 4 and Fig. 5 depict performance
of both control schemes: FLC control and simplified FLC control with rotor speed reference of
18 rad/s. In these figures measure speed, estimated speed, stator (q-d) currents and estimated
torque are illustrated.
Fig.6 and Fig. 7 present experimental results with 36 rad/s rotor speed reference.
Fig. 8 and Fig. 9 show FLC control and Simplified FLC with 45 rad/s rotor speed reference.
The above figures present experimental results for low rotor speed range of FLC control and
Simplified FLC control applying load torque. In accordance with the figures above, both
control schemes present similar performances in steady state. It is verified that both schemes
respond to compensated torque variations. With respect to Simplified FLC, it is necessary to
carrefully select fixed gains in order to guarantee the alignment of the rotor flux on the d axis.
9. Conclusion
Two different sensorless IM control schemes were proposed and developed based on
nonlinear control - FLC Control and Simplified FLC Control. These control schemes are
composed of a flux-speed controller, which is derived from a fifth-order IM model. In the
implementation of feedback linearization control (FLC), the control algorithm presents a large
number of computational requirements. In the simplified FLC scheme, a substitution of FLC
currents controllers by two PI controllers is proposed to generate the stator drive voltage.
In order to provide the rotor speed for both control schemes, a MRAS algorithm based on
reactive power is applied.
To correctly evaluate whether this Simplified FLC does not affect control performance, a
comparative experimental analysis of a FLC control and a simplified FLC control is presented.
Experimental results in DSP TMS 320F2812 platform show the performance of both systems
2
Current Id (A)
1.5
0.5
0
0 10 20 30 40 50 60 70
T ime(s)
(a) IM Stator Current Ids
2
Current Iq (A)
1.5
0.5
−0.5
0 10 20 30 40 50 60 70
T ime(s)
(b) IM Stator Current Iqs
20
Rotor Speed (rad/s)
15
10 ω
5 ω̂ k
0
−5
0 10 20 30 40 50 60 70
T ime(s)
(c) Rotor Speed - Estimated and Encoder Measurement
2
Load T orque (N.m)
−1
−2
0 10 20 30 40 50 60 70
T ime(s)
(d) Estimated Load Torque
Fig. 4. FLC control with 18 rad/s rotor speed reference

1.5
Current Id (A)
0.5
0
0 10 20 30 40 50 60 70
T ime(s)
1.5
Current Iq (A)
0.5
−0.5
0 10 20 30 40 50 60 70
T ime(s)
20
Rotor Speed (rad/s)
15
10 ω
5 ω̂ k
0
−5
0 10 20 30 40 50 60 70
T ime(s)
1.5
Load T orque (N.m)
0.5
−0.5
0 10 20 30 40 50 60 70
T ime(s)
Fig. 5. Simplified FLC control with 18 rad/s rotor speed reference

1.5
Current Id (A)
1
0.5
−0.5
0 10 20 30 40 50 60 70
T ime(s)
1.5
Current Iq (A)
0.5
−0.5
0 10 20 30 40 50 60 70
T ime(s)
40
Rotor Speed (rad/s)
30
20 ω
ω̂ k
10
0 10 20 30 40 50 60 70
T ime(s)
2
Load T orque (N.m)
−1
−2
0 10 20 30 40 50 60 70
T ime(s)

2.5
Current Id (A)
2
1.5
0.5
0
0 10 20 30 40 50 60 70
T ime(s)
2
Current Iq (A)
−1
0 10 20 30 40 50 60 70
T ime(s)
40
Rotor Speed (rad/s)
30
20 ω
ω̂ k
10
0 10 20 30 40 50 60 70
T ime(s)
2
Load T orque (N.m)
−1
0 10 20 30 40 50 60 70
T ime(s)

1.5
Current Id (A)
1
0.5
−0.5
−1
0 10 20 30 40 50 60 70
T ime(s)
2
Current Iq (A)
−1
0 10 20 30 40 50 60 70
T ime(s)
50
Rotor Speed (rad/s)
40
30
ω
20
ω̂ k
10
−10
0 10 20 30 40 50 60 70
T ime(s)
2
Load T orque (N.m)
−1
−2
0 10 20 30 40 50 60 70
T ime(s)

2.5
2
Current Id (A)
1.5
0.5
0
0 10 20 30 40 50 60 70
T ime(s)
2
Current Iq (A)
−1
0 10 20 30 40 50 60 70
T ime(s)
50
Rotor Speed (rad/s)
40
30
ω
20
ω̂ k
10
−10
0 10 20 30 40 50 60 70
T ime(s)
2
Load T orque (N.m)
−1
0 10 20 30 40 50 60 70
T ime(s)

in the 18rad/s, 36 rad/s and 45 rad/s rotor speed range. Both control schemes present
similar performance in steady-state. Hence, the proposed modification of the FLC control
allows a simplification of the control algorithm without deterioration in control performance.
However, it may necessary to carefully evaluate the gain selection in the simplified FLC
control, to guarantee rotor flux alignment on the d axis, as well as, to guarantee speed-flux
decoupling. Both control schemes indicate sensitivity with model parameter variation, and
one way to overcome this would be the is development of an adaptive FLC control laws on
FLC control.
10. References
Aström, K. & Wittenmark, B. (1997). Computer-Controlled Systems: Theory and Design,
Prentice-Hall.
Cardoso, R. & Gründling, H. A. (2009). Grid synchronization and voltage analysis based on
the kalman filter, in V. M. Moreno & A. Pigazo (eds), Kalman Filter Recent Advances
and Applications, InTech, Croatia, pp. 439–460.
De Campos, M., Caratti, E. & Grundling, H. (2000). Design of a position servo with induction
motor using self-tuning regulator and kalman filter, Conference Record of the 2000 IEEE
Industry Applications Conference, 2000.
Gastaldini, C. & Grundling, H. (2009). Speed-sensorless induction motor control with torque
compensation, 13th European Conference on Power Electronics and Applications, EPE ’09,
pp. 1 –8.
Krause, P. C. (1986). Analysis of electric machinery, McGraw-Hill.
Leonhard, W. (1996). Control of Electrical Drives, Springer-Verlag.
Marino, R., Peresada, S. & Valigi, P. (1990). Adaptive partial feedback linearization of
induction motors, Proceedings of the 29th IEEE Conference on Decision and Control, 1990,
pp. 3313 –3318 vol.6.
Marino, R., Tomei, P. & Verrelli, C. M. (2004). A global tracking control for speed-sensorless
induction motors, Automatica 40(6): 1071 – 1077.
Martins, O., Camara, H. & Grundling, H. (2006). Comparison between mrls and mras applied
to a speed sensorless induction motor drive, 37th IEEE Power Electronics Specialists
Conference, PESC ’06., pp. 1 –6.
Montanari, M., Peresada, S., Rossi, C. & Tilli, A. (2007). Speed sensorless control of induction
motors based on a reduced-order adaptive observer, IEEE Transactions on Control
Systems Technology 15(6): 1049 –1064.
Montanari, M., Peresada, S. & Tilli, A. (2006). A speed-sensorless indirect field-oriented
control for induction motors based on high gain speed estimation, Automatica
42(10): 1637 – 1650.
Orlowska-Kowalska, T. & Dybkowski, M. (2010). Stator-current-based mras estimator for a
wide range speed-sensorless induction-motor drive, IEEE Transactions on Industrial
Electronics 57(4): 1296 –1308.
Peng, F.-Z. & Fukao, T. (1994). Robust speed identification for speed-sensorless vector control
of induction motors, IEEE Transactions on Industry Applications 30(5): 1234 –1240.
Peresada, S. & Tonielli, A. (2000). High-performance robust speed-flux tracking controller for
induction motor, International Journal of Adaptive Control and Signal Processing, 2000.
Vieira, R., Azzolin, R. & Grundling, H. (2009). A sensorless single-phase induction motor drive
with a mrac controller, 35th Annual Conference of IEEE Industrial Electronics,IECON
’09., pp. 1003 –1008.
7
From Dynamic Modeling to Experimentation of

Induction Motor Powered by Doubly-Fed
Induction Generator by Passivity-Based Control
M. Becherif1 , A. Bensadeq2, E. Mendes3, A. Henni4 ,
P. Lefley5 and M.Y Ayad6
1 UTBM, FEMTO-ST/FCLab, UMR CNRS 6174, 90010 Belfort Cedex
2 AElectrical Power & Power Electronics Group, Department of Engineering
3 Grenoble INP - LCIS/ESISAR, BP 54 26902 Valence Cedex 9
4 Alstom Power - Energy Business Management
5 Electrical Power & Power Electronics Group, Department of Engineering
University of Leicester
6 IEEE Member
1,3,4,6 France
2,5 UK
1. Introduction
DFIG wind turbines are nowadays more widely used especially in large wind farms. The
main reason for their popularity when connected to the electrical network is their ability to
supply power at constant voltage and frequency while the rotor speed varies, which makes
it suitable for applications with variable speed, see for instance (10), (11). Additionally,
when a bidirectional AC-AC converter is used in the rotor circuit, the speed range can be
extended above its synchronous value recovering power in the regenerative operating mode
of the machine. The DFIG concept also provides the possibility to control the overall system
power factor. A DFIG wind turbine utilizes a wound rotor that is supplied from a frequency
converter, providing speed control together with terminal voltage and power factor control
for the overall system.
DFIGs have been traditionally used to convert mechanical power into electrical power
operating near synchronous speed. Some advantages of DFIGs over synchronous or squirrel
cage generators include the high overall efficiency of the system and the low power rating
of the converter, which is only rated by the maximum rotor voltage and current. In a
typical scenario the prime mover is running at constant speed, and the main concern is the
static optimization of the power flow from the primary energy source to the grid. A good
introduction to the operational characteristic of the grid connected DFIG can be found in (5).
We consider in this paper the isolated operation of a DFIG driven by a prime mover, with
its stator connected to a load—which is in this case an IM. Isolated generating units are
economically attractive, hence increasingly popular, in the new era of the deregulated market.
The possibility of a DFIG supplying an isolated load has been indicated in (6), (7) where some
mention is made of the steady–state control problem. In (8) a system is presented in which the
rotor is supplied from a battery via a PWM converter with experimental results from a 200W
prototype. A control system based on regulating the rms voltage of the DFIG is used which
results in large voltage deviations and very slow recovery following load changes. See also
(9; 12) where feedback linearization and sliding mode principles are used for the design of the
motor speed controller.
This paper presents a dynamic model of the DFIG-IM and proves that this system is
Blondel-Park transformable. It is also shown that the zero dynamics is unstable for a certain
operating regime. We implemented the passivity-based controller (PBC) that we proposed
in (3) to a 200W DFIG interconnected with an IM prototype available in IRII-UPC (Institute
of Robotics and Industrial Informatics - University Polytechnic of Catalonia). The setup
is controlled using a computer running RT-Linux. The whole system is decomposed in a
mechanical subsystem which plays the role of the mechanical speed loop, controlled by a
classical PI and an electrical subsystem controlled by the PBC where the model inversion was
used to build a reference model.
The proposed PBC achieves the tracking control of the IM mechanical rotor speed and flux
norm, the practical advantage of the PBC consists of using only the measurements of the two
mechanical coordinates (Motor and Generator positions). The experiments have shown that
the PBC is robust to variations in the machines’ parameters.
In addition to the PBC applied to the electrical subsystem, we proposed a classical PI
controller, where the rotor voltage control law is obtained via a control of the stator currents
toward their desired values, those latter are obtained by the inversion of the model.
In the sequel, and for the control of the electrical subsystem a combination of the PBC +
Proportional action for the control of the stator currents is applied. The last controller is a
combination of PBC + PI action for the control of the stator currents.
The stability analysis is presented. The simulations and practical results show the
effectiveness of the proposed solutions, and robustness tests on account of variations in
the machines’ parameters are also presented to highlight the performance of the different
controllers.
The main disadvantage of the DFIG is the slip rings, which reduce the life time of the
machine and increases the maintenance costs. To overcome this drawback an alternative
machine arrangement is proposed, in section 6, which is the Brushless Doubly Fed twin
Induction Generator (BDFTIG). The system is anticipated as an advanced solution to the
conventional doubly fed induction generator (DFIG) to decrease the maintenance cost and
develop the system reliability of the wind turbine system. The proposed BDFTIG employs two
cascaded induction machines each consisting of two wound rotors, connected in cascade to
eliminate the brushes and copper rings in the DFIG. The dynamic model of BDFTIG with two
machines’ rotors electromechanically coupled in the back-to-back configuration is developed
and implemented using Matlab/Simulink.
2. System configuration and mathematical model

The configuration of the system considered in this paper is depicted in Fig.1. It consists of a
wound rotor DFIG, a squirrel cage IM and an external mechanical device that can supply or
extract mechanical power, e.g., a flywheel inertia. The stator windings of the IM are connected
to the stator windings of the generator whose rotor voltage is regulated by a bidirectional
converter. The electrical equivalent circuit is shown in Fig. 2. The main interest in this
Induction Motor Powered by Doubly-Fed Induction Generator by Passivity-Based Control 115
configuration is that it permits a bidirectional power flow between the motor, which may
operate in regenerative mode, and the generator.
Battery bank
with converter Primary mechanical
energy source
DC-bus
Flywheel
ω mM
*
Controller Inverter
(PI+PBC)
ω mM Inertia
DFIG IM
Fig. 1. System configuration with speed controller.

DFIG MI
irG RrG LrG − LmG LsG − LmG RsG isG isM RsM LsM − LmM LrM − LmM irM
vrG λrG LmG vsG = vsM LmM RrM

λsG λsM λrM
Fig. 2. Equivalent circuit of the DFIG with IM.

In Fig. 3, we show a power port viewpoint description of the system. The DFIG is a three–port
system with conjugated power port variables1 prime mover torque and speed, (τLG , ω G ), and
rotor and stator voltages and currents, (vrG , irG ), (vsG , isG ), respectively. The IM, on the other
hand, is a two–port system with port variables motor load torque and speed, (τLM , ω M ), and
stator voltages and currents. The DFIG and the IM are coupled through the interconnection
vsG = vsM
isG = −isM . (1)
isM ωm
v rG i rG DFIG vsM = vsG IM τ LM
ωG τ LG
Fig. 3. Power port representation of the DFIG with IM.
To obtain the mathematical model of the overall system ideal symmetrical phases with
uniform air-gap and sinusoidally distributed phase windings are assumed. The permeability
1
The qualifier “conjugated power" is used to stress the fact that the product of the port variables has the
units of power.
of the fully laminated cores is assumed to be infinite, and saturation, iron losses, end winding
and slot effects are neglected. Only linear magnetic materials are considered, and it is further
assumed that all parameters are constant and known. Under these assumptions, the voltage
balance equations for the machines are
λ̇sG + RsG isG = vsG (2)

λ̇rG + RrG irG = vrG (3)
λ̇sM + RsM isM = vsM (4)
λ̇r M + Rr M ir M = 0 (5)
where λsG , λrG (λsM , λr M ) are the stator and rotor fluxes of the DFIG (IM, resp.), L sG , LrG , L mG
(L sM , Lr M , L mM ) are the stator, rotor, and mutual inductances of the DFIG (IM, resp.); RsG , RrG
(RsM , Rr M ) are the stator and rotor resistances of the DFIG (IM, resp.).
The interconnection (1) induces an order reduction in the system. To eliminate the redundant
coordinates, and preserving the structure needed for application of the PBC, we define
λsG M = λsG − λsM
which upon replacement in the equations above, and with some simple manipulations, yields
the equation
λ̇ + Ri = BvrG (6)
where we have defined the vector signals

⎡ ⎤ ⎡ ⎤
λrG irG
λ = ⎣ λsG M ⎦ , i = ⎣ isG ⎦ ,
λr M ir M
and the resistance and input matrices
RrG I2 , ( RsG + RsM ) I2 , Rr M I2

T
R = diag{ }, B= I2 0 ∈ IR6×2
R1 R2 R3
To complete the model of the electrical subsystem, we recall that fluxes and currents are
related through the inductance matrix by
λ = L (θ )i, (7)
where the latter takes in this case the form
⎡ ⎤
L rG I2 L mG e − Jn G θG 0
L ( θ ) = ⎣ L mG e Jn G θG ( LsG + LsM ) I2 − LmM e Jn M θ M ⎦ (8)
0 − LmM e − Jn Mθ M L rM I2
where n G , n M denote the number of pole pairs, θ G , θ M the mechanical rotor positions (with
respect to the stator) and to simplify the notation we have introduced

θG 0 −1 cos( x ) − sin( x )
θ= , J= = − J T , e Jx = = (e− Jx ) T .
θM 1 0 sin( x ) cos( x )
⎡ ⎤
[ LrM ( LsG + LsM ) − L2mM ] I2 − LmG LrM e − Jn G θG − LmG LmM e − J (n GθG −n M θ M )
1 ⎣ ⎦
L −1 ( θ ) = − LmG LrM e Jn G θG L rG L rM I2 L rG L mM e Jn Mθ M (9)
Δ
− LmG LmM e J (n GθG −n M θ M ) L rG L mM e − Jn Mθ M [ LrG ( LsG + LsM ) − L2mG ] I2
⎡ ⎤
L
L12 L13
1 ⎢ 11 T ⎥
⎣ L12 L22 L23 ⎦ (10)
Δ T T
L11 L23 L33
where
Δ = LrG [ Lr M ( L sG + L sM ) − L2mM ] − Lr M L2mG < 0 (11)
We recall that, due to physical considerations, R > 0, L (θ ) = L T (θ ) > 0 and L −1 ( θ ) =
T
L −1 (θ ) > 0.
A state–space model of the (6–th order) electrical subsystem is finally obtained replacing (7)
in (6) as
Σe : λ̇ + RL (θ )−1 λ = BvrG (12)
The mechanical dynamics are obtained from Newton’s second law and are given by
Σm : Jm θ̈ + Bm θ̇ = τ − τL (13)
where Jm = diag{ JG , J M } > 0 is the mechanical inertia matrix, Bm = diag{ BG , B M } ≥ 0

contains the damping coefficients, τL = [ τLG , τLM ] T are the external torques, that we will
assume constant in the sequel. The generated torques are calculated as usual from

τG 1 ∂ T
τ= =− λ [ L (θ )] −1 λ . (14)
τM 2 ∂θ
From (7), we obtain the alternative expression
1 ∂ T
τ= i L ( θ )i .
2 ∂θ
The following equivalent representations of the torques, that are obtained from direct
calculations using (7), (8) and (14), will be used in the sequel
⎡ T Je − JnG θ G i
⎤
− L mG irG sG
τ = ⎣ ⎦ (15)
T Je Jn M θ M i
− L mM isG rM
⎡ ⎤
− RsGn+GRsM λ̇sG Jn M θ M i
M J ( λ sG M − L mM e r M)
T
⎢ ⎥
= ⎣ ⎦ (16)
nM T
R rM λ̇r M Jλr M
2.1 Modeling of the DFIG-IM in the stator frame of the two machines
It has been shown in (4) and (3) that the DFIG-IM is Blondel–Park transformable using the
following rotating matrix:
⎡ ⎤
e( Jσ ) 0 0
⎢ ⎥
Rot(σ, θ G , θ M ) = ⎣ 0 e( J ( σ +nGθG )) 0 ⎦ (17)
0 0 e( J ( σ +nGθG −n M θ M ))
where σ is an arbitrary angle.

The model of the DFIG-IM in the stator frame of the two machines is given by (see (4) and (3)
for in depth details):
⎧ ⎡ ˙ ⎤ ⎡ ⎤⎡ ⎤ ⎡ ⎤
⎪
⎪
λrG aI2 − n G θ̇ G J bI2 0
λrG I2
⎪
⎪ ⎢ ⎥
⎨ ⎣ λ̇sMG ⎦+⎣ aI2 − n G θ̇ G J
⎪ eI2 − cI2 + n M θ̇ M J ⎦ ⎣ λsMG ⎦ =⎣ I2 ⎦ vrG
˙
λr M 0 − dI2 cI2 − n M θ̇ M J
λr M 0 (18)
⎪
⎪
⎪
⎪ JG ω̇ G BG 0 ωG
f λsMG J λrG
T − τLG
⎪
⎩ + + =
J M ω̇ M 0 BM ωM − f λsMGT Jλr M − τLM
or
⎧ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
⎪
⎪ ˙ rG
λ irG I2
⎪
⎪ ⎢ ⎥
⎪
⎨ ⎣ λ̇sMG ⎦ + ( R + L G n G θ̇ G + L M n M θ̇ M ) ⎣ isM ⎦ = ⎣ I2 ⎦ vrG
˙
λr M
ir M 0 (19)
⎪
⎪
⎪
⎪ J ω̇ B 0 ω f λ T J
λ − τLG
⎪
⎩ G G
+ G G
+ sMG rG
=
J M ω̇ M 0 BM ωM − f λsMG
T Jλr M − τLM
λsMG corresponds to the total leakage flux of the two machines referred to the stators of the
machines.
L sMG represent the total leakage inductance.
with
⎧ ⎡ ⎤
⎪ rG I2
R 0 0
⎪
⎪
⎪
⎪ R=⎣ R r M I2 ⎦
rG I2 ( RsG + RsM ) I2 − R
⎪
⎨
0 0
R
⎡ ⎤ ⎡r M I2 ⎤
⎪
⎪ − J J 0 0 0 0
⎪
⎪
⎪ ⎣ ⎦ et L M = L MM ⎣ 0 J ⎦
⎩ L G = L MG − J J 0
⎪ J
0 0 0 0 −J −J
with the positive parameters: a = R rG L −1 , b = R rG L −1 , c = R

r M L −1 , d = R
r M L −1 ,
MG sMG MM sMG
rG + RsG + RsM + Rr M ) L − −
e = (R sMG , f = L sMG , and the following transformations:
1 1
L mG JnG θG L mG JnG θG
λrG = e λrG , vrG = e vrG
LrG LrG
L mM Jn M θ M L
λr M = e λr M , irG = rG e JnG θG irG
Lr M L mG
3. Properties of the model

In this section, we derive some passivity and geometric properties of the model that will be
instrumental to carry out our controller design.
3.1 Passivity
An explicit power port representation of the DFIG interconnected to the IM is presented in
Fig.4
vrG irG
τLG -wG
− ∑G
isG isM
vsM = vsG
-τLM
∑ M
wM
Fig. 4. Explicit power port representation of the DFIG with IM.
Claim 1. The interconnection of the DFIG with the IM presented ⎡ in ⎤the explicit
⎡ power
⎤ port
vrG irG
representation in Fig.4 is a passive system2 with the passive map ⎣ τLG ⎦ → ⎣ − ω G ⎦.
− τLM ωM
⎡ ⎤ ⎡ ⎤
vrG irG
ΣG
Proof. Consider the Fig. 4, Σ = is passive ⇒ ⎣ τLG ⎦ → ⎣ − ω G ⎦ is passive ?
ΣM
− τLM ω
T M
For this purpose, we have to prove that vrG irG − τLG ω G − τLM ω M ≥ 0. We know that
each machine separately is passive (see [1])

Σ G is a passive system ⇔ T
vrG irG − τLG ω G + isM
T
vsM ≥ 0 (20)

Σ M is a passive system ⇔ T
vsG isG − τLM ω M ≥ 0 (21)
where equation (1) has been used in (20) and (21). Let’s consider

d T
vrG irG − τLG ω G + isM
T
vsM ≥ 0 (22)
T T
Using the energy conserving principle isM vsM = − isG vsG yields

d= T
vrG irG − τLG ω G − isGT
vsG (23)
From (21) we have

− T
isG vsG ≤ − τLM ω M (24)
Finally (23) and (24) yields

T
vrG irG − τLG ω G − τLM ω M ≥ d ≥ 0 (25)
Hence, the passivity of the DFIG interconnected to the IM is proven

2 Passive systems are defined here with no causality relation assumed among the port variables (13). This,
more natural, definition is more suitable for applications where power flow (and not signal behaviour)
is the primary concern.
4. Zero dynamics
For the IM speed control, we are interested in the internal behaviour of the system when the
motor torque τM is constant. In addition, for practical considerations, we are interested in the
control of the IM flux norm | λr M |, where | · | is the Euclidean norm.
For the study of the zero dynamics regarding these two outputs, we consider the DFIG-IM
model3 given by (18).
The control vrG is determined to obtain the desired equilibrium points of the DFIG-IM:
θ̈ G = θ̈ M = 0, θ̇ G = θ̇ G
d
= Constant, θ̇ M = θ̇ dM = Constant,
τLG = τLG0 = Constant, τLM = τLM0 = Constant

T
λr M λr M = β d = Constant > 0
2
(26)
The IM mechanical dynamics show that the desired equilibrium points are obtained if:
τM = τM
d
= τLM0 = Constant
Hence:
T
f λsMG J
λr M = τM = τM
d
= Constant (27)
The equation (27) can also be expressed by replacing λsMG by its value given by the third line
of the electrical subsystem (18):
f ˙ T
λr M + c
λr M − n M θ̇ dM J
λr M J
λr M = τM
d
Hence

f T f T
λ˙ r M J
λr M − n M θ̇ dM
λrTM
λr M = λ˙ r M J
λr M − n M θ̇ dM β2 = τM = τM
d
= cte (28)
d d
The relative degrees of the outputs y1 = β2 = λrTM

λr M and y2 = τM = f λsMG
T J
λr M , regarding
the input control vrG , are 2 and 1, respectively.
The zero dynamics of the mechanical subsystem (18) is stable, since the mechanical parameters
are positive.
Following on we will analyze the zero dynamics of the electrical subsystem considering the
equilibrium points such that (26), (27) and (28) are verified (we will omit the subscript d).
• Consequence of (26):
dβ2 T
˙ r M
=0 ⇒ λ λr M = 0 et λ˙ r M = 0
λrTM
dt
0 =
λrM λ˙ rM = d
T T
λrM λsMG − c T
λrM λrM + n M θ̇ M T
λrM J λrM
with the electrical
c
subsystem of (18), it comes: 0 = d T
λrM λsMG − λrM
d
Hence, with (26), it comes as solution of λsMG :
c
λsMG = λ + αJ
λr M , ∀α ∈ IR (29)
d rM
3 The zero dynamics analysis is independent from the chosen frame.
• Consequence of (27) and (28):

Replace (29) in (27):
c
τM = f
λrTM I2 − αJ J
λr M = f α
λrTM
λr M
d
Since
λrTM
λr M = β2 = Constant:
τM
τM = f β2 α ⇒ α=
f β2
and with τ̇M = 0:
τ̇M = f β2 α̇ ⇒ α̇ = 0
Recall of consequences of (26), (27) and (28):
At the equilibrium, the solutions of
λr M belong to the following set:

T d
λr M ∈ IR2 |
λr M = β2 > 0,
λrTM λrTM λ˙ r M J
λ˙ r M = 0, λr M = τM + n M θ̇ M β2 , β̇ = 0 (30)
f
At the equilibrium, the solutions of λsMG belong to the following set:

c τ
λsMG ∈ IR2 | λsMG = λr M + αJ
λr M , α = M2 , α̇ = 0 (31)
d fβ
Let take
λr M in the form :
β
λr M = e J ( ρ + n M θ M ) (32)
0
with the form (32) and the constrains of (30), it comes:
β̇ = 0 ⇒ λ˙ r M = (ρ̇ + n M θ̇ M ) J
λr M and λ˙ r M = 0
λrTM
T d d d τM
λ˙ r M J
λr M = τM + n M θ̇ dM β2 ⇒ (ρ̇ + n M θ̇ M ) β2 = τM + n M θ̇ M β2 ⇒ ρ̇ = = dα
f f f β2
with α given by (31).
Hence the vectors λsMG and λr M are completely defined by the outputs y1 and y2 .
Analyzing the behaviour of the dynamics of the state λrG :
the substraction of the two upper lines of (18) give:
˙ rG = (b + e)λsMG + λ̇sMG − c
λ λr M + n M θ̇ M J
λr M
By replacing λsMG by its value given by (31), the state becomes:

c
c(b + e − d)
λ˙ rG =
I2 + α(b + e) + n M θ̇ M J λr M + I2 + αJ λ˙ r M
d d
Using the general form (32) and its derivative under the constrains (30), yields:

˙ rG c(b + e − d) c
λ = − α(ρ̇ + n M θ̇ M ) I2 + α(b + e) + n M θ̇ M + (ρ̇ + n M θ̇ M ) J λr M
d d
= r M = M1 e Jγ
[ c1 I2 + c2 J ] λ λr M

with M1 = c21 + c22 = Constant and γ = arctan c2 π.
c1 = Constant if c1 = 0, else γ = 2
Then, with (32):

β 1
λ˙ rG = M1 e Jγ e J ( ρ+n M θ M )
= M1 βe J ( γ +ρ+n Mθ M )
0 0
Consequently,
λrG is of the form:

− M1 β 1
λrG = Je J ( γ +ρ+n M θ M ) + Constant
ρ̇ + n M θ̇ M 0
because ρ̈ = θ̈ M = 0, γ̇ = 0 and Ṁ1 = 0.

We can then conclude that the dynamics of λrG is stable if the desired operating point satisfies:
ρ̇ + n M θ̇ M = 0
Consequently, the zero dynamics (the dynamics of

λrG ) is unstable when the desired operating
point belongs to the slip line defined by:
ρ̇ = − n M θ̇ M
or in terms of the controlled outputs β2 and τM :
L2r M
τM = − 2
n M θ̇ M β2
L mM Rr M
With usual machine parameters, the operating point may belong to the slip line for very low
speed which is not the case with the considered operating points
5. Passivity-based controllers
The PBC achieves the IM speed and rotor flux norm control with all internal signals remaining
bounded under the condition ρ̇ + n M θ̇ M = 0. From a practical point of view it is interesting
to ensure the boundedness of the internal signals and in particular the stator current of the
two machines. For this purpose, two classical controllers (Proportional and Proportional plus
Integral) are applied or combined with the PBC on the stator current i sG .
In this section we address the stability analysis of the following controllers:
1 PBC : Passivity Based control;
2 PBC + P : Passivity Based control + Proportional action on the stator currents i sG ;
3 PBC + PI : Passivity Based control + Proportional plus Integral actions on the stator
currents isG .
As defined in (3) a nested loop control configuration is adopted for the PBC control of the
DFIG with the IM system. We propose to design first a torque tracking PBC for Σe , and then
add a speed tracking loop around it. This leads to the nested-loop scheme depicted in Fig.
5, where Cil is the inner-loop torque tracking PBC and Col is an outer-loop speed controller,
which generates the desired torque, and will be taken as a simple PI controller. The reader is
referred to (1) for motivation and additional details on this control configuration.
ω Md u
Col τ Md Cil τM ωM
Controller
(Mechanical)
Controller
(Electrical)
∑ e ∑ m
Fig. 5. Nested-loop control configuration.
5.1 PBC
To derive the torque tracking PBC we will shape the storage function Hλ (λ), which has a
minimum at zero, to take the form
1
Hλd = λ̃ T R−1 λ̃ ≥ 0 (33)
2
where λ̃ = λ − λd , with λd a signal to be defined. As suggested in (1), we propose to establish
the following relationship between λd and vrG :
BvrG = λ̇d + RL −1 (θ )λd . (34)

Comparing with (12) we see that this, so–called implicit representation of the controller, is
a “copy" of the electrical subsystem but evaluated along some desired trajectories. We will
prove now that this control action indeed shapes the storage function as desired. Combining
(34) with (12) yields the error equation for the fluxes
λ̃˙ + RL −1 (θ )λ̃ = 0. (35)
The derivative of the desired energy function (33) along the trajectory of (35) is
Ḣλd = − λ̃ T L −1 λ̃ ≤ 0 (36)
Hence, λ̃(t) → 0 exponentially.
To complete our torque tracking design there are two remaining issues:
(i) find an explicit representation for the controller (34);
∗ (t ), we have
(ii) select λd such that, for any given desired trajectory τM
∗ ( t );
λ̃(t) → 0 ⇒ τM (t) → τM
5.2 PBC + P
The PBC + P controller is given by the equation below:
BvrG = λ̇d + RL −1 (θ )λd + BK p (isG − isG

d
) (37)
where K p is a proportional positive gain. We have:
1

isG − isG
d
= L21 (λrG − λrGd
) + L22 (λsG M − λdsG M ) + L23 (λr M − λrdM ) (38)
Δ
1

= L21 L22 I
L23 (λ − λd ) = 0 I2 0 L −1 (θ )λ̃ (39)
Δ 2
λ̃ P
with L ij (i = 1,¯ 3, j = 1,¯ 3) and Δ are given by (10) and (11), respect. Then,
BvrG = λ̇d + RL −1 (θ )λd + K p BPL −1 (θ )λ̃ (40)

The closed loop error dynamic can be written as the following:
λ̃˙ = − RL −1 (θ )λ̃ − K p BPL −1 (θ )λ̃ (41)
Consider the desired energy function given by (33), its derivative along the trajectories of (41)
is:
Ḣλd = λ̃ T R−1 λ̃˙

= − λ̃ T L −1 (θ ) λ̃ − λ̃ T R−1 BPL −1 (θ ) λ̃

>0 Q
To prove that the error dynamic (41) is stable, it’s enough to prove that Q is a positive
semi-definite matrix:
⎡ T ⎤
L12 L22
L23
1
Q = R−1 BPL −1 (θ ) = ⎣ 0 0 0 ⎦ (42)
R1 Δ
0 0 0
Q is positive semi-definite if:

X T QX ≥ 0 X ∈ 6×1
⎡ T ⎤

L12 L22
L23
1 T⎣
T
X QX = X 0 0 0 ⎦X (43)
R1 Δ
0 0 0
⎡ T
⎤
1 1
L12 2 L22 2 L23
1 T ⎢ 1 T ⎥
= X ⎣ 2 L22 0 0 ⎦X (44)
R1 Δ 1 LT 0 0
2 23
Q
In order to prove that Q is positive semi-definite (since Δ < 0), it’s enough to prove that Q is
negative semi-definite.
⎡ ⎤
− L mG Lr M e JnG θG 1L L
2 rG r M I2
1L L
2 rG mM e
Jn M θ M
⎣ ⎦
Q = 1L L
2 rG r M I2 0 0 (45)
1 − Jn M θ M
2 LrG L mM e 0 0
We can see that all the sub-determinant of Q are negative, hence the exponential stability of
the PBC + P controller is proven.
5.3 PBC + PI
The PBC + PI controller is given by the equation below:

−1
BvrG = λ̇ + RL (θ )λ + B K p (isG − isG ) + Ki (isG − isG )
d d d d
(46)
where K p and Ki are the proportional and integral positive gains. We have:

−1
(isG − isG
d
) = (λ − λdsG M ) (47)
R2 sG M
−1

= 0 I2 0 (λ − λd ) (48)
R2
P λ̃
Then,
Ki
BvrG = λ̇d + RL −1 (θ )λd + K p BPL −1 (θ )λ̃ − BP λ̃ (49)
R2
The closed loop error dynamic is:
K
λ̃˙ = − RL −1 (θ )λ̃ − K p BPL −1 (θ )λ̃ + i BP λ̃ (50)
R2
Consider the desired energy function given by (33), it’s derivative along the trajectories of (50)
is:
V̇ = λ̃ T R−1 λ̃˙
K
= − λ̃ T L −1 (θ )λ̃ − K p λ̃ T R−1 BPL −1 (θ )λ̃ + i λ̃ T R−1 BP λ̃
R2

K
= − λ̃ T L −1 (θ ) λ̃ + λ̃ T − K p R−1 BPL −1 (θ ) + i R−1 BP λ̃
R2
>0
M
To show that the error dynamic (50) is stable, it’s enough to prove that M is a negative
semi-definite matrix:
⎡ ! ⎤
K Kp Kp
1 ⎢
− |Δp| LmG LrM e Jn G θG L L
| Δ | rG rM
+ Ki
I2 L L e Jn M θ M
| Δ | rG mM
R2 ⎥
M= ⎣ 0 0 0 ⎦ (51)
R1
0 0 0
M is a negative semi-definite matrix if: X T MX ≤ 0 X ∈ 6×1

⎡ ! ⎤
K Kp Kp
− |Δp| LmG LrM e Jn G θG L L
2| Δ | rG rM
+ Ki
I2 L L e Jn Mθ M
2| Δ | rG mM
1 ⎢ ! 2R2
⎥
X T MX= X T⎢ ⎥X
Kp Ki
L L + 2R I2 0 0 (52)
R1 ⎣ 2| Δ | rG rM 2 ⎦
Kp
L L e − Jn Mθ M
2| Δ | rG mM
0 0

M
To prove that M is a negative semi-definite matrix, it’s enough to prove that M is a negative
semi-definite matrix, the calculus of the sub-determinants of this latter show that M is a
negative semi-definite matrix.
Hence the exponential stability of the PBC + PI controller is proven.
6. The construction for BDFTIG

To establish the complete mathematical representation of the dynamic behaviour of the
BDFTIG it is first necessary to clarify the kind of the electromechanical interconnection that
exists between the cascaded machines. One of the simplest ways to connect these two
machines is in the back-to-back method with no phase inversion on the rotor side, as shown
by Figure 6.
Fig. 6. BDFTIG Back-to-Back Connection
By this connection, the rotor currents produced by the two machines join in the subtractive
style, and the rotor voltages have the same signs, i.e. Ir p = − Irc and Vr p = Vrc . The chosen
connection really affects the distribution of the magnetic fields and flux inside the BDFTIG,
producing the two counter-rotating torques as will be discussed in the following sections.
6.1 Equivalent circuit analysis of the BDFTIG

Figure 7- shows the equivalent circuit of the BDFTIG from which the electrical system
equations can be derived.
Fig. 7. Equivalent Circuit of the BDFTIG
To simplify the controller algorithm, the machine quantities should be expressed in the d-q
frame by employing Park’s and Clark’s transformation. The reason of this transformations is
to remove as many time-varying quantities from the system as possible. By converting the
three-phase machine to its two-phase equivalent and selecting the suitable reference frame,
all the time-varying inductances in both the stator and the rotor are eliminated, allowing for
a simple however complete dynamic model of the electric machine. From these equivalent
circuits the electrical equations of BDFTIG can be determined as shown in the next section.
6.2 Electrical system equations for BDFTIG

Starting with the power machine, the general form of the vector equations of the BDFTIG can
be written as:
q q
q q disp dir p
vsp = Rsp isp + L sp + ω p L sp isp
d
+ L mp + ω p L mp irdp
dt dt
q q q q
vsp = Rsp isp + ( L sp isp + L mp ir p )s + ( L sp isp
d
+ L mp irdp )ω p
Fig. 8. Equivalent Circuits of d-q BDFTIG
The flux linkage current relations are:

q q q
Ψsp = L sp isp + L mp ir p
Ψdsp = d
L sp isp + L mp irdp (53)
q
q q dΨsp
vsp = Rsp isp + + ω p Ψdsp (54)
dt
q q
q q dir p disp
vr p = Rr p ir p + Lr p + ωr Lr p irdp + L mp + ωr L mp isp
d
dt dt
q q q q
vr p = Rr p ir p + ( Lr p ir p + L mp isp )s + ( Lr p irdp + L mp isp
d
)ωr
We have also:
q q q
Ψr p = Lr p ir p + L mp isp
Ψrdp = Lr p irdp + L mp isp
d
(55)
q
q q dΨr p
vr p = Rr p isp + + ω p Ψrdp (56)
dt
dispd dirdp
q q
vdsp = d
Rsp isp + L sp − ω p L sp isp + L mp − ω p L mp ir p
dt dt
q q
vdsp = d
Rsp isp + ( L sp isp
d
+ L mp irdp )s − ( L sp isp + L mp ir p )ω p
dΨdsp q
vdsp = d
Rsp isp + − ω p Ψsp (57)
dt
dirdp q dispd
q
vrdp = Rr p irdp + Lr p + ωr Lr p ir p + L mp + ωr L mp isp
dt dt
q q
vrdp = Rr p irdp + ( Lr p irdp + L mp isp
d
)s + ( Lr p ir p + L mp isp )ωr
dΨrdp q
vrdp = d
Rr p isp + + ω p Ψr p (58)
dt
Electrical system equations for control machine:
q q
q q disc di
vsc = Rsc isc + L sc + ω c L sc isc
d
+ L mc rc + ω c L mc irc d
dt dt
q q q q
vsc = Rsc isc + ( L sc isc + L mc irc )s + ( L sc isp
d
+ L mc irc
d
)ω c

q q q
Ψrc = L sc isc + L mc irc
Ψdsc = d
L sc isp + L mc irc
d
(59)
q
q q dΨsc
vsc = Rsc isc + + ω c Ψdsc (60)
dt
q q
q q di di
vrc = Rrc irc + Lrc rc + ωr Lrc irc d
+ L mc sc + ωr L mc isc d
dt dt
q q q q
vrc = Rrc irc + ( Lrc irc + L mc isc )s + ( Lr p irdp + L mc isc
d
)ωr
and:
q q q
Ψrc = Lrc irc + L mc isc
d
Ψrc = d
Lrc irc + L mc isc
d
(61)
q
q q dΨrc
vrc = Rrc irc + + ωr Ψrcd
(62)
dt
di d q di d q
vdsc = d
Rrc isc + L sc sc − ω c L sc isc + L mc rc − ω c L mc irc
dt dt
q q
vdsc = d
Rsc isc + ( L sc isc
d
+ L mc irc
d
)s − ( L sc isc + L mc irc )ω c
dΨdsc q
vdsc = d
Rsc isc + − ω c Ψsc (63)
dt
di d q di d q
d
vrc = d
Rrc irc + Lrc rc + ω c Lrc irc + L mc sc + ωr L mc isc
dt dt
q q
d
vrc = d
Rrc irc + ( Lrc irc
d
+ L mc isc
d
)s + ( Lr p ir p + L mc isc )ωr
d
dΨrc q
d
vrc = d
Rrc irc + + ωr Ψrc (64)
dt
As mentioned before, for the BDFTIG with the back-to-back configuration and with no phase
inversion, the rotor currents of the individual machines have the opposite signs, the fluxes
inside the rotor combine to produce the essential rotor flux, hence; ir p = −irc = ir , Ψr =
Ψr p − Ψrc , vr p = vrc , 0 = vr p − vrc
q q q
q q dir disp q di
0r = Rr p ir + Lr p + ωr Lr p ird + L mp + ωr L mp isp
d
+ Rrc ir + Lrc r + ωr Lrc ird
dt dt dt
q
disc
− L mc − ωr L mc isc
d
dt
But Lr = Lr p + Lrc and Rr = Rr p + Rrc
q q q
q q dir disp di
0r = Rr ir + Lr + ωr Lr ird + L mp + ωr L mp ispd
− L mc sc − ωr L mc isc
d
dt dt dt
q q q q q
0r = Rr ir + ( Lr ir + L mp isp − L mc isc )s + ( Lr ird + L mp isp
d
− L mc isc
d
)ωr

q q q q
Ψr = Lr ir + L mp isp − L mc isc
Ψrd = Lr ird + L mp isp
d
− L mc isc
d
(65)
q
q q dΨr
0r = Rr ir + + ωr Ψrd (66)
dt
d
q dird q disp q di d q
0r = Rr ird + Lr + ωr Lr ir + L mp + ωr L mp isp − L mc sc − ωr L mc isc
dt dt dt
q q q q
0r = Rr ird + ( Lr ird + L mp isp
d
− L mc isc
d
)s + ( Lr ir + L mp isp − L mc isc )ωr
q dΨrd q
0r = Rr ird + + ω r Ψr (67)
dt
The electrical torque creation in the power machine is governed by the same principles that
apply to any induction machine. The general equation of the electrical torque in this case is
simply:
3 P
Te = Ψm Ir (68)
2 2
In the d-q reference frame, however, the last equation is rearranged to show the torque as a
function of certain control parameter. As the power machine is grid connected, it will have a
constant voltage. The torque could be:

3 Pp q d q
Tp = (Ψsp isp − Ψdsp isp ) (69)
2 2
It is clear from the above equation that the only control variables are the d-q components of
the stator current, because the power machine stator fluxes are almost constant. Furthermore,
when the controller reference frame is aligned with one of the flux components, the number of
the control variables is reduced. To derive the electrical torque for the control machine, we can
use the same general equation for the electrical torque. This case cannot be simplified because
the stator fluxes of the control machine will be variable. The control machine torque must be
expressed as a function of the excitation current and the purpose in this research is to provide
a flexible power control of the BDFIG. So the next equation is the control machine torque, and
it is given in terms of the future control quantities.

3 Pc d q q
Tc = − L mc (isc ir − isc ird ) (70)
2 2
The option of the rotor current as the second variable is clearly shown and that there exists
an electric coupling between the two stators of the BDFTIG, which is achieved through the
common rotor current. This reflects the behaviour of the inner workings of the BDFTIG. The
total electric torque (Te) for the BDFTIG is the sum of the individual electrical torques of both
machines:
3 !
q d q d q q
Te = − Pp (Ψsp isp − Ψdsp isp ) + Pc L mc (isc ir − isc ird ) (71)
4
The electric torque equation is defined by the friction and total inertia of the power and control
machines:
p p dω m
Te = TL + ( BF + BFc )ω m + ( js + jsc ) (72)
dt
Rearranging the last equation to derive the shaft speed:
dω m 1 !
p
= p Te − TL − ( B F + B c
F ) ω m
dt js + jsc
Hence, the shaft speed is
Te − TL
ωm = p p (73)
( BF + BFc ) + ( js + jsc )s
6.3 Simulation of the BDFTIG Model

The BDFTIG model was tested to determine if it was a true representation of the actual
generator. Using Matlab/Simulink to test the BDFTIG, the main tests consisted of disabling
one side of the BDFTIG and applying a constant AC voltage on the opposite side, at the
same time as changing the load torque to allow both motoring and generation modes of
operation. The short circuit test consisted of shorting the stator side of the control machine,
and a natural speed of 900 rpm was recorded, because both machines have four poles each,
as shown in Figure 9. For the next test, the load torque was decreased at time 2.25s to put
the BDFTIG into the generation mode as shown in Figure 10. The system responded as
expected by increasing its speed and moving into the super-synchronous mode of operation,
the electrical torque changed at the same time as the load demand. In this section, the dynamic
model of the generator was developed based on the selected d-q reference frame. The model
was implemented and tested in MATLAB/Simulink. The simulation results verified that the
model can correctly describe the dynamic behaviour of BDFTIG design.
Fig. 9. Speed-Torque Curve of BDFTIG with short circuit test.
Fig. 10. Generation Mode of BDFTIG.

In this section, the dynamic model of the generator was developed based on the selected
d-q reference frame. The model was implemented and tested in MATLAB/Simulink. The
simulation results verified that the model can correctly describe the dynamic behaviour of
BDFTIG design.
7. Experimentation
In order to validate the new controllers, experiments were conducted on a real system. The
following controllers were implemented: PBC, PBC+Proportional action on stator currents, PI
controller on stator currents, and a combination of PBC and PI control. The experiments were
done in the IRII-UPC (Institute of Robotics and Industrial Informatics - University Polytechnic
of Catalonia) where a 200W DFIG interconnected with an IM prototype is available (see Fig.
(11)). The setup was controlled using a computer working under RT-Linux operating system.
With the PBC, only the position sensors of the Generator and the Induction machine were
used for the control. For the Proportional and PI controllers of the electrical subsystem,
measurement of the two stator currents were also needed. In order to show the behaviour
of the system under different load conditions, a non-measured load torque was applied.
Pentium 4; 1,8 GHz; 512 MB RAM Promax 3 Jeulin 188 016
AD215BY
PCI DAS 4020/12 Isolation Amplifier
SERVO AMPLIFIER
80% of 46V 75% of 42V
Uc Advanced Motion Control

DAC 1:1
+ 5V
PCIDAS4020
Ramp Braking DC Motor
6N137 Three Phase Inverter

Optocouplers Promax 2
Vbus
#SD +
DIO
ADC - 12BNCs 74HC244

Buffer Non-Inverting Jeulin 188 019
C’ W,A,V
#PWME
PCI8133 Bridge Protection B’ 1V 100mV
Off System
U+ (16)
V+ (17)
(Salicrú) A’ 1A 250mV
W+(18) 3 3 1W 10mV
PWMs 3
U- (34)
Rotor Stator
V- (35)
W-(36)
3 3 DFIM Induction
3
nA2(23)
nB2(24)
nA1(20)
nB1(21)
DC Motor Generator Motor 3ph Brake

A2(5)
B2(6)
A1(2)
B1(3)
Vbus
-
Promax 1 Promax 2
Encoder 360 Encoder 100
2 2 2 2 pulses/revol. pulses/revol.
M speed G speed
M position G position POWERBOX 100V-10A
AD215BY Isolation Amplifier

1:13
2
12
1A-250mV Hall Sensor EH050
sw Board Channel Signal 2
dada1 0 0 74HC14 Inverter
A,B A,B
dada2 0 1 V1
2
dada3 0 2 I4 V2
notA notB 2
2
dada4 0 3 I3 I1
I2 1000rpm
dada5 1 0 I2 1V
I3
dada6 1 1 I1 DL10050
dada7 1 2
V3 AD215BY Isolation Amplifier
dada8 1 3 1:13,4
V4
dada9 2 0 V4 2
I4
dada10 2 1 V3 X
dada11 2 2 V2 X 1A-250mV Hall Sensor EH050
dada12 2 3 V1 2
74HC14 Inverter
A,B A,B
2 notA notB 2
Select 12 DAQs 2
1000rpm DL10050
1V
Fig. 11. Experimental setup
Since a load torque sensor was not available for the acquisition, we built an estimator of the
resistive torque based on the measurement of the mechanical IM speed.
7.1 Estimation of the load torque

The mechanical dynamics of the IM is given by:
J M θ̈ M = τM − τLM − B M θ̇ M (74)
Since the asymptotic stability of the electrical subsystem Σe is proven we can consider that in
the steady state τM → τMd (exponentially). Then,we have in the steady state the following:
J M θ̈ M = d
τM − τLM − B M θ̇ M (75)

τML
Hence, a linear load torque observer was designed (with l1 , l2 are design parameters):

ω̂˙ mM = d
τM − τ̂ML /J M + l1 (ω̂ mM − ω mM ) (76)
τ̂˙ ML = l2 (ω̂ mM − ω mM ) (77)
7.2 PBC
(rpm)
1500
mM
&ω
1000
ref
mM
500
ω
184 186 188 190 192 194 196 198 200 202 204
(a) t(s)
3000
ωmG(rpm)
2000
1000
184 186 188 190 192 194 196 198 200 202 204
(b) t(s)
6
θG & θM (rad)
193.74 193.76 193.78 193.8 193.82 193.84 193.86 193.88 193.9 193.92 193.94
(c) t(s)
0.4
τG(N.m)
0.2
0
184 186 188 190 192 194 196 198 200 202 204
(d) t(s)
2
(N.m)
1
0
Md
−1
τ
−2
184 186 188 190 192 194 196 198 200 202 204
(e) t(s)
Fig. 12. PBC-(a) Regulated Motor speed and its reference. (b)Generator speed. (c) DFIG & IM
rotor position. (d) Generator torque (e) Motor desired torque.
Figure 12 presents the mechanical IM speed and its smooth reference, the mechanical DFIG
speed, the DFIG and IM rotor positions, the DFIG torque τG and the IM desired torque τMd .
The real IM speed tracks the reference very well, i.e. low overshoot and no steady state error
are observed. Figure 13 shows the stator currents isa and isb , and their references over a
suitable period of time. The stator currents do not track exactly their desired values but are
bounded. This is because the goal of the PBC is to track the IM speed and to keep internal
signals bounded.
Figure 14 shows the DFIG rotor currents irGa and irGb, and their references over a period of
time. Again, these currents are sinusoidal and bounded.
Figure 15 presents the DFIG rotor voltages vrGa and vrGb, the IM rotor speed ω mM and its
estimation ω̂ mM , the estimated IM load torque τ̂ML , and the estimated IM speed, given by
20
10
(A)
sGb
0
&i
sGa
i
−10
−20
193.74 193.76 193.78 193.8 193.82 193.84 193.86 193.88 193.9 193.92 193.94
(a) t(s)
20
10
(A)
sGa
& id
0
sGa
i
−10
−20
193.74 193.76 193.78 193.8 193.82 193.84 193.86 193.88 193.9 193.92 193.94
(b) t(s)
20
10
(A)
sGb
& id
0
sGb
i
−10
−20
193.74 193.76 193.78 193.8 193.82 193.84 193.86 193.88 193.9 193.92 193.94
(c) t(s)
d , i (c) i d , i
Fig. 13. PBC-(a) isa , isb (b) isa sa sb sb
.
15
10
(A)
5
rGb
0
&i
rGa
−5
i
−10
−15
193.74 193.76 193.78 193.8 193.82 193.84 193.86 193.88 193.9 193.92 193.94
(a) t(s)
50
& id (A)
rGa
0
rGa
i
−50
193.75 193.8 193.85 193.9 193.95 194 194.05
(b) t(s)
50
& id (A)
rGb
0
rGb
i
−50
193.75 193.8 193.85 193.9 193.95 194 194.05
(c) t(s)
d ,i
Fig. 14. PBC-(a) irGa, irGb (b) irGa d
rGa (c) irGb , irGb .
60
40
vrGa & vrGb(V)

20
−20
−40
−60
193.75 193.8 193.85 193.9 193.95 194 194.05
(a) t(s)
1600
1400
(rpm)
1200
mM
1000
ωmM & ω
800
600
400
184 186 188 190 192 194 196 198 200 202 204
(b) t(s)
1
ττLM(N.m)
−1
−2
−3
184 186 188 190 192 194 196 198 200 202 204
(c) t(s)
Fig. 15. PBC-(a) vrGa, vrGb (b) ω mM , ω̂ mM (c) τ̂ML .
(76)-(77), is tracking the real speed. Hence, a good estimation of the real IM load torque is
obtained. It has to be noticed that the IM rated torque is 0.7Nm.
It can be concluded that the PBC provides good practical performance even when the applied
load torque is twice the magnitude of the nominal load torque of the IM.
7.3 PBC + P
(rpm)
1500
mM
&ω
1000
ref
mM
500
ω
55 60 65 70 75 80
(a) t(s)
3000
ωmG(rpm)
2000
1000
55 60 65 70 75 80
(b) t(s)
6
θG & θM (rad)
68.85 68.9 68.95 69 69.05 69.1

(c) t(s)
0.3
0.2
τG(N.m)
0.1
0
55 60 65 70 75 80
(d) t(s)
2
(N.m)
1
0
Md
−1
τ
−2
55 60 65 70 75 80
(e) t(s)
Fig. 16. PBC+P-(a) Regulated Motor speed and its reference. (b)Generator speed. (c) DFIG &
IM rotor position. (d)Generator torque (e) Motor desired torque.
As with the PBC alone, the results obtained with the PBC+P are given in figures 16-19. On the
whole, the system behaviour is the same as the PBC alone. One difference that is noticeable is
15
10
(A)
5
sGb
0
&i
sGa
−5
i
−10
−15
68.85 68.9 68.95 69 69.05 69.1

(a) t(s)
15
10
(A)
5
sGa
& id
0
sGa
−5
i
−10
−15
68.85 68.9 68.95 69 69.05 69.1
(b) t(s)
15
10
(A)
5
sGb
& id
0
sGb
−5
i
−10
−15
68.85 68.9 68.95 69 69.05 69.1

(c) t(s)
d ,i
Fig. 17. PBC+P-(a) isGa, isGb (b) isGa d
sGa (c) i sGb , i sGb .
15
10
(A)
5
rGb
0
&i
rGa
−5
i
−10
−15
68.85 68.9 68.95 69 69.05 69.1
(a) t(s)
60
40
& id (A)
20
rGa
0
rGa
−20
i
−40
68.85 68.9 68.95 69 69.05 69.1 69.15 69.2 69.25 69.3

(b) t(s)
60
40
& id (A)
20
rGb
0
rGb
−20
i
−40
68.85 68.9 68.95 69 69.05 69.1 69.15 69.2 69.25 69.3

(c) t(s)
d ,i
Fig. 18. PBC+P-(a) irGa, irGb (b) irGa d
the small error between the desired stator currents and the real ones thanks to the proportional
controller.
The PBC+P controller exhibits good practical performance but not significantly better than
those obtained with the PBC alone.
7.4 PBC + PI
Again, as for the PBC and the PBC+P controllers, figures 20-23 show the results. It can be
seen in figure 21 that the integral actions on the stator currents do not decrease the error
significantly between the real and desired values in comparison with the results for the PBC+P
60
40
vrGa & vrGb(V)

20
−20
−40
−60
68.85 68.9 68.95 69 69.05 69.1 69.15 69.2 69.25 69.3
(a) t(s)
1600
(rpm)
1400
1200
mM
ωmM & ω
1000
800
600
400
55 60 65 70 75 80
(b) t(s)
1
ττLM(N.m)
−1
−2
−3
55 60 65 70 75 80
(c) t(s)
Fig. 19. PBC+P-(a) vrGa, vrGb (b) ω mM , ω̂ mM (c) τ̂ML .

(rpm)
1500
mM
&ω
1000
ref
mM
500
ω
40 45 50 55 60 65
(a) t(s)
3000
ωmG(rpm)
2000
1000
40 45 50 55 60 65
(b) t(s)
6
θG & θM (rad)
52.6 52.65 52.7 52.75 52.8 52.85

(c) t(s)
0.3
0.2
τG(N.m)
0.1
40 45 50 55 60 65
(d) t(s)
2
(N.m)
1
0
Md
−1
τ
−2
40 45 50 55 60 65
(e) t(s)
Fig. 20. PBC+PI-(a) Regulated Motor speed and its reference. (b)Generator speed. (c) DFIG &
IM rotor position. (d)Generator torque (e) Motor desired torque.
isGa & isGb(A)

10
0
−10
52.6 52.65 52.7 52.75 52.8 52.85

(a) t(s)
isGa & isGa(A)
10
d
0
−10
52.6 52.65 52.7 52.75 52.8 52.85

(b) t(s)
isGb & isGb(A)
10
d
0
−10
52.6 52.65 52.7 52.75 52.8 52.85

(c) t(s)
d ,i
Fig. 21. PBC+PI-(a) isGa, isGb (b) isGa d
irGa & irGb(A)
10
0
−10
52.6 52.65 52.7 52.75 52.8 52.85

(a) t(s)
irGa & irGa(A)
50
d
−50
52.6 52.65 52.7 52.75 52.8 52.85 52.9 52.95 53
(b) t(s)
50
irGb & irGb(A)
d
−50
52.6 52.65 52.7 52.75 52.8 52.85 52.9 52.95 53
(c) t(s)
d ,i
Fig. 22. PBC+PI-(a) irGa, irGb (b) irGa d
controller (see fig. 17). This is due to the fact that the reference values are sinusoidal and that
the bandwidth of the PI controllers cannot be increased sufficiently experimentally.
It can be concluded that the PI action on the stator currents does not improve significantly the
performance obtained with the PBC+P controller.
7.5 PI
The PI control law (with K p and Ki are proportional and integral gains) is given below:

BvrG = B K p (isG − isG
d
) + Ki (isG − isG
d
) (78)
(V)
50
rGb
&v
0
rGa −50
v
52.6 52.65 52.7 52.75 52.8 52.85 52.9 52.95 53
(a) t(s)
ωmM & ωmM(rpm)
1500
1000
500
40 45 50 55 60 65
(b) t(s)
2
ττ (N.m)
0
LM
−2
40 45 50 55 60 65
(c) t(s)
Fig. 23. PBC+PI-(a) vrGa, vrGb (b) ω mM , ω̂ mM (c) τ̂ML .

(rpm)
1500
mM
&ω
1000
ref
500
mM
230 235 240 245 250 255

ω
2500 t(s)
(rpm)
2000
1500
mG
1000
ω
500
230 235 240 245 250 255
θG & θM (rad)
6 t(s)
4
2
241.35 241.4 241.45 241.5 241.55

t(s)
τ (N.m)
0
−0.2
G
−0.4
230 235 240 245 250 255
t(s)
(N.m)
2
0
Md
−2
τ
230 235 240 245 250 255

t(s)
Fig. 24. PI-(a) Regulated Motor speed and its reference. (b)Generator speed. (c) DFIG & IM
rotor position. (d) Generator torque (e) Motor desired torque.
10
(A)
5
sGb
0
&i
sGa
−5
i
−10
241.35 241.4 241.45 241.5 241.55
t(s)
10
(A)
5
sGa
& id
0
sGa
−5
i
−10
241.35 241.4 241.45 241.5 241.55
t(s)
10
(A)
5
sGb
& id
0
sGb
−5
i
−10
241.35 241.4 241.45 241.5 241.55
t(s)
d ,i
Fig. 25. PI-(a) isGa, isGb (b) isGa d
20
(A)
rGb
&i
0
rGa
i
−20
241.35 241.4 241.45 241.5 241.55
t(s)
50
& id (A)
rGa
0
rGa
−50
i
241.35 241.4 241.45 241.5 241.55

t(s)
50
& id (A)
rGb
0
rGb
−50
i
241.35 241.4 241.45 241.5 241.55

t(s)
d ,i
Fig. 26. PI-(a) irGa, irGb (b) irGa d
Finally, in order to obtain a significant comparison between controllers, a PI-based control

has been designed without a PBC, i.e. there is one PI controller for each stator current.
Figures 24-27 show the results. These results show clearly that the system behaviour is much
deteriorated in comparison with the results obtained with the previous controllers. Even if
there is no IM speed error in the steady state, the speed does not track its reference during
transients, and there is a speed error when a load torque is applied. This is mainly due to the
saturation of the desired IM torque at a value four times its nominal value. Consequently, the
stator currents are very large, i.e. their magnitude is about twice those currents with the PBC,
and so significant stator losses can be expected.
50
(V)
rGb
&v
0
rGa
v
−50
241.35 241.4 241.45 241.5 241.55 241.6 241.65 241.7 241.75
t(s)
(rpm)
1500
hat
mM
1000
&ω
mM
500
ω
230 235 240 245 250 255

t(s)
1.2
(N.m)
1
hat
0.8
LM
τ
0.6
0.4
230 235 240 245 250 255
t(s)
Fig. 27. PI-(a) vrGa, vrGb (b) ω mM , ω̂ mM (c) τ̂ML .
These results show that the PI control alone of the stator currents is not efficient for the
control of the DFIG+IM system. The PBC, with or without P or PI actions, shows much better
performance.
7.6 Robustness tests

In order to highlight the performances of the controllers, and check their behaviour in the
presence of machine parameter variations, a change in the DFIG and IM rotor and stator
resistances is applied. In the real case, the resistances of a machine increase with temperature.
In this case, all the resistances of the two machines used in the controllers are decreased by
40% when the "Switch on Parameters" signal value goes from 0 to 1 (see figure 31). This test
has been carried out with the four controllers (i.e. PBC, PBC+P, PBC+PI and PI). The results
show that all the controllers are robust to a large change in machine resistances. To be brief,
only the results obtained with the PBC are reported here.
Figure 28 presents the mechanical IM speed and its smooth reference, the mechanical DFIG
speed, the DFIG and IM rotor positions, the DFIG torque τG and the IM desired torque τMd .
The real IM speed tracks very well the reference, i.e. low overshoot and no steady state error
are observed. Figure 29 shows the stator currents i sa and isb , and their references over a period
of time. The stator currents do not track exactly the desired values but are bounded. This is
because the goal of the PBC is to track the IM speed and to keep internal signals bounded.
Figure 30 shows the DFIG rotor currents irGa and irGb, and their references over a period of
time. Again, these currents are sinusoidal and bounded.
Figure 31 presents the control signals vrGa and vrGb , the rotor IM speed ω mM and its estimation
ω̂ mM , and the "Switch on Parameters" signal. These results illustrate the robustness of the PBC
when the parameters are varied.
8. Conclusion
Speed–torque tracking controllers for an IM powered by a DFIG have been presented. The
joint system extracts energy from a primary mechanical source that is transformed by the
(rpm)
1500
mM
&ω
1000
ref
mM 500
70 75 80 85 90 95
ω
t(s)
(rpm)
2500
2000
1500
mG
1000
ω
500
70 75 80 85 90 95
θ & θ (rad)
6 t(s)
4
M
2
G
83.5 83.55 83.6 83.65 83.7

t(s)
0.1
τG(N.m)
0
−0.1
70 75 80 85 90 95
2 t(s)
τMd (N.m)
−2
70 75 80 85 90 95
t(s)
Fig. 28. PBC-robustness test-(a) Regulated Motor speed and its reference. (b)Generator
speed. (c) DFIG & IM rotor position. (d) Generator torque (e) Motor desired torque.
10
isGa & isGb(A)
5
0
−5
−10
83.5 83.55 83.6 83.65 83.7

t(s)
10
isGa & idsGa(A)
5
0
−5
−10
83.5 83.55 83.6 83.65 83.7

t(s)
10
isGb & idsGb(A)
5
0
−5
−10
83.5 83.55 83.6 83.65 83.7

t(s)
d ,i
Fig. 29. PBC -robustness test-(a) i sGa, isGb (b) isGa d
20
irGa & irGb(A)

10
0
−10
−20
83.5 83.55 83.6 83.65 83.7 83.75 83.8 83.85 83.9 83.95
t(s)
50
irGa & idrGa(A)
−50
83.5 83.55 83.6 83.65 83.7 83.75 83.8 83.85 83.9 83.95
t(s)
50
irGb & idrGb(A)
−50
83.5 83.55 83.6 83.65 83.7 83.75 83.8 83.85 83.9 83.95
t(s)
d ,i
Fig. 30. PBC-robustness test-(a) irGa, irGb (b) irGa d
50
(V)
rGb
0
&v
rGa
v
−50
83.5 83.55 83.6 83.65 83.7 83.75 83.8 83.85 83.9 83.95
t(s)
(rpm)
1500
hat
mM
1000
&ω
mM
500
ω
70 75 80 85 90 95
t(s)
Switch on Parameters
0.5
0
70 75 80 85 90 95
t(s)
Fig. 31. PBC-robustness test-(a) vrGa , vrGb (b) ω mM , ω̂ mM (c) Switch.
DFIG, which at the same time controls the speed of the IM making use of the rotor voltage of
the DFIG as a control variable. A complete stability proof for inner loop control is given. The
proof of the overall scheme including the outer speed loop follows verbatim from (1) and is
omitted here for brevity.
The main advantage of the PBC is that it requires the measurement of only two mechanical
positions for the speed tracking. The PI controller applied to the inner loop provides good
performance but saturation in the transient state can be observed. Robustness tests were
performed to observe the behaviour of the controllers to machine parameter variations. All
the proposed controllers were found to be robust towards variation in machine resistances.
Also, a power flow analysis can be undertaken between the generator, the IM and the grid
in order to optimize the efficiency of the overall system. A comparison of the experimental
results of the proposed PBC, PI, PBC+P and PBC+PI algorithm is presented in Table 2. It is
based on the performance obtained practically with the different controllers. In addition to the
comparison criteria of Table 2, it is proposed to check the following to see what their effects
are on the controllers’performance:
!2
• eω M = nT 1
∑ni=1 ω M (i ) − ω Re f M (i ) indication about the IM speed tracking error. Where
n is the length of the sampled data and T is the sampling time;
!2
• eisGa = nT1
∑ni=1 isGa (i ) − isGa
d (i ) indication about the stator current tracking error in the
phase a;
• Observed magnitude of i sGa;
• PavgG = n1 ∑ni=1 [ τG (i )ω G (i )] indication about the rotor average value of the instantaneous
absorbed power in the DIFG;
• Pavg M = n1 ∑ni=1 [τM (i )ω M (i )] indication about the rotor average value of the instantaneous
absorbed power in IM;
Rs ( Ω) Rr ( Ω) L s ( mH ) L r ( mH ) L m ( mH ) J ( Nm2 /rad)
DFIG 0.365 0.559 0.938 0.938 12.975 4.358 × 10−3
IM 0.5 0.2 1.2 1.2 9.00 1.1 × 10−3
Table 1. The parameters for DFIG and IM
PBC PBC+P PBC+PI PI

ω Re f M 500→1000 500→1000 500→1000 500→1000
[rpm] →1400→800 →1400→800 →1400→800 →1400→800
(1st order filter) (1st order filter) (1st order filter) (1st order filter)
τLM [N.m] 0.5 → 1.45 → 0.5 0.5 → 1.4 → 0.5 0.5 → 1 → 0.5 0.5 → 1.15 → 0.5
settling time
of ω Re f M 0.4s 0.4s 0.4s 0.4s
settling time
of ω M 0.1s 0.1s 0.1s 2s
ωG [rpm] 1500 1500 1500 1500
e ω M x 105 2.9 4.8 5.7 38.6
e i sGa x 103 3.7 2.65 2.68 0.25
Observed magnitude
of isGa [A] 5 7 8 10
PavgG [W] 4.7 5.6 5 9
Pavg M [W] 58.9 74.4 70.7 177.9
Table 2. Comparison table of experimental results

If we take in account the problem of speed tracking of the IM interconnected to the DFIG and
according to the robustness tests and the experimental results presented in Table 2 we can say
that the PBC controller provided the best performance.
In addition, this paper has provided the detailed analysis of operational principles of the
BDFTIG.
9. Acknowledgment
The authors would like to express their gratitude to Jordi Riera, Enric Fossas and Miguel Allué
from IRII-UPC, Barcelona, Spain for their help with the practical experiments.
10. References
[1] R. Ortega, A. Loria, P.J. Nicklasson, and H. Sira-Ramirez, “Passivity-based
control of Euler-Lagrange systems,” in Communications and Control Engineering.
Berlin,Germany:Spring-Verlag, 1998.
[2] Liu, X., G. Verghese, J. Lang and M. Önder, Generalizing the Blondel-Park
Transformation of Electrical Machines: Necessary and Sufficient Conditions, IEEE Trans.
Circ. Syst., Vol. 36, No. 8, pp. 1085-1067, 1989.
[3] M. Becherif, R. Ortega, E. Mendes and S. Lee, “Passivity-based control of a doubly-fed
induction generator interconnected with an induction motor,” in CDC 2003.
[4] M. Becherif, “Contribution aux techniques de façonnement dé́nergie: Application à la
commande des systèmes électromécaniques,” PhD thesis, Université de Paris XI, France,
2004.
[5] W. Leonhard, “Control of electrical drives,” Spring-Verlag, 1985.
[6] M.S. Vicatos and J.A. Tagopoulos, “Steady state analysis of a doubly-fed induction
generator under sychronous operation,” IEEE Trans. on Energy Conversion, vol.4, no.3,
pp.495-501, 1989.
[7] F. Bogalecka, “Dynamics of the power control of a double fed induction generator
connected to the soft power grid,” ISIE International Symposium on Industrial Electronics,
Budapest, pp.509-513, 1993.
[8] A. Mebarki and R.T. Lipczynsky, “Novel variable speed constant frequency generation
system with voltage regualtion,” EPE European Conference on Power Electronics and
Applications, vol.2, pp.465-471, 1995.
[9] P. Caratozzolo, E. Fossas and J. Riera “Robust nonlinear control of an isolated motion
system,” CIEP International Power Electronics Congress, Mexico, 2002.
[10] R. Datta and V.T. Ranganathan, “Variable-speed wind power generation using doubly
fed wound rotor induction machine-a comparision with alternative schemes,” IEEE
Trans. on energy conversion, vol.17, no.3, pp.414-421, 2002.
[11] S. Muller, M. Deicke, and Rik W. De Doncker, “Doubly fed induction generator systems
for wind turbines,” IEEE Industry Applications Magazine, vol., no., pp.26-33, May/June,
2002.
[12] P. Caratozzolo, E. Fossas, and J. Riera, “Nonlinear control of an isolated motion system
with DFIG,” IFAC International Federation of Automatic Control , 2002.
[13] A. J. van der Schaft, “L2 –Gain and Passivity Techniques in Nonlinear Control,”
Springer–Verlag, Berlin, 1999.
8
A RMRAC Parameter Identification Algorithm

Applied to Induction Machines
Rodrigo Z. Azzolin1 , Cristiane C. Gastaldini2 , Rodrigo P. Vieira3 and
Hilton A. Gründling4
1,2,3,4Federal University of Santa Maria
1 Federal University of Rio Grande
3 Federal University of Pampa
Brazil
1. Introduction
This chapter deals with the problem of parameter identification of electrical machines to
achieve good performance of a control system. The development of an identification algorithm
is presented, which in this case is applied to Single-Phase Induction Motors, but could easily
be applied to other electrical machine. This scheme is based on a Robust Model Reference
Adaptive Controller and measurements of stator currents of a machine with standstill rotor.
From the obtained parameters, it is possible to design high performance controllers and
sensorless control for induction motors.
Single-phase induction motors (SPIM) are widely used in fractional and sub-fractional
horsepower applications, usually in locations where only single-phase energy supply is
available. In most of these applications the machine operates at constant frequency and is
fed directly from the AC grid with an ON/OFF starting procedure. In recent years, several
researchers have shown that the variable speed operation can enhance the SPIM’s efficiency
(Blaabjerg et al., 2004; Donlon et al., 2002). In addition, other researchers have developed high
performance drives for SPIM’s, using Field-Oriented Control (FOC) and sensorless techniques
(de Rossiter Correa et al., 2000; Vaez-Zadeh & Reicy, 2005; Vieira et al., 2009b). However,
the FOC associated with the sensorless technique demands accurate knowledge of electrical
motor parameters to achieve good performance.
A good deal of research has been carried out in the last several years on parameter estimation
of induction motors, mainly with regards to three-phase induction machines (Azzolin et al.,
2007; Koubaa, 2004; Ribeiro et al., 1995; Velez-Reyes et al., 1989).
However, few methods have been proposed for automatic estimation in single-phase
induction motors. One of them uses a classical method for electrical parameter identification
(Ojo & Omozusi, 2001), its implementation is onerous. In (Vieira et al., 2009a) a Recursive
Least Squares (RLS) identification algorithm is used to obtain machine parameters. The
identification results are good, although this method requires the design of filters to obtain
the variable derivative, which can be deteriorated by noises.
In order to solve these problems, this chapter details a closed-loop algorithm to estimate
the electrical parameters of a single-phase induction motor based on (Azzolin & Gründling,
Nq
Rsq Llsq Llrq Nd
ωrfrd
+ -
+ isq
vsq Lmq Rrq
-
Nd
Rsd Llsd Llrd Nq ωrfrq
+ isd + -
vsd Lmd Rrd
-
Fig. 1. Equivalent circuit of a SPIM.
2009). In (Azzolin & Gründling, 2009) a Robust Model Reference Adaptive Controller
(RMRAC) algorithm was used for parameter identification of three-phase induction motors.
The RMRAC algorithm eliminates the use of filters to obtain derivatives of the signals. Thus,
the objectives of this chapter are to apply the RMRAC algorithm in the electrical parameter
identification of a SPIM and make use of the robustness of the system in dealing with of noise
measurement.
As a result, the proposed parameter estimation procedure is divided into three steps:
(I) identification of the RMRAC controller gains;
(II) estimation of the transfer function coefficients of the induction motor at standstill;
(III) calculation of the stator resistance Rsi , rotor resistance Rri , stator inductance L si , rotor
inductance Lri and mutual inductance L mi using steps (I) and (II), where the index ”i ”
express the axes q or d.
This chapter is organized as follows: Section 2 presents the induction machine model. A short
review of the RMRAC algorithm applied to the identification system is presented in sections 3
and 4. Section 5 describes the assumptions and equations of Model Reference Control (MRC)
while section 6 shows the proposed parameter identification algorithm. Sections 7 and 8
present the simulation and experimental results. Finally, chapter conclusions are presented
in section 9.
2. Mathematical model of a single phase induction motor

The SPIM equivalent circuit without the permanent split-capacitor can be represented by an
asymmetrical two-phase induction motor as shown in Fig. 1. In this figure L lsi and L lri are
the stator and rotor leakage inductances, ωr is the speed rotor, φri is the electromagnetic flux
and Ni is the number of turns for auxiliary winding or axis d and for main winding or axis q.
The stator and rotor inductances are relationship with the leakage and mutual inductance as
a L si = L lsi + L mi and Lri = L lri + L mi , respectively.
From Fig. 1 and from (Krause et al., 1986) it is possible to derive the dynamical model of a
SPIM. The SPIM dynamical model in a stationary reference frame can be described by 1. In this
A RMRAC Parameter Identification Algorithm Applied to Induction Machines 147
equation σ̄q = L sq Lrq − L2mq , σ̄d = L sd Lrd − L2md , p is the poles pairs and n is the relationship
between the number of turns for auxiliary and for main winding Nd /Nq .
⎡ ⎤
Rsq Lrq 1 L mq L md Rrq L mq 1 Lrd L mq
⎢ − − pω − pω
n σ̄q ⎥
r r
⎡ • ⎤ ⎢ σ̄q n σ̄q σ̄q ⎥
isq ⎢ L L L R L L R L md ⎥
⎢ pωr n mq md
− rd sd
pω n
rq md rd ⎥
⎢ isd ⎥ ⎢ σ̄d σ̄d
r
σ̄d σ̄d ⎥
⎢ ⎥=⎢ ⎥
⎣ irq ⎦ ⎢ L mq Rsq 1 L sq L md L sq Rrq 1 L sq Lrd ⎥ . (1)
⎢ pωr − pωr ⎥
ird ⎢ σ̄q n σ̄q σ̄q n σ̄q ⎥
⎢ ⎥
⎣ L sd L mq L md Rsd L sd Lrq L R ⎦
− pωr n − pωr n − sd rd
σ̄d σ̄d σ̄ σ̄d
⎡d ⎤
Lrq
⎢ 0 ⎥
⎡ ⎤ ⎢ σ̄q ⎥
isq ⎢ L ⎥
⎢ 0 rd ⎥
⎢ isd ⎥ ⎢ σ̄ ⎥ vsq
⎢ ⎥
.⎣ ⎦+⎢ L ⎢ d ⎥
irq ⎥ vsd
⎢ − mq 0 ⎥
ird ⎢ σ̄q ⎥
⎢ ⎥
⎣ L md ⎦
0 −
σ̄d
From equation 1 it is possible to obtain the transfer functions in the axes q and d at standstill
rotor (ωr = 0), where these equations are decoupled and presented in 2 and 3.
Lrq Rrq
+
s
isq (s) σ̄q σ̄q
Hq (s) = = , (2)
vsq (s) R rq R sq
s2 + spq +
σ̄q
Lrd R
s + rd
i (s) σ̄d σ̄d
Hd (s) = sd = , (3)
vsd ( s) R R
s2 + spd + rd sd
σ̄d
where
Rsq Lrq + Rrq L sq R L + Rrd L sd

pq = and pd = sd rd . (4)
σ̄q σ̄d
3. Identification system
The proposed electrical parameter identification system is shown in Fig. 2. This system is
based on a RMRAC algorithm used to generate the control action vsq by the difference between
the measured current isq and the reference isq ∗ at standstill rotor. In this test the auxiliary
winding (or axis d) is open while the main winding (or axis q) is identified.
The dotted box in Fig. 2 is detailed in Fig. 3 where the RMRAC control law applied to q axis
is shown. In Fig. 3, the reference model and plant are given by
Zm ( s )
Wm (s) = k m , (5)
Rm (s)
isq* vsq
RMRAC q
vsd = 0
IM
isq
isd = 0
Fig. 2. Block diagram of the RMRAC identification system.
y m (isqm )
(isq* ) Wm ( s )
- e1
r
( vsq ) (isq )
yp +
1 + up
Gp (s )
θ4 +
++ T
θ 1T w1 a (s)
θ4 Λ (s )
T
θ 2T w2 a ( s )
θ4 Λ(s )
up ü θ3
yp ï
ï θ4
ω1 ý
ω2 ï Pξ
e 1 ïþ θ& = -s Pθ - 2 ε
m
Fig. 3. RMRAC structure.
Z p (s)
G p (s) = k p . (6)
R p (s)
where Zm (s), Rm (s), Z p (s) and R p (s) are monic polynomials and k m and k p are high frequency
gains.
The general idea behind RMRAC is to create a closed-loop controller with gains that can be
updated and change the response of the system. The output of the system y p is compared to a
desired response from a reference model ym . The controller gains vector θ = [ θ1 θ2 θ3 θ4 ]
is updated by a gradient algorithm based on e1 error. The goal is that the gains converge to
ideal values that cause the plant response to match the response of the reference model. These
gains can be obtained by a gradient algorithm presented in (Ioannou & Sun, 1996; Ioannou &
Tsakalis, 1986) and described as follows. More details of RMRAC algorithms can be seen in
(Câmara & Gründling, 2004).
The procedure described in Fig. 2 and Fig. 3 is applied to parameter identification of the q axis
or transfer function 2. However, the same procedure is used for parameter identification of
the d axis or transfer function 3.
4. RMRAC gains adaptation algorithm

The gradient algorithm used to obtain the control law gains is given by
Pξ
θ̇ = − σPθ − ε, (7)
m2
with

δ
ṁ = δ0 m + δ1 u p + y p + 1 , m(0) > 1 , δ1 ≥ 1, (8)
δ0
and
ξ = Wm (s)Iw, (9)
w = [ w1 w2 y p u p ] , (10)

w1 , w2 are auxiliary vectors, δ0 , δ1 are positive constants and δ0 satisfies δ0 + δ2 ≤min( p0 , q0 ),
q0 ∈ + is such that the Wm (s − q0 ) poles and the (F − q0 I) eigenvalues are stable and δ2 is a
positive constant. The sigma modification σ in 7 is given by
⎧
⎪
⎪ 0 if θ < M0
⎨ θ
σ = σ0 −1 if M0 ≤ θ < 2M0 , (11)
⎪
⎪ M0
⎩
σ0 if θ ≥ 2M0
where M0 > θ∗ and σ0 > 2μ −2 /R2 , R, μ ∈ + are design parameters. In this case, the
parameters used in the implementation of the gradient algorithm are
⎧
⎪
⎪ δ0 = 0.7
⎪
⎪
⎨ δ1 = 1
δ2 = 1 , (12)
⎪
⎪
⎪
⎪ σ = 0.1
⎩ 0
M0 = 10
More details of design of the gradient algorithm can be seen in (Ioannou & Tsakalis, 1986). As
defined in (Lozano-Leal et al., 1990), the modified error in 7 is given by
ε = e1 + θ ξ − Wm θ w, (13)
or
ε = φ ξ + μη. (14)
When the ideal values of gains are identified and the plant model is well known, the plant can
be obtained by equation analysis of MRC algorithm described in the next section.
r 1 + å up yp
Gp (s)
θ4 +
++ θ T1 a ( s )
θ4 Λ(s )
θ 2T a ( s )
θ4 Λ(s )
θ3
θ4
Fig. 4. MRC structure.
5. MRC analysis
The Model Reference Control (MRC) shown in Fig. 4 can be understood as a particular case
of RMRAC structure, which is presented in Fig. 3. This occurs after the convergence of the
controller gains when the gradient algorithm changes to the steady-state. It is important to
note that this analysis is only valid when the plant model is well known and free of unmodeled
dynamics and parametric variations.
To allow the analysis of the MRC structure, the plant and reference model must satisfy some
assumptions as verified in (Ioannou & Sun, 1996). These suppositions, which are also valid for
RMRAC, are given as follow:
Plant Assumptions:
P1. Z p (s) is a monic Hurwitz polynomial of degree m p ;
P2. An upper bound n of the degree n p of R p (s);
P3. The relative degree n ∗ = n p − m p of G p (s);
P4. The signal of the high frequency gain k p is known.
Reference Model Assumptions:
M1. Zm (s), Rm (s) are monic Hurwitz polynomials of degree q m , pm , respectively, where
pm ≤ n;
M2. The relative degree n ∗m = pm − q m of Wm (s) is the same as that of G p (s), i.e.,
n ∗m = n ∗ .
In Fig. 4 the feedback control law is
θ1 α (s) θ α (s) θ 1

up = up + 2 y p + 3 y p + r, (15)
θ4 Λ ( s ) θ4 Λ ( s ) θ4 θ4
and
Δ
α (s) = αn−2 (s) = sn−2 , sn−3 , . . . , s, 1 for n ≥ 2,
Δ
(16)
α (s) = 0 for n = 1,
θ3 , θ4 ∈ 1 ; θ1 , θ2 ∈ n−1 are constant parameters to be designed and Λ(s) is an arbitrary
monic Hurwitz polynomial of degree n − 1 that contains Zm (s) as a factor, i.e.,
Λ ( s ) = Λ 0 ( s ) Zm ( s ) , (17)
which implies that Λ0 (s) is monic, Hurwitz and of degree n0 = n − 1 − q m . The controller
parameter vector

θ = θ1 θ2 θ3 θ4 ∈ 2n , (18)
is given so that the closed loop plant from r to y p is equal to Wm (s). The I/O properties of the
closed-loop plant shown in Fig. 4 are described by the transfer function equation
y p = Gc (s)r, (19)
where
k p Z p Λ2
Gc ( s ) =

, (20)
Λ θ4 Λ − θ1 α R p − k p Z p θ2 α + θ3 Λ
Now, the objective is to choose the controller gains so that the poles are stable and the
closed-loop transfer function Gc (s) = Wm (s), i.e.,
k p Z p Λ2 Zm

= k m . (21)
Λ θ4 Λ − θ1 α R p − k p Z p θ2 α + θ3 Λ R m
Thus, considering a system free of unmodeled dynamics, the plant coefficients can be known
by the MRC structure, i.e., k p , Z p (s) and R p (s) are given by 21 when the controller gains θ1 ,
θ2 , θ3 and θ4 are known and Wm (s) is previously defined.
6. Parameter identification using RMRAC

The proposed parameter estimation method is executed in three steps, described as follows:
6.1 First step: Convergence of controller gains vector

The proposed parameter identification method is shown in Fig. 2. In this figure the parameter
identification of q axis is shown, but the same procedure is performed for parameter
identification of d axis, one procedure at a time.
∗ is applied at q axis of SPIM at standstill rotor.
A Persistent Excitant (PE) reference current i sq
The current i sq is measured and controlled by the RMRAC structure while i sd stays at null
value. The controller structure is detailed in Fig. 3. When e1 goes to zero, the controller gains
go to an ideal value. Subsequently, the gradient algorithm is put in steady-state and the system
looks like the MRC structure given by Fig. 4. Therefore, the transfer function coefficients can
be found using equation 21.
6.2 Second step: Estimation of k pi , h0i , a1i and a0i

This step consists of the determination of the Linear-Time-Invariant (LTI) model of the
induction motor. The machine is at standstill and the transfer functions given in 2 and 3 can
be generalized as follows
isi Z pi (s) s + h0i

= k pi = k pi 2 , (22)
vsi R pi (s) s + sa1i + a0i
where
Lri R R R
k pi = , h0i = ri , a1i = pi and a0i = si ri , (23)
σ̄i Lri σ̄i
The reference model given by 5 is rewritten as
Zm s + z0
Wm (s) = k m = km 2 , (24)
Rm s + p1 s + p0
and from the plant and reference model assumptions results

m p = 1, n p = 2, n ∗ = 1,
(25)
q m = 1, pm = 2, n ∗m = 1,
The upper bound n is chosen equal to n p because the plant model is considered well known
and with n = n p only one solution is guaranteed for the controller gains. Thus, the filters are
given by

Λ ( s ) = Zm ( s ) = s + z 0 ,
(26)
α ( s ) = z0 ,
Assuming the complete convergence of controller gains, the plant coefficients are obtained
combining the equations 22, 24 and 26 in 21 and are given by
⎧
⎪
⎪ k pi = k m θ4i ,
⎪
⎪
⎪
⎪
⎪ z0
,
⎨ h0i = θ4i − θ1i
θ4i
(27)
⎪
⎪ =
a
⎪ 1i
⎪
p 1 + k m θ3i ,
⎪
⎪
⎪
⎩ a0i = p0 + k m z0 θ + θ3i .
2i
6.3 Third step: Rsi , Rri , L si , Lri and L mi calculation

Combining the equations 4, 23 and using the values obtained in 27 after the convergence of
the controller gains, we obtain the parameters of the induction motor:
⎧ a0i
⎪
⎪ R̂si = ,
⎪
⎪ k
⎪
⎪ pi h0i
⎪
⎪
⎪
⎪ a
⎪
⎪ R̂ri = 1i − R̂si ,
⎪
⎨ k pi
R̂ (28)
⎪
⎪ L̂ si = L̂ri = ri ,
⎪
⎪
⎪
⎪ h0i
⎪
⎪
⎪
⎪
⎪
⎪ R̂si R̂ri
⎪
⎩ L̂ mi = L̂ si −
2 .
a0i
In the numerical solution it-is considered that stator and rotor inductances have the same
values in each winding.
7. Simulation results
Simulations have been performed to evaluate the proposed method. The machine model given
by 1 was discretized by Euller technique under frequency of f s = 5kHz. The SPIM was
performed with a square wave reference of current and standstill rotor. The SPIM used is
a four-pole, 368W, 1610rpm, 220V/3.4A. The parameters of this motor obtained from classical
no-load and locked rotor tests are given in Table 1.
Rsq Rrq L mq L sq
7.00Ω 12.26Ω 0.2145H 0.2459H
Rsd Rrd L md L sd
20.63Ω 28.01Ω 0.3370H 0.4264H
Table 1. Motor parameter obtained from classical tests.
1.5
Output of reference model and plant
0.5
−0.5
−1 yp (isq)
ym (isqm)
−1.5
650 650.1 650.2 650.3 650.4 650.5 650.6
Time (s)
Fig. 5. Plant and reference model output.
The reference model Wm (s) is chosen so that the dynamic will be faster than plant output i sq .
Thus, the reference model is given by
s + 45
, Wm (s) = 180 (29)
s2 + 180s + 8100
The induction motor is started in accordance with Fig. 2 with a Persistent Excitant reference
current signal. A random noise was simulated to give nearly experimental conditions. Fig. 5
show the plant and reference model output after convergence of gains.
Fig. 6 shows the convergence of controller gains for parameter identification of q axis.
This figure shows that gains reach a final value after 600s, demonstrating that parameter
identification is possible. The gain convergence of d axis is shown in Fig. 7. Table 2 presents
the final value of controller gains for the q and d axes, respectively.
θ1q θ2q θ3q θ4q
-0.0096 -1.0925 0.8332 -0.0950
θ1d θ2d θ3d θ4d
-0.0164 -0.6429 0.6885 -0.0352
Table 2. Final value of controller gains obtained in simulation.
The parameters of SPIM are obtained by combining the final value of controller gains from
Table 2 with the equations 27, 28 and the reference model coefficients previously defined in
Convergence of controller gains

0.5
−0.5
−1 θ1q
θ2q
−1.5 θ3q
θ4q
−2
0 100 200 300 400 500 600 700
Time (s)
Fig. 6. Convergence of controller gains vector for q axis.
1
0.5
−0.5
−1 θ1d
θ2d
−1.5 θ3d
θ4d
−2
0 100 200 300 400 500 600 700
Time (s)
Fig. 7. Convergence of controller gains vector for d axis.
equation 29. The results are shown in Table 3. It is possible to observe in simulation that the
electrical parameters converge to machine parameters, even with noise in the currents.
Rsq Rrq L mq L sq
7.07Ω 12.21Ω 0.2150H 0.2462H
Rsd Rrd L md L sd
20.22Ω 27.67Ω 0.3312H 0.4292H
Table 3. Motor parameter identified in simulation.
0.1

0
−0.1
−0.2
−0.3
θ1q
−0.4
θ2q
−0.5 θ3q
θ4q
0 100 200 300 400 500

Time (s)
Fig. 8. Experimental convergence of controller gains for q axis.
This section presents experimental results obtained from the induction motor described in
simulation, whose electrical parameters obtained by classical no-load and locked rotor tests
are shown in Table 1. The drive system consists of a three-phase inverter controlled by
a TMS320F2812 DSP controller. The sampling period is the same used in the preceding
simulation.
Unlike the simulation, unmodeled dynamics by drive, sensors and filters, among others, are
included in the implementation. This implies that the plant model is a little different from
physical plant. As a result, there is a small error that is proportional to plant uncertainties
and is defined here as a residual error. The tracking error e1 can be minimized by increasing
the gradient gain P. However, increasing P in order to eliminate the residual error can cause
divergence of controller gains and the system becomes unstable.
To overcome this problem a stopping condition was defined for the gain convergence. The
identified stator resistance R̂si was compared to measured stator resistance Rsi obtained from
measurements. Thus, the gradient gain P must be adjusted until the identified stator resistance
is equal to the stator resistance measurement.
Figure 8 presents the convergence of controller gains for q axis. The gains reach a final value
after 400s. Figure 9 presents the convergence of controller gains for d axis. The value of the
gain that resulted in R̂si = Rsi was P = 20I.
The plant output isq and reference model output i sqm are shown in Fig. 10, after controller gain
convergence, where it is possible to see the residual error between the two curves. The final
values of controller gains, for axes q and d, are shown in Table 4.
The parameters of SPIM are obtained by combining the final value of controller gains of Table
4 with the equations 27, 28 and the reference model coefficients previously defined in equation
29. The results are shown in Table 5.
0.5

0
−0.5
−1
θ1d
θ2d
−1.5
θ3d
θ4d
−2
0 100 200 300 400 500
Time (s)
Fig. 9. Experimental convergence of controller gains for d axis.
2
Output of reference model and plant
1.5
0.5
−0.5
−1
−1.5 yp (isq)
ym (isqm)
−2
401 401.2 401.4 401.6 401.8 402
Time (s)
Fig. 10. Experimental plant and reference model output.
θ1q θ2q θ3q θ4q
0.0136 -0.558 -0.0555 -0.0423
θ1d θ2d θ3d θ4d
0.0042 -0.6345 0.1560 -0.0207
Table 4. Final value of controller gains obtained in experimentation.
Rsq Rrq L mq L sq
6.9105Ω 15.4181Ω 0.1821H 0.2593H
Rsd Rrd L md L sd
20.9438Ω 34.9016Ω 0.4926H 0.6447H
Table 5. Motor parameter identified in experimentation.
isq(measured)
1.5
Measured and simulated currents
isq (simulated)
1
0.5
−0.5
−1
−1.5
−2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Time (s)
Fig. 11. Transient of comparison among measured and simulated currents from q axis.
8.1 Comparison for parameter validation

A comparative study was performed to validate the parameters obtained by the proposed
technique. The physical SPIM was fed by steps of 50V and 30Hz. The currents of main
winding, auxiliary winding and speed rotor were measured.
Then, the dynamic model of SPIM given by 1 was simulated, using the parameters from Table
5, under the same conditions experimentally, i.e., steps of 50V and 30Hz. The measured rotor
speed was used in the simulation model to make it independent of mechanical parameters.
Thus, the simulated currents, from q and d windings, were compared with measured currents.
Figures 11 and 12 shows the transient currents from axis q and d, respectively, while Figures
13 and 14 shows the steady-state currents from axis q and d, respectively.
From Figures 11-14, it is clear that the simulated machine with the proposed parameters
presents similar behavior to the physical machine, both in transient and steady-state.
9. Conclusions
This chapter describes a method for the determination of electrical parameters of single phase
induction machines based on a RMRAC algorithm, which initially was used in three-phase
induction motor estimation in (Azzolin & Gründling, 2009). Using this methodology, it is
possible to obtain all electrical parameters of SPIM for the simulation and design of an high
performance control and sensorless SPIM drives. The main contribution of this proposed work
is the development of automated method to obtain all electric parameters of the induction
machines without the requirement of any previous test and derivative filters. Simulation
1
isd(measured)

isd (simulated)
0.5
−0.5
−1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

Time (s)
Fig. 12. Transient of comparison among measured and simulated currents from d axis.
1.5
0.5
−0.5
−1
isq(measured)
−1.5 isq (simulated)
0.5 0.52 0.54 0.56 0.58 0.6 0.62 0.64 0.66 0.68 0.7
Time (s)
Fig. 13. Steady-state of comparison among measured and simulated currents from q axis.
results demonstrate the convergence of the parameters to ideal values, even in the presence of
noise. Experimental results show that the parameters converge to different values in relation
to the classical tests shown in Table 1. However, the results presented in Figures 11-14 show
that the parameters obtained by proposed method present equivalent behavior to physical
machine.
0.8

0.6
0.4
0.2
−0.2
−0.4
−0.6
isd(measured)
−0.8
isd (simulated)
−1
0.5 0.52 0.54 0.56 0.58 0.6 0.62 0.64 0.66 0.68 0.7
Time (s)
Fig. 14. Steady-state of comparison among measured and simulated currents from d axis.
10. References
Azzolin, R. & Gründling, H. (2009). A mrac parameter identification algorithm for three-phase
induction motors, Electric Machines and Drives Conference, 2009. IEMDC ’09. IEEE
International, pp. 273 –278.
Azzolin, R., Martins, M., Michels, L. & Gründling, H. (2007). Parameter estimator of an
induction motor at standstill, Industrial Electronics, 2009. IECON ’09. 35th Annual
Conference of IEEE, pp. 152 – 157.
Blaabjerg, F., Lungeanu, F., Skaug, K. & Tonnes, M. (2004). Two-phase induction motor drives,
Industry Applications Magazine, IEEE 10(4): 24 – 32.
Câmara, H. & Gründling, H. (2004). A rmrac applied to speed control of an induction motor
without shaft encoder, Decision and Control, 2004. CDC. 43rd IEEE Conference on, Vol. 4,
pp. 4429 – 4434 Vol.4.
de Rossiter Correa, M., Jacobina, C., Lima, A. & da Silva, E. (2000). Rotor-flux-oriented control
of a single-phase induction motor drive, Industrial Electronics, IEEE Transactions on
47(4): 832 –841.
Donlon, J., Achhammer, J., Iwamoto, H. & Iwasaki, M. (2002). Power modules for appliance
motor control, Industry Applications Magazine, IEEE 8(4): 26 –34.
Ioannou, P. & Sun, J. (n.d.). Robust Adaptive Control, Prentice Hall.
Ioannou, P. & Tsakalis, K. (1986). A robust direct adaptive controller, Automatic Control, IEEE
Transactions on 31(11): 1033 – 1043.
Koubaa, Y. (2004). Recursive identification of induction motor parameters, Simulation
Modelling Practice and Theory 12(5): 363 – 381.
URL: http://www.sciencedirect.com/science/article/B6X3C-4CJVR4N-1/2/bcc4bdf719a9c7
ad5b99750423f5ff23
Krause, P., Wasynczuk, O. & Sudhoff, S. (n.d.). Analysis of Electric Machinery, NJ: IEEE Press.
Lozano-Leal, R., Collado, J. & Mondie, S. (1990). Model reference robust adaptive control
without a priori knowledge of the high frequency gain, Automatic Control, IEEE
Transactions on 35(1): 71 –78.
Ojo, O. & Omozusi, O. (2001). Parameter estimation of single-phase induction machines,
Industry Applications Conference, 2001. Thirty-Sixth IAS Annual Meeting. Conference
Record of the 2001 IEEE.
Ribeiro, L., Jacobina, C. & Lima, A. (1995). Dynamic estimation of the induction machine
parameters and speed, Power Electronics Specialists Conference, 1995. PESC ’95 Record.,
26th Annual IEEE, Vol. 2, pp. 1281 –1287 vol.2.
Vaez-Zadeh, S. & Reicy, S. (2005). Sensorless vector control of single-phase induction motor
drives, Electrical Machines and Systems, 2005. ICEMS 2005. Proceedings of the Eighth
International Conference on, Vol. 3, pp. 1838 –1842 Vol. 3.
Velez-Reyes, M., Minami, K. & Verghese, G. (1989). Recursive speed and parameter estimation
for induction machines, Industry Applications Society Annual Meeting, 1989., Conference
Record of the 1989 IEEE, pp. 607 –611 vol.1.
Vieira, R., Azzolin, R. & Gründling, H. (2009a). Parameter identification of a single-phase
induction motor using rls algorithm, Power Electronics Conference, 2009. COBEP ’09.
Brazilian.
Vieira, R., Azzolin, R. & Gründling, H. (2009b). A sensorless single-phase induction motor
drive with a mrac controller, Industrial Electronics, 2009. IECON ’09. 35th Annual
Conference of IEEE, pp. 1003 –1008.
9
Swarm Intelligence Based Controller

for Electric Machines and Hybrid
Electric Vehicles Applications
Omar Hegazy1, Amr Amin2, and Joeri Van Mierlo1
1Facultyof Engineering Sciences Department of ETEC- Vrije Universiteit Brussel,
2Power and Electrical Machines Department, Faculty of Engineering – Helwan University,
1Belgium
2Egypt
1. Introduction
Swarm Intelligence in the form of Particle Swarm Optimization (PSO) has potential
applications in electric drives. The excellent characteristics of PSO may be successfully
used to optimize the performance of electric machines and electric drives in many
aspects. It is estimated that, electric machines consume more than 50% of the world
electric energy generated. Improving efficiency in electric drives is important, mainly,
for two reasons: economic saving and reduction of environmental pollution. Induction
motors have a high efficiency at rated speed and torque. However, at light loads, the
iron losses increase dramatically, reducing considerably the efficiency. Swarm
intelligence is used to optimize the performance of three applications; these
applications are represented as follows:
• Losses Minimization of two asymmetrical windings induction motor
• Maximum efficiency and minimum operating cost of three-phase induction motor
• Optimal electric drive system for fuel cell hybrid electric vehicles.
In this chapter, a field-oriented controller that is based on Particle Swarm Optimization is
presented. In this system, the speed control of two asymmetrical windings induction motor
is achieved while maintaining maximum efficiency of the motor. PSO selects the optimal
rotor flux level at each operating point. In addition, the electromagnetic torque is also
improved while maintaining a fast dynamic response. A novel approach is used to evaluate
the optimal rotor flux level by using Particle Swarm Optimization. PSO method is a member
of the wide category of Swarm Intelligence methods (SI). The swarm intelligence is based on
real life observations of social animals (usually insects), it is more flexibility and robust than
any traditional optimization methods. PSO algorithm searches for global optimization for
nonlinear problems with multi-objective. There are two speed control strategies explained in
the next sections. These are field-oriented controller (FOC), and FOC based on PSO. The
strategies are implemented mathematically and experimental. The simulation and
experimental results have demonstrated that the FOC based on PSO method saves more
energy than the conventional FOC method.
In this chapter, another application of PSO for losses and operating cost minimization
control is presented for the induction motor drives. Two strategies for induction motor
speed control are proposed. These strategies are maximum efficiency strategy (MES), based
PSO, and minimum operating cost Strategy. The proposed technique is based on the
principle that the flux level in a machine can be adjusted to give the minimum amount of
losses and minimum operating cost for a given value of speed and load torque.
In the demonstrated systems, the powertrain components sizing and the power control
strategy are the only adjustable parameters to achieve optimal power sharing between
sources and optimal design with minimum cost, minimum fuel consumption, and
maximum efficiency for Electric Vehicles (EVs) and Hybrid Electric Vehicles (HEVs). Their
selection greatly influences the performance of the drive system in Hybrid Electric Vehicles
applications. In this section, the design and power management control are investigated and
optimized by using Particle Swarm Optimization.
2. Losses minimization of two asymmetrical windings induction motor

In this section, a field orientation based on Particle Swarm Optimization (PSO) is applied to
control the speed of two-asymmetrical windings induction motor. The maximum efficiency
of the motor is obtained by the evaluation of optimal rotor flux at each operating point. In
addition, the electro-magnetic torque is also improved while maintaining a fast dynamic
response. In this section, a novel approach is used to evaluate the optimal rotor flux level.
This approach is based on Particle Swarm Optimization (PSO). This section presents two
speed control strategies. These are field-oriented controller (FOC) and FOC based on PSO.
The strategies are implemented mathematically and experimental. The simulation and
experimental results have demonstrated that the FOC based on PSO method saves more
energy than the conventional FOC method [Hegazy, 2006; Amin et al., 2007; Amin et al.,
2009]. The two asymmetrical windings induction motor is treated as a two-phase induction
motor (TPIM). It is used in many low power applications, where three–phase supply is not
readily available. This type of motor runs at an efficiency range of 50% to 65% at rated
operating conditions.
The conventional field-oriented controller normally operates at rated flux at any values with
its torque range. When the load is reduced considerably, the core losses become so high
causing poor efficiency. If significant energy savings are required, it is necessary to optimize
the efficiency of the motor. The optimum efficiency is obtained by the evaluation of the
optimal rotor flux level. This flux level is varied according to the torque and the speed of the
operating point.
PSO is applied to evaluate the optimal flux. It has the straightforward goal of minimizing
the total losses for a given load and speed. It is shown that the efficiency is reasonably close
to optimal.
2.1 Mathematical model of the motor

The d-q model of an unsymmetrical windings induction motor in a stationary reference
frame can be used for a dynamic analysis. This model can take in account the core losses.
The d-q model as applied to TPIM is described in [Hegazy, 2006; Amin et al., 2009]. The
equivalent circuit is shown in Fig. 1.
Swarm Intelligence Based Controller for
Electric Machines and Hybrid Electric Vehicles Applications 163
rM LlM Llr (1/k)ω rλ dr

+
iqs iqr + -
Vqs rr
Rqfe iqfe Lmq
rA LlA LlR
kω rλ qr
+
ids idr - +
rR
Vds Rdfe idfe Lmd
-
Fig. 1. The d-q axes two-phase induction motor Equivalent circuit with iron losses.
The machine model may be expressed by the following voltage and flux equations:
Voltage Equations (1):
vqs = rm i qs + pλ qs (1)
vds = ra i ds + pλds (2)
0 = rr i qr − (1 / k ) * ωr λ dr + pλ qr (3)
0 = rR i ds + k * ωr λqr + pλdr (4)
0 = − i qfe R qfe + L mq ( p iqs + p i qr − p i qfe ) (5)
0 = − i dfe R dfe + L md ( p ids + p i dr − p i dfe ) (6)
Flux Equations:
λ qs = L lm i qs + L mq ( i qs + i qr − i qfe ) (7)
λ ds = L la i ds + L md ( i ds + i dr − i dfe ) (8)
λ qr = L lr i qr + L mq ( i qs + i qr − i qfe ) (9)
λ dr = L lR i dr + L md ( i ds + i dr − i dfe ) (10)
Electrical torque equation is expressed as:
P 1
Te = ( k Lmq idr ( iqs + iqr − iqfe ) − L md iqr ( ids + idr − iqfe ) (11)
2 k
Te − T l = j m p ω r + B mω r (12)
2.2 Field-Oriented Controller [FOC]

The stator windings of the motor are unbalanced. The machine parameters differ from the d
axis to the q axis. The waveform of the electromagnetic torque demonstrates the unbalance
of the system. The torque in equation (11) contains an AC term; it can be observed that two
values are presented for the referred magnetizing inductance. It is possible to eliminate the
AC term of electro-magnetic torque by an appropriate control of the stator currents.
However, these relations are valid only in linear conditions. Furthermore, the model is
implemented using a non-referred equivalent circuit, which presumes some complicated
measurement of the magnetizing mutual inductance of the stator and the rotor circuits.
The indirect field-oriented control scheme is the most popular scheme for field-oriented
controllers. It provides decoupling between the torque and the flux currents. The electric
torque must be a function of the stator currents and rotor flux in synchronous reference
frame [Popescu & Navrapescu, 2000]. Assuming that the stator currents can be imposed as:
i sds = i sds 1 (13)
i s qs = k i s qs 1 (14)
Where: k= Msrd / Msrq
P ⎡
Te = Msqr i s qs λ s dr − Msdr i s ds λ s qr ⎤ (15)
2 Lr ⎣ ⎦
By substituting the variables ids, and iqs by auxiliary variables ids1, and iqs1 into (15) the torque
can be expressed by
P Msdr ⎡ s
Te = i λ s − i s ds 1 λ s qr ⎤ (16)
2 Lr ⎣ qs 1 dr ⎦
In synchronous reference frame, the electromagnetic torque is expressed as:
P Msdr ⎡ e
Te = i λ e − i e ds 1 λ e qr ⎤ (17)
2 Lr ⎣ qs 1 dr ⎦
P Msdr ⎡ e
Te = i λe ⎤ (18)
2 Lr ⎣ qs 1 r ⎦
λ er
i e ds 1 = (19)
Msdr
Msdr
ωe − ωr = i e qs 1 (20)
τ r * λr
2.3 Synchronous reference frame for losses model

It is very complex to find the losses expression for the two asymmetrical windings induction
motor with losses model. In this section, a simplified induction motor model with iron
losses will be developed. For this purpose, it is necessary to transform all machine variables
to the synchronous reference frame. The voltage equations are written in expanded form as
follows [Hegazy, 2006; Amin et al., 2006; Amin et al., 2009]:
e e
e e
diqs diqm
vqs = rmiqs + Llm + Lmq + ωe (Llaidse + Lmd idm
e
) (21)
dt dt
e
e e dids di e
vds = raids + Lla + Lmd dm − ωe (Llmiqse + Lmq iqm
e
) (22)
dt dt
e e
e
diqr diqm ωsl e e
0 = rr iqr + Llr + Lmq + (LlRidr + Lmd idm ) (23)
dt dt k
e
e didr di e e e
0 = rR idr + LlR + Lmd dm − k * ωsl (Llr iqr + Lmq iqm ) (24)
dt dt
e e e e
iqs + iqr = iqfe + iqm (25)
e e e e
ids + idr = idfe + idm (26)
e
ωe Llr Lmqs e
vdm =− iqs (27)
Lr
e e
vqm = ωe Lmdsids (28)
Where:
e
veqm e
e
vdm
i dfe = ; i dfe =
R qfe R dfe
The losses in the motor are mainly:

a. Stator copper losses,
b. Rotor copper losses,
c. Core losses, and
d. Friction losses
The total electrical losses can be expressed as follows
Plosses = Pcu1 + Pcu2 +Pcore (29)

Where:
Pcu1 : Stator copper losses
Pcu2 : Rotor copper losses
Pcore : Core losses
The stator copper losses of the two asymmetrical windings induction motor are caused by
electric currents flowing through the stator windings. The core losses of the motor are
produced from the hysteresis and eddy currents in the stator. The total electrical losses of
motor can be rewritten as:
e2 2
e2 e 2 e 2 e 2 vqm e
vdm
Plosses = rmiqs + raids + rr iqr + rR idr + + (30)
Rqfe Rdfe
The total electrical losses are obtained as follows:
⎛ ⎞
⎡ ⎜ ⎟ ⎛
rr L2mqs ωe 2 L2lr L2mqs ⎤ ⎜ Te2 L2r ⎟ + ⎜ r + ωe Lmds ⎞⎟ λr
2 2 2
Plosses = ⎢rm + 2 + ⎥ (31)
L2r Rdfe ⎥⎦ ⎜ 2 ⎛ Lmds ⎞2 2 ⎟ ⎜⎝ Rqfe ⎟⎠ L2mds
a
⎣⎢ Lr
⎜⎜ P ⎜ ⎟ λr ⎟⎟
⎝ ⎝ K ⎠ ⎠
2Te * rr
ωsl = (32)
P * λr2
Where:
ωe = ωr+ ωsl, and ωsl is the slip speed (rad/sec).
Equation (31) is the electrical losses formula, which depends on rotor flux (λr) according to
operating point (speed and load torque). The total losses of the motor ( TP losses ) are given
as follows:
TP losses = Plosses + PFric = Pin - Pout (33)
Pout
Efficiency (η ) = (34)
Pout + TP losses
Where:
Friction power losses = F ∗ωr2, and
Output power (Pout) = TL∗ωr.
2.4 Losses minimization control scheme

The equation (33) is the cost function, the total losses, which depends on rotor flux (λr)
according to the operating point. Figure 2 presents the distribution of losses in motor and its
variation with the flux. As the flux reduces from the rated value, the core losses decrease,
but the motor copper losses increase. However, the total losses decrease to a minimum value
and then increase again. It is desirable to set the rotor flux at the optimal value, so that the
efficiency is optimum [Hegazy, 2006; Amin et al., 2006; Amin et al., 2009].
Fig. 2. Losses variation of the motor with varying flux

PSO is applied to evaluate the optimal flux that minimizes the motor losses. The problem
can be formulated as follows:
minimize TP losses = Γ (λ r,TL , ωr ) (35)
2.4.1 Particle Swarm Optimization (PSO)

Particle swarm optimization (PSO) was originally designed and introduced by Eberhart and
Kennedy [Ebarhart, Kennedy, 1995; Ebarhart, Kennedy, 2001]. Particle Swarm Optimization
(PSO) is an evolutionary computation technique (a search method based on a nature
system). It can be used to solve a wide range of optimization problems. Most of the
problems that can be solved using Genetic Algorithms (GA) could be solved by PSO. For
example, neural network training and nonlinear optimization problems with continuous
variables can be easily achieved by PSO [Ebarhart, Kennedy, 2001]. It can be easily
expanded to treat problems with discrete variables.
The system initially has a population of random solutions. Each potential solution, called a
particle. Each particle is given a random velocity and is flown through the problem space.
The particles have memory and each particle keeps track of its previous best position (call
the pbest) and with its corresponding fitness. There exit a number of pbest for the respective
particles in the swarm and the particle with greatest fitness is called the global best (gbest)
of the swarm. PSO can be represented by the concept of velocity and position. The Velocity
of each agent can be modified by the following equations: (36 & 38):
vk + 1 = w vi k + c1 r1 * ( pbest − si k ) + c 2 r2 * ( gbest − si k ) (36)

Start
Read motor parameters
Determination of the new

operation conditions
(speed and torque)
Run the Simulik model of two asymmetrical

windings induction motor with losses
varying the rotor

Calculation of motor currents flux level using PSO
calculation of the cost

function
Is
the value of
No
the cost
function is
minimum
?
Yes
set this value as optimum point

and record the corresponding
optimum values of the efficiency
and the rotor flux load
No Are
all possible operating
conditions optimized
Yes
End
Fig. 3. The flowchart of the execution of the PSO [Hegazy, 2006)

ωmax − ωmin
ω = ωmax − * iter (37)
itermax
Using the above equations, a certain velocity can be calculated that gradually gets close to
(pbest) and (gbest). The current position (searching point in the solution space) can be
modified by the following equation:
Si k + 1 = Si k + vi k + 1 (38)
Where:
νk : Current velocity of agent i at iteration.
ν i k+1 : Modified velocity of agent i
r1 , r2 : random number distributed [0,1],
Ski : current position of agent i,
ω : weight function for velocity of agent i,
c1, c2 : positive constants; [c1+ c2< 4].
ωmax : Initial weight,
ωmin : Final weight,
itermax : Maximum iteration number,
iter : Current iteration number.
• In (35), the losses formula is the cost function of the PSO. The particle swarm
optimization (PSO) technique is used for minimizing this cost function.
• The PSO is applied to evaluate the optimal rotor flux that minimizes the motor losses at
any operating point. Figure 3 presents the flowchart of the execution of PSO, which
evaluates the optimal flux by using MATLAB /SIMULINK.
The optimal flux is the input of the indirect rotor flux oriented controller. The indirect field-
oriented controller generates the required two reference currents to drive the motor
corresponding to the optimal flux. These currents are fed to the hysteresis current controller
of the two-level inverter. The switching pattern is generated according to the difference
between the reference current and the load current through the hysteresis band. Figure 4
shows a whole control diagram of the proposed losses-minimization control system.
Vdc Vdc
main
Te* i qs * =i a ref. Rotor
Four Switch
Hysteresis-band
FO Inverter aux
ϕ* i ds *= i b ref. current control [FSI]
flux optimal
ib actual
ia actual
PSO
M/C
Tl Nr Paramerters
Fig. 4. The proposed losses minimization control system.


In this section, the proposed application is implemented numerically using MATLAB-
SIMULINK to validate the performance of the proposed control strategy. The motor used in
this study has the following parameters, which were measured by using experimental tests.
Table 1 shows motor parameters. The used parameters of the PSO are shown in Table 2.
Rated power 750 w

V 220 v
F 50 Hz
rM 4.6 Ω
rA 10.6 Ω
X Lm 4.31 Ω
X La 7.1472 Ω
rr 3.455 Ω
X Lr 4.284 Ω
X mq 89.65Ω
Xmd 169.43Ω
Rqfe 1050Ω
Rdfe 1450Ω
J 0.005776 kg.m2
B 0.00328N.m.sec/r
Pole pair 2
Table 1. Motor Parameters
The optimal rotor flux provides the maximum efficiency at any operating point. There are
six-cases of the motor operation are studied by using FOC based on PSO. PSO will evaluate
the optimal rotor flux level. This flux is fed to the FOC module. Figure 5 shows the
performance of the motor at case (1) (TL=0.25 PU, Nr=0.5 Nrated), when PSO is applied side-
by-side FOC as shown in Fig.4.
Population size 10
Max. iter 50
c1 0.5
c2 0.5
Max. weight 1.4
Min. weight 0.1
r1 [ 0,1]
r2 [ 0,1]
Lower Bound 0.2
Upper Bound 2
Table 2. PSO Algorithm parameter
(a) (b)
(c) (d)
(e) (f)
Fig. 5. Simulation results of the motor at case (1).Speed-time curve, (b) Torque-time curve,
(c) The stator current in q-axis, (d) The stator current in d-axis, (e) Total Losses against
iterations, (f) Efficiency against iterations
Figure 6 illustrates the comparison between FOC and FOC based PSO control methods at
different operating points. Figure 7 presents the optimal flux, which is obtained by applying
PSO. Table 3 presents the summary of the results of FOC and FOC based PSO methods.
Fig. 6. the comparison between FOC and FOC based PSO
Fig. 7. The optimal flux at different load torque
FOC FOC based PSO

TL Nr
Cases λ λ Optimal Improvement
(PU) (rpm) η (%) η (%)
(PU) (PU) (%)
(1) 0.25 0.5 Nrated 1 33.85 0.636 46.11 36.22

(2) 0.375 0.5 Nrated 1 36.51 0.6906 49.15 34.62
(3) 0.5 0.5 Nrated 1 48.21 0.722 57.11 18.46
(4) 0.6125 0.5 Nrated 1 55.15 0.761 62.34 13.04
(5) 0.75 0.5 Nrated 1 60.175 0.8312 65.31 8.53
(6) 1 0.5 Nrated 1 63.54 0.8722 68.15 7.26
Table 3. Summary of the results of the two controllers
In practical system, the flux level based on PSO at different operating points (torque and
speed) is calculated and stored in a look up table (LUT). The use of look up table will enable
the system to work in real time without any delay that might be needed to calculate the
optimal point. The proposed controller would receive the operating point (torque and
speed) and get the optimum flux (λoptimal) from the look up table. It will generate the
required reference current. It is noticed that, the efficiency with the FOC based on PSO
method is higher than the efficiency with the FOC method only.
2.6 Experimental results

To verify the validity of the proposed control scheme, a laboratory prototype is built and
tested [Hegazy, 2006; Amin et al., 2006; Amin et al., 2009]. The basic elements of the
proposed experimental scheme are shown in Fig. 8 and Fig. 9. The experimental results of
the motor are achieved by coupling the motor to an eddy current dynamometer. The
experimental results are achieved using two control methods:
• Field-Oriented Control [FOC], and
• Field-Oriented Control [FOC] based on PSO.
The reference and the actual motor currents are fed to the hysteresis current controller. The
switching pattern of the two-level four-switch inverter [FSI] is generated according to the
difference between the reference currents and the load currents. Figure 10 shows the
experimental results of the motor with FOC at case (1), where the motor is loaded by TL =
0.25 PU. Figure 11 shows the experimental result of the motor with FOC based on PSO at
case (1). The cases are summarized in Table 4.
Two dc-supply
Switching
Pattern
Inverter
isqs
4
ωref I a_ref
T*e
INTER-FACE
Indirect ia
+ PI dq-->ab I aact
- Field-Oriented
s
transfor- Hystersis
i ds me
φ* current
I b_ref
Rotor
ib controller I bact
PSO
ωactual Shaft
encoder
ωactual
M/c
paramaters
Tl Nr
Fig. 8. Block diagram of the proposed experimental scheme [Hegazy, 2006; Amin et al., 2009]
FOC FOC with PSO

Improvement
Cases λ Power η λ Power η (%)
(PU Input (W) (%) (PU Input (W) (%)
(1) 1 235 32.3 0.636 169 44.92 39.07
(2) 1 323 35.2 0.690 243 47.06 33.69
Table 4. The summary of the two-cases
D D
Rs Rs
S1 D1 S2 D2
Cs Cs
Edc G1 G2 S
S iqs
main
D rotor
Edc D
S2 D3
Rs S4 D4
Rs
Cs G4 S Cs
G2
S ids aux
Fig. 9. The power circuit of Four Switch Inverter [FSI]
(a)
(b)
(c)
Fig. 10. Experimental results of FOC method; the reference and actual speed, (b) the
reference and actual current in q-axis, (c) The reference and actual current in d-axis
(a)
(b)
Fig. 11. Experimental results of FOC method based on PSO. (a) The reference and actual
current in q-axis, the reference and actual current in d-axis
Finally, these results demonstrate that, the FOC based on PSO method saves more energy
than conventional FOC method. Thus, the efficiency with PSO is improved than it's at FOC.
3. Maximum efficiency and minimum operating cost of induction motors

This section presents another application of PSO for losses and operating cost minimization
control in the induction motor drives. Two control strategies for induction motor speed
control are proposed. Those two strategies are based on PSO and called Maximum Efficiency
Strategy and Minimum Operating Cost Strategy [A. Hamid et al. 2006]. The proposed
technique is based on the principle that the flux level in the machine can be adjusted to give
the minimum amount of losses and minimum operating cost for a given value of speed and
load torque. The main advantages of the proposed technique are; its simple structure. It is a
straightforward maximization of induction motor efficiency and its operating cost for a given
load torque. As was demonstrated, PSO is efficient in finding the optimum operating
machine's flux level. The optimum flux level is a function of the machine operating point.
The main induction motor losses are usually split into five components: stator copper losses,
rotor copper losses, iron losses, mechanical losses, and stray losses [Kioskeridis & Margaris,
1996].
The efficiency that decreases with increasing losses can be improved by minimizing the losses.
Copper losses reduce with decreasing the stator and the rotor currents, while the core losses
essentially increase with increasing air-gap flux density. A study of the copper and core losses
components reveals that their trends conflict. When the core losses increase, the copper losses
tends to decrease. However, for a given load torque, there is an air-gap flux density at which
the total losses is minimized. Hence, electrical losses minimization process ultimately comes
down to the selection of the appropriate air-gap flux density of operation. Since the air-gap
flux density must be variable when the load is changing, control schemes in which the (rotor,
air-gap) flux linkage is constant will yield sub-optimal efficiency operation especially when the
load is light. Then to improve the motor efficiency, the flux must be reduced when it operates
under light load conditions by obtaining a balance between copper and iron losses.
The challenge to engineers, however, is to be able to predict the appropriate flux values at
any operating points over the complete torque and speed range which will minimize the
machines losses, hence maximizing the efficiency. In general, there are three different
approaches to improve the induction motor efficiency especially under light-load.
a. Losses Model Controller (LMC)
This controller depends on a motor losses model to compute the optimum flux analytically.
The main advantage of this approach is its simplicity and it does not require extra hardware.
In addition, it provides smooth and fast adaptation of the flux, and may offer optimal
performance during transient operation. However, the main problem of this approach is
that it requires the exact values of machine parameters. These parameters include the core
losses and the main inductance flux saturation, which are unknown to the users and change
considerably with temperature, saturation, and skin effect. In addition, these parameters
may vary due to changes in the operating conditions. However, with continuing
improvement of evolutionary parameter determination algorithms, the disadvantages of
motor parameters dependency are slowly disappearing.
b. Search Controller (SC)
This controller measures the input power of the machine drive regularly at fixed time
intervals and searches for the flux value, which results in minimum power input for given
values of speed and load torque. This particular method does not demand knowledge of the
machine parameters and the search procedure is simple to implement.
However, some disadvantages appear in practice, such as continuous disturbances in the
torque, slow adaptation (7sec.), difficulties in tuning the algorithm for a given application,
and the need for precise load information. In addition, the precision of the measurements
may be poor due to signal noise and disturbances. This in turn may cause the SC method to
give undesirable control performance. Moreover, nominal flux is applied in transient state
and is tuned after the system reaches steady state to an optimal value by numerous
increments, thus lengthening the optimization process. Therefore, the SC technique may be
slow in obtaining the optimal point. In addition, in real systems, it may not reach a steady
state and so cause oscillations in the air gap flux that result in undesirable torque
disturbances. For these reasons, this is not a good method in industrial drives.
c. Look Up Table Scheme
It gives the optimal flux level at different operating points. This table, however, requires
costly and time-consuming prior measurements for each motor. In this section, a new
control strategy uses the loss model controller based on PSO is proposed. This strategy is
simple in structure and has the straightforward goal of maximizing the efficiency for a given
load torque. The resulting induction motor efficiency is reasonably close to optimal. It is
well known that the presence of uncertainties, the rotor resistance, for instance makes the
result no more optimal. Digital computer simulation results are obtained to demonstrate the
effectiveness of the proposed method.
3.1 Definition of operating strategies

The following definitions are useful in subsequent analyses. Referring to the analysis of the
induction motor presented in [A. Hamid et al. 2006], the per-unit frequency is
ω e ωs + ωr (39)
a= =
ωb ωb
The slip is defined by
ωs ωs
s= = (40)
ωb ωs + ωr
The rotor current is given by
' φm (41)
Ir = 2
⎛ '⎞
⎜ r r ⎟ + X'
2
⎜ sa ⎟ lr
⎝ ⎠
The electromagnetic torque is given by
⎛ '⎞
r
⎜ r⎟
⎜ sa ⎟
⎜ ⎟ (42)
⎝ ⎠
φ
2
Te = 2
⎛ ⎞
' m
r
⎜ r⎟
⎜ sa ⎟ + X
' 2
lr
⎜ ⎟
⎝ ⎠
The stator current is related to the air gap flux and the electromagnetic torque as:
2
⎛
I s = ⎜ S1 φm + S 2 φ
3
+ S2 φ
5⎞
2
⎟ + CL
T e
(43)
φ
2
⎝ m m⎠
m
Where
'
CL = 1 + 2 ×
x lr
x m
The air gap flux is related to the electromagnetic torque as:
2
⎛ '⎞
φm =
sa ⎜ r ⎟
'
r
⎜ sa ⎟ + x
' 2
Te
(44)
lr
r ⎜
r ⎝
⎟
⎠
The efficiency is defined as the output power divided by the electric power supplied to the
stator (inverter losses are included):
Pout
Efficiency (η ) = (45)
Pin
3.1.1 Maximum efficiency strategy

In MES (Maximum Efficiency Strategy), the slip frequency is adjusted so that the efficiency
of the induction motor drive system is maximized [A. Hamid et al. 2006].
The induction motor losses are the following:
1. Copper losses: these are due to flow of the electric current through the stator and rotor
windings and are given by:
2 ' '2
Pcu = rs I s + r r I r (46)
2. Iron losses: these are the losses due to eddy current and hysteresis, given by
( )
Pcore = K e 1 + S 2 a2 ϕ + K h ( 1 + S ) aϕ
2
m
2
m
(47)
3. Stray losses: these arise on the copper and iron of the motor and are given by:
'2
Pcu = C str ω r I r
2
(48)
4. Mechanical losses: these are due to the friction of the machine rotor with the bearings
and are given by:
Pfw = C fw + ω r
2
(49)
5. Inverter losses: The approximate inverter loss as a function of stator current is given by:
2
Pinv = K 1inv i s + K 2 inv i s (50)
Where: K1inv, K2inv are coefficients determined by the electrical characteristics of a switching
element where: K1inv= 3.1307e-005, K2inv=0.0250.
The total power losses are expressed by:
⎡ 2 ' ' 2⎤
P losses = Pcu + Pcore + Ps + Pfw + Pinv = ⎢⎣rs I s + r r I r ⎥⎦ +
( )
⎡K 1 + S 2 a2 2 ⎤ + ⎡K ( 1 + S ) a 2 ⎤ + ⎡C
ϕ m ⎥⎦ ⎢⎣ str ω r2 I 'r ⎤⎥⎦
2
⎢⎣ e ϕ m ⎥⎦ ⎢⎣ h (51)
+ ⎡⎢K 1inv i s + K 2inv i s ⎦⎤⎥

2
⎣
The output power is given by:
Pout = TL × ω r (52)
The input power is given by:
⎡ ' 2⎤
P in = P out + P losses = Pcu + Pcore + Ps + Pfw + Pinv = ⎢⎣rs I s + r r I r
2 '
⎥⎦ +
( )
⎡K 1 + S 2 a2 2 ⎤ + ⎡K ( 1 + S ) a 2 ⎤ + ⎡C
ϕ m ⎦⎥ ⎣⎢ str ω r2 I 'r ⎤⎦⎥
2
⎣⎢ e
ϕ m ⎦⎥ ⎣⎢ h (53)
+ ⎡⎢K 1inv i s + K 2 inv i s ⎦⎤⎥ + TL × ωr

2
⎣
The efficiency is expressed as:
TL × ω r
η=
⎡⎡
⎢ ⎢rs
2
I +r I
s
'
r
' 2⎤
r
⎡
( 2 2
⎥ + ⎢Ke 1 + S a ) φ 2⎤
m⎥
⎡
+ ⎢ K h (1 + S ) a φ
2⎤
m⎥
⎡
+ ⎢C str ωI
2 ' 2 ⎤⎤
r ⎥⎥ (54)
⎢⎣ ⎦ ⎣ ⎦ ⎣ ⎦ ⎣ r
⎦⎥
⎢ ⎡ 2 ⎤ ⎥
⎢
⎣⎢
+⎢
⎣ K 1inv i +K
s 2inv i ⎥⎦ + T
s L ×ω ⎥
⎥⎦
The efficiency maximization of the induction motor problem can be formulated as follows:
Maximize η (TL , ω s , ω r ) (55)
The maximization should observe the fact that the amplitude of the stator current and flux
cannot exceed their specified maximum point.
3.1.2 Minimum operating cost strategy

In Minimum Operating cost Strategy (MOCS), the slip frequency is adjusted so that the
operating cost of the induction motor is minimized. The operating cost of the induction
machine should be calculated over the whole life cycle of the machine. That calculation can
be made to evaluate the cost of the consumed electrical energy. The value of average energy
cost considering the power factor penalties can be determined by the following stages [A.
Hamid et al. 2006]:
1. If 0 ≤ PF < 0.7
⎡ ⎛ 0.9 − PF ⎞ 1 ⎤
C = C0 ⎢1 + ⎜ ⎟Χ ⎥ (56)
⎣ ⎝ 0.01 ⎠ 100 ⎦
2. If 0.7 ≤ PF ≤ 0.92, If PF ≥ 0.9, PF = 0.9
⎡ ⎛ 0.9 − PF ⎞ 0.5 ⎤ (57)

C = C0 ⎢1 + ⎜ ⎟Χ ⎥
⎣ ⎝ 0.01 ⎠ 100 ⎦
3. If 0.9 ≤ PF ≤ 1, If 0.95 ≤ PF ≤ 1, PF = 0.95
⎡ ⎛ 0.9 − PF ⎞ 0.7 ⎤ (58)

C = C0 ⎢1 + ⎜ ⎟Χ ⎥
⎣ ⎝ 0.01 ⎠ 100 ⎦
If the average energy cost C is calculated, it can be used to establish the present value of
losses. The total cost of the machine is the sum of its initial cost plus the present worth value
of losses and maintenance costs.
⎡1 ⎤ (59)
PWL = C ×T × N × Pout × ⎢ − 1⎥
⎣η ⎦
Where:
PWL = present worth value of losses
C0 = energy cost per kWh,
C = modified energy cost per kWh
T = running time per year (Hrs / year)
N = evaluation life (years)
Pout = the output power (kW)
η = the efficiency
The operating cost minimization of the induction motor problem can be formulated as
follows:
(
Minimize PWL TL , ω s , ωr ) (60)

The simulation is carried out on a three-phase, 380 V, 1-HP, 50 Hz, and 4-pole, squirrel cage
induction motor. The motor parameters are Rs=0.0598, Xls=0.0364, Xm=0.8564, Xlr=0.0546,
Rr=0.0403, Ke=0.0380, Kh=0.0380, Cstr =0.0150, Cfw=0.0093, S1=1.07, S2=-0.69, S3=0.77. For cost
analysis, the following values were assumed: C0=0.05, N=15, T=8000. The task of PSO
controller is to find that value of slip at which the maximum efficiency occurs. At certain
load torque and rotor speed, the PSO controller determines the slip frequency ωs at which
the maximum efficiency and minimum operating cost occur. The block diagram of the
optimization process based on PSO is shown in Fig.12. To observe the improvements in
efficiency using the suggested PSO controller, Fig. 13 shows the efficiency of the selected
machine for all operating conditions using conventional methods (constant voltage to
frequency ratio, field oriented control strategy) and using the proposed PSO controller at
different rotor speed levels, Wr = 0.2 PU, and Wr = 1 PU respectively [A. Hamid et al. 2006].
This figure shows that a considerable energy saving is achieved in comparison with the
conventional method (field oriented control strategy and constant voltage to frequency
ratio) especially at light loads and small rotor speed. Figure 14 compares the efficiency of
the induction motor drive system under the maximum efficiency strategy with the
minimum operating cost strategy at Wr = 1 PU. It is obvious from the figure that the
efficiency is almost the same for both strategies for all operating points.
Fig. 12. The proposed drive system based on PSO controller

Fig. 13. The efficiency of the induction motor using the maximum efficiency strategy
compared with the efficiency using the conventional methods at (a) Wr = 0.2 PU, (b) Wr= 1
PU [A. Hamid et al. 2006].
Fig. 14. The efficiency of the induction motor using the maximum efficiency strategy
compared with the efficiency using minimum operating cost strategy at Wr= 1 PU
Table 5 shows the efficiency comparison using few examples of operating points.
Figure 15 compares the power factor of the induction motor drive system under the
maximum efficiency strategy with the minimum operating cost strategy at Wr = 1 PU.
Finally, the proposed PSO-controller adaptively adjusts the slip frequency such that the
drive system is operated at the minimum loss and minimum operating cost. It was found
that the optimal system slip changes with variations in speed and load torque. When
comparing the proposed strategy with the conventional methods field oriented control
strategy and constant voltage to frequency ratio). It was found that a significant efficiency
improvement especially at light loads for all speeds. On the other hand, small efficiency
improvement is achieved at near rated loads (see Fig.13, and Fig.15).
Table 5. Some examples of efficiency comparison under different Load torque levels and Wr
= 1 PU [A. Hamid et al. 2006].
Fig. 15. The power factor of the induction motor using the maximum efficiency strategy
compared with the efficiency using minimum operating cost strategy at Wr= 1 PU
[A. Hamid et al. 2006]
4. Optimal electric drive system for fuel cell hybrid electric vehicles
Although there are various FC technologies available for use in vehicular systems, the
proton exchange membrane FC (PEMFC) has been found to be a prime candidate, since
PEMFC has higher power density and lower operating temperatures when compared to the
other types of FC systems. A stand-alone FC system integrated into an automotive
powertrain is not always sufficient to satisfy the load demands of a vehicle. Although FC
systems exhibit good power capability during steady-state operation, the response of fuel
cells during transient and instantaneous peak power demands is relatively poor. Thus, the
FC system can be hybridized with supercapacitors (SC) or batteries to meet the total power
demand of a hybrid electric vehicle (HEV) [Van Mierlo et. al, 2006; Paladini et. al, 2007].
In this section, a new control strategy based PSO algorithm is proposed for the Fuel
Cell/Supercapacitor hybrid electric vehicles to optimize the electric drive system [Hegazy &
Van Mierlo, 2010]. Many factors influence on the performance of the electric drive system.
These factors are mass, volume, size, efficiency, fuel consumption and control strategy.
Therefore, the PSO is proposed to minimize the cost, the size and the mass of the powertrain
sources (Fuel cell, and supercapacitor) as well as minimum fuel consumption and improves
the efficiency of the system. PSO algorithm searches for global optimization for nonlinear
problems with multi-objective. For a given driving cycle, the size and the cost of fuel cell
and supercapacitor are minimized by identifying the best number of units of each,
respectively. Three methods have been designed to achieve the optimal sizing. These are
conventional method, trial and error, as was mentioned in [Wu & Gao, 2006], GA, and PSO.
In addition, the hydrogen consumption is minimized by the evaluation of the optimal
power distribution between fuel cell (main source) and supercapacitor (auxiliary source).
Three control strategies are implemented to minimize the hydrogen consumption and
maintain the state of charge (SOC) of the supercapacitor ( SOCinitial =SOCfinal), which are
control strategy based on Efficiency Map (CSEM), Control strategy based on PSO (CSPSO),
and control strategy based on GA (CSGA).
4.1 System description

The power system configuration is illustrated in Fig.16. A hybrid fuel cell/supercapacitor
vehicle utilizes a PEM fuel cell as the main power source and a supercapacitor as the
auxiliary power source. A multiple-input power electronic converter (MIPEC) is proposed
to interface the traction drive requirements. In the MIPEC, the FC is connected to DC Bus
Fig. 16. The drive system of the Fuel Cell/Supercapacitor Hybrid Electric Vehicle
via a Boost DC/DC converter (ηB= ηconv) and the supercapacitor is connected to DC Bus via
a Buck/Boost converter (ηB/B= ηconv). The desired value of the DC-Bus voltage is chosen to
be 400 V with variations of ± 10% are permissible. The power supplied by the powertrain
has to be obtaining from the power demand predicted by the dynamics of the vehicle. The
efficiency of each component in the hybrid powertrain is taken into account. A detailed
model of the powertrain is built in MATLAB /SIMULINK.
4.1.1 Modeling of the vehicle power demand

The load force of the vehicle consists of gravitational force, rolling resistance, aerodynamic
drag force, and acceleration force. Hereby, the load power required for vehicle acceleration
can be written as [Hegazy & Van Mierlo, 2010; Hegazy et. al 2010]
(F + F +F +F ) *V
g roll AD acc
P = (61)
load η
GB
F = M . g . sin(α ) (62)
g
F = M . g . f . cos(α ) (63)
roll r
2
F = 0.5 ρ .C . A .V (64)
AD a D F
dV
F = M. (65)
acc dt
V =ω .r (66)
w w
The total electric power required from sources can be expressed as:
P
= load
P
req η .η
(67)
.η
m Inv Conv
The parameters of the vehicle are given in Table 6. The analysis of FCHEV is performed
with two standard driving cycles:
1. The Federal Test Procedure (FTP75) Urban;
2. The New European Driving Cycle (NEDC)
Suppose that the efficiencies of the motor (ηm), inverter (ηInv), and MIPEC (ηConv= ηB= η B/B)
are 0.90, 0.94 and 0.95, respectively.
M Vehicle mass (kg) 1450 Af Front Area (m2) 2.13

Radius of the
fr Rolling Resistance Coefficient 0.013 rω 0.28
wheel (m)
Aerodynamic Drag Air density
CD 0.29 ρa 1.202
Coefficient (CD) (kg/m3)
Table 6. Vehicle Parameters [Wu & Gao, 2006]
4.2 Optimal powertrain design

The first goal of optimization algorithm, PSO, is to minimize the cost, the mass, and the
volume of the fuel cell (FC) and supercapacitor (SC). It is assumed that, the cost, the mass
and the volume of the fuel cell and supercapacitor are a function of the number of the
parallel units Nfcp and Nscp, respectively. The multi-objective criterion should be aggregated
in a single objective function if the design objective is to embody a unique solution. The
objective function can be formulated as follows:
F ( x) = w1 cos t + w2 mass + w3volume

(68)
cos t = C1. Nfcs. Nfcp + C 2. Nscs. Nscp (69)
The coefficients of the terms in F(x) were chosen to reflect the importance of minimizing the
cost, the mass and the volume. Suppose that w1, w2, and w3 are 0.35, 0.35, and 0.3,
respectively. Figure 17 presents the flowchart of the execution of PSO, which evaluates the
optimal number of the FC units and the supercapacitor units by using MATLAB
/SIMULINK. The layout of the fuel-cell stack and layout of the supercapacitor system are
shown in Fig.18 (a) and (b), respectively. The constraints of the optimization problems are
mentioned in [Hegazy & Van Mierlo, 2010].
Fig. 17. The flowchart of the execution of PSO [Hegazy et. al 2010]
(a)
(b)
Fig. 18. (a) Layout of the FC; (b) Layout of the SC

Based on minimizing the objective function F(x) in (68), the results of the optimal design and
components sizing of the FC/SC powertrain are shown in Fig.19. The analyses and
parameters of the FC and the SC are mentioned in [Hegazy & Van Mierlo, 2010].
(a) The optimal numbers of cells of FC and SC
(b) The cost of the FC/SC components

(c) The mass of the FC/SC components
(d) The mass of the FC/SC components
Fig. 19. The Comparative of the optimal design between different methods for FC/SC HEV
4.3 Optimal Power Control (OPC)

The second goal of the PSO is to minimize the vehicle fuel, hydrogen, consumption while
maintaining the supercapacitor state of charge. As a hybrid powertrain is under
consideration, a power management strategy is required to define what both the FC and SC
powers are. The global optimization algorithms, such as GA and dynamic programming
(DP), achieve an optimal power control for FC/SC hybrid electric vehicle, which leads to the
lowest hydrogen consumption and maintains the supercapacitor SOC [Sinoquet et. al 2009;
Sundstrom & Stefanopoulou 2006].
In this study, the optimal power control can be achieved by using PSO and GA for a given
driving cycle. Suppose that the degree of hybridization of the fuel cell is Kfc at time t and
Ksoc, Proportional controller gain, which used to adapt the SOC during charging from the
FC. A balance equation can naturally be established, since the sum of power from both
sources has to be equal to the required power at all times:
Preq (t ) = Pfc(t ) + Psc(t ) (70)
Kfc(t ) =
Pfc(t ) (71)
Preq (t )
The net energy consumed from the FC at time t can be computed as follows:
t Pfc (t )
Efc(t ) = ∫ dt
(72)
η ( Pfc (t ))
0
The cost function can be expressed as follows:
N PfcOpti (k )
1 (73)
F2 ( x) = ∑ ΔT
Elow K = 0 η ( PfcOpti (k ))
The Optimal fuel cell power output, PfcOpti, is calculated based on the SOC of the
supercapacitor and power demand, Preq, as follows:
⎡ SOC ref − SOC (k ) ⎤ (74)

PfcOpti (k ) = Kfc (k ) Preq (k ) + Ksoc (k ) ( Pfc max − Pfcmin ) ⎢ ⎥
⎢ ( SOC max − SOC min ) / 2 ⎥
⎣ ⎦
Fig. 20. The block diagram of the Optimal power Control
Where: N= T/ΔT is number of samples during the driving cycle, and ΔT=1s is the sampling
time.
The block diagram of the optimal power control based on optimization algorithm is shown
in Fig.20.
Based on minimizing the objective function F2(x) in (73), the results of the optimal power
sharing based PSO and the comparative study for the FC/SC powertrain are summarized in
Fig.21 [Hegazy et. al 2010].
(a) The power sharing between FC and SC on NEDC driving cycle
(b) The power sharing between FC and SC on FTP75 driving cycle

(c) The Comparative of the hydrogen consumption between control strategies
(d) The Hydrogen improvements with respect to pure fuel cell without SC
Fig. 21. The results of the optimal power Control for FC/SC
5. Conclusion
This chapter deals with the applicability of swarm intelligence (SI) in the form of particles
swarm optimization (PSO) used to achieve the best performance for the electric machines
and electric drives. In addition, by analyzing and comparing the results, it is shown that
control strategy based on PSO is more efficient than others control strategies to achieve the
optimal performance for fuel cell/supercapacitor hybrid electric vehicles (FCHEV).
It is very important to note that, these applications were achieved without any additional
hardware cost, because the PSO is a software scheme. Consequently, PSO has positive
promises for a wide range of variable speed drive and hybrid electric vehicles applications.
6. Index I
List of principal symbols
ωe : synchronous speed
ωr : rotor speed
p : differential operator
rm , ra : main, auxiliary stator windings resistance
rr : rotor winding resistance
Rfeq,d : equivalent iron-loss resistance(d and q axis)
Llm ,Lla : main, auxiliary stator leakage inductance
Lmd ,Lm q : magnetizing inductance (d& q axis)
Llr : rotor leakage inductance
K : turns ratio auxiliary/main windings
Te : electromagnetic torque
J : inertia of motor
λds,qs : stator flux(d and q axis)
λdr,qr : rotor flux(d and q axis)
Vds,qs : stator voltage (d and q axis)
ids,qs : stator current (d and q axis)
M : mutual inductance
7. References
Amin. A. M. A., Korfally. M. I., Sayed. A. A. and Hegazy. O.T. M., (2009), Efficiency
Optimization of Two Asymmetrical Windings Induction Motor Based on Swarm
Intelligence, IEEE Transactions on Energy Conversion, Vol. 24, No. 1, March
2009
Amin. A. M. A., Korfally. M. I., Sayed. A. A. and Hegazy. O.T. M., (2006), Losses
Minimization of Two Asymmetrical Windings Induction Motor Based on Swarm
Intelligence, Proceedings of IEEE- IECON 06 , pp 1150 – 1155, Paris , France , Nov.
2006 .
Amin. A. M. A., Korfally. M. I., Sayed. A. A. and Hegazy. O.T. M., (2007), Swarm
Intelligence-Based Controller of Two-Asymmetrical Windings Induction Motor,
accepted for IEEE. EMDC07, pp 953 –958, Turkey, May 2007.
Eberhart. R, Kennedy. J, (1995), A New Optimizer Using Particles Swarm Theory, Proc.
Sixth International Symposium on Micro Machine and Human Science (Nagoya,

Japan), IEEE Service Center, Piscataway, NJ, pp. 39-43,
A. Hamid Radwan H., Amin Amr. M. A., Ahmed Refaat S., and El-Gammal Adel A. A.
,(2006), New Technique For Maximum Efficiency And Minimum Operating Cost
Of Induction Motors Based On Particle Swarm Optimization (PSO)”, Proceedings
of IEEE- IECON 06 , pp 1029 – 1034, Paris , France , Nov. 2006.
Hegazy Omar, (2006), Losses Minimization of Two Asymmetrical Windings Induction
Motor Based on Swarm Intelligence, M.Sc., Helwan University, 2006.
Hegazy Omar, and Van Mierlo Joeri, (2010), Particle Swarm Optimization for Optimal
Powertrain Component Sizing and Design of Fuel cell Hybrid Electric Vehicle, 12th
International Conference on Optimization of Electrical and Electronic Equipment,
IEEE OPTIM 2010
Hegazy Omar, Van Mierlo Joeri, Verbrugge Bavo and Ellabban Omar, (2010), Optimal
Power Sharing and Design Optimization for Fuel Cell/Battery Hybrid Electric
Vehicles Based on Swarm Intelligence, The 25th World Battery, Hybrid and Fuel
Cell Electric Vehicle Symposium & Exhibition © EVS-25 Shenzhen, China, Nov. 5-9,
2010.
Kennedy. J and Eberhart .R, (2001), Swarm Intelligence, Morgan Kaufmann Publishers, Inc.,
San Francisco, CA
Kioskeridis, I; Margaris, N., (1996), Losses minimization in scalar-controlled induction
motor drives with search controllers" Power Electronics, IEEE Transactions,
Volume: 11, Issue: 2, March 1996 Pages: 213 – 220
Popescu. M, Navrapescu. V, (2000) ,A method of Iron Loss and Magnetizing Flux
Saturation Modeling in Stationary Frame Reference of Single and Two –Phase
Induction Machines”, IEE 2000, Conf. power Elec. & Variable Speed Drives, 140-
146
Sundstrom Olle and Stefanopoulou Anna, (2006), Optimal Power Split in Fuel Cell Hybrid
Electric Vehicle with different Battery Sizes, Drive Cycles, and Objectives,
Proceedings of the 2006 IEEE International Conference on Control Applications
Munich, Germany, October 4-6, 2006.
Van Mierlo Joeri, Cheng Yonghua, Timmermans Jean-Marc and Van den Bossche Peter,
(2006), Comparison of Fuel Cell Hybrid Propulsion Topologies with Super-
Capacitor, IEEE, EPE-PEMC 2006, Portorož, Slovenia
Wu Ying, Gao Hongwei, (2006) ,Optimization of Fuel Cell and Supercapacitor for Fuel-Cell
Electric Vehicles, IEEE Transactions On Vehicular Technology, Vol. 55, No. 6,
November 2006.
10
Operation of Active Front-End Rectifier in

Electric Drive under Unbalanced Voltage Supply
Miroslav Chomat
Institute of Thermomechanics AS CR, v.v.i.
Czech Republic
1. Introduction
Non-standard conditions in the power network such as voltage unbalance can negatively
affect operation of electric drives. The unbalance can be caused by a failure in the network
or by an unbalanced load in the electric vicinity of the affected drive. Unsymmetrical
voltages at the input of a voltage source inverter cause pulsations in the DC link voltage
when not properly taken care of. This may result in significantly reduced power capabilities
and, therefore, limited controllability of the drive. This text deals with the effects of
unbalanced voltage supply on the DC-link voltage pulsations, methods to address this
problem and the additionally imposed constraints in operating regions of the rectifier.
2. Control method
A simplified scheme of the drive under investigation is shown in Fig. 1. The front-end
controlled rectifier is connected to the mains through input filter inductors. The output
current of the rectifier supplies the DC current to the output inverter and maintains the
voltage across the DC-link capacitors constant at the same time. The value of this current
can be controlled by suitable switching of solid-state elements in the front-end stage.
i DC
power input front end DC bus inverter electric
network impedance machine
VA iA
L R
CDC VDC
iB
VB L R
N 0
M
iC
VN VA L R V0
CDC VDC
Fig. 1. Scheme of system under investigation.

Suitable control of the front-end AC/DC converter can be employed in order to draw
constant input power from the power network even at unbalanced voltage supply
(Stankovic & Lipo, 2001; Lee et al., 2006; Cross et al., 1999; Song & Nam, 1999). The
switching functions for the front-end AC/DC converter are generated so that a constant
voltage across the DC bus is maintained. Series combinations of inductance and resistance
are considered at the input terminals of the inverter.
The system can be electrically described by the following set of ordinary differential
equations (Chomat & Schreier, 2005):
diA
vA − L − RiA − vSA + vN − v0 = 0 , (1)
dt
diB
vB − L − RiB − vSB + vN − v0 = 0 , (2)
dt
diC
vC − L − RiC − vSC + vN − v0 = 0 , (3)
dt
where
vSA = s A ⋅ VDC , (4)
vSB = sB ⋅ VDC , (5)
vSC = sC ⋅ VDC (6)
are the voltages at the input of the inverter. The functions sA, sB, and sC are the
corresponding unit switching functions of the particular phases of the front-end stage,
which represent the fundamental harmonic components of the pulse width modulated
output. Sinusoidal switching functions with the nominal frequency are considered
throughout this paper, whereas the higher harmonics that would arise in a real power
converter are neglected in the calculation for simplification. VDC represents one half of the
overall DC-link voltage here. The voltage vN is the electric potential of the neutral of the
mains and v0 is the electric potential of the centre point of the capacitor bank in the DC bus.
The DC-link current can be calculated from the phase currents and the switching functions
according to
1
iDC = ( sAi A + sBiB + sC iC ) . (7)
2
The coefficient ½ takes into account the fact that currents in both positive and negative
directions that flow through different current paths in the DC bus are produced by the
rectifier.
An unbalanced system of phase quantities can advantageously be represented by phasors of
positive and negative rotating sequences. It is not necessary to take zero-sequence quantities
into account here as no neutral wire is considered in the system and, therefore, no zero-
sequence current can develop. The resulting rotating vector of such a quantity may then be
written as
x = X P e jωt + X N e − jωt . (8)

Operation of Active Front-End Rectifier in Electric Drive under Unbalanced Voltage Supply 197
The subscripts P and N denote the positive and negative rotating sequences, respectively.
Based on these assumptions, (1) - (3) and (4) – (6) may be rewritten in phasor form as
VP − ( R + jωL ) I P − SPVDC = 0 , (9)
VN − ( R − jωL ) IN − SNVDC = 0 . (10)
The solution of (9) and (10) for positive and negative sequence currents is
VP − SPVDC
IP = , (11)
R + jωL
VN − SNVDC
IN = (12)
R − jωL
and the corresponding rotating vector of the input currents is therefore
i = IP e jωt + IN e − jωt . (13)
Similarly, we can formally introduce a rotating vector of the switching functions
s = SP e jωt + SN e − jωt . (14)
Then the resulting instantaneous value of the current supplied into the DC link by the front-
end converter from (7) can be written in the vector form as
13
iDC = Re {i ⋅ s} , (15)
22
where the bar over the symbol denotes the complex conjugate value. The term 3/2 appears
in (15) due to the transformation from rotating vector form to instantaneous quantities.
The resulting relation obtained after substituting (13) and (14) into (15) can be written as the
sum of two separate current components and given as
iDC = iDC ( avg ) + iDC ( 2ωt ) , (16)
where
3 ⎧⎪ VP − SPVDC V − SNVDC ⎪⎫
iDC ( avg ) = Re ⎨ SP + N SN ⎬ , (17)
4 ⎩⎪ R + jωL R − jωL ⎭⎪
3 ⎧⎪ VP − SPVDC V − SNVDC ⎪⎫
iDC ( 2ωt ) = Re ⎨ SN e j 2ωt + N SP e − j 2 ω t ⎬ . (18)
4 ⎪⎩ R + jωL R − jωL ⎭⎪
The first component, (17), represents a DC component and the second, (18), represents a
pulsating component with the frequency twice as high as that of the mains. The pulsating
component is only produced when the negative sequence of either the input voltages or the
switching functions is present.
From (18), a condition for the elimination of the pulsating component in the DC link can be
derived
⎪⎧ V − SPVDC ⎪⎫ ⎪⎧ V − SNVDC ⎪⎫
Re ⎨ P SN e j 2ωt ⎬ = − Re ⎨ N SP e − j 2 ω t ⎬ . (19)
⎪⎩ R + jωL ⎪⎭ ⎪⎩ R − jωL ⎭⎪
As the real part of a complex number equals the real part of its conjugate value, (18) can also
be written as
⎧⎪ V − SPVDC ⎫⎪ ⎧⎪ V − SNVDC ⎫⎪
Re ⎨ P SN e j 2ωt ⎬ = − Re ⎨ N SP e j 2 ω t ⎬ . (20)
⎪⎩ R + jωL ⎭⎪ ⎩⎪ R + jωL ⎭⎪
For the above equation to be satisfied at any time, the following must hold providing that
there is non-zero input impedance
( VP − SPVDC ) SN = − ( VN − SNVDC ) SP . (21)
If the input voltages are known and the control is free to choose the positive sequence
component of the switching functions, the negative sequence of the switching functions
obtained from (21) is
SP VN
SN = . (22)
2SPVDC − VP
It should be noted that the relation does not contain values of input resistance and
inductance and is, therefore, the same for pure inductance as well as for pure resistance
connected to the front end of the inverter.
As the amplitudes of the individual switching functions need to be less than or equal
to one, the range of practical combinations of SP and SN is constrained. A simple, and
rather conservative, condition to keep the switching functions in allowable limits can be
written as
SP + S N ≤ 1 . (23)
For more precise evaluation of the constraints, we need to evaluate magnitudes of switching
vectors in individual phases
S A = SP + SN , (24)
S B = SP a + SN a 2 , (25)
SC = SP a 2 + SN a (26)
and limit the magnitude of each of them
( SA ≤ 1 ) ∧ ( SB ≤ 1 ) ∧ ( SC ≤ 1 ) . (27)
For its operation, the above discussed control method requires to monitor the instantaneous
values of the input phase voltages and of the DC-link voltage. Based on this information, a
convenient combination of values of SP and SN can be chosen to produce the required value
of the DC-link current and to satisfy the conditions in (22) and (23) at the same time. From
SP and SN, the switching functions for the individual legs of the rectifier are computed and
switching pulses are generated for individual switching devices based on a particular pulse
width modulation algorithm. The switching devices are considered to be transistors or
thyristors with forced commutation.
3. Operation of drive under unbalanced voltage supply

3.1 Numerical simulation of drive under unbalanced voltage supply
Operation of the described system has been numerically simulated under various types of
unbalanced voltage supply in order to investigate the effect of the unbalance on the system
behavior and the influence of certain circuit parameters. The reference parameters of the
input impedance were chosen to be R = 0.1 Ω and L = 10 mH. The input phase voltages had
nominal voltage amplitudes of 230 V, nominal frequency of 50 Hz, and mutual phase shifts
of 120° to form a three-phase voltage system in the case of the symmetrical system. The
DC-link voltage was set to 560 V and the capacitor of 1000 µF was used in the DC bus
(Chomat et al., 2007).
First, operation of the investigated system under symmetrical voltage supply was simulated
to obtain the reference case to compare with unbalanced operation. Figure 2 shows input
phase voltages and currents and Figure 3 shows the DC-link current and voltage. It can be
seen that both electrical quantities in the DC bus are smooth with no visible pulsations.
Fig. 2. Phase voltages and currents under symmetrical voltage supply.

Second, the unbalance caused by setting the magnitude of the voltage in phase A to
200 VRMS was investigated. Figures 4 and 5 show the corresponding quantities at the input of
the rectifier and in the DC bus when no measures are taken to eliminate the pulsations by
suitable modification of switching in the active front-end rectifier. The DC-link current and
voltage contain significant pulsations that would make control of the drive more
complicated. When the switching functions are modified in order to eliminate the effect of
the supply voltage unbalance, the pulsations are nearly entirely eliminated, Figures 6 and 7.
This has also an effect on the input phase currents compared to the previous case.
40
30
iDC [A]
20
10
580
570
vDC [V]
560
550
540
0 0.005 0.01 0.015 0.02
time [s]
Fig 3. DC-link voltage and current under symmetrical voltage supply.
Fig 4. Phase voltages and currents under unbalanced voltage supply without compensation.
40
30
iDC [A]
20
10
580
570
vDC [V]
560
550
540
0 0.005 0.01 0.015 0.02
time [s]
Fig. 5. DC-link voltage and current under unbalanced voltage supply without
compensation.
Fig. 6. Phase voltages and currents under unbalanced voltage supply with compensation.
40
30
[A]
20
DC
i
10
580
570
[V]
DC
560
v
550
540
0 0.005 0.01 0.015 0.02
time [s]
Fig. 7. DC-link voltage and current under unbalanced voltage supply with compensation.
When DC-link capacitor is reduced to have the capacity of only 500 µF instead of 1000 µF,
Figures 8 and 9, the DC-link voltage pulsations are increased twice as could be expected in
case with no measures to eliminate the effect of the unbalance in the front-end rectifier. No
change appears in the case when switching functions are modified to eliminate the effect of
the unbalance, Figures 10 and 11.
Fig. 8. Phase voltages and currents under unbalanced voltage supply with reduced DC-link
capacitance without compensation.
40
30
iDC [A]
20
10
580
vDC [V] 570
560
550
540
0 0.005 0.01 0.015 0.02
time [s]
Fig. 9. DC-link voltage and current under unbalanced voltage supply with reduced DC-link
capacitance with compensation.
40
30
[A]
20
DC
i
10
580
570
[V]
DC
560
v
550
540
0 0.005 0.01 0.015 0.02
time [s]
Fig. 11. DC-link voltage and current under unbalanced voltage supply with reduced DC-
link capacitance with compensation.
The change of the input inductance from 10 mH to 5 mH leads to an adequate increase in
the input phase currents as well as in the DC-link current, Figures 12 to 15. The relative
amount of pulsations remain at about the same levels.
Fig. 12. Phase voltages and currents under unbalanced voltage supply with reduced input
inductance without compensation.
40
30
iDC [A]
20
10
580
570
vDC [V]
560
550
540
0 0.005 0.01 0.015 0.02
time [s]
Fig. 13. DC-link voltage and current under unbalanced voltage supply with reduced input
inductance with compensation.
40
30
iDC [A]
20
10
580
570
vDC [V]
560
550
540
0 0.005 0.01 0.015 0.02
time [s]
Finally, the unbalance caused by shifting the voltage phasor of phase A by 10° was
investigated. Corresponding results due to the changes in circuit parameters are illustrated
in Figures 16 to 27. It can be noted that the effects are similar to the previous case of
unbalance.
Fig. 16. Phase voltages and currents under unbalanced voltage supply without
compensation.
40
30
iDC [A]
20
10
580
570
vDC [V]
560
550
540
0 0.005 0.01 0.015 0.02
time [s]
Fig. 17. DC-link voltage and current under unbalanced voltage supply without
compensation.
Fig. 18. Phase voltages and currents under unbalanced voltage supply with compensation.
40
30
iDC [A]
20
10
580
570
vDC [V]
560
550
540
0 0.005 0.01 0.015 0.02
time [s]
Fig. 19. DC-link voltage and current under unbalanced voltage supply with compensation.
The effect of reduction of the DC-link capacitor from 1000 µF to 500 µF is shown in Figures
20 through 23.
40
30
iDC [A]
20
10
580
570
vDC [V]
560
550
540
0 0.005 0.01 0.015 0.02
time [s]
link capacitance without compensation.
capacitance with compensation.
40
30
iDC [A]
20
10
580
570
vDC [V]
560
550
540
0 0.005 0.01 0.015 0.02
time [s]
link capacitance with compensation.
The corresponding situation for reduced input inductance from 10 mH to 5 mH is illustrated
in Figures 24 to 27.
40
30
iDC [A]
20
10
580
570
vDC [V]
560
550
540
0 0.005 0.01 0.015 0.02
time [s]
40
30
iDC [A]
20
10
580
570
vDC [V]
560
550
540
0 0.005 0.01 0.015 0.02
time [s]
3.2 Limitation of control range due to unbalanced voltage supply

The necessity to generate the negative sequence component of the switching functions in
order to eliminate the effect of the supply-voltage unbalance on the DC-link voltage
pulsations reduces the control range for the positive sequence component of the switching
functions (Chomat et al., 2009). This is due to the fact that in individual phases the
maximum of the switching function can only reach one at most at any given time. Another
constraint results from the current rating of the converter. The resulting constraints depend
on the value and type of the unbalance.
Analysis of the limitation corresponding to various types of unbalanced supply voltages has
been carried out. The reference parameters of the input impedance were chosen to be
R = 0.1 Ω and L = 10 mH. The input phase voltages had nominal voltage amplitudes of
230 V, nominal frequency of 50 Hz, and mutual phase shifts of 120° to form a three-phase
voltage system in the case of the symmetrical system. The DC-link voltage was set to 400 V.
The choice of the positive sequence component of the switching functions from the available
control range affects both the magnitude of the DC-link current and the currents in
individual input phases. Figure 28 shows what magnitudes of the DC-link current
correspond to the coordinates from the available control range. The unbalance was formed
Fig. 28. DC-link current under unbalanced voltage supply (L = 10 mH, R = 0.1 Ω,
Vdc = 400 V).
by setting the magnitude of the voltage in phase A to 0.75 p.u. The corresponding maximal
input phase current magnitude, calculated as the maximum of all the phase currents, is
shown in Figure 29. It can be seen from Figure 28 that the resulting DC-link current
decreases in the vertical direction of the operating region, whereas the maximal input
current in Figure 29 decreases in the horizontal direction. The corresponding measure of the
current unbalance is depicted in Figure 30 and the average power factor of all the three
input phases is depicted in Figure 31.
Fig. 29. Maximal input phase current under unbalanced voltage supply (L = 10 mH,
R = 0.1 Ω, Vdc = 400 V).
Fig. 30. Input current unbalance under unbalanced voltage supply (L = 10 mH, R = 0.1 Ω,
Vdc = 400 V).
Fig. 31. Power factor under unbalanced voltage supply (L = 10 mH, R = 0.1 Ω, Vdc = 400 V).
If we change the value of the input inductance from 10 mH to 1 mH, the constraints caused
by the switching functions remain the same as can be seen from Figures 32 through 35.
However, both the DC-link current and the input current increased nearly ten times as the
input reactance represents the main limiting factor for the currents entering the rectifier. The
excessive values of the currents would, in a case of a real rectifier, impose additional
restrictions to the operating regions resulting from current stress of electronic components
in the bridge. This can also be considered in the shape of new borders of operating regions.
Vdc = 400 V).
Fig. 33. Maximal input phase current under unbalanced voltage supply (L = 1 mH, R = 0.1 Ω,
Vdc = 400 V).
Vdc = 400 V).
A different situation arises when the input resistance is increased ten times to 1 Ω. The
corresponding electrical quantities are shown in Figures 36 through 39. The increase in the
DC-link and input phase currents is not as dramatic as the resistance plays less significant
role in limiting the currents than the inductance. The values of the currents are similar to the
ones in the first case.
Fig. 36. DC-link current under unbalanced voltage supply (L = 1 mH, R = 1 Ω, Vdc = 400 V).
Fig. 37. Maximal input phase current under unbalanced voltage supply (L = 1 mH, R = 1 Ω,
Vdc = 400 V).
Fig. 38. Input current unbalance under unbalanced voltage supply (L = 1 mH, R = 1 Ω,
Vdc = 400 V).
Fig. 39. Power factor under unbalanced voltage supply (L = 1 mH, R = 1 Ω, Vdc = 400 V).
A change in the DC-link voltage introduces, on the other hand, a noticeable change in the
shape of constraints caused by the limitation of the switching functions. Figures 40 through
43 show the situation for the decrease in the DC-link voltage from 400 V to 200 V and
Figures 45 through 47 show the situation for the increase to 600 V. In the latter case, a rise of
an isolated restricted area in the right hand side of the figure completely surrounded by
available control space can be noticed.
Vdc = 200 V).
R = 0.1 Ω, Vdc = 200 V).
Vdc = 200 V).
Fig. 43. Power factor under unbalanced voltage supply (L = 10 mH, R = 0.1 Ω,
Vdc = 200 V).
Vdc = 600 V).
R = 0.1 Ω, Vdc = 600 V).
Vdc = 600 V).
Measurements on an experimental system identical to the simulated one have been carried
out in order to verify the investigated method. The scope traces in Figure 48 show the
measured current in phase A and the DC link current when the negative-sequence in the
supply voltage is not compensated for by the control method and the DC link current,
therefore, contains significant component pulsating with a frequency of 100 Hz, twice the
fundamental network frequency. The case when unbalanced voltage system is compensated
by the investigated control method is illustrated in Figure 49. It can be seen that the
pulsating component of the DC link current has been effectively eliminated by the
investigated method.
Fig. 48. Phase A current and DC-link current under unbalanced voltage supply without
elimination of pulsating component.
Fig. 49. Phase A current and DC-link current under unbalanced voltage supply with
elimination of pulsating component.
7. Conclusion
It has been shown in the article that it is possible to effectively compensate for the unbalanced
voltage source at the input of a solid-state converter so that constant power flow into the DC
bus is maintained. The results of simulations show that the choice of the operating point of
front end converter may significantly affect the impact of the rectifier on the supplying power
grid. It is possible to select the optimal operating point according to the chosen optimization
criteria, which can be e.g. maximal power factor or current unbalance.
8. Acknowledgment
This work was supported by the Grant Agency of the Czech Republic under research grant
No. 102/09/1273 and by the Institutional Research Plan AV0Z20570509.
9. References
Stankovic, A. V. & Lipo, T. A. (2001). A Novel Control Method for Input Output Harmonic
Elimination of the PWM Boost Type Rectifier Under Unbalanced Operating
Conditions, IEEE Trans. on Power Electronics, 16, pp. 603-611, ISSN: 0885-8993.
Stankovic, A. V. & Lipo, T. A. (2001). A Generalized Control Method for Input-Output
Harmonic Elimination of the PWM Boost Type Rectifier Under Simultaneous
Unbalanced Input Voltages and Input Impedances, Power Electronics Specialists
Conference, pp. 1309-1314, ISBN: 0-7803-7067-8, Vancouver, Canada, June 2001.
Lee, K.; Jahns, T. M.; Berkopec, W. E. & Lipo, T. A. (2006). Closed-form analysis of
adjustable-speed drive performance under input-voltage unbalance and sag
conditions, IEEE Trans. on Industry Applications, vol. 42, no. 3., pp. 733-741,
ISSN: 0093-9994.
Cross, A. M.; Evans, P. D. & Forsyth, A. J. (1999). DC Link Current in PWM Inverters with
Unbalanced and Non Linear Loads, IEE Proc.-Electr. Power Appl., vol. 146, no. 6,
pp. 620-626, ISSN: 1350-2352.
Song, H. & Nam, K. (1999). Dual Current Control Scheme for PWM Converter Under
Unbalanced Input Voltage Conditions, IEEE Trans. on Industrial Electronics, 46,
pp. 953-959, ISSN: 0278-0046.
Chomat, M. & Schreier, L. (2005). Control Method for DC-Link Voltage Ripple Cancellation
in Voltage Source Inverter under Unbalanced Three-Phase Voltage Supply, IEE
Proceedings on Electric Power Applications, vol. 152, no. 3, pp. 494 – 500,
ISSN: 1350-2352.
Chomat, M.; Schreier, L. & Bendl, J. (2007). Operation of Adjustable Speed Drives under
Non Standard Supply Conditions, IEEE Industry Applications Conference/42th
IAS Annual Meeting, pp. 262-267, ISBN: 978-1-4244-1259-4, New Orleans, USA,
September 2007.
Chomat, M.; Schreier, L. & Bendl, J. (2009). Influence of Circuit Parameters on Operating
Regions of PWM Rectifier Under Unbalanced Voltage Supply, IEEE International
Electric Machines and Drives Conference, pp. 357-362, ISBN: 978-1-4244-4251-5,
Miami, USA, May 2009.
Chomat, M.; Schreier, L. & Bendl, J. (2009). Operating Regions of PWM Rectifier under
Unbalanced Voltage Supply, International Conference on Industrial Technology,
pp. 510 – 515, ISBN: 978-1-4244-3506-7, Gippsland, Australia, February 2009.
11
Space Vector PWM-DTC Strategy for

Single-Phase Induction Motor Control
Ademir Nied1, José de Oliveira1, Rafael de Farias Campos1,
Seleme Isaac Seleme Jr.2 and Luiz Carlos de Souza Marques3
1StateUniversity of Santa Catarina
2FederalUniversity of Minas Gerais
3Federal University of Santa Maria
Brazil
1. Introduction
Single-phase induction motors are widely used in fractional and sub-fractional horsepower
applications, mostly in domestic and commercial applications such as fans, refrigerators, air
conditioners, etc., operating at constant speed or controlled by an on/off strategy which can
result in poor efficiency and low-power factor. In terms of construction, these types of motors
usually have a main and an auxiliary stator winding, are asymmetrical and are placed 90
degrees apart from each other. The rotor is usually the squirrel-cage type. The asymmetry
presented in the stator windings is due to the fact that these windings are designed to be
electrically different so the difference between the stator windings currents can produce a
starting torque (Krause et al., 1995). Since it has main and auxiliary stator windings, the single-
phase induction motor is also known as a two-phase asymmetric induction motor.
In recent years, with the growing concern about low-cost operation and the efficient use of
energy, the advance in motor drive control technology made it possible to apply these
motors to residential applications with more efficiency. Different inverter topologies have
been proposed to drive single-phase induction motors, providing ways to save energy. In
dos Santos et al. (2010) different ac drive systems are conceived for multiple single-phase
motor drives with a single dc-link voltage to guarantee installation cost reduction and some
individual motor controls. In Wekhande et al. (1999) and Jabbar et al. (2004), Campos et al.
(2007a) and Campos et al. (2007b), two topologies are considered. One is a Half-bridge
inverter and the other is a three-leg inverter. The cost difference between the two topologies
lays in the fact that the H-bridge inverter needs two large capacitors in the dc link rated for
dc link voltage. Also, there is a need of two large resistors connected in parallel with the
capacitors to balance the voltage of the capacitors.
Despite the fact that the three-leg inverter has more switches, the development of power
modules and the need for just one capacitor in the dc link have decreased the topology cost.
Along with the reduced cost, a more efficient use of the dc link voltage is achieved.
Besides the effort for developing more efficient driving topologies, many strategies to
control single-phase motors have been proposed. In Jacobina et al. (1999), rotor-flux control,
stator-flux control and direct torque control (DTC) (Takahashi and Noguchi, 1986) are
analyzed. The main drawback of the two first strategies is that they use an encoder to obtain
the speed signal. Since there is no need for speed and position signals, a DTC scheme
appears to be a suitable solution. But it has some disadvantages such as current and torque
distortions, variable switching frequency and low-speed operation problems (Buja and
Kazmierkowski, 2004). In Neves et al. (2002), a DTC strategy is applied for a single-phase
motor and the performance is improved with the use of pulse width modulation.
Along with control strategies and driver topologies, many researchers have investigated
ways to optimize modulation techniques applied in single-phase induction motor drives. In
Jabbar et al. (2004), space-vector modulation (SVPWM) is used to reduce the torque ripple
and alleviate the harmonic content at the terminals of the single-phase induction motor
being driving by a three-leg inverter. In Chaumit and Kinnares (2009) the proposed SVPWM
method controls the two-phase voltage outputs of an unbalanced two-phase induction
motor drive by varying the modulation index and voltage factors.
In this chapter, the authors are interested in studying the DTC strategy combined with the
SVPWM applied to a three-leg inverter topology to drive a single-phase induction motor.
2. Single-phase induction motor model

A single-phase induction motor with main and auxiliary windings is designed to be
electrically different. In order to make the motor self-starting, a capacitor is connected in
series with the auxiliary winding.
When the windings of a single-phase induction motor are fed independently (i.e., using a
voltage source inverter) one can consider a single-phase induction motor an example of an
unsymmetrical two-phase induction motor.
In this section, the mathematical model of a single-phase induction motor will be derived.
As is commonly done, the derivation of the motor model is based on classical assumptions:
• The stator and rotor windings are in space quadrature;
• The rotor windings are symmetrical;
• The magnetic circuit is linear and the air-gap length is constant;
• A sinusoidal magnetic field distribution produced by the motor windings appears in
the air gap;
• The motor is a squirrel-cage type. Therefore the rotor voltages are zero.
Since the single-phase induction motor will be considered as acting as a two-phase system,
to derive the dynamic motor model of the two-phase system, a common reference frame (a-
b) will be used, as shown in Fig. 1.
Fig. 1. Common reference frame (a-b).

Space Vector PWM-DTC Strategy for Single-Phase Induction Motor Control 219
Since the stator windings are in space quadrature, there is no magnetic coupling between
them. The same consideration is applied to the rotor windings. According to Krause et al.
(1995), the relations between the fluxes and currents can be established as:
⎡ λas ⎤ ⎡Lasas Lasbs Lasar Lasbr ⎤ ⎡ ias ⎤

⎢ λ ⎥ ⎢L Lbsbs Lbsar Lbsbr ⎥⎥ ⎢⎢ ibs ⎥⎥
⎢ bs ⎥ = ⎢ bsas (1)
⎢λar ⎥ ⎢Laras Larbs Larar Larbr ⎥ ⎢iar ⎥
⎢ ⎥ ⎢ ⎥⎢ ⎥
⎣λbr ⎦ ⎣Lbras Lbrbs Lbrar Lbrbr ⎦ ⎣ibr ⎦
In Equation (1), Lasas(bsbs) is the stator windings self-inductance; Larar(brbr) is the rotor windings
self-inductance; Lasbr(bras), Larbs(bsar) and Lasar(aras), Lbsbr(brbs) are the mutual inductance between the
stator and rotor windings. Since the stator windings are in space quadrature and
asymmetric, and the rotor windings are in space quadrature and symmetric, the following
relations can be written:
Lasas = Las (2)
Lbsbs = Lbs (3)
Lasbs = Lbsas = 0 (4)
Larbr = Lbrar = 0 (5)
Larar = Lbrbr = Lr (6)
The self-inductances of stator and rotor are composed of a leakage inductance and a
magnetizing inductance. That way, a new set of equations can be derived:
Las = Llas + Lmas (7)
Lbs = Llbs + Lmbs (8)
Lr = Llr + Lmr (9)
where (Llas, Llbs) and (Lmas, Lmbs) indicate the stator leakage inductance and magnetizing
inductance, respectively, and Llr and Lmr indicate the rotor leakage inductance and
magnetizing inductance, respectively. Since the rotor windings are assumed to be
symmetric, Equation (9) expresses the rotor windings.
As shown in Fig. 1, there is an angular displacement between the stator and rotor windings
establishing a magnetic coupling between them which results in a mutual inductance. The
equation for the mutual inductances may be expressed in matrix form
⎡ Lsra cosθ r −Lsra sin θ r ⎤

Lsr = ⎢ (10)
⎣ Lsrb sin θ r Lsrb cosθ r ⎥⎦
where Lsra and Lsrb are the amplitude of the mutual inductances.
Thus, the Equation (1) can be rewritten as
⎡ λas ⎤ ⎡ Las 0 Lsra cosθ r −Lsra sin θ r ⎤ ⎡ ias ⎤

⎢λ ⎥ ⎢ 0 Lbs Lsrb sin θ r Lsrb cosθ r ⎥⎥ ⎢⎢ ibs ⎥⎥
⎢ bs ⎥ = ⎢ (11)
⎢ λar ⎥ ⎢ Lsra cosθ r Lsrb sin θr Lr 0 ⎥ ⎢iar ⎥
⎢ ⎥ ⎢ ⎥⎢ ⎥
⎣ λbr ⎦ ⎣ −Lsra sin θ r Lsrb cosθ r 0 Lr ⎦ ⎣ibr ⎦
The voltage equations expressed in the common reference frame may be written as
⎡ vas ⎤ ⎡ras 0 ⎤ ⎡ias ⎤ d ⎡λas ⎤

⎢v ⎥ = ⎢ 0 + (12)
⎣ bs ⎦ ⎣ rbs ⎥⎦ ⎢⎣ibs ⎥⎦ dt ⎢⎣λbs ⎥⎦
⎡0 ⎤ ⎡rr 0 ⎤ ⎡iar ⎤ d ⎡λar ⎤

⎢0 ⎥ = ⎢ 0 + (13)
⎣ ⎦ ⎣ rr ⎥⎦ ⎢⎣ibr ⎥⎦ dt ⎢⎣λbr ⎥⎦
Inserting Equation (11) in Equations (12) and (13), the single-phase induction motor model
acting as a two-phase system can be mathematically expressed by
⎡ d d d ⎤
⎢ ras + dt Las 0 Lsra cosθ r − Lsra sin θ r ⎥
dt dt
⎡ vas ⎤ ⎢ d d d
⎥ ⎡i ⎤
⎢v ⎥ ⎢ Lsrb cosθ r ⎥ ⎢ ⎥
as
0 rbs + Lbs Lsrb sin θ r
⎢ bs ⎥ = ⎢⎢ dt dt dt ⎥ ⎢ ibs ⎥
⎥⎢ ⎥ (14)
⎢0 ⎥ d d d ⎥ iar
⎢ ⎥ ⎢⎢ Lsra cosθ r Lsrb sin θr rr + Lr 0
⎥ ⎢⎣ibr ⎥⎦
⎣ ⎦0 dt dt dt
⎢ d d d ⎥
⎢ − Lsra sin θ r Lsrb cosθ r 0 rr + Lr ⎥
⎣ dt dt dt ⎦
The torque can be determined as
T
⎡i ⎤ d(Lsr )T ⎡ias ⎤
Te = ⎢ ar ⎥ ⎢i ⎥ (15)
⎣ibr ⎦ dθ r _ mech ⎣ bs ⎦
Since θr = pθr_mech , where p is the number of pole pairs and θr is the electrical angle, and
θr_mech is the actual angular displacement of the rotor, the Equation (15) can be rewritten
T
⎡i ⎤ d(Lsr )T ⎡ias ⎤
Te = p. ⎢ ar ⎥ ⎢i ⎥ (16)
⎣ibr ⎦ dθ r ⎣ bs ⎦
In expanded form, (16) becomes
Te = p[ −iasLsra ( iar sin θ r + ibr cosθ r ) + ibsLsrb (iar cosθ r − ibr sin θ r )] (17)
According to Krause et al. (1995), in order to obtain the motor mathematical model with
constant parameters, it is necessary to transform all the variables to the stationary reference
frame (d-q) as shown in Fig. 2. That way θr = 0.
A transformation matrix is necessary in order to establish the new reference frame. Thus
⎡ cosθ r sin θr ⎤ ⎡ cosθr − sin θ r ⎤

T=⎢ ; T −1 = ⎢ (18)
⎣ − sin θ r cosθ r ⎥⎦ ⎣ sin θ r cosθ r ⎥⎦
Fig. 2. Stationary reference frame (d-q).

Applying the transformation matrix to Equations (14) and (17), then
⎡ dLds dLsrd ⎤
⎢rds + dt 0
dt
0 ⎥
⎡ vds
s ⎤ ⎢ ⎥ ⎡i s ⎤
dLqs dLsrq ⎥ ⎢ ds ⎥
⎢ s ⎥ ⎢ 0 rqs + 0 s
⎢ vqs ⎥ ⎢ dt dt ⎥⎥ ⎢iqs ⎥
⎢ ⎥ = ⎢ ⎢s ⎥ (19)
dLsrd dLr
⎢0 ⎥ ⎢ pωr Lsrq rr + pωr Lr ⎥⎥ ⎢idr ⎥
⎢0 ⎥ ⎢ dt dt ⎢s ⎥
⎣ ⎦ ⎢ ⎥ ⎣⎢iqr ⎦⎥
⎢ − pω L dLsrq dL
− pωr Lr rr + r ⎥⎥
⎣⎢
r srd
dt dt ⎦
s s s s
Te = p[Lsrq iqs idr − Lsrd ids iqr ] (20)
where the superscript s denotes the stationary frame. The torque can also be expressed as
dωr
p(Te − Tm ) = J + Bωr (21)
dt
This analytical approach represents the single-phase induction motor as an asymmetric two-
phase motor. The dq voltages, currents, and fluxes for stator and rotor are,
s s s s s s s s
respectively: vds , vqs , ids , iqs , idr , iqr , λds , λqss , λdr , λqrs . The terms Lds , Lqs , Lr , Lsrd , Lsrq denote
the stator and rotor self-inductance and their respective mutual inductance. The stator and
rotor resistance are denoted by rds , rqs , rr . The motor electromagnetic-torque and the load
torque are indicated by Te and Tm , respectively. The moment of inertia, viscous friction
coefficient and the motor speed are, respectively: J , B and ωr .
One of the drawbacks of dealing with an asymmetric motor lies in the pulsating
electromagnetic torque that occurs due to the unbalance between the stator variables, as can
be seen in (19) and (20). To overcome this drawback and to create a symmetric model, a
transformation of the stator variables employing the mutual inductances was proposed by
Correa et al. (2004). The transformation matrix and its application can be written as
⎡1 0 ⎤
S=⎢ ⎥ (22)
⎣0 n ⎦
⎡ vds
s ⎤ ⎡ vds
s ⎤
1
⎢ s ⎥ = S⎢ s ⎥ (23)
⎢⎣ vqs ⎥⎦ ⎢⎣ vqs 1 ⎥⎦
⎡ ids
s ⎤ ⎡idss ⎤
1
⎢ s ⎥ = S −1 ⎢ s ⎥ (24)
⎢⎣ iqs ⎥⎦ ⎢⎣ iqs 1 ⎥⎦
⎡ λds
s ⎤ ⎡λdss ⎤
1
⎢ s ⎥ = S −1 ⎢ s ⎥ (25)
⎢⎣ λqs ⎥⎦ ⎢⎣ λqs 1 ⎥⎦
Lsrd
The transformation element can be written as n = .
Lsrq
When these transformations are applied to the mathematical model of the motor, the
unbalance between the main and auxiliary windings stator variables is eliminated. The new
symmetrical model of the single-phase induction motor acting as a two-phase system can be
given by the following equations:
⎡ vds
s ⎤
1 ⎡ rds 0 ⎤ ⎡ids s ⎤
1 d ⎡λds 1 ⎤
s
⎢ s ⎥=⎢ ⎥.⎢ s ⎥ + ⎢ s ⎥ (26)
⎢⎣ vqs 1 ⎥⎦ ⎢⎣ 0 r 'qs ⎥⎦ ⎢iqs 1 ⎥ dt ⎢λqs 1 ⎥
⎣ ⎦ ⎣ ⎦
⎡ vdr
s ⎤
⎡r 0 ⎤ ⎡ idr ⎤ d ⎡λdr ⎤
s s
⎡ 0 1 ⎤ ⎡λdr ⎤
s
⎢ s ⎥=⎢r ⎥ . ⎢ s ⎥ + ⎢ s ⎥ + ωr ⎢ ⎥ .⎢ s ⎥ (27)
⎢⎣ vqr ⎥⎦ ⎣ 0 rr ⎦ ⎢ iqr ⎥ dt ⎢λqr ⎥
⎣ ⎦ ⎣ ⎦ ⎣ −1 0 ⎦ ⎢⎣λqr ⎥⎦
⎡ λds
s ⎤
1 ⎡ Lds 0 ⎤ ⎡ ids s ⎤
1 ⎡ Lsrd 0 ⎤ ⎡idr ⎤
s
⎢ s ⎥=⎢ ⎥.⎢ s ⎥ + ⎢ ⎥ .⎢ s ⎥ (28)
⎢⎣ λqs 1 ⎥⎦ ⎢⎣ 0 L 'qs ⎥⎦ ⎢ iqs 1 ⎥ ⎣ 0
⎣ ⎦
Lsrd ⎦ ⎢iqr ⎥
⎣ ⎦
⎡ λdr
s ⎤
⎡L 0 ⎤ ⎡ idr ⎤ ⎡Lsrd
s
0 ⎤ ⎡ids 1 ⎤
s
⎢ s ⎥=⎢ r ⎥ .⎢ s ⎥ + ⎢ ⎥ .⎢ s ⎥ (29)
⎢⎣ λqr ⎥⎦ ⎣ 0 Lr ⎦ ⎢ iqr ⎥ ⎣ 0
⎣ ⎦
Lsrd ⎦ ⎢iqs 1 ⎥
⎣ ⎦
s s s s
Te = PLsrd (iqs 1idr − ids 1iqr ) (30)
where L 'qs = n2 Lqs and r 'qs = n 2 rqs .
According to Correa et al. (2004), if the asymmetry that appears in the motor model depends
only on the number of turns of each stator winding, then the ratio presented by n will
corresponds approximately to refer the auxiliary winding variables to the main winding.
Also, if the inductances are slightly different, then L 'qs ≈ Lds .
Observing Equation (26) there is an asymmetry between the stator resistances. This term can
be isolated
⎡ vds
s ⎤
1 ⎡r 0 ⎤ ⎡ids 1 ⎤ d ⎡λds 1 ⎤
s s
⎢ s ⎥ = ⎢ ds . ⎢ ⎥ + ⎢ ⎥ + Δr (31)
⎢⎣ vqs 1 ⎥⎦ ⎣ 0 rds ⎥⎦ ⎢iqs
s
dt ⎢λqss 1 ⎥
⎣ 1 ⎥⎦ ⎣ ⎦
The term Δr is given by
⎡ 0 ⎤
Δr = ⎢ 2 s ⎥ (32)
⎢⎣(n rqs − rds )iqs 1 ⎥⎦
When dealing with motor control, it is necessary to write the dynamic equations of the
induction motor in an arbitrary reference frame. This transformation allows the d-q variables
to be treated as dc signals, which is commonly used in control theory.
If Δr is considered negligible, a symmetric model of the single-phase induction motor in the
arbitrary reference frame can be derived as shown in Fig. 3.
Fig. 3. Revolving reference frame.

The transformation matrix is given by
⎡ cos θ a − sin θ a ⎤ ⎡ cosθ a sin θ a ⎤

Ta = ⎢ ; T −1 = ⎢ (33)
⎣ sin θ a cosθ a ⎥⎦ a ⎣ − sin θ a cosθ a ⎥⎦
When applying the transformation to the motor model based on the stationary reference
frame, given by (26) and (27), the resulting equations can be written in a different matrix
form:
⎡ vDs
a ⎤
1 ⎡r 0 ⎤ ⎡iDs 1 ⎤ d ⎡λDs 1 ⎤
a a
⎡0 −1⎤ ⎡λDs 1 ⎤
a
⎢ a ⎥=⎢ s . ⎢ ⎥ + ⎢ ⎥ + ω ⎥ a ⎥
. ⎢ (34)
⎢⎣ vQs 1 ⎥⎦ ⎣ 0 rs ⎥⎦ ⎢iQs
a a
dt ⎢λQs
a⎢
⎣ 1 0 ⎦ ⎢⎣λQs 1 ⎥⎦
⎣ 1 ⎥⎦ ⎣ 1 ⎥⎦
0 ⎤ ⎡iDr ⎤ d ⎡λDr ⎤ ⎡0 −1⎤ ⎡λDr ⎤

a a a
⎡0 ⎤ ⎡rr
⎢0 ⎥ = ⎢ 0 ⎥ . ⎢ a ⎥ + ⎢ a ⎥ + (ωa − pωr ) ⎢ ⎥ .⎢ a ⎥ (35)
⎣ ⎦ ⎣ rr ⎦ ⎢iQr ⎥ dt ⎢λQr ⎥ ⎣ 1 0 ⎦ ⎢⎣λQr ⎥⎦
⎣ ⎦ ⎣ ⎦
⎡λDs
a ⎤
1 ⎡L 0 ⎤ ⎡iDs 1 ⎤ ⎡Lsrd
a
0 ⎤ ⎡iDr ⎤
a
⎢ a ⎥ = ⎢ ds ⎥ .⎢ a ⎥ + ⎢ ⎥ .⎢ a ⎥ (36)
⎢⎣λQs 1 ⎥⎦ ⎣ 0 Lds ⎦ ⎢iQs 1 ⎥ ⎣ 0
⎣ ⎦
Lsrd ⎦ ⎢iQr ⎥
⎣ ⎦
⎡λDr
a ⎤
⎡L 0 ⎤ ⎡iDr ⎤ ⎡Lsrd
a
0 ⎤ ⎡iDs 1 ⎤
a
⎢ a ⎥=⎢ r ⎥ .⎢ a ⎥ + ⎢ ⎥ .⎢ a ⎥ (37)
⎢⎣λQr ⎥⎦ ⎣ 0 Lr ⎦ ⎢iQr ⎥ ⎣ 0
⎣ ⎦
Lsrd ⎦ ⎢iQs 1 ⎥
⎣ ⎦
The torque equation yields
a a a a
Te = pLsrd [ iQs 1iDr − iDs 1iQr ] (38)
3. Conventional DTC scheme for single-phase induction motor

The basic DTC scheme consists of a logical switching table, which calculates the right
voltage vector to be applied by the inverter to obtain as fast a torque response as possible at
every instant. The control is based on error between the references and estimated values of
torque and flux magnitude and also uses the position of the estimated flux vector. The
torque and flux magnitude error signals are the inputs to the torque and flux hysteresis
controllers, respectively. That way, both the stator flux magnitude and the developed torque
can be directly controlled by proper selection of stator voltage space vectors in order to
reduce the torque and flux errors within the prefixed limits. The hysteresis determines the
inverter switching frequency, which varies with the synchronous speed and load conditions.
This technique achieves robust and fast torque response. Fig 4 shows the block diagram for
the conventional DTC scheme applied to a single-phase induction motor.
To determine the proper voltage vectors to be applied, a logical switching table must be made.
This can be achieved by dividing the dq plane into six sectors. A power inverter will be
employed in a three-leg configuration. That way the dq plane has the form of an asymmetric
hexagon, as shown in Fig 5. There are six active vectors and two zero vectors (V7 and Vo). The
zero vectors are used to reduce the torque. Since a single-phase induction motor can be viewed
as a two-phase induction motor, there is no need for three-to-two phase transformation for the
primary voltage and current. Another aspect concerning this method of control is the absence
of field orientation which makes the control task a less complex issue.
So for the conventional DTC, the following equations are derived from (26) to estimate the
stator flux
λsds = ∫ ( vsd
s s
− rsd isd )dt (39)
λsqs = ∫ ( vsq
s s
− rsq isq )dt (40)
The electromagnetic torque as a function of stator variables can be written as:
s s s s
Te = p(isq λsd − isd λsq ) (41)
Estimation of the stator flux by pure integration can present instability, especially at low
speeds. This problem can be overcome by using a low-pass filter instead of a pure integrator
or other filtering techniques presented in Hu (1998).
Fig. 4. Block diagram of the basic DTC scheme.
Fig. 5. Space-vectors hexagon.

Since DTC uses hysteresis control resulting in a so-called bang-bang control of the torque
and flux, the result is a fast response of the control commands. However, the steady state
performance is characterized by undesirable ripples in current, flux and torque. To avoid
such effects, a high switching frequency should be delivered by the hysteresis loops. But the
amplitude of the hysteresis band has a strong effect on those undesirable ripples mentioned
above (Noguchi and Takahashi, 1997). To eliminate these issues, instead of using a
switching table, a pulse width modulator can be used. Basically, the DTC scheme can be
implemented by means of a closed-loop PI controller which will calculate the required stator
voltage vector and then will be synthesized by a PWM technique (Jabbar et al., 2004).
Therefore, the pulse width modulator is used to optimize the steady state drive
performance.
4. Space vector modulation technique

To control the three-leg inverter driving a single-phase induction motor indirectly, space
vector PWM can be employed. This approach is known to deliver less harmonic distortion in
the output voltages applied to the phases of the induction motor. Consequently, the ripples in
the stator current are diminished and a smooth flux and torque waveform is obtained.
In Correa et al. (2002), Tomaselli et al. (2004) and Jabbar et al. (2004), a space vector
modulation applied to a three-leg inverter driving a single-phase motor is presented. As
shown in Fig. 6, the spatial disposition of the space vectors differs from the one found when
the same inverter system drives a three-phase induction motor. In the latter, the space vector
displays a symmetric hexagon (Zhou, 2002). For the single-phase system, the space vectors
form an asymmetric hexagon (Jabbar et al., 2004). Since the voltages supplied by the three-
leg inverter to the motor phases are sinusoidal waves 900 degrees apart from each other, the
stator voltage vector can be decomposed into real and imaginary parts,
v * = vds
s* s*
+ jvqs (42)
The reference vector v * is used to determine the switching signals that will drive the three-
s* s*
leg inverter. The adjacent vectors vds and vqs display the magnitude of the resulting
switching vectors. Similar to the three-phase inverter feeding a three-phase induction motor,
in a three-phase inverter feeding a two-phase induction motor there are eight voltage
vectors to be applied (six active vectors and two zero vectors). Table I presents the eight
voltage vectors that can be applied to the motor (Vn indicates the resulting space vectors
with n = 0,1,2,…,7).
From the asymmetric hexagon, one can see that there are four vectors with amplitude E
(V1,V3,V4,V6), two vectors with amplitude 2 E (V2,V5) and two null vectors (V0,V7).
Fig. 6. Space Vectors spatial disposition.

s s
Vn S1 S2 S3 vds vqs
V0 0 0 0 0 0
V1 1 0 0 E 0
V2 1 1 0 E E
V3 0 1 0 0 E
V4 0 1 1 -E 0
V5 0 0 1 -E -E
V6 1 0 1 0 -E
V7 1 1 1 0 0
Table 1. Switching states and the adjacent vectors amplitude.
According to Jabbar et al. (2004), the reference vector and the adjacent vectors can be related
by the following equation:
TpwmV * = tds ⋅ Vdss * + tqs ⋅ Vqss * (43)
In Equation (19), tds and tqs are the durations in time in which the vectors Vdss * and Vqss * are
within period Tpwm . The period of duration of the zero vectors can be defined by
t0 = Tpwm − tds − tqs (44)
5. SVPWM-DTC Proposed Scheme

As mentioned in Section 3, to optimize the steady state performance and to diminish the
switching harmonics, a pulse width modulator can be applied. An improvement can be
achieved when torque and stator flux magnitude are controlled by PI controllers using a
closed-loop. The DTC strategy adapted to single-phase induction motors was discussed in
Jacobina et al. (1999) and Neves et al. (2002). Fig. 7 shows the proposed scheme. The essence of
DTC is kept, since its principle of accelerating the flux vector to increase torque is maintained,
and there is no need for speed or position signals. The output signals of the PI controllers can
be viewed as stator voltage components operating in Cartesian coordinates. After the reference
frame transformation, the PWM is fed with the stator voltage components in the stationary
frame. The control strategy relies on the stator flux orientation; therefore, the arbitrary
reference frame should be aligned with the stator flux vector, θ a = θ sf .
The condition to achieve the field orientation can be expressed by
sf sf (45)
λDs 1 = λs ; λQs 1 = 0
To determine the dynamic equations for the proposed technique, some algebraic
manipulations must be done. Taking the Equations (34), (35), (37), aligned with the stator
reference frame, the designed control signals can be derived:
sf λs dλ Lsrd ⎡ Lr L L sf sf ⎤ (46)
vDs 1 = + s − ⎢ λs − r sd iDs 1 + Lsrd iDs 1 ⎥
τ dsσ s dt τ dsσ s Lr ⎣ Lsrd Lsrd ⎦
sf sf
sf sf diQs 1 Lsrd dλQr
vQs 1 = rqs iQs 1 + Lqsσ s + (47)
dt Lr dt
where,
σ s = 1 − L2srd /(Lr Lsd ) (48)
Fig. 7. SVPWM-DTC proposed scheme.

By manipulating (46) and (47) in the synchronous frame, the designed control signals can be
obtained as:
sf λs dλ
vDs 1 = + s + eds (49)
τ dsσ s dt
sf
sf sf diQs 1
vQs 1 = rqs iQs 1 + Lqsσ s + eqs (50)
dt
assuming the terms eds and eqs as feed-forward elements and given by (51) and (52), and
considering the terms that indicate the asymmetry and disturbance negligible:
Lsrd ⎡ Lr L L sf sf ⎤
eds = − ⎢ λs − r ds isD 1 + Lsrd isD1 ⎥ (51)
τ dsσ s Lr ⎣ Lsrd Lsrd ⎦
Lsrd ⎡ rr Lsd sf ⎤
eqs = ⎢ isQ 1 ⎥ (52)
Lr ⎣ Lsrd ⎦
The torque as function of stator flux and currents is given by:
sf sf sf sf
Te = p(iQs 1λDs 1 − iDs 1λQs 1 ) (53)
A 4 poles, ¼ HP, 110 V, 60 Hz, asymmetrical 2-phase induction machine was used with the
following parameters expressed in ohms (Krause et al., 1995):
rds = 2.02;
Xld = 2.79; Xmd = 66.8;
rqs = 7.14;
Xlq = 3.22; Xmq = 92.9;
r´r = 4.12;
X´lr = 2.12.
The total inertia is J = 1.46 × 10-2 kgm2 and Nsd/Nsq = 1.18, where Nsd is the number of turns
of the main winding and Nsq is the number of turns of the auxiliary winding. It was
considered a squirrel cage motor type with only the d rotor axis parameters.
In terms of stator-flux field-orientation
sf
Te = p(iQs 1λs ) (54)
sf
According to (49) and (50), the stator flux control can be accomplished by vDs 1 and torque
sf sf * sf *
control by vQs 1 . The stator voltage reference values vDs 1 and vQs 1 are produced by two PI
controllers. The stator flux position is used in a reference frame transformation to orient the
dq stator currents. Although there is a current loop to decouple the flux and torque control,
the DTC scheme is seen as a control scheme operating with closed torque and flux loops
without current controllers (Jabbar et al., 2004).
6. Simulation results
Some simulations were carried out in order to evaluate the control strategy performance.
The motor is fed by an ideal voltage source. The reference flux signal is kept constant at 0.4
Wb. The reference torque signal is given by: (0,1,-1,0.5)Nm at (0,0.2,0.4,0.6)s, respectively.
The SVPWM method used produced dq axes voltages. The switching frequency was set to
5kHz. Fig. 8 shows the actual value of the motor speed. In Fig. 9 and Fig. 10, the torque
Fig. 8. Motor speed (rpm).

Fig. 9. Commanded and estimated torque (Nm).
Fig. 10. Commanded and estimated flux
Fig. 11. Stator currents in stator flux reference frame.

waveform and the flux waveform are presented. Although the torque presents some
oscillations, the flux control is not affected. The good response in flux control can be seen.
Fig. 11 shows the relation between the d stator current component to the flux production
and the q stator current component to the torque production.
7. Conclusion
The investigation carried out in this paper showed that DTC strategy applied to a single-
phase induction motor represents an alternative to the classic FOC control approach. Since
the classic direct torque control consists of selection of consecutive states of the inverter in a
direct manner, ripples in torque and flux appear as undesired disturbances. To minimize
these disturbances, the proposed SVPWM-DTC scheme considerably improves the drive
performance in terms of reduced torque and flux pulsations, especially at low-speed
operation. The method is based on the DTC approach along with a space-vector modulation
design to synthesize the necessary voltage vector.
Two PI controllers determine the dq voltage components that are used to control flux and
torque. Like a field orientation approach, the stators currents are decoupled but not
controlled, keeping the essence of the DTC.
The transient waveforms show that torque control and flux control follow their commanded
values. The proposed technique partially compensates the ripples that occur on torque in
the classic DTC scheme. The proposed method results in a good performance without the
requirement for speed feedback. This aspect decreases the final cost of the system. The
results obtained by simulation show the feasibility of the proposed strategy.
8. References
Buja, G. S. and Kazmierkowski, M. P. (2004). Direct Torque Control of PWM Inverter-Fed
AC Motors - A Survey, IEEE Transactions on Industrial Electronics, vol. 51, no. 4,
pp. 744-757.
Campos, R. de F.; de Oliveira, J; Marques, L. C. de S.; Nied, A. and Seleme Jr., S. I. (2007a).
SVPWM-DTC Strategy for Single-Phase Induction Motor Control, IEMDC2007,
Antalya, Turkey, pp. 1120-1125.
Campos, R. de F.; Pinto, L. F. R.; de Oliveira, J.; Nied, A.; Marques, L. C. de S. and de Souza,
A. H. (2007b). Single-Phase Induction Motor Control Based on DTC Strategies,
ISIE2007, Vigo, Spain, pp. 1068-1073.
Corrêa, M. B. R.; Jacobina, C. B.; Lima, A. M. N. and da Silva, E. R. C. (2004).Vector Control
Strategies for Single-Phase Induction Motor Drive Systems, IEEE Transactions on
Industrial Electronics, vol. 51, no. 5, pp. 1073-1080.
Charumit, C. and Kinnares, V. (2009). Carrier-Based Unbalanced Phase Voltage Space
Vector PWM Strategy for Asymmetrical Parameter Type Two-Phase Induction
Motor Drives, Electric Power Systems Research, vol. 79, no. 7, pp. 1127-1135.
dos Santos, E.C.; Jacobina, C.B.; Correa, M. B. R. and Oliveira, A.C. (2010). Generalized
Topologies of Multiple Single-Phase Motor Drives, IEEE Transactions on Energy
Conversion, vol. 25, no. 1, pp. 90-99.
Jabbar, M. A.; Khambadkone, A. M. and Yanfeng, Z. (2004). Space-Vector Modulation in a
Two-Phase Induction Motor Drive for Constant-Power Operation, IEEE
Transactions on Industrial Electronics, vol. 51, no. 5, pp. 1081-1088.
Jacobina, C. B.; Correa, M. B. R.; Lima, A. M. N. and da Silva, E. R. C. (1999). Single-phase

Induction Motor Drives Systems, APEC´99, Dallas, Texas, vol. 1, pp. 403-409.
Krause, P. C.; O. Wasynczuk, O. and Sudhoff, S. D. (1995). Analysis of Electric Machinery.
Piscataway, NJ: IEEE Press.
Neves, F. A. S.; Landin, R. P.; Filho, E. B. S.; Lins, Z. D.; Cruz, J. M. S. and Accioly, A. G. H.
(2002). Single-Phase Induction Motor Drives with Direct Torque Control,
IECON´02, vol.1, pp. 241-246.
Takahashi, I. and Noguchi, T. (1986). A New Quick-Response and High-Efficiency Control
Strategy of an Induction Motor, IEEE Transactions on Industry Applications, vol.
IA-22, no. 5, pp.820-827.
Noguchi, T. and Takahashi, I. (1997), High frequency switching operation of PWM inverter
for direct torque control of induction motor, in Conf. Rec. IEEE-IAS Annual
Meeting, pp. 775–780.
Wekhande, S. S.; Chaudhari, B. N.; Dhopte, S. V. and Sharma, R. K. (1999). A Low Cost
Inverter Drive For 2-Phase Induction Motor, IEEE 1999 International Conference on
Power Electronics and Drive Systems, PEDS’99, July 1999, Hong Kong.
Hu, J. and Wu, B. (1998). New Integration Algorithms for Estimating Motor Flux over a
Wide Speed Range. IEEE Transactions on Power Electronics, vol. 13, no. 5, pp. 969-
977.
12
The Space Vector Modulation PWM Control

Methods Applied on Four Leg Inverters
Kouzou A, Mahmoudi M.O and Boucherit M.S
Djelfa University and ENP Algiers,
Algeria
1. Introduction
Up to now, in many industrial applications, there is a great interest in four-leg inverters for
three-phase four-wire applications. Such as power generation, distributed energy systems
[1-4], active power filtering [5-20], uninterruptible power supplies, special control motors
configurations [21-25], military utilities, medical equipment[26-27] and rural electrification
based on renewable energy sources[28-32]. This kind of inverter has a special topology
because of the existence of the fourth leg; therefore it needs special control algorithm to fulfil
the subject of the neutral current circulation which was designed for. It was found that the
classical three-phase voltage-source inverters can ensure this topology by two ways in a way
to provide the fourth leg which can handle the neutral current, where this neutral has to be
connected to the neutral connection of three-phase four-wire systems:
1. Using split DC-link capacitors Fig. 1, where the mid-point of the DC-link capacitors is
connected to the neutral of the four wire network [34-48].
Ta Tb Tc
Sa Sb Sc
C
Va
Vg N Vb
VcN
Vc
C VbN
Ta Tb Tc
VaN
Fig. 1. Four legs inverter with split capacitor Topology.

Ta Tb Tc Tf
Sa Sb Sc Sf
Va
Vg Vb
Vcf
Vc Vbf
Vaf
Ta Vf
Ta Tb Tc Tf
Fig. 2. Four legs inverter with and additional leg Topology.

2. Using a four-leg inverter Fig. 2, where the mid-point of the fourth neutral leg is
connected to the neutral of the four wire network,[22],[39],[45],[48-59].
It is clear that the two topologies allow the circulation of the neutral current caused by the
non linear load or/and the unbalanced load into the additional leg (fourth leg). But the first
solution has major drawbacks compared to the second solution. Indeed the needed DC side
voltage required large and expensive DC-link capacitors, especially when the neutral
current is important, and this is the case of the industrial plants. On the other side the
required control algorithm is more complex and the unbalance between the two parts of the
split capacitors presents a serious problem which may affect the performance of the inverter
at any time, indeed it is a difficult problem to maintain the voltages equally even the voltage
controllers are used. Therefore, the second solution is preferred to be used despite the
complexity of the required control for the additional leg switches Fig.1. The control of the
four leg inverter switches can be achieved by several algorithms [55],[[58],[60-64]. But the
Space Vector Modulation SVM has been proved to be the most favourable pulse-width
modulation schemes, thanks to its major advantages such as more efficient and high DC link
voltage utilization, lower output voltage harmonic distortion, less switching and conduction
losses, wide linear modulation range, more output voltage magnitude and its simple digital
implementation. Several works were done on the SVM PWM firstly for three legs two level
inverters, later on three legs multilevel inverters of many topologies [11],[43-46],[56-57],[65-
68]. For four legs inverters there were till now four families of algorithms, the first is based
on the αβγ coordinates, the second is based on the abc coordinates, the third uses only the
values and polarities of the natural voltages and the fourth is using a simplification of the
two first families. In this chapter, the four families are presented with a simplified
mathematical presentation; a short simulation is done for the fourth family to show its
behaviours in some cases.
2. Four leg two level inverter modelisation

In the general case, when the three wire network has balanced three phase system voltages,
there are only two independents variables representing the voltages in the three phase
system and this is justified by the following relation :
Vaf + Vbf + Vcf = 0 (1)
Whereas in the case of an unbalanced system voltage the last equation is not true:
Vaf + Vbf + Vcf ≠ 0 (2)
And there are three independent variables; in this case three dimension space is needed to
present the equivalent vector. For four wire network, three phase unbalanced load can be
expected; hence there is a current circulating in the neutral:
I La + I Lb + I Lc = I n ≠ 0 (3)
I n is the current in the neutral. To built an inverter which can response to the requirement
of the voltage unbalance and/or the current unbalance conditions a fourth leg is needed,
this leg allows the circulation of the neutral current, on the other hand permits to achieve
unbalanced phase-neutral voltages following to the required reference output voltages of
The Space Vector Modulation PWM Control Methods Applied on Four Leg Inverters 235
the inverter. The four leg inverter used in this chapter is the one with a duplicated additional
leg presented in Fig.1. The outer phase-neutral voltages of the inverter are given by:
Vif = ⎡⎣Sa − S f ⎤⎦ ⋅ Vg where : i = a , b , c (4)
f designed the fourth leg and S f its corresponding switch state.

The whole possibilities of the switching position of the four-leg inverter are presented in
Table 1. It resumes the output voltages of different phases versus the possible switching states
Vaf Vbf Vcf

Vector SaSbSc S f
Vg Vg Vg
V1 1111 0 0 0
2
V 0010 0 0 +1
V3 0100 0 +1 0
4
V 0110 0 +1 +1
5
V 1000 +1 0 0
V6 1010 +1 0 +1
V7 1100 +1 +1 0
8
V 1110 +1 +1 +1
9
V 0001 −1 −1 −1
10
V 0011 −1 −1 0
11
V 0101 −1 0 −1
12
V 0111 −1 0 0
13
V 1001 0 −1 −1
14
V 1011 0 −1 0
15
V 1101 0 0 −1
V 16 0000 0 0 0
Table 1. Switching vectors of the four leg inverter

Equation (4) can be rewritten in details:
⎡Vaf ⎤ ⎡ 1 0 0 −1⎤ ⎡⎢ a ⎤⎥
S
⎢ ⎥ ⎢ ⎥ Sb
⎢Vbf ⎥ = ⎢0 1 0 −1⎥ ⋅ ⎢⎢ ⎥⎥ ⋅ Vg (5)
⎢ ⎥ ⎢0 0 1 −1⎥ ⎢ Sc ⎥
⎢⎣Vcf ⎥⎦ ⎣ ⎦ S
⎢⎣ f ⎥⎦
Where the variable Si is defined by:
⎧ 1 if the upper switch of the leg i isclosed

Si = ⎨ where : i = a , b , c , f
⎩0 if the upper switch of the leg i is opened
3. Three dimensional SVM in a − b − c frame for four leg inverters

The 3D SVM algorithm using the a − b − c frame is based on the presentation of the
switching vectors as they were presented in the previous table [34-35],[69-72]. The vectors
were normalized dividing them by Vg . It is clear that the space which is containing all the
space vectors is limited by a large cube with edges equal to two where all the diagonals pass
by (0,0,0) point inside this cube Fig. 3, it is important to remark that all the switching vectors
are located just in two partial cubes from the eight partial cubes with edges equal to one Fig.
4. The first one is containing vectors from V 1 to V 8 in this region all the components
following the a , b and c axis are positive. The second cube is containing vectors from V 9 to
V 16 with their components following the a , b and c axis are all negative. The common
point (0,0,0) is presenting the two nil switching vectors V 1 and V 16 .
c Axis
a Axis
V 8 V6
V4 +1
V2
V5
V7 +1
b Axis V3 V 1 = V 16 V 14
+1 −1
V 12
−1 V 10
V 15
−1 V 13
11
V V9
V9
Fig. 3. The large space which is limiting the switching vectors
c Axis
a Axis
V8 V6
+1
V4 V2 a Axis
V5
V7 +1
b Axis V 3
V 1 = V 16 V 14
+1 −1
12
V
−1 V 10
V 15 V 13
−1
11 9
V V
V9
Fig. 4. The part of space which is limiting the space of switching vectors
c Axis
V8 V6
+1
V4 V2 a Axis
V7 V5
+1
b Axis V 3
V 1
V 16 −1
+1 V 14
12
V
−1 V 10
−1
V 15 V 13
V 11
V9
Fig. 5. The possible space including the voltage space vector (the dodecahedron) .
The instantaneous voltage space vector of the reference output voltage of the inverter travels
following a trajectory inside the large cube space, this trajectory is depending on the degree
of the reference voltage unbalance and harmonics, but it is found that however the
trajectory, the reference voltage space vector is remained inside the large cube. The limit of
this space is determined by joining the vertices of the two partial cubes. This space is
presenting a dodecahedron as it is shown clearly in Fig. 5. This space is containing 24
tetrahedron, each small cube includes inside it six tetrahedrons and the space between the
two small cubes includes 12 tetrahedrons, in Fig. 6 examples of the tetrahedrons given. In
this algorithm a method is proposed for the determination of the tetrahedron in which the
reference vector is located. This method is based on a region pointer which is defined as
follows:
6
RP = 1 + ∑ C i ⋅ 2(
i − 1)
(6)
i =1
Where:
C i = Sign( INT ( x(i ) + 1)) i = 1 : 6 (7)
The values of x(i ) are:
⎡ Varef ⎤
⎢ V ⎥
⎢ bref ⎥
⎢ Vcref ⎥
x=⎢ ⎥
⎢Varef − Vbref ⎥
⎢Vbref − Vcref ⎥
⎢ ⎥
⎣⎢Varef − Vcref ⎦⎥
Where the function Sign is:
⎧+1 if V > 1
⎪
Sign(V ) = ⎨ −1 if V < 1 (8)
⎪
⎩ 0 if V = 1
V8 V5 V8 V8 V8 V5
V6 V6 V6 V6
V2 V2 V2 V2
V4 V4 V4 V4
V5 V5 V5 V5
V7 V7 V7 V7
V1 , V16 V1 ,V16 V1 , V16 V1 , V16

V14 V14 V14 V14
V3 V12 V3 V12 V3 V3 V12
V12
V10 V10 V10 V10
V15 V13 V15 V13 V15 V13 V15V1 ,V16 V13
V11
V11 V9 V11 V9 V11 V9 V9
Fig. 6. The possible space including the voltage space vector (the dodecahedron).
RP V1 V2 V3 RP V1 V2 V3
9 10 12 9 13
1 V V V 41 V V V 14
5 V2 V 10 V 12 42 V5 V 13 V 14
7 V2 V4 V 12 46 V5 V6 V 14
2 4 8 5 6
8 V V V 48 V V V8
9 V9 V 10 V 14 49 V9 V 11 V 15
13 V2 V 10 V 14 51 V3 V 11 V 15
2 6 14 3 7
14 V V V 52 V V V 15
16 V2 V6 V8 56 V3 V7 V8
17 V9 V 11 V 12 57 V9 V 13 V 15
3 11 12 5 13
19 V V V 58 V V V 15
23 V3 V4 V 12 60 V5 V7 V 15
24 V3 V4 V8 64 V5 V7 V8
Table 2. The active vector of different tetrahedrons
Each tetrahedron is formed by three NZVs (non-zero vectors) confounded with the edges
and two ZVs (zero vectors) ( V 1 , V 16 ). The NZVs are presenting the active vectors
nominated by V1 , V2 and V3 Tab. 2. The selection of the active vectors order depends on
several parameters, such as the polarity change, the zero vectors ZVs used and on the
sequencing scheme. V1 , V2 and V3 have to ensure during each sampling time the equality
of the average value presented as follows:
Vref ⋅ Tz = V1 ⋅ T1 + V2 ⋅ T2 + V3 ⋅ T3 + V01 ⋅ T01 + V016 ⋅ T016 Tz = T1 + T2 + T3 + T01 + T016 (9)
The last thing in this algorithm is the calculation of the duty times. From the equation given
in (9) the following equation can be deducted:
⎡Varef ⎤ ⎡V V2 a V3 a ⎤ ⎡T1 ⎤ ⎡T1 ⎤

⎢ ⎥ ⎢ 1a ⎥ 1 ⎢ ⎥ 1 ⎢ ⎥
⎢Vbref ⎥ = ⎢V1b V2 b V3b ⎥ ⋅ ⋅ ⎢T2 ⎥ = M ⋅ ⋅ ⎢T2 ⎥ (10)
⎢ ⎥ ⎢ T Tz
⎥ z ⎢⎣T3 ⎥⎦ ⎢⎣T3 ⎥⎦
⎢⎣Vcref ⎥⎦ ⎣V1c V2 c V3c ⎦
M
Then the duty times:
⎡T1 ⎤ ⎡Varef ⎤
⎢T ⎥ = T ⋅ M −1 ⋅ ⎢V ⎥ (11)
⎢ 2⎥ z ⎢ bref ⎥
⎢⎣T3 ⎥⎦ ⎢ ⎥
⎣⎢Vcref ⎦⎥
4. 3D-SVM in α − β − γ coordinates for four leg inverter

This algorithm is based on the representation of the natural coordinates a , b and c in a new
3-D orthogonal frame, called α − β − γ frame [72-80], this can be achieved by the use of the
Edit Clark transformation, where the voltage/current can be presented by a vector V :
⎡Vα ⎤ ⎡Va ⎤ ⎡ Iα ⎤ ⎡Ia ⎤

⎢ ⎥ ⎢ ⎥
V = ⎢Vβ ⎥ = C ⋅ ⎢⎢Vb ⎥⎥ I = ⎢ I β ⎥ = C ⋅ ⎢⎢ I b ⎥⎥ (12)
⎢Vγ ⎥ ⎢⎣Vc ⎥⎦ ⎢ Iγ ⎥ ⎢⎣ I c ⎥⎦
⎣ ⎦ ⎣ ⎦
C represents the matrix transformation:
⎡ 1 −1 2 −1 2 ⎤
2 ⎢ ⎥
C = ⋅⎢ 0 3 2 − 3 2⎥ (13)
3 ⎢
⎣1 2 12 1 2 ⎥⎦
When the reference voltages are balanced and without the same harmonics components in
the three phases, the representations of the switching vectors have only eight possibilities
which can be represented in the α − β plane. Otherwise in the general case of unbalance
and different harmonics components the number of the switching vectors becomes sixteen,
where each vector is defined by a set of four elements ⎡⎣Sa , Sb , Sc , S f ⎤⎦ and their positions in
the α − β − γ frame depend on the values contained in these sets Tab. 3.
Each vector can be expressed by three components following the three orthogonal axes as
follows:
⎡Vαi ⎤
⎢ ⎥
V = ⎢Vβi ⎥
i
⎢ i⎥
⎣⎢Vγ ⎦⎥
Where:
i = 1,16 (12)
It is clear that the projection of these vectors onto the αβ plane gives six NZVs and two
ZVs; these vectors present exactly the 2D presentation of the three leg inverters, it is
explained by the nil value of the γ component where there is no need to the fourth leg.
On the other side Fig. 7 represents the general case of the four legs inverter switching
vectors. The different possibilities of the switching vectors in the α − β − γ frame are shown
clearly, seven vectors are localised in the positive part of the γ axis, while seven other
vectors are found in the negative part, the two other vectors are just pointed in the
( 0,0,0 ) coordinates, this two vectors are very important during the calculation of the
switching times.
Vector SaSbSc S f Vγ Vα Vβ
V9 0001 −1 0 0
V 10 0011 −1 −1
3 3
2
V 11 0101 − −1 +1
3 3 3
V 13 1001 +2 0
3
V 12 0111 −2 0
3
1 −1
V 14 1011 − +1
3 3 3
V 15 1101 +1 +1
3 3
16 0000 0 0 0
V
Vector SaSbSc S f Vγ Vα Vβ
1 0 0 0
V 1111
V2 0010 −1 −1
3 3
1
V 3 0100 + −1 +1
3 3 3
V5 1000 2 0
3
V4 0110 −2 0
3
2 +1 −1
V6 1010 + 3 3
3
V 7 1100 +1 +1
3 3
9
V 1110 +1 0 0
Table 3. Switching vectors in the αβγ frame

γ − axis
Vγ
β − axis
1 ⋅ Vg
V8
2 V7
+ ⋅ Vg
3
V4 V
3 V6
1 V5
+ ⋅ Vg
3
V2
α − axis
0 ⋅ Vg V1 V16
V15
V12
1
− ⋅ Vg V14
3
V11
2 V13
− ⋅ Vg
3 V10
V9
− 1 ⋅ Vg
Fig. 7. Presentation of the switching vector in the αβγ frame
V aref a, b, c
V α ref
V bref
V cref α,β V β ref
+
V α ref ⋅ V β ref
+ +
V β ref V β ref
True True True True

V α ref ≥ 3 ⋅ V β ref V α ref ≥ 3 ⋅ V β ref V α ref ≥ 3 ⋅ V β ref V α ref ≥ 3 ⋅ V β ref
P1 P2 P4 P5 P3 P2 P6 P5
Fig. 8. Determination of the prisms

The position of the reference space vector can be determined in two steps.
1. Determination of the prism, in total there are six prisms. The flowchart in Fig.8 explains
clearly, how the prism in which the reference space vector is found can be determined.
2. Determination of the tetrahedron in which the reference vector is located. Each prism
contains four tetrahedrons Fig. 9, the determination of the tetrahedron in which the
reference space vector is located is based on the polarity of the reference space vector
components in a − b − c frame as it is presented in Tab. 4.
γ − axis γ − axis γ − axis γ − axis γ − axis

β − axis β − axis β − axis β − axis β − axis
V8 V8 V8 V8 V8
V6 V6 V6 V6 V6
V2 V2 V2 V2 V2
β − axis β − axis β − axis β − axis β − axis
V1 V16 V1 V16 V1 V16 V1 V16 V1 V16
V14 V14 V14 V14 V14
V9 V9 V9 V9 V9
Reference vector
Active vectors
Prism Tetrahedron components
V1 V1 V2 V2 V3 V3 Vaf Vbf Vcf
T1 V 15
1101 V 13
1001 V 5
1000 ≥ ≺ ≺
T2 V5 1000 V7 1100 V 15 1101 ≥ ≥ ≺
P1
T13 V9 0001 V 13 1001 V 15 1101 ≺ ≺ ≺
T14 V8 1110 V7 1100 V5 1000 ≥ ≥ ≥
T3 V3 0100 V7 1100 V 15 1101 ≥ ≥ ≺
T4 V 15 1101 V 11 0101 V3 0100 ≺ ≥ ≺
P2
T15 V 9
0001 V 11
0101 V 15 1101 ≺ ≺ ≺
T16 V8 1110 V7 1100 V3 0100 ≥ ≥ ≥
T5 V 12 0111 V 11 0101 V3 0100 ≺ ≥ ≺
T6 V3 0100 V4 0110 V 12 0111 ≺ ≥ ≥
P3
T17 V 9
0001 V 11
0101 V 12 0111 ≺ ≺ ≺
T18 V8 1110 V4 0110 V3 0100 ≥ ≥ ≥
T7 V 2
0010 V 4
0110 V 12 0111 ≺ ≥ ≥
T8 V 12 0111 V 10 0011 V2 0010 ≺ ≺ ≥
P4
T19 V9 0001 V 10 0011 V 12 0111 ≺ ≺ ≺
T20 V8 1110 V4 0110 V2 0010 ≥ ≥ ≥
T9 V 14
1011 V 10
0011 V2 0010 ≺ ≺ ≥
T10 V2 0010 V6 1010 V 14 1011 ≥ ≺ ≥
P5
T21 V9 0001 V 10 0011 V 14 1011 ≺ ≺ ≺
T22 V8 1110 V6 1010 V2 0010 ≥ ≥ ≥
T11 V 14
1011 V 13
1001 V5 1000 ≥ ≺ ≺
T12 V5 1000 V6 1010 V 14 1011 ≥ ≺ ≥
P6
T23 V 9
0001 V 13
1001 V 14 1011 ≺ ≺ ≺
T24 V8 1110 V6 1010 V5 1000 ≥ ≥ ≥
Table 4. Tetrahedron determination.

By the same way, using (9) the duty times of the active vectors can be calculated using the
following expression:
⎡Vα ref ⎤ ⎡V V2α V3α ⎤ ⎡T1 ⎤ ⎡T1 ⎤

⎢ ⎥ ⎢ 1α ⎥ 1 1
⎢Vβ ref ⎥ = ⎢V1 β V2 β V3 β ⎥ ⋅ ⋅ ⎢⎢T1 ⎥⎥ = N ⋅ ⋅ ⎢⎢T1 ⎥⎥ (13)
⎢ ⎥ ⎢ T Tz
⎢⎣Vγ ref ⎥⎦ ⎣V1γ V2γ V3γ ⎥⎦ z ⎢⎣T1 ⎥⎦ ⎢⎣T1 ⎥⎦
N
Finally:
⎡T1 ⎤ ⎡Vα ref ⎤

⎢T ⎥ = T ⋅ N −1 ⋅ ⎢V ⎥ (14)
⎢ 1⎥ z ⎢ β ref ⎥
⎢⎣T1 ⎥⎦ ⎢ ⎥
⎣⎢Vγ ref ⎦⎥
5. 3D-SVM new algorithm for four leg inverters

A new method was recently proposed for the identification of tetrahedron and the three
adjacent nonzero vectors [81]. It exposes the relationship between the reference voltages and
the corresponding tetrahedron, on the other side the relationship between the three adjacent
vectors and their duty times in each sampling period. This method is based on the idea that
the three adjacent vectors are automatically in a tetrahedron, but it is not required to
identify this tetrahedron. The authors of this method proposed two algorithms for the
implementation of 3-D SVPWM where the phase angle is necessary to be determined. Each
of the tetrahedrons is appointed by T ( x , y , z ) , it is composed of three non-zero vectors Vx ,
Vy and Vz as it is exposed in the other methods. On the other side the authors of this
method have noticed that the shape of sliced prisms in two methods have the same shapes
but with differences of scale and spatial position Fig. 10. On this basis the initial
transformation used between the a − b − c frame and α − β − γ frame is decomposed to
three matrixes:
⎡Uα ⎤ ⎡2 1 1 ⎤ ⎡U a ⎤ ⎡U a ⎤
⎢U ⎥ = 1 ⋅ ⎢ 0 3
⎥ ⎢ ⎥
− 3 ⎥ ⋅ ⎢Ub ⎥ = T1 ⋅ T2 ⋅ T3 ⋅ ⎢⎢Ub ⎥⎥ (15)
⎢ β⎥ 3 ⎢
⎢⎣ U0 ⎥⎦ ⎢1 1 1 ⎥⎦ ⎢⎣Uc ⎥⎦ ⎢⎣Uc ⎥⎦
⎣
Where:
⎡ 2 3 0 0 ⎤ ⎡ 2 3 0 1 3⎤ ⎡1 0 0 ⎤
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
T1 = ⎢ 0 2 3 0 ⎥ , T2 = ⎢ 0 3 0 ⎥ , T3 = ⎢0 12 − 1 2⎥ (16)
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢⎣ 0 0 1 3 ⎥⎦ ⎢⎣ − 1 3 0 2 3 ⎥⎦ ⎣⎢0 12 1 2 ⎦⎥
The first matrix rotates the a − b − c coordinates around the a axis by an angle of 45 °. Then
the second matrix rotates the a − b − c coordinates around the b axis by an angle of 36.25 °.
Finally, the third matrix modifies its scale by multiplying the a and b axis by 2 3 and
1 3 respectively. After this transformation, it was noticed that the vector used in each
tetrahedron are the same in either frames a − b − c and α − β − γ , of course with two
different spatial positions, Hence, it is deducted that the duration of the adjacent vectors are
independent of coordinates.
γ − axis
Vγ c − axis
β − axis
b − axis
1 ⋅ Vg
V8 a − axis
V7
2 V4
+ ⋅Vg
3
V6
V3
1
+ ⋅Vg V5
3
V2
0 ⋅ Vg V1 α − axis
V16
V15
1 V12
− ⋅ Vg
3 V11 V14
2
− ⋅ Vg V13
3
V10
V9
−1 ⋅ V g
γ − axis
c Axis
α − axis
V8
V8 V6
V7
V4 +1
V3 V6 V4 V2 a Axis
V5
V2 V7 V5
+1
b Axis V3 V1 V 16 −1
V1 V16 β − axis
+1 V 14
V15
V 12 − 1 V 10
V12
V11 −1
V 15 V 13
V13
V10 11
V
V9
V9

The determination of the tetrahedron can be extracted directly by comparing the relative
values of U a , Ub , Uc and zero. The Zero value is used in the comparison for determining
the polarity of the voltages in three phases. If the voltages U a , Ub , Uc and zero are
ordered in descending order, the possible number of permutations is 24 which is equal to
the number of Tetrahedron. Tab. 5 shows the relationship between Terahedrons and the
order of U a , Ub , Uc and zero. Therefore the tetrahedron T ( x , y , z ) can be determined

without complex calculations. These elements are respectively denoted U1 , U 2 , U 3 and U 4
in descending order.
U1 ≥ U2 ≥ U3 ≥ U 4 (17)
For example for: U a ≥ 0 ≥ Uc ≥ Ub
It can be found that: U 1 = U a , U 2 = 0 , U 3 = Uc , U 4 = Ub .
Tetrahedron Vecteurs U 1 ≥ U 2 ≥ U 3 ≥ U 4 Tetrahedron Vecteurs U1 ≥ U2 ≥ U3 ≥ U 4

1 T ( 1, 3,7 ) 0 ≥ Uc ≥ Ub ≥ U a 13 T ( 4, 5,7 ) U b ≥ 0 ≥ Uc ≥ U a
2 T ( 1, 3,11 ) 0 ≥ Uc ≥ U a ≥ Ub 14 T ( 4, 5,13 ) U b ≥ 0 ≥ U a ≥ Uc
3 T ( 1, 5,7 ) 0 ≥ U b ≥ Uc ≥ U a 15 T ( 4,6,7 ) U b ≥ Uc ≥ 0 ≥ Uc
4 T ( 1, 5,13 ) 0 ≥ U b ≥ U a ≥ Uc 16 T ( 4,6,14 ) U b ≥ U c ≥ Uc ≥ 0
5 T ( 1,9,11 ) 0 ≥ U a ≥ Uc ≥ Ub 17 T ( 4,12,13 ) U b ≥ U a ≥ 0 ≥ Uc
6 T ( 1,9,13 ) 0 ≥ U a ≥ U b ≥ Uc 18 T ( 4,12,14 ) U b ≥ U a ≥ Uc ≥ 0
7 T ( 2, 3,7 ) Uc ≥ 0 ≥ Ub ≥ U a 19 T ( 8,9,11 ) U a ≥ 0 ≥ Uc ≥ Ub
8 T ( 2, 3,11 ) Uc ≥ 0 ≥ U a ≥ Ub 20 T ( 8,9,13 ) U a ≥ 0 ≥ U b ≥ Uc
9 T ( 2,6,7 ) Uc ≥ Ub ≥ 0 ≥ U a 21 T ( 8,10,11 ) U a ≥ Uc ≥ 0 ≥ Ub
10 T ( 2,6,14 ) Uc ≥ U b ≥ U a ≥ 0 22 T ( 8,10,14 ) U a ≥ Uc ≥ Ub ≥ 0
11 T ( 2,10,11 ) Uc ≥ U a ≥ 0 ≥ Ub 23 T ( 8,12,13 ) U a ≥ U b ≥ 0 ≥ Uc
12 T ( 2,10,14 ) Uc ≥ U a ≥ U b ≥ 0 24 T ( 8,12,14 ) U a ≥ U b ≥ Uc ≥ 0
Table 5. Determination of tetrahedron vectors

If an equality occurs between two elements, then the reference voltage is in the boundary
between two neighboring tetrahedrons. If two neighboring equalities occur, then the
reference voltage is within the boundary of six Tetrahedrons. If an equality is occurs
between the first and the second element and at the same time an equality occurs between
the third and fourth element, the reference voltage is within the boundary of four
Tetrahedron.if three equalities occur, this means that the space vector is passing in the
point (0,0,0) connecting all the tetrahedrons. For example:
1. For Uc ≥ U a ≥ Ub = 0 the reference voltage is located in the interface of T ( 2,10,11 ) and
T ( 2,10,14 ) , which contains the two vectors V2 and V10 .
2. For U b ≥ Uc = 0 = U a the reference voltage is parallel to V4 and it is located in the
interface among T ( 4, 5,7 ) , T ( 4, 5,13 ) , T ( 4,6,7 ) , T ( 4,6,14 ) , T ( 4,12,13 ) and
T ( 4,12,14 ) .
3. For U b = Uc ≥ 0 = U a the reference voltage is parallel to V6 and is located in the
interface among T ( 2,6,7 ) , T ( 2,6,14 ) , T ( 4,6,7 ) and T ( 4,6,14 ) .
4. For Uc = U b = U a = 0 the reference voltage is nil.
It is clear, that as the other methods the determination of the tetrahedron T ( x , y , z ) allows
the selection of the three vectors Vx , Vy and Vz , and the calculation of the application
duration of the switching states. These switching states have a binary format x , y and z .
Using the relationship between the tetrahedrons and the voltages U a , Ub , Uc and 0 Tab. 5.
The rule for the determination of switching states is derived as follows:
x = 2i , y = x + 2 j , z = y + 2 k (18)
i , j and k are determined from the elements U1 , U 2 and U 3 . Similarly the parameter r
can be deduced, this parameter is used subsequently for the calculation of the application
durations of the three vectors.
⎧0 U1 = 0 ⎧0 U1 = 0 ⎧0 U1 = 0 ⎧0 U1 = 0
⎪1 U 1 = Uc ⎪1 U 1 = Uc ⎪1 U 1 = Uc ⎪1 U 1 = Uc
⎪ ⎪ ⎪ ⎪
i=⎨ , j=⎨ , k=⎨ , r=⎨ (19)
⎪2 U1 = Ub ⎪2 U1 = Ub ⎪2 U1 = Ub ⎪2 U1 = Ub
⎪⎩ 3 U1 = U a ⎪⎩ 3 U1 = U a ⎪⎩ 3 U1 = U a ⎪⎩ 3 U1 = U a
The determination of the duration of each vector is given by:
⎡T1 ⎤ ⎡ a1 a4 a7 ⎤ ⎡U a ⎤
⎢ ⎥ Tp ⎢ ⎥ ⎢ ⎥
⎢T2 ⎥ = U ⎢ a2 a5 a8 ⎥ ⋅ ⎢Ub ⎥ (20)
⎢⎣T3 ⎥⎦ dc ⎢⎣ a3 a6 a9 ⎥⎦ ⎢⎣Uc ⎥⎦
This method can be applied in both frames a − b − c and α − β − γ in the same way. The
switching states x , y and z or the voltage vectors Vx , Vy and Vz are independent of the
coordinates and are determined only from the relative values of U a , Ub and Uc . All matrix
elements ai take the values 0, 1 or -1. Therefore, the calculations need only the addition and
subtraction of U a , Ub and Uc except the coefficient Tp U dc . The ai values are determined
from the following relations where they can be presented as a function of elementary
relative voltages:
⎧ 1 U a = U1 ⎧ 1 U a = U2 ⎧ 1 Ua = U3
⎪ ⎪ ⎪
a1 = ⎨ −1 U a = U 2 , a2 = ⎨ −1 U a = U 3 , a3 = ⎨ −1 U a = U 4
⎪ 0 otherwise ⎪ 0 otherwise ⎪ 0 otherwise
⎩ ⎩ ⎩
⎧ 1 Ub = U1 ⎧ 1 Ub = U2 ⎧ 1 Ub = U 3
⎪ ⎪ ⎪
a4 = ⎨ −1 U b = U 2 , a5 = ⎨ −1 Ub = U 3 , a6 = ⎨ −1 U b = U 4 (21)
⎩ ⎩ ⎩
⎧ 1 Uc = U 1 ⎧ 1 Uc = U 2 ⎧ 1 Uc = U 3
⎪ ⎪ ⎪
a7 = ⎨ −1 Uc = U 2 , a8 = ⎨ −1 Uc = U 3 , a9 = ⎨ −1 Uc = U 4
⎩ ⎩ ⎩
If we substitute these values in (20) and according to the given definitions of i , j , k and
r , the application duration of adjacent vectors, can be expressed in (22), it shows that they
are only depending on the relative voltage vectors U1 , U 2 , U 3 and U 4 .
⎡T1 ⎤ ⎡U1 − U 2 ⎤
⎢T ⎥ = Tp ⋅ ⎢⎢U 2 − U 3 ⎥⎥ (22)
⎢ 2⎥ U
⎣⎢T3 ⎦⎥ ⎣⎢U 3 − U 4 ⎦⎥
dc
6. 3D-SVM new algorithm for four leg inverters

A new algorithm of tetrahedron determination applied to the SVPWM control of four leg
inverters was presented by the authors in [82]. In this algorithm, a new method was
proposed for the determination of the three phase system reference vector location in the
space; even the three phase system presents unbalance, harmonics or both of them. As it was
presented in the previous works the reference vector was replaced by three active vectors
and two zero vectors following to their duty times [34-35], [69-80]. These active vectors are
representing the vectors which are defining the special tetrahedron in which the reference
vector is located.
In the actual algorithm the numeration of the tetrahedron is different from the last works,
the number of the active tetrahedron is determined by new process which seems to be more
simplifiers, faster and can be implemented easily. Form (5) and (12) the voltages in the αβγ
frame can be presented by:
⎡Vα ⎤ ⎡Sa − S f ⎤
⎢ ⎥ ⎢ ⎥
⎢Vβ ⎥ = C ⋅ ⎢Sb − S f ⎥ ⋅ Vg (23)
⎢Vγ ⎥ ⎢ ⎥
⎣ ⎦ ⎢⎣Sc − S f ⎥⎦
It is clear that there is no effect of the fourth leg behaviours on the values of the components
in the α − β plane. The effect of the fourth leg switching is remarked in the γ component.
The representation of these vectors is shown in Fig. 12 –c-.
6.1 Determination of the truncated triangular prisms

As it is shown in the previous sections, the three algorithms are based on the values of the
a − b − c frame reference voltage components. In this algorithm there is no need for the
calculation of the zero (homopolar) sequence component of the reference voltage. Only the
values of the reference voltage in a − b − c frame are needed. The determination of the
truncated triangular prism (TP) in which the reference voltage space vector is located is
based on four coefficients. These four coefficients are noted as C 0 , C 2 , C 3 and C 4 . Their
values can be calculated via two variables x and y which are defined as follows:
Vα
x= (24)
V
γ − axis
V8
γ − axis
α − axis
V8
V7
V4 α − axis
V7
β − axis
V3 V6
V4
V3 V6
V5
V2 V5
V2
V1 V16 β − axis TP = 2
V1 V16 β − axis
V12
V15
V15 TP = 3 TP = 1 α − axis
V12
V11
V14
V11
V14
TP = 4 TP = 6
V13 V13
V10 V10 TP = 5
V9
V9
(a) (b) (c)

Fig. 12. Presentation of the possible switching vectors in a − b − c
Vβ
y= (25)
V
Where:
V = Vα2 + Vβ2 (26)

The coefficients can be calculated as follows:
⎡ 1 ⎤
⎡C 0 ⎤ ⎢⎢ ⎛ 5 ⎞
⎥
⎢C ⎥ ⎢ INT ⎜ − x − ε ⎟ ⎥⎥
⎢ 1⎥ = ⎢ ⎝2 ⎠ (27)
⎢C 2 ⎥ ⎢ INT ( 1 − y − ε ) ⎥⎥
⎢ ⎥ ⎢
⎣C 3 ⎦ ⎢ INT ⎛ 5 + x + ε ⎞ ⎥⎥
⎜ ⎟
⎣⎢ ⎝2 ⎠ ⎦⎥
ε is used to avoid the confusion when the reference vector passes in the boundary between
two adjacent triangles in the αβ plane, the reference vector has to be included at each
sampling time only in one triangle Fig. 12-c-. On the other hand, as it was mentioned in the
first family works, the location of the reference vectors passes in six prism Fig. 12-b-, but
effectively this is not true as the reference vector passes only in six pentahedron or six
truncated triangular prism (TP) as the two bases are not presenting in parallel planes
following to the geometrical definition of the prism Fig. 12-a-. The number of the truncated
prism TP can be determined as follows:
2
TP = 3C 2 + ∑ ( −1 ) C iC i + 1
i
(28)
i =0
6.2 Determination of the tetrahedrons

In each TP there are six vectors, these vectors define four tetrahedrons. Each tetrahedron
contains three active vectors from the six vectors found in the TP. The way of selecting the
tetrahedron depends on the polarity changing of each switching components included in
one vector. The following formula permits the determination of the tetrahedron in which the
voltage space vector is located.
3
Th = 4 (TP − 1 ) + 1 + ∑ ai (29)
1
Where:
ai = 1 if Vi ≥ 0 else ai = 0
i = a, b , c
To clarify the process of determination of the TP and Th for different three phase reference
system voltages cases which may occurred. Figures 13 and 14 are presenting two general
cases, where:
• Figures noted as ‘a’ present the reference three phase voltage system;
• Figures noted as ‘b’ present the space vector trajectory of the reference three phase
voltage system ;
• Figures noted as ‘c’ present the concerned TP each sampling time, where the reference
space vector is located;
• Figures noted as ‘d’ present the concerned Th in which the reference space vector is
located.
Case I: unbalanced reference system voltages
300
Vb
Va
Vc
200 30
20
Voltage Magnitude (V)
100
10
Vgama (V)
0
0
-10
-20
-100
-30
400
-200 200 400
0 200
Vb
eta
(V -200
0
ha (V)
-300 )
-400
-200
Valp
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02 -400
Time (s)
(a) (b)
6 24
5.5 22
20
5
18
4.5
16
The Number of TP
The Number of Th
4 14
3.5 12
3 10
8
2.5
6
2
4
1.5 2
1 0
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02 0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02
Time (s) Time (s)
(c) (d)
Fig. 13. Presentation of instantaneous three phase reference voltages, reference space vector,
TP and Th
Case II Unbalanced reference system voltages with the presence of unbalanced harmonics
500
400
300 150
100
200
Volatge Magnitude (V)
50
Vgama (V)
100
0
0
-50
-100
-100
-200
-150
400
-300
200 400
-400 Vb 0 200
eta 0 v)
(V) -200 ha (
-500 -200 Valp
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02 -400 -400
Time (s)
(a) (b)
6 25
5.5
5 20
4.5
The Number of the TP
The Number of TH
4 15
3.5
3 10
2.5
2 5
1.5
1 0
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02 0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02
Time (s) Time (s)
(c) (d)
Fig. 14. Presentation of instantaneous three phase reference voltages, reference space vector,
TP and Th
6.3 Calculation of duty times

To fulfill the principle of the SVPWM as it is mentioned in (9) which can be rewritten as
follows:
3
Vref ⋅ Tz = ∑ Ti ⋅ Vi (30)
i =0
Where:
3
Tz = ∑ Ti (31)
i =0
In this equation the a − b − c frame components can be used, either than the use of the
α − β − γ frame components of the voltage vectors for the calculation of the duty times, of
course the same results can be deduced from the use of the two frames. The vectors V1 , V2
and V3 present the edges of the tetrahedron in which the reference vector is lying. So each
vector can take the sixteen possibilities available by the different switching possibilities. On
the other hand these vectors have their components in the α − β − γ frame as follows:
⎡Vα i ⎤ ⎡Sai − S fi ⎤
⎢ ⎥ ⎢ ⎥
Vi = ⎢Vβ i ⎥ = C ⋅ ⎢Sbi − S fi ⎥ ⋅ Vg (32)
⎢Vγ i ⎥ ⎢ ⎥
⎣ ⎦ ⎢⎣ Sci − S fi ⎥⎦
From (30), (31) and (32) the following expression is deduced:
⎡Sai − S fi ⎤
3 ⎢ ⎥ 1
∑ Ti ⋅ ⎢Sbi − S fi ⎥ = V ⋅ C −1 ⋅ Vref ⋅ Tz (33)
1 ⎢ ⎥ g
⎢⎣ Sci − S fi ⎥⎦
In the general case the following equation can be used to calculate the duty time for the
three components used in the same tetrahedron:
( ) (( )( ) ( )( )) ⎤⎥
t
⎡ S −S ⋅ S −S ⋅ S −S − S −S ⋅ S −S
⎢ ai fi bj fj ck fk bk fk cj fj ⎡Vrefa ⎤
⎢ ⎥
⎢
( ) (( )( ) (
Ti = σ ⋅ ⎢ Sbi − S fi ⋅ Sak − S fk ⋅ Scj − S fj − Saj − S fj ⋅ Sck − S fk )( )) ⎥⎥ ⋅ ⎢Vrefb ⎥ (34)
⎢ ⎥ ⎢ ⎥
( ) (( )( ) (
⎢ Sci − S fi ⋅ Saj − S fj ⋅ Sbk − S fk − Sak − S fk ⋅ Sbj − S fj
⎣ )( ) )⎥⎦ ⎢⎣Vrefc ⎥⎦
Where:
1
σ= 3
(35)
∑ ( Sai − S fi ) ⋅ ⎡⎣( Sbj − S fj ) ⋅ ( Sck − S fk ) − ( Sbk − S fk ) ⋅ ( Scj − S fj )⎤⎦
1
Variable j and k are supposed to simplify the calculation where:
j = i + 1 − 3 ⋅ INT (i / 3) ; k = i + 2 − 3 ⋅ INT (( i + 1 ) / 3) i = 1, 2, 3
A question has to be asked. From one tetrahedron, how the corresponding edges of the
existing switching vectors can be chosen for the three vectors used in the proposed SVPWM.
Indeed the choice of the sequence of the vectors used for V1 , V2 and V3 in one tetrahedron
depends on the SVPWM sequencing schematic used [108],[115], in one sampling time it is
recommended to use four vectors, the fourth one is corresponding to zero vector, as it was
shown only two switching combination can serve for this situation that is V 16 (0000) and V 1
(1111). On the other hand only one changing state of switches can be accepted when passing
from the use of one vector to the following vector. For example in tetrahedron 1 the active
vectors are: V 11 (1000), V 3 (1001) and V 4 (1101), it is clear that if the symmetric sequence
schematic is used and starts with vector V 1 then the sequence of the use of the other active
vectors can be realized as follow:
V 1 , V 11 , V 3 , V 4 , V 10 , V 4 , V 3 , V 11 , V 1
Active
V1 V 11 V3 V4 V 10 V4 V3 V 11 V1
vector
Sa 1 1 1 1 0 1 1 1 1
Sb 1 1 0 0 0 0 0 1 1
1 0 0 0 0 0 0 0 1
Sc
1 1 1 0 0 0 1 1 1
Sf
1
Ta ..
0
1
Tb ..
0
1
Tc ..
0
1
T f ..
0
t0 t1 t2 t3 t0 t3 t2 t1 t0
4 2 2 2 2 2 2 2 4
Otherwise, if it starts with vector V 16 then the sequence of the active vectors will be
presented as follow Tab.9:
V 16 , V 4 , V 3 , V 11 , V 1 , V 11 , V 3 , V 4 , V 16
Active
V 16 V4 V3 V 11 V1 V 11 V3 V4 V 16
vector
Sa 0 1 1 1 1 1 1 1 0
Sb 0 0 0 1 1 1 0 0 0
0 0 0 0 1 0 0 0 0
Sc
0 0 1 1 1 1 1 0 0
Sf
1
Ta ..
0
1
Tb ..
0
1
Tc ..
0
1
T f ..
0
t0 t1 t2 t3 t0 t3 t2 t1 t0
4 2 2 2 2 2 2 2 4
6.4 Applications
To finalize this chapter two applications are presented here to show the effectiveness of the
four-leg inverter. The first application is the use of the four-leg inverter to feed a balanced
resistive linear load under unbalanced voltages. The second application is the use of the
four-leg inverter as an active power filter, where the main aim is to ensure a sinusoidal
balanced current circulation in the source side. In the two cases an output filter is needed
between the point of connection and the inverter, in the first case an “L” filter is used, while
for the second case an “LCL” filter is used as it is shown in Fig. 15 and Fig. 20.
6.4.1 Applications1
Ta Tb Tc Tf
Sa Sb Sc Sf
R F 1 LF 1 I La
Va
Vcf I Lb 3-phase
Vg Vb balnced
Vbf I Lc linear load
Vc
Vaf I Ln
Ta Vf
RFf LFf
Ta Tb Tc Tf
Vγ
3D-SVM γ − axis
β − axis
1 ⋅V g
V8
2 V7
+ ⋅V g
3
Sa Sb Sc S f
V4
V3 V6
1 V5
+ ⋅ Vg
3
V2
α − axis
0 ⋅Vg V1 V16
V15
V12
1
− ⋅Vg
3 V14
V11
2 V13
− ⋅Vg
3 V10
V9
−1 ⋅ Vg
Fig. 15. Four-leg inverter is used as a Voltage Source Inverter ‘VSI’ for feeding balanced
linear load under unbalanced voltages.
In this application, the reference unbalanced voltage and the output voltage produced by
the four leg inverter in the three phases a, b and c are presented in Fig. 16. The currents in
the four legs are presented in Fig. 17, it is clear that because of the voltage unbalance the
fourth leg is handling a neutral current. To clarify the flexibility of the four leg inverter and
the control algorithm used, Fig. 18 shows the truncated prisms and the tetrahedron in which
the reference voltage space vector is located.
500
Vrefabc
-500
0 0.01 0.02 0.03 0.04 0.05 0.06
1000
Vsa
-1000
0 0.01 0.02 0.03 0.04 0.05 0.06
1000
Vsb
-1000
0 0.01 0.02 0.03 0.04 0.05 0.06
1000
Vsc
-1000
0 0.01 0.02 0.03 0.04 0.05 0.06
Time (s)
Fig. 16. Presentation of three phase reference voltages and the output voltage of the three legs.
ILa
0
-5
0 0.01 0.02 0.03 0.04 0.05 0.06
5
ILb
-5
0 0.01 0.02 0.03 0.04 0.05 0.06
5
ILc
-5
0 0.01 0.02 0.03 0.04 0.05 0.06
5
ILn
-5
0 0.01 0.02 0.03 0.04 0.05 0.06
Time (s)
Fig. 17. Presentation of instantaneous load currents generated by the four legs
6 12
5.5
10
5
Truncated Prism TP
4.5
8
Tetrahedron Th
4
3.5 6
3
4
2.5
2
2
1.5
1 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0 0.01 0.02 0.03 0.04 0.05 0.06
Time (s) Time (s)
Fig. 18. Determination of the Truncated Prism TP and the tetrahedron Th in which the
reference voltage space vector is located.
The presentation of the reference voltage space vector and the load current space vector are
presented in the both frames α − β − γ and a − b − c ,where the current is scaled to compare
the form of the current and the voltage, just it is important to keep in mind that the load is
purely resistive.
200 200
100 100
C u rre n t
Vgama
C axic
C u rre n t
0 0
-100 -100
Vo l ta g e
Vo l tag e
-200 -200
40 40
20 40 20 40
Vb 0 20 Ba 20
eta 0 x is
0
0
ha
Va lp is
-20 -20
-40 -40
-20
-40 -40
-20
A ax
Fig. 19. Presentation of the instantaneous space vectors of the three phase reference system
voltages and load current in α − β − γ and a − b − c frames ( the current is multiplied by 10,
to have the same scale with the voltage)
6.4.2 Applications2
The application of the fourth leg inverter in the parallel active power filtering has used in
the last years, the main is to ensure a good compensation in networks with four wires,
where the three phases currents absorbed from the network have to be balanced, sinusoidal
and with a zero shift phase, on the other side the neutral wire has to have a nil current
circulating toward the neutral of power system source. Figures 21, 22, 23 and 24 show the
behavior of the four leg inverter to compensation the harmonics in the current. The neutral
current of the source in nil as it is shown in Fig 24. Finally the current space vectors of the
load, the active filter and the source in the both frames α − β − γ and a − b − c are presented.
3-phase unbalanced non-linear load
Power Supply 3-phase

Rs Ls I sa I La Ra La Non-linear
esa load
I Fa
I sb I Lb Rb Lb
esb 1-phase
Non-linear
I Fb load
I sc I Lc Rc Lc
esc
3-phase
I Fc unbalnced
I sN I LN RN LN linear load
I FN
LF 2
Ta Tb Tc Tf RF 2
Sa Sb Sc Sf
RF 1 LF1 RFC C F
Va
Vcf
Vg Vb
Vbf
Vc
V af
Ta Vf
RFf LFf
Ta Tb Tc Tf
Vγ
3D-SVM γ − axis
β − axis
1 ⋅Vg
V8
2 V7
+ ⋅V
3 g
S a Sb Sc S f
V4 V
3 V6
1 V5
+ ⋅Vg
3
V2
α − axis
0 ⋅ Vg V1 V16
V15
V12
1
− ⋅ Vg V14
3
V11
2 V13
− ⋅ Vg
3 V10
V9
−1 ⋅ Vg
Fig. 20. Four-leg inverter is used as a Parallel Active Power Filter ‘APF’ for ensuring a
sinusoidal source current.
ILa Iaref
50 30
20
10
0 0
-10
-20
-50 -30
1.1 1.11 1.12 1.13 1.14 1.15 1.16 1.17 1.18 1.19 1.2 1.1 1.11 1.12 1.13 1.14 1.15 1.16 1.17 1.18 1.19 1.2
Time (s) Time (s)
Ias Iaf
30 30
20 20
10 10
0 0
-10 -10
-20 -20
-30 -30
1.1 1.11 1.12 1.13 1.14 1.15 1.16 1.17 1.18 1.19 1.2 1.1 1.11 1.12 1.13 1.14 1.15 1.16 1.17 1.18 1.19 1.2
Time (s) Time (s)
Fig. 21. Presentation of the instantaneous currents of Load, reference, active power filter and
source of phase ‘a’
ILb Ibref
40 30
30
20
20
10
10
0 0
-10
-10
-20
-20
-30
-40 -30
1.1 1.11 1.12 1.13 1.14 1.15 1.16 1.17 1.18 1.19 1.2 1.1 1.11 1.12 1.13 1.14 1.15 1.16 1.17 1.18 1.19 1.2
Time (s) Time (s)
Ibs Ibf
30 30
20 20
10 10
0 0
-10 -10
-20 -20
-30 -30
1.1 1.11 1.12 1.13 1.14 1.15 1.16 1.17 1.18 1.19 1.2 1.1 1.11 1.12 1.13 1.14 1.15 1.16 1.17 1.18 1.19 1.2
Time (s) Time (s)
source of phase ‘b’
ILc Icref
30 20
15
20
10
10
5
0 0
-5
-10
-10
-20
-15
-30 -20
1.1 1.11 1.12 1.13 1.14 1.15 1.16 1.17 1.18 1.19 1.2 1.1 1.11 1.12 1.13 1.14 1.15 1.16 1.17 1.18 1.19 1.2
Time (s) Time (s)
Ics Icf
40 30
30 20
20
10
10
0
0
-10
-10
-20 -20
-30 -30
1.1 1.11 1.12 1.13 1.14 1.15 1.16 1.17 1.18 1.19 1.2 1.1 1.11 1.12 1.13 1.14 1.15 1.16 1.17 1.18 1.19 1.2
Time (s) Time (s)
source of phase ‘c’
ILn Ifref
15 15
10 10
5 5
0 0
-5 -5
-10 -10
-15 -15
1.1 1.11 1.12 1.13 1.14 1.15 1.16 1.17 1.18 1.19 1.2 1.1 1.11 1.12 1.13 1.14 1.15 1.16 1.17 1.18 1.19 1.2
Time (s) Time (s)
-14
x 10 Isn Iff
1.5 15
1 10
0.5 5
0 0
-0.5 -5
-1 -10
-1.5 -15
1.1 1.11 1.12 1.13 1.14 1.15 1.16 1.17 1.18 1.19 1.2 1.1 1.11 1.12 1.13 1.14 1.15 1.16 1.17 1.18 1.19 1.2
Time (s) Time (s)
source of the fourth neutral leg ‘f’
40 8
30 6
4
20
Load 2 Load
Gama axis
10
C axis
0
0
-2
-10
-4
-20 -6 Source
-30 Filter -8 Filter
40 Source 50
20 50 50
Be
B 0 ta
ax a0x s
is 0
xi s is 0 a xi
-20 Aa Al p ha
-40 -50 -50 -50
Fig. 25. Presentation of the instantaneous currents space vectors of the load, active power
filter and the source in α − β − γ and a − b − c frames
7. Conclusion
This chapter deals with the presentation of different control algorithm families of four leg
inverter. Indeed four families were presented with short theoretical mathematical
explanation, where the first one is based on α − β − γ frame presentation of the reference
space vector, the second one is based on a − b − c frame where there is no need for matrix
transformation. The third one which was presented recently where the determination of the
space vector is avoided and there is no need to know which tetrahedron is containing the
space vector, it is based on the direct values of the three components following the three
phases, the duty time can be evaluated without the passage through the special location of
the space vector. The fourth method in benefiting from the first and second method, where
the matrix used for the calculation of the duty time containing simple operation and the
elements are just 0,1 and -1. As a result the four methods can lead to the same results; the
challenge now is how the method used can be implemented to ensure low cost time
calculation, firstly on two level inverters and later for multilevel inverters. But it is
important to mention that the SVMPWM gave a great flexibility and helps in improving the
technical and economical aspect using the four leg inverter in several applications.
8. References
[1] Ionel Vechiu, Octavian Curea, Haritza Camblong, “Transient Operation of a Four-Leg
Inverter for Autonomous Applications With Unbalanced Load,” IEEE
TRANSACTIONS ON POWER ELECTRONICS, VOL. 25, NO. 2, FEBRUARY 2010
[2] L. Yunwei, D. M. Vilathgamuwa, and L. P. Chiang, “Microgrid power quality
enhancement using a three-phase four-wire grid-interfacing compensator,” IEEE
Trans. Power Electron., vol. 19, no. 1, pp. 1707–1719, Nov./Dec. 2005.
[3] T. Senjyu, T. Nakaji, K. Uezato, and T. Funabashi, “A hybrid power system using
alternative energy facilities in isolated island,” IEEE Trans. Energy Convers, vol. 20,
no. 2, pp. 406–414, Jun. 2005.
[4] M. N. Marwali, D. Min, and A. Keyhani, “Robust stability analysis of voltage and current
control for distributed generation systems,” IEEE Trans. Energy Convers., vol. 21,
no. 2, pp. 516–526, Jun. 2006.
[5] C. A. Quinn and N. Mohan, “Active filtering of harmonic currents in three-phase, four-
wire systems with three-phase and single-phase nonlinear loads,” in Proc. IEEE-
APEC’93 Conf., 1993, pp. 841–846.
[6] A. Campos, G.. Joos, P. D. Ziogas, and J. F. Lindsay, “Analysis and design of a series
voltage unbalance compensator based on a three-phase VSI operating with
unbalanced switching functions,” IEEE Trans. Power Electron., vol. 10, pp. 269–274,
May 1994.
[7] S.-J. Lee and S.-K. Sul, “A new series voltage compensator scheme for the unbalanced
utility conditions,” in Proc. EPE’01, 2001.
[8] D. Shen and P. W. Lehn, “Fixed-frequency space-vector-modulation control for three-
phase four-leg active power filters,” in Proc. Inst.Elect. Eng., vol. 149, July 2002, pp.
268–274.
[9] Zhihong Ye; Boroyevich, D.; Kun Xing; Lee, F.C.; Changrong Liu “Active common-mode
filter for inverter power supplies with unbalanced and nonlinear load” Thirty-
Fourth IAS Annual Meeting. Conference Record of the 1999 IEEE, Vol., pp. 1858-
1863, 3-7 Oct. 1999.
[10] A. Julian, R. Cuzner, G. Oriti, and T. Lipo, “Active filtering for common mode
conducted EMI reduction in voltage source inverters ” Applied Power Electronics
Conference APEC 98 , Anaheim, CA, 1998, pp. 934–939.
[11] Z. Lin, L.Mei, Z. Luowei, Z. Xiaojun, and Y. Yilin, “Application of a fourleg ASVG based
on 3D SVPWM in compensating the harmful currents of unbalanced system,” in
Proc. IEEE Power Syst. Technol., 2002, vol. 2, pp. 1045–1050.
[12] P. Lohia, M. K. Mishra, K. Karthikeyan, and K. Vasudevan, “A minimally switched
control algorithm for three-phase four-leg VSI topology to compensate unbalanced
and nonlinear load,” Trans. Power Electron., vol. 23, no. 4, pp. 1935–1944, Jul. 2008.
[13] C. A. Quinn, N. Mohan, “Active Filtering of harmonic Currents in Three-Phase, Four-
Wire Systems with Three-Phase and Single-Phase Non-Linear Loads”, IEEE
Applied Power Electronics Conference (APEC), pp. 829-836, 1992.
[14] A. Nava-Segura, G. Mino-Aguilar, “Four-Branches-Inverter-Based-Active-Filter for
Unbalanced 3-Phase 4-Wires Electrical Distribution Systems”, IEEE Industry
Applications Conference (IAS), pp. 2503-2508, 2000.
[15] P. Rodriguez, R. Pindado, J. Bergas, “Alternative Topology For Three-Phase Four-Wire
PWM Converters Applied to a Shunt Active Power Filter”, IEEE Proceedings of
Industrial Electronics Society (IECON), pp. 2939-2944, 2002.
[16] N. Mendalek, “Modeling and Control of Three-Phase Four-Leg Split-Capacitor Shunt
Active Power Filter,” ACTEA '09. International Conference on Advances in
Computational Tools for Engineering Applications, pp. 121-126. July 15-17, 2009
Zouk Mosbeh, Lebanon.
[17] Seyed Hossein Hosseini, Tohid Nouri2 and Mehran Sabahi, “A Novel Hybrid Active
Filter for Power Quality Improvement and Neutral Current Cancellation,”
International Conference on Electrical and Electronics Engineering, ELECO 2009,
pp. I-244 - I-248.2009.
[18] Valdez, A.A.; Escobar, G.; Torres-Olguin, R.E, “A novel model-based controller for a
three-phase four-wire shunt active filter,” 37th IEEE Power Electronics Specialists
Conference, 2006. PESC '06.
[19] M. Aredes, K. Heumann, J. Hafner, “A Three-phase Four-Wire Shunt Active Filter
Employing a Conventional Three-Leg Converter”, European Power Electronics
Journal, Vol. 6, no 3-4, pp 54-59, December, 1996.
[20] Verdelho P., Marques G., “A Neutral Current Electronic Compensator”, Industrial
Electronics Conference, 1998.
[21] Fabien Meinguet, Johan Gyselinck, “Control Strategies and Reconfiguration of Four-Leg
Inverter PMSM Drives in Case of Single-Phase Open-Circuit Faults,” , pp. 299-
304,2009
[22] Fabien Meinguet*, Eric Semail§, lohan Gyselinck, “Enhanced Control of a PMSM
Supplied by a Four-Leg Voltage Source Inverter Using the Homopolar Torque,”
Proceedings of the 2008 International Conference on Electrical Machines,2008
[23] R.L.A Ribeirol, C.B. Jacobina, A.M.N. Lima, E.R.C. da Silva, “A Strategy for Improving
Reliability of Motor Drive Systems Using a Four-Leg Three-phase Converter,” ,pp.
385-390,2001
[24] Drazen Dujic, Martin Jones, Member, “A General PWM Method for a (2n + 1)-Leg
Inverter Supplying n Three-Phase Machines,” IEEE Transaction on Industrial
electronics, Vol. 56, No. 10,pp. 4107-4118, October 2009
[25] A. Bouscayrol, S. Siala*, M. Pietrzak-David, B. deFomel, “Four-Legged PWM inverters
feeding two induction motors for a Vehicle drive applications,”, Power
Elecironi~zsa nd Variable-Speed Drivers, pp. 700-705,26 - 28 October 1994,
[26] Tomáš Glasberger, Zdenek Peroutka, “Control of Power Supply Unit for Military
Vehicles Based on Four-Leg Three-Phase VSI with Proportional-Resonant
Controllers,” 2008 13th International Power Electronics and Motion Control
Conference (EPE-PEMC ), pp. 1268-1273,2008
[27] Desheng Liu Miao Guan Lei Zhang Yong Wang, “ Research on the Three-Phase Four-
Leg Aeronautical Static Inverter Based on Three-dimensional Space Vector
Modulation in abc Coordinates,” The Ninth International Conference on Electronic
Measurement & Instruments, (ICEMI),pp. 729-732,2009
[28] Zdenĕk Peroutka, Tomáš Glasberger and Jan Molnár, “Design of New Diesel-Electric
Power Supply Unit for Military Vehicles”, 14th International Power Electronics and
Motion Control Conference, EPE-PEMC 2010,pp. 101-108,2010
[29] Jing Li, Fang Zhuo, Jinjun Liu, Xianwei Wang, Bo Wen, Lin Wang, Song Ni, “Study on
Unified Control of Grid-connected Generation and Harmonic Compensation in
Dual-stage High-capacity PV system,”,pp. 3336-3342,2009
[30] Said El-Barbari and W. Hofmann, “Control of a 4 Leg Inverter for Standalone
Photovoltaic Systems,”,pp.348-354,2001
[31] Yogesh K. Chauhan, Sanjay K. Jain, and Bhim Singh, “Transient Performance of Three-
Phase Four-Wire Stand Alone Supply System with Static Converter Employed for
Industrial Loads,”2008
[32] Salem Fouzey, Said El-Barbari4 W. Hofmann, Caner Unsalver, “A new space vector
modulation scheme for three phase four wire inverter for standalone photovoltaic
systems,”
[33] Egon Ortjohann, Arturo Arias, Danny Morton, Alaa Mohd, Nedzad Hamsic, Osama
Omari, “Grid-Forming Three-Phase Inverters for unbalanced loads in hybrid power
system,”,pp. 2396-2399,2006
[34] Manuel A. Perales,M.M. Prats,Ramón Portillo,José L.Mora,José I.León, and Leopoldo
G.Franquelo, “Three-Dimensional Space Vector Modulation in abc
CoordinatesforFour-Leg Voltage Source Converters”, IEEE Power Elect.
Letters,Vol.1, No.4, pp 104-109, December 2003.
[35] M. M. Prats, L. G. Franquelo. J. I Leon. R. Portillo,E. Galvan and J. M. Carrasco, "A SVM-
3D generalized algorithm for multilevel converters".. Proc. IEEE IECON.'03, 2003.
pp. 24-29.
[36] M. Aredes, J. Hafner, and K. Heumann, “Three-phase four-wire shunt active filter
control strategies,” IEEE Trans. Power Electron., vol. 12, no. 2, pp. 311–318, Mar.
1997.
[37] C. J. Zhan, A. Arulampalam, and N. Jenkins, “Four-wire dynamic voltage restorer based
on a three-dimensional voltage space vector PWM algorithm,” IEEE. Trans. Power
Electron., vol. 18, no. 4, pp.1093–1102, Jul. 2003.
[38] R. Faranda and I. Valade, “UPQC compensation strategy and design aimed at reducing
losses,” in Proc. IEEE Int. Symp. Ind. Electron., Jul. 2002, vol. 4, pp. 1264–1270.
[39] Jun Liang, Tim C. Green, Chunmei Feng, and George Weiss, “Increasing Voltage
Utilization in Split-Link Four-Wire Inverters,” IEEE TRANSACTIONS ON POWER
ELECTRONICS, VOL. 24, NO. 6,pp. 1562-1569, JUNE 2009
[40] Changjiang Zhan, Atputharajah Arulampalam, Nicholas Jenkins, “Four-Wire Dynamic
Voltage Restorer Based on a Three-Dimensional Voltage Space Vector PWM
Algorithm,” IEEE Trans. On Power electronics, Vol. 18, No. 4,pp.1093-1102, July
2003
[41] H. Pinheiro, F. Botterbn, C. Rech, L. Schuch, R. F. Camargo, H. L. Hey, H. A. Griindling,
J. R. Pinheiro, “Space Vector Modulation for Voltage-Source Inverters: A Unified
Approach,”,pp. 23-29,2002
[42] Ning-Yi Dai, Chi-Seng Lam, Man-Chung Wong , Ying-Duo Han, “Application of 3D
Direct PWM in Parallel Power Quality Compensators in Three-phase Four-wire
Systems,”, pp. 3220-3225,2008
[43] T.H. Nguyen , Paul K.W. Chan, Y. Shrivastava, S.Y.R. Hui, “A Three-Dimensional
Space Vector Modulation Scheme for Three-Level Three-wired Neutral Point
Clamped Converters,”,pp. 2307-2314,2005
[44] Jing Tang, Man-Chung Wong , Yingduo Hun, “Novel Five-Level Inverter PWM Control
in 3-Phase 4-Wire System for Power Quality,” ,pp. 579-584,2002
[45] Ning-Yi Dai , Man-Chung Wong, Yin-Duo Han, “Controlling Tri-level Center-Split
Power Quality Compensator by 3-Dimensional Space Vector Modulation,” , pp.
1692-1697,2003
[46] Man-Chung Wong, Jing Tang, Ying-Duo Han, “Cylindrical Coordinate Control of
Three-Dimensional PWM Technique in Three-Phase Four-Wired Trilevel Inverter,”
IEEE Transaction On Power Electronics, Vol. 18, No. 1,pp. 208-220, January 2003
[47] Ning-Yi Dai,Man-Chung Wong,Ying-Duo Han, “Three-Leg Center-Split Inverter
Controlled by 3DSVM under dc Variation,”,pp. 1362-1367
[48] Salem M. Ali, Marian P. Kazmierkowski, “PWM Voltage and Current Control of Four-
Leg VSI,”,pp. 196-201,1998
[49] Ojo, O.; Kshirsagar, P.M. “Concise modulation strategies for four-leg voltage source
inverters” IEEE 33rd Annual Power Electronics Specialists Conference, pesc 02.
vol.1, pp. 238 – 243, 2002.
[50] Ojo, O.; Kshirsagar, P.M. “Concise modulation strategies for four-leg voltage source
inverters” Power Electronics, IEEE Transactions on , Vol. 19, Issue: 1 , pp. 46 –53,
Jan 2004.
[51] Liu, C., Peng, D., Lai, J., Lee, F.C., Boroyevich, D., Zhang, R. “Fourlegged converter 3-D
SVM scheme over-modulation study” Applied Power Electronics Conference and
Exposition, APEC Fifteenth Annual IEEE , Vol.1 , pp. 562 -568 , 2000.
[52] Gan Dong, Olorunfemi Ojo, “Design Issues of Natural Reference Frame Current
Regulators with Application to Four Leg onverters”, Proc. Of IEEE-PESC ’05, pp.
1370-1376.
[53] Gan Dong, Olorunfemi Ojo, “Current Regulation in Four-Leg Voltage-Source

Converters”, IEEE Transactions on Industrial Electronics, vol.54, no 4, Aug. 2007,
pp. 2095-2105.
[54] Fanghua Zhang, Yangguang Yan, “Selective Harmonic Elimination PWM Control
Scheme on a Three-Phase Four-Leg Voltage Source Inverter,” IEEE
TRANSACTIONS ON POWER ELECTRONICS, VOL. 24, NO. 7, JULY 2009.
[55] M. J. Ryan, R. W. De Doncker, and R. D. Lorenz, “Decoupled control of a 4-leg inverter
via a new 4 4 transformation matrix,” in Proc. IEEE Power Electron. Spec. Conf.,
1999, pp. 187–192.
[56] Jang-Hwan Kim, Seung-Ki Sul,Prasad N. Enjeti, “A Carrier-Based PWM Method with
Optimal Switching Sequence for a Multi-level Four-leg VSI,” IAS,pp. 99-105, 2005
[57] Maxim A, Dybko, Sergey S. Tumaev, Sergey V. Brovanov, “A Power Losses Calculation
in a Four-legged Three-level Voltage Source Inverter,” 10th International
Conference and seminar EDM, Sections VI, pp. 365-369, JULY 1-6, ERLAGOL,2009
[58] Worrajak Muangjai, Suttichai Premrudeepreechacharn, “Implementation of a Carrier-
based Three-dimensional Space Vector PWM Technique for Three-phase Four-leg
Voltage Source Converter with Microcontroller,” ICIEA,pp. 837-841, 2009
[59] Praveen Kumar M, Mahesh Kumar Mishra, Sandeep Joseph, “Switching Minimization
of Three-phase Four-leg Dynamic Voltage Restorer,” 2009
[60] Salem M. Ali, Marian P. Kazmierkowski, “Current Regulation of Four-Leg PWM-
VSI,”,pp. 1853-1858,1998
[61] Verdelho, P.; Marques, G.D, “A current control system based in αβ0 variables for a four-
leg PWM voltage converter,” Proceedings of the 24th Annual Conference of the
IEEE Industrial Electronics Society, 1998. IECON '98. Vol. 3, pp. 1847 – 1852.
[62] Reza Nasiri, Student, Ahmad Radan, “Pole-Placement Control Strategy for 4-Leg
Voltage-Source Inverters,” 1st Power Electronic & Drive Systems & Technologies
Conference,pp.74-79,2010
[63] V. Yaramasu, J. Rodriguez, B. Wu, M. Rivera, A. Wilson and C. Rojas, “A Simple and
Effective Solution for Superior Performance in Two-Level Four-Leg Voltage Source
Inverters: Predictive Voltage Control,”
[64] Nikhil Prabhakar, Mahesh K. Mishra, “Dynamic Hysteresis Current Control to
Minimize Switching for Three-Phase Four-Leg VSI Topology to Compensate
Nonlinear Load,” IEEE Trans ON Power Electronics, Vol. 25, No. 8,pp. 1935-1942,
August 2010
[65] N.V. Nho and M.J. Youn, “Carrier PWM algorithm with optimised switching loss for
three-phase four-leg multilevel inverters,” ELECTRONICS LETTERS Vol. 41 No. 1,
6th January 2005
[66] N.V. Nho and M.J. Youn, “Carrier PWM algorithm with optimised switching loss for
three-phase four-leg multilevel inverters,” ELECTRONICS LETTERS Vol. 41 No. 1,
6th January 2005
[67] Salvador Ceballos, Josep Pou, Jordi Zaragoza, José L. Martín, Eider Robles, Igor Gabiola,
Pedro Ibáñez, “Efficient Modulation Technique for a Four-Leg Fault-Tolerant
Neutral-Point-Clamped Inverter,” IEEE Trans On Ind. Elect.Vol. 55, No. 3,pp. 1067-
1074, March 2008
[68] O´ scar Lo´pez, Jacobo A´ lvarez, Jesu´s Doval-Gandoy, Francisco Freijedo, Alfonso
Lago and Carlos M. Penãlver, “Four-Dimensional Space Vector PWM Algorithm
for Multilevel Four-Leg Converters,” 2008
[69] Fernando Botteron, Robinson F. de Camargo, Humberto Pinheiro Hilton A. Griindling,
Jose R. Pinheiro and Helio L. Hey, “On the Space Vector Modulation and Limiting
Algorithms for Three-phase Four-Leg Voltage Source Inverter in abc Coordinates,”

, pp. 12-17,2002
[70] Manuel A. Perales, M. M. Prats, Ramon Portillo, Jose L. Mora, Leopoldo G. Franquelo,
“Three Dimensional Space Vector Modulation for four-leg inverters using natural
coordinates,”, pp. 1129-1134,2004
[71] Rui Wu, Donghua chen, shaojun xie, “ A Three-dimentsional Space Vector Modulation
Algorithm in A-B-C Coordinate Implementation By FPGA”,pp. 1017-1075,2005
[72] Man.Wong, J.Tang and Y.Han “Three-Dimensional Pulse-Width Modulation Technique
in Three-level Power Inverters for Three-Phase Four-Wired System”, IEEE Trans.
on Power Electronics, 16(3), pp.418 -427, May 2001.
[73] A.Kouzou, Mahmoudi M.O, Boucherit. M.S, “Comparative Study of the SVPWM 3D
Algorithms Used for four-leg inverters”, SSD09,6th International Multi-Conference
on Systems, signals and Devices. Djerba Tunisia. 23-26 March 2009.
[74] Richard Zhang, V. Himamshu Prasad, Dushan Boroyevich and FredC.Lee, “Three-
Dimensional Space Vector Modulation for Four-Leg Voltage-Source Converters”,
IEEE Transaction on Power Electronics, Vol.17, No.3, pp 314-326, May2002
[75] R. Zhang, D. Boroyevich, H. Prasad, H. Mao, F. C. Lee, and S Dubovsky, “A three-phase
inverter with a neutral leg with space vector modulation,” in Proc. IEEE-APEC’97
Conf., 1997, pp. 857–863.
[76] R. Zhang, H. Prasad, D. Boroyevich, and F. C. Lee, “Analysis and design of a three-
phase inverter with a neutral leg,” in Proc. EPE’97 Conf., 1997, pp. 1.170–1.175.
[77] H. Prasad, D. Boroyevich, and R. Zhang, “Analysis and comparison of space vector
modulation schemes for a four-leg inverter,” in Proc. IEEEAPEC’ 97 Conf., 1997,
pp. 864–871.
[78] E. Ortjohann, A. Mohd, N. Hamsic, A. Al-Daib, M.Lingemann, “ three-Dementional
Space Vector Modulation Algorithm for Three-leg Four Wire voltage source
Inverters,” POWERENG,pp. 605-610, April 12-14,2007, Setubal, Portugal
[79] M. G. Villala, E. Ruppert F, « 3-D Space vector PWM for Three-leg four-wire Source
Inverter,” 35th annual IEEE Power electronics Specialists conference,pp. 3946-3951,
Aachen, Germany,2004
[80] E. Ortjohann, A. Mohd, N. Hamsic, M. Lingemann, W. Sinsukthavorn, D. Morton, “A
Novel Space Vector Modulation Control Strategy for Three-leg Four-Wire Voltage
Source Inverters,”
[81] Xiangsheng Li, Zhiquan Deng, Zhida Chen, and Qingzhao Fei, “Analysis and
Simplification of Three-dimensional Space Vector PWM for Three-phase Four-leg
Inverters,”
[82] Kouzou A, Mahmoudi M.O, Boucherit M.S, “A new 3D-SVPWM algorithm for Four-leg
Inverters,” IEEE International Electric Machines and Drives Conference, IEMDC
'09.pp. 1674 – 1681, May 2009. Miami, Florida, USA
This book was distributed courtesy of:
For your own Unlimited Reading and FREE eBooks today, visit:
http://www.Free-eBooks.net
Share this eBook with anyone and everyone automatically by selecting any of the
options below:
To show your appreciation to the author and help others have

wonderful reading experiences and find helpful information too,
we'd be very grateful if you'd kindly
post your comments for this book here.
COPYRIGHT INFORMATION
Free-eBooks.net respects the intellectual property of others. When a book's copyright owner submits their work to Free-eBooks.net, they are granting us permission to distribute such material. Unless
otherwise stated in this book, this permission is not passed onto others. As such, redistributing this book without the copyright owner's permission can constitute copyright infringement. If you
believe that your work has been used in a manner that constitutes copyright infringement, please follow our Notice and Procedure for Making Claims of Copyright Infringement as seen in our Terms
of Service here:
http://www.free-ebooks.net/tos.html

Electric Machines and Drives PDF

Uploaded by

Copyright:

Available Formats

Electric Machines and Drives PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Electric Machines and Drives PDF

Uploaded by

Copyright:

Available Formats

ELECTRIC MACHINES

Copyright © 2011 InTech

Publishing Process Manager Katarina Lovrecic

First published February, 2011

A free online edition of this book is available at www.intechopen.com

Electric Machines and Drives, Edited by Miroslav Chomat

Chapter 1 Premium Efficiency Motors 1

Chapter 2 Electric Motor Performance Improvement Using

Chapter 3 Magnetic Reluctance Method for Dynamical

Chapter 4 Minimization of Losses in Converter-Fed

Chapter 5 Sensorless Vector Control of Induction Motor Drive

Chapter 6 Feedback Linearization of Speed-Sensorless

Chapter 7 From Dynamic Modeling to Experimentation

Chapter 8 A RMRAC Parameter Identification

Chapter 9 Swarm Intelligence Based Controller for Electric

Chapter 10 Operation of Active Front-End Rectifier

Chapter 11 Space Vector PWM-DTC Strategy

Chapter 12 The Space Vector Modulation PWM Control

Premium Efficiency Motors

BP 182, El Harrach, 16200 Alger

Montréal, Québec, Canada H2X 1X6

2. Energy, climate change and electricity

the level of pre-industrial times is likely to require emissions reductions up to 90 % below

Fig. 1. Electricity generation by fuel

3. Motor losses segregation and efficiency

ΔPstr = Pin – Pout – (ΔPel1 + ΔPel2 + ΔPcore + ΔPmech) (1)

Fig. 2. Induction motor losses distribution

5. Minimum energy performance standard MEPS and efficiency motor

5.1 MEPS in North America

Fig. 3. Efficiency trend 2001-2006 and efficiency expected 2008-2013

5.2 New IEC standard

• Super Premium efficiency level IE4

5.3 MEPS in EU and BRIC countries

Efficiency Class IEC USA/Canada CEMEP China

6. Life cycle cost Premium motors

Fig. 4. Lifetime motor cost

7. Barriers to high efficiency motors market penetration

8. Incentive policies to overcome barriers

Fig. 5. Incentives, voluntary and mandatory measures impact

9. Experimental setup and results

Fig. 6. Measurement setup with instrumentation

Fig. 7. Stator Premium motors

Fig. 8. Rotor Premium motors

Fig. 9. Manufacturing technology impact Fig. 10. Viltage feeding impact

Fig. 11. Drive impact Fig. 12. Drive impact

Fig. 13. Frequency impact Fig. 12. Frequency & manufacturing

Rating Poles Rating Rating Poles

Annex 3. CEI efficiencies

Rating Poles Rating Rating Poles

Annex 3. CEI efficiencies

Annex 3. CEI efficiencies

Electric Motor Performance Improvement Using

Fig. 1. High Power factor induction machines-windings arrangement

2.2 Auxiliary windings connections

2.3 Variable Capacitors

Fig. 3. Auxiliary windings: “single –phase connection”

2.3.1 Thyristor-based variable capacitor

2.3.2 IGBT-based variable capacitor

Fig. 5. IGBT in bidirectional topology