Fulltext

Modeling and Simulation of Hydropower
Plant in Scilab
Erik Rognaldsen
Master of Energy Use and Energy Planning

Submission date: June 2017
Supervisor: Bjørnar Svingen, EPT
Norwegian University of Science and Technology

Department of Energy and Process Engineering
Abstract
A hydro power plant is modeled in a computer application called Scilab, an open source
MATLAB clone with capabilities to create block diagrams and do system simulations. The
system is designed to do both frequency regulation when running in island operation and
power regulation in a stiff power network. The conduit system are set up with a dynamic
penstock model, surge shaft and a rigid head-race tunnel. Draft tube and a lower surge
shaft are also been implemented. It is used a linear turbine model which utilize linearized
turbine characteristics of a designed turbine to get a detailed description of the turbine
behavior. The linearization are performed with a computer program based on modified
Euler’s turbine equations deduced by Torbjørn Nielsen. The model is then evaluated up
against a hydro power simulation tool called LVTrans, a program coded in LabVIEW
based on the method of characteristics.
The scilab model shows good performance for simulations around its operating point.
Simulation in island mode shows a frequency deviation between the models during a 10%
load drop of 0.25Hz, or 0.5%. Simulating in stiff grid both models responds similarly
to a frequency change of +0.01Hz decreasing the turbine power output by 16.7% with a
droop of 6%. The system simulation performance of mass oscillations and water hammer
is also tested. The simulations was performed with a change in the friction factor to a
more representative number for an turbulent and flow fluctuating situations. This gave
good results compared to LVTrans simulations, but it gave too much friction at the end of
the event as the mass flow fluctuations reaches zero.
The surge shaft model is evaluated up against a u-pipe test rig located at the water power
laboratory. Comparing simulations to the rig tests gave indications that the model is suited
for simulating transient pressure pulsations in form of mass oscillations. It also shows that
the friction model brakes the pulsations too fast resulting in a decrease in amplitude and
the time to reach steady state gets too short.
Keywords: Hydropower modeling, block diagram, linearized Francis turbine, dynamic

penstock, model verification.
i
Sammendrag
Modelering av et vannkraft verk er utført i et program kalt Scilab. Scilab gir mulighet for
numerisk kalkulering og system-modellering likt som MATLAB, men er uten kostnader og
har åpen kildekode. Det modelerte anlegget er designet for å utføre både frekvenssimu-
leringer i isolert nett samt effektregulering mot stivt nett. Systemet er modelert med en
dynamisk rørgate, øvre og nedre svingesjakt, sugerør samt inngangstunnel. Det er benyt-
tet en linearisert fransis turbin modell basert på modifiserte Euler turbin ligninger utviklet
av Torbjørn Nielsen.
Modellen er blitt testet opp mot et program utviklet i labVIEW kalt LVTrans. Dette pro-
grammet er basert på karakteristikk metoden og er et utprøvd program av bla. statkraft med
gode resultater. Dette vil fungere som en mal for hvordan scilab modellen presterer og blir
evaluert der etter. Kjøringer i øydrift avslører at turbinen har et begrenset arbeidsomrøde
der den gir gode resultater. Lastavslag fra 85% til 75% gir veldig liten forskjell i transient
frekvens økning (0.25Hz). Kjøringer med større last avslag som fra 100% til 50% gir en
forskjell på 4Hz som er ca. halvparten av hva LVTrans simulerte. Simulering i stivt nett gir
de beste resultatene der det er svært lite forskjell mellom modellene. Begge responderte
likt ved en frekvensendring på +0.01Hz og strupet produksjonen med 16.7% ca 15.7MW
med en statikk på 6%.
Systemets ytese for trykktransienter og massesvinginger er også testet. Scilab modellen

viser gode resultater, da på tross av enkel statisk friksjonsmodel, men ved bruk av reléer i
modellen er det muligå endre friksjonskoeffisienter under kjøringen noe som kompenserer
for manglende samsvar mellom modellene. Det er også utført trykksjakt valitering med
urør-rig ved vannkraft laboratoriet på NTNU. Kjøringene er validert opp mot trykks-
jakten på riggen og simuleringene vier tilfredsstillende resultater, men med noe avvik.
Kjøringene syneligjør den enkle friksjonsmodellen som scilab modellen tar i bruk og avvik
forkommer ved at simuleringer gir lavere trykkamplituder, og massesvingningene dør ut
for kjapt.
Oppsummert, viser modellen lovende resultater sammenliknet med LVTrans, men den har
sine restriksjoner da turbinen er linearisert og friksjonsmodellen er statisk.
ii
Preface
The master thesis has been written at the Waterpower Laboratory under the Department of
Energy and Process Engineering at the Norwegian University of Science and Technology.
During the last semester, a simulation tool for hydropower simulations has been developed
in a program called Scilab. The process of system modelling and increasing the quality
of simulation has been some of the central challenges. The master thesis is a continuation
and extension of the project work conducted during the autumn semester 2016.
I would like to thank my supervisor Bjørnar Svingen for supporting, and for motivating
me during the whole last year at NTNU. Special thanks to PhD candidates at the lab for
their availability and good discussions. Also special thanks to my girlfriend Kathrine who
has been supportive and patient during the busy times working with the thesis. Thanks to
my parents for being motivational, and supporting me in every way. Also thanks to the
department of energy and process engineering for allowing me to write my master thesis
at the water power laboratory.
Last but no least, thanks to all my fellow students at the Waterpower Laboratory for making
the last semesters amazingly great, and the trip to Nepal unforgettable. Working in such
competent and enjoyable environment has made the last year one of the best ones so far.
Erik Rognaldsen
Trondheim, June 15, 2017
iii
iv
Table of Contents
Summary i
Sammendrag ii
Preface iii
Table of Contents ix
List of Tables xi
List of Figures xvi
Nomenclature xvii
1 Introduction 1
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Thesis Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 System Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Previous Work 5
v
2.1 Project Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Paper Presented at the 7th International Symposium on Current Research

in Hydraulic Turbines . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3 Theory 7
3.1 General Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.2 Transient Fudamentals . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.2.1 Pressure wave velocity . . . . . . . . . . . . . . . . . . . . . . . 9
3.2.2 Wave reflection time . . . . . . . . . . . . . . . . . . . . . . . . 10
3.2.3 Mass Forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.3 Friction Losses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.4 Surge shaft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.4.1 Surge shaft Oscillations . . . . . . . . . . . . . . . . . . . . . . 17
3.4.2 Surge Shaft Transfer Function . . . . . . . . . . . . . . . . . . . 18
3.5 Penstock and Tunnels . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.5.1 Dynamic penstock . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.5.2 Approximation of Tanh . . . . . . . . . . . . . . . . . . . . . . . 20
3.5.3 RLC Penstock Model . . . . . . . . . . . . . . . . . . . . . . . 21
3.5.4 Static Tunnel Model . . . . . . . . . . . . . . . . . . . . . . . . 23
4 Modelling and Simulation Tools 25
4.1 Scilab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.2 LVTrans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5 Hydropower Modelling 29
5.1 Modeling scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.2 Hydraulic System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
vi
5.2.1 Penstock . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
5.2.2 Head Race Tunnel and Draft Tube . . . . . . . . . . . . . . . . . 34
5.3 Surge Shaft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5.4 Surgeshaft validation model . . . . . . . . . . . . . . . . . . . . . . . . 37
5.5 Turbine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.6 PID - Regulator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.7 Servo System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.8 PID Tuning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5.8.1 Method 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5.8.2 PID tuning by frequency analysis . . . . . . . . . . . . . . . . . 45
6 Simulation and Analysis 47
6.1 System Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
6.1.1 Scilab Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
6.1.2 LVTrans Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
6.2 Functional Requirement . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
6.3 Stability Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
6.4 Evaluation of Hydraulic Systems in Scilab . . . . . . . . . . . . . . . . . 60
6.4.1 Comparing Dynamic and Static Penstock . . . . . . . . . . . . . 60
6.4.2 The Effect of Extending Penstock Length . . . . . . . . . . . . . 62
6.4.3 The effect of adding a draft tube and lower surge shaft . . . . . . 63
6.5 Model Validation with LVTrans . . . . . . . . . . . . . . . . . . . . . . 64
6.5.1 Frequency regulation in island mode. . . . . . . . . . . . . . . . 64
6.5.2 Power Regulation in Stiff Grid Mode. . . . . . . . . . . . . . . . 65
6.5.3 Water Hammer and Mass Oscillations . . . . . . . . . . . . . . . 67
vii
7 Laboratory Work 71
7.1 Laboratory Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
7.2 Results From the Rig . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
7.3 Rig results V.S Scilab model . . . . . . . . . . . . . . . . . . . . . . . . 76
8 Discussion 79
8.1 Model Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
8.2 Model Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
8.3 Scilab VS LVTrans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
8.3.1 Frequency regulation in Island mode . . . . . . . . . . . . . . . . 82
8.3.2 Power regulation in Stiff Grid Mode . . . . . . . . . . . . . . . . 82
8.3.3 Mass oscillation and water hammer simulations . . . . . . . . . . 83
8.4 Surgeshaft Verification With Test Rig . . . . . . . . . . . . . . . . . . . 84
9 Conclusion 85
10 Further Work 87
Bibliography 89
Appendix A
A A-1
B B-1
C C-1
D D-1
viii
E E-1
F F-1
G G-1
ix
x
List of Tables
3.1 Typical surface roughness on different materials. [15] . . . . . . . . . . . 14
3.2 RLC component description . . . . . . . . . . . . . . . . . . . . . . . . 23
5.1 Hydraulic system parameter values. . . . . . . . . . . . . . . . . . . . . 30
5.2 Turbine spesifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5.3 Proposal for PID tuning by frequency analysis [9] . . . . . . . . . . . . . 45
6.1 Values from comparing dynamic and static penstock . . . . . . . . . . . . 62
6.2 Values inserted in the scilab model during the simulations . . . . . . . . . 63
7.1 Losses in the test pipe under different flow scenarios. . . . . . . . . . . . 73
7.2 Values inserted in the scilab model during the simulations . . . . . . . . . 76
10.1 Parameter values for the head race tunnel . . . . . . . . . . . . . . . . . A-1
10.2 Parameter values for the penstock . . . . . . . . . . . . . . . . . . . . . A-3
10.3 Values for the parameters, a and b based on earlier experiences. . . . . . B-2
xi
xii
List of Figures
1.1 General layout of the Norwegian island operated turbine. . . . . . . . . . 3
1.2 Simulation set up for stiff grid mode. . . . . . . . . . . . . . . . . . . . 4
2.1 Block diagram made in the project work fall 2016. . . . . . . . . . . . . 5
3.1 A representation of pressure propagation in a conduit.[5] . . . . . . . . . 10
3.2 Description of mass forces during transient state. . . . . . . . . . . . . . 11
3.3 The moody diagram, used to obtain friction factor f [10] . . . . . . . . . 14
3.4 Cross section of the inlet tunnel . . . . . . . . . . . . . . . . . . . . . . . 15
3.5 Cross section sketch of the conduit system at the susrge shaft and the forbay. 17
3.6 A penstock with lenght a L can be represented with this equivalent circuit
schematic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.1 Interface of scilab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.2 Interface of LVTrans . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5.1 Cross section of the system. . . . . . . . . . . . . . . . . . . . . . . . . 29
5.2 General layout for the system in grid mode with frequency regulation. . . 31
xiii
5.3 General layout for the system in stiff network mode with power regulation. 31
5.4 General layout for the hydraulic system used in the simulations with scilab. 32
5.5 Block diagram from scilab describing the penstock used in the simulations. 33
5.6 Transfer function block representing the RLC penstock model . . . . . . 34
5.7 Transfer function block from scilab describing the head-race tunnel used
in the simulations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
5.8 Transfer function block from scilab describing the draft tube tunnel used
in the simulations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
5.9 Transfer function block from scilab describing the surge shaft used in the
simulations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5.10 Block diagram for simulation of the test rigg at the laboratory. . . . . . . 37
5.11 The model for surge shaft validation with scilab . . . . . . . . . . . . . . 37
5.12 Picture of interface of the programe used to linearize the turbine charac-
teristics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.13 Transfer function block from scilab describing the surge shaft used in the
simulations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5.14 PID regulator for speed regulation and power regulation, as modelled in
scilab. [17] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
5.15 The ratelimiter block used for waterhammer simulations. . . . . . . . . . 43
5.16 Examples of unstable and stable systems based on phase and gain margin
[11]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
6.1 Layout of the block diagram for frequency regulation in island mode. . . 49
6.2 Layout of the block diagram for load regulation connected to a stiff load. . 50
6.3 The effect of three different droop settings . . . . . . . . . . . . . . . . . 50
6.4 The hydraulic system used in the simulations. . . . . . . . . . . . . . . . 51
6.5 Analytic functions for scilab model . . . . . . . . . . . . . . . . . . . . . 52
6.6 Changes in the infrastructure due to water hammer simulations. . . . . . . 53
6.7 set context tool to inset system variables. . . . . . . . . . . . . . . . . . . 53
xiv
6.8 Analytic model with water hammer sim and friction modification. . . . . 54
6.9 System in LVTrans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
6.10 Bode plot for the open loop transfer function. Kp = 3, Ti = 7 and time
delay = 0.01s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
6.11 Gain and phase response due to changes in Kp . . . . . . . . . . . . . . . 58
6.12 Gain and phase response due to changes in Ti . . . . . . . . . . . . . . . 58
6.13 Frequency response for a load drop of 50% . . . . . . . . . . . . . . . . 59
6.14 Gate opening, responding to a drop in load demand of 50% . . . . . . . . 59
6.15 Comparing the frequency dynamics of two penstock models, when simu-
lating a drop in load of 10 % in island-mode. . . . . . . . . . . . . . . . . 60
6.16 Comparing the pressure dynamics in front of the turbine when simulating
a drop in load of 50 % in island-mode. . . . . . . . . . . . . . . . . . . . 61
6.17 Simulation of pressure and frequency, comparing the penstock models iso-
lated. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
6.18 Pressure pulsation effect for different penstock lengths . . . . . . . . . . 62
6.19 Frequency response of system start up and during a load rejection of 10%
during full load. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
6.20 Frequency regulation during a drop from 100% to 50% in load demand . . 65
6.21 Frequency regulation during a drop from 85% to 75% in load. . . . . . . 65
6.22 Power regulation in stiff grid mode with a drop of 0.01% in grid frequency 66
6.23 Power regulation in stiff grid mode with a increase of 0.01% in grid fre-
quency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
6.24 Water hammer simulations done in scilab and LVTrans . . . . . . . . . . 67
6.25 A close look at the first transient from figure 6.24. . . . . . . . . . . . . . 67
6.26 Simulation of a full system stop, with changes in the internal friction fac-
tors in the scilab model. . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
7.1 Picture of the rig. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
7.2 A scetch of the test ring at the laboratory . . . . . . . . . . . . . . . . . . 72
xv
7.3 Rig test result #1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
7.4 Rig test result #2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
7.5 Rig test result #3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
7.6 Comparison of all three test results from the rig. . . . . . . . . . . . . . . 75
7.7 Comparing test 1 with the scilab model . . . . . . . . . . . . . . . . . . 76
10.1 Cross section of the head race tunnel . . . . . . . . . . . . . . . . . . . . A-2
10.2 Cross section of the penstock . . . . . . . . . . . . . . . . . . . . . . . . A-3
10.3 Velocity triangles for a Francis turbine. [2] . . . . . . . . . . . . . . . . . B-1
10.4 Calculations of transfer coefficients in reference to induced torque by the

turbine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C-1
10.5 Calculations of transfer coefficients in reference to flow through the tur-

bine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C-2
10.6 Power regulation in a stiff network. . . . . . . . . . . . . . . . . . . . . . D-2
10.7 Frequency regulation in island mode. . . . . . . . . . . . . . . . . . . . . D-3
10.8 The hydraulic system used in the simulations. . . . . . . . . . . . . . . . D-4
10.9 Parameters used in the LVTrans simulations . . . . . . . . . . . . . . . . E-2
10.10Risk assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . F-1
xvi
Nomenclature
µ = Dynamic viscosity
kg
ρ = Mass density [ m 3]
ηbep = Efficiency at best point [-]
ω = Rotational speed [RPM]
= Signal deviation [pu]
a = Wave Velocity [ sm2 ]
A = Area [m2 ]
At = Proportionality factor [-]
As = Surge shaft area [m]
Ap = Penstock area
cf luid = Velocity of sound in a fluid [ m

s ]
D = d = Pipe diameter [m]
Dh = Hydraulic diameter
D = Machine damping constant [-]
e = typical surface roughness [m]

N
E = Young’s Modulus [ m 2]
xvii
f = frequency [Hz], may also be defined in [%]
fn = Nominal frequency [Hz] or [pu]
ft = Thickness of the pipe wall [m]
fD = Friction coefficient [-]
Re = Reynolds number
F = Force [N]
g = Acceleration of gravity [ sm2 ]
G = Guide vane opening [%]
hf = Friction losses [mH2O ]
Hr = Rated head [m]
kf = friction factior [-]

N
K = Bulk Modulus [ m2]
Kp = Proportionality factor
l = L = Length [m]
m = Mass [kg]
mt = Mechanical torque [pu]

1
M = Mannings number [ ms3 ]
n = Speed [%]
nr = Rated speed [%], may be defined as [rpm] (rotations per minute)
N bep = Best efficiency speed [rpm]
p = Whetted perimeter [m]
Pm = Mechanical power [M W ]
pr = Rated power [%], may be defined as [M W ]
pe = Grid power [%], may be defined as [M W ]
pt = Transient pressure [mH2O ]
xviii
Pd = Power demand
Pg = Power generation
q = Flow [pu]
3
Q = Flow [ ms ]
3
Q1 = Flow in the head race tunnel [ ms ]
3
Q2 = Flow into the upper surge shaft [ ms ]
3
Q3 = Flow in the penstock [ ms ]
3
Qbep = Best efficiency point flow [ ms ]
3
Qr = Rated flow [ ms ]
r = Radius [m]
r1 = Inlet runner radius [m]
r2 = Outlet runner radius [m]
rh = hydraulic resistance [-]
Rq = Constant for RPT [-]
Rm = Angular velocity damping constant [-]
Rd = Design constant koeffisient [-]
s = Laplace operator
t = time [s]
T bep = Best efficiency torque [Nm]
Ts = Surge shaft filling time [s]
Te = Wave travel time
Tw = Water time constant [s]
Tr = Pressure reflection time [s]
Tc = Closure time [s]
xix
T d = Time derivative constant [s]
T i = Time integrate constant [s]
u = output signal
v = Speed [ m
s ]
z = Upper surge shaft water level [m]
z0 = Head loss in the head race tunnel [m]
Zp = Hydraulic surge impedance of the penstock [-]
xx
Chapter 1
Introduction
1.1 Motivation
To achieve a satisfying power system with a high security of supply, robustness and stabil-
ity, power generation with good regulation and governing is necessary. Implemented in the
production of active and reactive energy, governor systems capable of regulating voltage
and frequency is a prerequisite to achieving the goals for stability and safety of supply [8].
Frequency variations on the grid are a direct consequence of a difference in energy demand
and power generation. A common analogy is an example of riding a tandem bike. When
the road suddenly gets steeper somebody must push harder, or the speed will decrease.
The same counts for power systems, as load demand increases, the power generation must
follow, or else the frequency will decrease. Depending on the size and type of system, all
hydropower plants must have regulation where power and speed can be regulated to some
extent.
In Norway aggregates larger than 10 MVA must have the opportunity to do frequency reg-
ulation, this is also applicable for smaller systems, where it is possible to do so. System’s
that have reserves to do frequency regulations must regulate within a 1% mark. So for
power systems that are being supplied with a base frequency of 50Hz, the generation stay
within a range of 49.5 to 50.5 Hz. This counts as 1% droop, or higher dependent on the
energy reserves [20].
Due to these strict regulations, simulation of hydropower plants is a valuable and money
saving process to anticipate the dynamics of the system, compared to real life experiment.
When setting up a new power plant or making smaller parameter changes, predictability is
important and can be achieved with system simulations.
1
Chapter 1. Introduction
1.2 Thesis Description
The thesis is based on the project work done in the first semester at the last year at NTNU.
During the project work a simple model of a hydropower plant was made and a literature
study of modelling turbines and hydraulic systems. The master thesis is based on the same
system but has been heavily modified.
Seven primary objectives were given in the thesis:
1. Add a surge shaft and an inlet tunnel to the existing model.
2. Verify the surge shaft with experiments in the laboratory.
3. Research the opportunities to add an elastic penstock model to the system.
4. Theoretical deduction of the partial derivates for use in the xcos model, based on
Torbjørn Nielsen’s turbine model.
5. Simulate the system with both an isolated and stiff grid.
6. Look for possibilities to do a frequency response analysis.
7. A short paper shall be written and presented at the 7th International Symposium on
Current Research in Hydraulic Turbines (CRHT-VII) at Katmandu University, April
2017.
8. Write a report.
The plant is based on a Francis turbine with a rated power output of 94MW. It is designed
3
for a gross head of 200m and a flow rate of 50 ms . The power plant is modelled in block
diagrams based on mathematical models and transfer functions. The turbine is based on a
linearized model, which gives an accurate description of the turbine behaviour for a given
operating point. The plant is controlled by a PI regulator providing frequency stability and
eliminates the deviation between process and reference values.
By increasing the level of detail in the model, better simulation results are expected. A
more detailed system gives better flexibility which provides more simulation possibilities
and situations. Four systems have been made in the thesis. Three in island mode with a
single machine, single load, where the first one is build up with a block having a standard
input configuration directly in the blocks. The second one is the one called analytic model
where the equations are inserted in the different blocks, and the input values are defined
through a context file. The third model is for water hammer simulation and is based on
the analytic model but with relays forcing a fast closure of the guide vanes and changing
the friction characteristics of the system. The fourth model is the stiff grid operation with
turbine power regulation, and with frequency as an input variable.
2
1.3 System Introduction
All the simulation models share the same turbine model and hydraulic system. In this
thesis, all the systems will be evaluated up against LVTrans, and the surge shaft model
will be tested towards a test rig at the water power laboratory.
1.3 System Introduction
The plant design is based on a typical Norwegian layout represented in a block diagram
seen in figure 1.1.
Figure 1.1: General layout of the Norwegian island operated turbine.
The block diagrams consist of a frequency reference value, sending the signal to a sum-
mation block. The summation block subtracts the process frequency value which becomes
and sends it into the PID regulator. The regulator then transfers the signal into what is
then represented as the guide vane opening ∆G. Through turbine dynamics, the amount of
flow ∆Q is then given by the guide vane opening and sent into the hydraulic system. The
hydraulic system is then transferring the signal into pressure head, ∆H. This pressure is
then processed through turbine dynamics one last time giving turbine power as an output.
The speed is then represented after the torque (Tt )/power (Pt ) signal has gone through
the generator dynamics also called electromechanical system. The torque provided to the
generator is turbine power subtracted with power demands (PL ) giving ∆P or ∆T . Since
the system is in per unit values, the values can easily get converted by multiplying with
the base value of either torque or power. The same counts for the speed which can be
multiplied with both the base value of the grid frequency (50Hz) or the rated generator
speed (333,33RPM).
For simulations in a stiff grid environment, the model has to change some of the layouts
as seen in figure 1.2.
3
Chapter 1. Introduction
Figure 1.2: Simulation set up for stiff grid mode.
In the figure, 1.2 much of the same signal processing occurs but the regulator loop is power
dependant, and the governor regulates towards a power reference and droop characteristics.
Changing the value of grid frequency provides a change in the system stability. The system
will then automatically regulate the power output from the turbine to the value decided by
the droop.
4
Chapter 2
Previous Work
2.1 Project Work
During the fall of 2016 the project work called ”comparison of block diagrams for hy-
dropower plants” was done. It was a literature study combined with modelling and sim-
ulation of an elementary hydropower plant. The literature study gave insight and knowl-
edge of system regulation, modelling of different components in a hydropower plant from
a mathematical approach.
The model created in the project work consisted of a linear turbine model regulated by
a simple PI regulator, a simple rigid penstock modelled as a static water column, and
generator. The system can be seen in figure 2.1
Figure 2.1: Block diagram made in the project work fall 2016.
The model was made to perform simulations in island mode with frequency regulation
during load acceptance and rejection. Simulations gave reasonably good results compared
to simulations conducted in LVTrans, but there were some deviations between them. Lack
5
Chapter 2. Previous Work
of system dynamics and simulation flexibility made it desirable to develop the system
further.
By giving the system more scenarios for simulation such as power regulation during fre-
quency variation in a stiff network environment, and create systems for transient pressure
simulations increases the usability.
2.2 Paper Presented at the 7th International Symposium

on Current Research in Hydraulic Turbines
During the last semester at NTNU, the students at the water power laboratory attended
at the international symposium on current research in hydraulic turbines at Kathmandu
University. In that occasion a presentation and a paper where to be delivered, presenting
the thesis and the work to come.
In Appendix G, the paper provided to Kathmandu University is found. It was written in an

early stage of the thesis and lacks extensively of system model and simulations.
Some of the larger improvements and implementations that have been done post paper
delivery are:
• New turbine transfer coefficients are calculated.
• New friction coefficients

• New PI settings for the regulator
• Draft tube and lower surge shaft are implemented
• The ”Analytic” model has been made
The presentation was performed during the trip to Nepal and was successfully conducted
at the conference.
6
Chapter 3
Theory
This section includes different mathematical and physical definitions directly linked to
understanding the modelling, and the various physical situations that could occur during
simulation of a hydropower system. At first, some general equations for fluid mechanics
as well as energy, and mass behaviour will be shown and is found in section 3.1. A method
for calculating forces in transient states such as water hammer phenomenon are included in
3.2. Later in the chapter, different approaches for calculating friction, and the various parts
of the hydropower plant will be presented, found in section 3.3 to 3.5. Most of the theory
found in this chapter are found in books such as: ”Modelling and controlling Hydropower
plants” [12], ”Vassdragsteknikk II” [5], and the paper by J. Riera [14].
7
Chapter 3. Theory
3.1 General Equations
A string of water must comply with the equation of motion and the equation of continuity.
The continuity equation constrains the law of mass flow inside a control volume. For a
given amount of volume flow getting into the system. The same amount must come out
plus the eventual rest inside the volume. The equation of continuity is defined mathemati-
cally as:
∂H a2 ∂v
+ · =0 (3.1)
∂t g ∂x
Where a is the velocity of sound in water, approximately around 1200 m s . This velocity
is defined in equation 3.6, and it is dependent on the thickness of the pipe walls, pipe
diameter, the mass density of water, and the water compressibility module.
The equation of motion describes the forces acting on the control volume. All the forces
on it must be zero wich leads to the following equation:
∂H ∂v v|v|
g· + +λ· =0 (3.2)
∂t ∂t 2D
8
3.2 Transient Fudamentals
The hydraulic transients are recognised as the change in pressure or pressure fluctuations
in a fluid caused by a change in flow conditions. Pressure waves travel with the velocity
of sound which is given by:
s
K
cf luid = (3.3)
ρ
Where ρ is the density and K is the bulk modulus of the fluid given by:
dP N
K =ρ· → KH2O = 2.2 · 109 2 (3.4)
dV m
There are many different events that lead to a change in the mean flow of the system,
and some of the more usual are: system start-up and shut down, changes in frequency,
vibrations of a runner, and different sorts of dam failure.
3.2.1 Pressure wave velocity

s
K
a= (3.5)
ρ(1 + DK
ft E )
Where D is the internal penstock diameter, ft is the wall thickness of the penstock walls,
N
and E is Young’s modulus of pipe wall material. For steel this value is 2.22 · 1011 m 2 . For
a perfectly rigid penstock, assuming E to be infinite simplifies the equation to 3.3:
s
K
a= = cf luid (3.6)
ρ
9
Chapter 3. Theory
3.2.2 Wave reflection time
After a rapid closure of the turbine valves, a high pressure occurs in front of the turbine
as the water masses is forced to decelerate, (F=ma).This overpressure will immediately
propagate up towards the nearest water surface like a wave. The amount of time in seconds
for the wave to reach the open surface is given by:
L
Te = (3.7)
a
Where L is the length of the penstock [m] and a is the wave speed [ m s ]. In figure 3.1 the
pressure propagation in two cycles are shown. note that in the figure, the wave velocity is
given by C and not a.
Figure 3.1: A representation of pressure propagation in a conduit.[5]
Pressure reflection time is given by the time for the wave to reach surface and back again to
the closing valve. This naturally becomes that Tr = 2 · Te . To avoid damage to the turbine
and the rest of the system, the closure time (Tc ) for the valve should always be larger than
Tr . The closing time is seen as the simplest ways to avoid high-pressure transients, but
again it’s delimiting the power plans ability to regulate fast.
10
3.2.3 Mass Forces
Mass forces for the pressure transients in front of the turbine is essential to know when
designing a power system. Pressure transients during rapidly closed valves could harm
the pipelines and the penstock. In this subsection, the equation for the change in pressure
given in mWC is deduced.
Figure 3.2: Description of mass forces during transient state.
Using pulse rate to the system:
F · dt = d(mv) = m · dv + v · dm (3.8)
Knowing that F = dp·A and the changes in mass for the control volume is dm = ρ·A·c·dt
gives:
dp = ρ · c · dv + ρ · v 2 (3.9)
At an instantaneous change in mean flow gives dv = v. Also in practical senarios wave

velocity, c is in magnitude of 100 → 1000 m
s , wich results in v << c
Giving the simpler form:
dp = ρ · c · v (3.10)
and in the form of height [m]:
c · ∆v
dh = (3.11)
g
This equation applies for instantaneously change in the fluid velocity. When using closure
times to the turbine valves a modification to equation 3.11 is given by:
11
Chapter 3. Theory
c∆v Tr ∆Q L
∆h = · =2· · (3.12)
g Tc Tc gA
12
3.3 Friction Losses
3.3 Friction Losses
Losses due to friction inside the conduit wall can be found by using Darcy Weisbach’s
empirical equation. There are other factors that lead to head losses in a hydro power plant,
but in this thesis, only friction due to sheer forces inside the conduit is taken account.
L V |v|
hf = fD · · (3.13)
D 2g
Where: hf = head loss, fD = Darcy weisbach’s friction factor, v = the average velocity of
the fluid, D = hydraulic diameter, L = the length of the pipe.
For laminar flow with Re < 2320 in a circular pipe, Darcy Weisbach’s friction coefficient
fD is defines as follows:
64
fD = (3.14)
Re
where Reynolds number is defined by:
ρ · v · Dh
Re = (3.15)
µ
ρ is the mass density of the fluid, v is the speed, L is the characteristic length and µ is the
dynamic viscosity.
For calculating friction during turbulent flow, the Colebrook-White Equation can be used.
An implicit equation that requires numerical solutions to be solved. The equation is de-
fined as:
1 e 9.35
√ = 1.14 − 2log10 · ( + √ ) (3.16)
f d Re f
Valid for flow with Re > 4000. de is the relative roughness of the material, where e is
the typical surface roughness, and d is the diameter. In table 3.1 one can find values for
roughness over different materials.
13
Chapter 3. Theory
Table 3.1: Typical surface roughness on different materials. [15]
Material Nature of Material Roughness [mm]

Steel Drawn, new 0.02 - 0.1
- Welded, new 0.05 - 0.1
- galvanized, new 0.15
- Used, cleand 0.15 - 0.2
- Lightly corroded 0.1 - 0.4
- Severly corroded 0.4 - 3
- Light scaling 1 - 1.5
- Heavily scaling 1.5 - 4
- Bitumed coated 0.05
Cast - Iron pipe New 0.25 - 1
- Corroded 1-2
- With scaling 1-4
Concrete pipe Smooth finish 0.3 - 1
- Rough 1-3
Sheet steel Smooth 0.07
Glass,lead,copper,brass - 0.0001 - 0.0015
When the relative roughness and Reynold’s number are known the friction factor can also
be found in a Moody diagram, presented in figure 3.3.
Figure 3.3: The moody diagram, used to obtain friction factor f [10]
14
3.3 Friction Losses
Friction losses can also be found using Manning’s formula which is often used to calculate
friction losses in blasted shafts [5].
L · v|v|
hf = 4 (3.17)
M 2 · Rh3
Where L = tunnel length, v = fluid velocity, M = Mannings number, Rh = hydraulic

radius. Manning’s number varies with tunnel area and the roughness of the tunnel surface.
For blasted shafts, Maning’s number is around 34, but are assumed lower in newer shafts
as the roughness has increased over the years. For concrete Manning’s number are found
to be around 80. The hydraulic radius can be known from the formula:
4A
Dh = (3.18)
p
Where A is the area of the duct, and p is the wetted perimeter of the cross-section. Hy-
draulic radius is then defined as the ratio of area and wetted perimeter:
A
Rh = (3.19)
p
Cross section of the inlet tunnel can be seen in figure 3.4
Figure 3.4: Cross section of the inlet tunnel
For the given cross section the wet perimeter is found by:
p = (4 + π)r (3.20)
This gives leads to a general equation for the tunnelshape in figure3.4 :
15
Chapter 3. Theory
r
8·A
Dh = (3.21)
4+π
and for the radius:
A
Rh = (3.22)
(4 + π)r
For typical hydro power plants, the friction coefficient is set to 0.05 for the head race tunnel
and 0.01 for the penstock [21]. These are the values that are used in the power system.
Friction factors for the test rig are calculated using Colebrook-White Equation 7.1. This is
due to the requirements of preciseness for the friction loss when comparing simulation to
real life tests.
16
3.4 Surge shaft
3.4 Surge shaft
Figure 3.5: Cross section sketch of the conduit system at the susrge shaft and the forbay.
During steady state operation Z0 represents height losses due to frictional forces in the
headrace tunnel. If a change in the fluid mass flow occurs the system equilibrium ceases,
resulting in oscillations for Z0 .
3.4.1 Surge shaft Oscillations
Since the oscillations in the surge shaft appear to be slow, the water within the surge
shaft considers being inelastic. This simplifies the equations regarding mass oscillations
of water in the surge shaft. The force which creates an unbalanced fluid level in a U-pipe
is defined as:
F = ρgA · 2z (3.23)
Amount of mass to be accelerated:
m = ρAl (3.24)
Where l = fluid column length [m]. For a ideal fluid with and neglecting friction we have:
dv
F = ma = m (3.25)
dt
Where v = fluid flow speed [ m
s ]. Inserting 10.6 into 10.7:
dv
ρgA · 2z = ρAl (3.26)
dt
17
Chapter 3. Theory
Introducing fluctuations in away from the level of reference, where:
dz
v=− (3.27)
dt
This gives the harmonic swing equation, given in a differential form.:
d2 z 2g
+ z=0 (3.28)
dt2 l
The swing equation has the following solution:
z = k · cos(ωt + φ) (3.29)
The angular velocity are described as inversely proportional to the swing
time constant T :
r
2g 2π
ω= = (3.30)
l T
Identical to a physical pendulum with length l/2

s
l
T = 2π (3.31)
2g
3.4.2 Surge Shaft Transfer Function
The surge tank can be derived from the equation of continuity, and in this case, the losses
are neglected. The equation for the changes in the surge shaft is given below 3.32. The
changes of water level inside the surge shaft are dependent on the mass flow in the pipe
system:
AHr dz
· = Q2 = Q1 − Q3 (3.32)
Qr dt
With the equation linearized and Laplace transformed gives the following equation:
18
3.4 Surge shaft
1
s·z = (Q1 − Q3) · Qr (3.33)
As
where zs is the head level inside the surge shaft and Q1 − Q3 = dQ gives the transfer
function for a change in the head due to a change in flow:
dH Qr 1
= (3.34)
dQ h · As s
Surge shaft filling time is there fore expressed as:
AHr
Ts = (3.35)
Qr
Giving the final transfer function for the surge shaft component:
dH 1
= (3.36)
dQ Ts · s
19
Chapter 3. Theory
3.5 Penstock and Tunnels
In this section, the principles for modelling the different penstock types are given. There
are used three different methods in this thesis to model both penstocks and tunnels. The
first pipe or tunnel is described as the head race tunnel, and it utilizes the simplified pen-
stock model for a static pipe with no water elasticity.
Later in the hydraulic system comes the penstock. For the hydro power system simulations,
there are only used two different types. The dynamic pen stock model and the penstock
model from the electrical equivalent circuit. At last the draft tube is modelled with the
same model as the head race. This was chosen due to to its short length and slow speed.
For the surge shaft verification model, there are only used simplified penstock model due
to the short pipe length of 11m.
3.5.1 Dynamic penstock
The model shown in this section is described in literature such as Modelling and Control-
ling Hydropower plants, [12]. The transfer function for a penstock given as a dynamic
system with compressible water effect and dynamic pipe walls are provided by the follow-
ing equation:
∆H(s)
= Zp · tanh(s · Te + hf ) (3.37)
∆Q(s)
Where the hydraulic surge impedance of the penstock, Zp is given by:
Tw
Zp = (3.38)
Te
Where Tw is the water starting time in the penstock. This parameter describes the time for
the water masses to accelerate up from Q = 0 to Q = Qr [22]. The value is given by:
Lp Qr
Tw = · (3.39)
Ap · g H r
3.5.2 Approximation of Tanh
The model is deduced from the equations of continuity and motion which lead to the
following relationship:
20
The hyperbolic tangent function can be written as:
ex − e−x sinh(x)
= = tanh(x) (3.40)
ex + e−x cosh(x)
Due to complications when solving for a irrational terms in transfer functions, the tanh
function is rather used. By inserting x = Te · s gives the finite approximation for tanh
function with the use of Maclaurin series [14]:
∞
" #
Y s · Te 2
s · Te 1+( )
n=1
n·π
tanh(Te · s) = ∞ " # (3.41)
Y 2s · Te 2
1+( )
n=1
(2n − 1)π
For approximations n=0 ,1 and 2, tanh function becomes:
tanh(Te · s) |n=0 ≈ s · Te (3.42)
s · Te · (1 + ( s·Te 2
π ) )
tanh(Te · s) |n=1 ≈ (3.43)
(1 + ( 2·s·T e 2
π ) )
s · Te · (1 + ( s·Te 2 s·Te 2
π ) ) · (1 + ( 2·π ) )
tanh(Te · s) |n=2 ≈ (3.44)
(1 + ( 2·s·Te 2 2·s·Te 2
π ) ) · (1 + ( 3·π ) )
This approximation is necessary when modelling the system in block diagrams. Due to
solver issues and iteration problems when directly inserted as a simple timedelay in form
of a trignometric function ”tanh” or as a time delay block represented as e−timedelay[s] .
3.5.3 RLC Penstock Model
The RLC model also called ”Penstock Model From Electrical Equivalent Circuit” are using
electrical analogy to represent a dynamic hydraulic line. The constants and simulation
results are only valid during water hammer simulation and under severe pressure changes
within the system.
The electric RLC equivalent circuit for a hydraulic pipeline can be described using the
momentum and mass conservation equations. The equations are in advanced linearized
around an operating point at Hr and Qr . The model is used in literature such as [6] and
21
Chapter 3. Theory
[18].
∂∆H rh 1 ∂∆Q
= · ∆Q + · (3.45)
∂x A g·A ∂t
∂∆Q Ag ∂∆H
= 2 · (3.46)
∂x a ∂t
Where A = pipe cross section, rh = hydraulic resistance, g= acceleration of gravty, D =

pipe diameter. The equation for pressure wave velocity, ”a” can be found under section
3.2.1. Hydraulic resistance is define as:
rh = A · 2Kr · Qr (3.47)
and
∂H0
Kr = (3.48)
∂x · Q2r
A constant level of head and dx = L gives:
H=
Kr = (3.49)
L · Qr
The electrical equivalent circuit can be ssen in figure 3.6.
Figure 3.6: A penstock with lenght a L can be represented with this equivalent circuit schematic
The components are defined as:
Introducing the Thevenin equivalent impedance Z2 for the circuit in figure 3.2 gives the
impedence seen from the turbine. This value is defined by the transferfunction 3.52.
22
Table 3.2: RLC component description
Component Definition Formula

rh
Resistor ZR AL
AgL
Capacitor ZC 2a2
L
Inductor ZL gA
∆H2
(s) = −Z2 (s) (3.50)
∆Q2
Where Z2 is can be found form parallel component theory:
ZC (s) · (ZL + ZR )(s)

Z2 (s) = (3.51)
ZC (s) + (ZL + ZR )(s)
This gives:
s · (2 · L · a2 ) + 2 · a2 · rh · L · g
Z2 (s) = (3.52)
s2 · (L2 Ag) + s · rh · A(g · L)2 + 2a2 · g · A
3.5.4 Static Tunnel Model
The static model is based on the dynamic penstock model found in section 3.5.1, seen in
equation 5.1
H(S) Tw
=− · tanh(Te · S + hf ) (3.53)
Q(S) Te
When assuming incompressible fluid with the conduit system defined as rigid and short,
the time-traveling effect is negligible and elasticity is insignificant. The equation 5.1 is
reduced to [12]:
H(S)
= −Tw · s − hf (3.54)
Q(S)
23
Chapter 3. Theory
24
Chapter 4
Modelling and Simulation Tools
In this thesis, there are used two different computer programs to generate and simulate the
hydropower system. At first, a MATLAB clone called Scilab is used to create the block
diagram of the system. Then a program created in LabVIEW called LVTrans is used to
verify and compare the block diagram. Another programme from now called ”hill chart
TN” is used to deduce turbine transfer coefficients, implemented in the Francis, turbine
model. Other programs such as Excel and MATLAB are utilised during the thesis to
handle simulation data and generate graphs but are not included in this chapter. The thesis
is written in an online latex client called sharelatex.
25
Chapter 4. Modelling and Simulation Tools
4.1 Scilab
To create block diagrams, a program handling this feature is used. MATLAB were consid-
ered but Scilab is preferred due to its implementation in the industry and that it interfaces
with LabVIEW. Xcos, a built-in feature is used to create the system, which is suitable for
generating block diagrams and numeric system simulation.
The program interface is similar to Simulink, but it is free to use, open source software
which makes it easy for programmers to access source codes and further develop the pro-
gram. Figure 4.1 shows the main interface of scilab to the right. Top left is Xcos program-
ming window. Blocks can be added from the palette browser (bottom left) by drag and
drop. [19]
Figure 4.1: Interface of scilab
26
4.2 LVTrans
4.2 LVTrans
LVTrans is a program made in the LabVIEW environment, created to give high usability
when it comes to modelling and simulating hydropower plants. The power system is built
up of pre-made blocks representing pipes, draft tubes, turbines, PID regulators and valves.
It is possible to run the simulation in real time and make adjustments at the same time,
which makes the program quite versatile.
In this thesis, the program will be used to compare the model made in Scilab. Due to
LVTrans preciseness, a fair comparison and model evaluation is achievable. Figure 4.2
shows the interface and how the system is build up from blocks. The left window shows
the simulation settings, such as time steps and simulation speed. The window to the right
is the programming schematics of the power plant. A closer view of the system can be
found in section 10
Figure 4.2: Interface of LVTrans
27
Chapter 4. Modelling and Simulation Tools
28
Chapter 5
Hydropower Modelling
Figure 5.1: Cross section of the system.
The modelled system is a self-designed power plant based on a typical Norwegian layout
and can be seen in figure 5.1. Starting from the left, a fore-bay giving a constant head to
the system which leads to the head-race tunnel. There are not taken account for any losses
at the inlet gate.
The head race tunnel is given the length L1 , and leads to the penstock and the surge shaft,
29
Chapter 5. Hydropower Modelling
with respective lengths L2 and Ls . At the end of the penstock lies the turbine, placed
below the lower reservoir to avoid air getting sucked into the system. Pipe areas for head
race tunnel, penstock the surge shaft and the draft tube are A1 , A2 , As and A3 . The draft
tube is in this case simplified and not designed or specially designed. The cross section is
chosen out from try and failure.
Frictional forces are taken account for in the tunnel penstock, and draft tube and they are
calculated bu the use of Darcy-Weisbach friction equation. Calculations for the tunnel
system are found in Appendix A and is summarised in table 5.1.
Table 5.1: Hydraulic system parameter values.
Part Description Value Unit

L1 Head-race length 2000 m
L2 Penstock length 250 m
L3 Draft tube length 25 m
Dh1 Hydraulic diameter L1 5.292 m
Dhs1 Hydraulic diameter, surge shaft 7.137 m
Dhs2 Hydraulic diameter, Lower surge shaft 3 m
A1 Head-race tunnel area 25 m2
A2 Penstock area 12 m2
A3 Draft tube area 12 m2
Ath1 Thoma area 26.48 m2
Ath2 Thoma area, lower surge shaft 7.08 m2
As 1 Surge shaft area 40 m2
As 2 Lower Surge shaft area 11 m2
fD1 Head race friction factor 0.05 -
fD2 Penstock friction factor 0.01 -
fD3 Draft tube friction factor 0.05 -
hf 1 Head race loss 3.91 m
hf 2 Penstock loss 0.57 m
hf 2 Draft tube loss 0.28 m
Hr Rated head 200 m
m3
Qr Rated flow 50 s
Tw1 Water mass response time L1 2.04 s
Twtotal Water mass response time Ltot 2.62 s
Ts1 Surge shaft time constants(higher) 160 s
Ts2 Surge shaft time constants(lower) 28 s
30
5.1 Modeling scenarios
5.1 Modeling scenarios
In figure 5.2, a general representation of the hydro power system with frequency regulation
in a isolated grid with one machine is shown. The governor in this master thesis is in both
systems (frequency and power regulated plants) based on a single PID regulator but blocks
for pilot servo are also included in some simulations scenario. This will be specified during
the analysis part. The system layout can be found in [13].
Figure 5.2: General layout for the system in grid mode with frequency regulation.
The system connected to a stiff grid in power regulation mode can be seen in figure 5.3.
Inputs such as power reference and grid frequency are given in this model. The feedback
loop is taken from generated power from the system to a summation point. From there the
deviation signal is controlled by a permanent droop gain given by the size and specifica-
tions of the plant. The permanent droop controls the given amount of change in machine
power output due to a certain amount of change in frequency.
Figure 5.3: General layout for the system in stiff network mode with power regulation.
31
5.2 Hydraulic System
The hydraulic system is divided into three different parts. The head race tunnel that leads
water from the reservoir or fore bay to the penstock. The penstock is the last piping
between the head race tunnel and the turbine spiral casing. The surge shaft is included in
the model since mot of the hydropower plant has one implemented. The upper surge shaft
lowers the wave pressure reflection time which could lead to better stability, and regulation
performance. This is important due to flexibility and maintaining good production quality.
System requirements and suggestions due to plant regulation and stability can be found in
section 6.2.
As seen in figure 5.4 the dynamics are dependent of all the three transfer functions. The
functions are deduced from the the equation of continuity and equation of momentum.
The hydraulic system are designed to give an realistic representation of the water column
transients from the fore bay to the turbine via an surge shaft. Typical outputs from the
hydraulic system that are useful designing a hydro power plant are the forces applied in
transient state due to water-hammer effect, such as pressure pulsations. The head pressure
dynamics also directly impacts the torque delivered by the turbine which influences its
speed. The values inserted in the models are calculated in Appendix A. A full overview of
the scilab hydraulic system is found in appendix D.
Figure 5.4: General layout for the hydraulic system used in the simulations with scilab.
5.2.1 Penstock
The penstock model used in the simulations is called the hyperbolic model. Due to the
surge shaft and head race tunnel layout, the RLC model could not be integrated into the
32
hydraulic system. The model is however included as a part of the theory since it can be
used in simulations isolated from surge shaft and tunnels.
5.2.1.1 Hyperbolic Model
The transfer function representing the elastic water column model is given in equation 5.1
and is found in Hernandez, [12]:
H(S) Tw
=− · tanh(Te · S + hf ) (5.1)
Q(S) Te
where elasticity time constant is given by assuming wave velocity during transients are
1388 m
s . This value is a little high for elastic wave simulation and should be corrected
during later simulations. A suggestion for a new value of Te is given in Appendix A-6
l 250m
Te = = = 0.18s (5.2)
a 1388 m
s
During the simulations Due to complications starting the iterations when using tanh func-
tion directly in xcos the series expansion of tanh is used. Expanding to n = 2 is done to
get a higher order function, giving a realistic penstock behaviour:
s · Te · (1 + ( s·Te 2 s·Te 2
π ) ) · (1 + ( 2·π ) )
tanh(Te · s) |n=2 ≈ (5.3)
(1 + ( 2(s·T
π
e) 2
) ) · (1 + ( 2(s·Te) 2
3·π ) )
Inserting values for the penstock seen in table 5.1 gives the total block diagram for the
penstock seen in figure 5.5
Figure 5.5: Block diagram from scilab describing the penstock used in the simulations.
33
5.2.1.2 Penstock Model From Electrical Equivalent Circuit
Transfer function for the electrical equivalent model is given by equation 5.4 and is de-
duced in section 3.5.3.
∆H2
(s) = −Z2 (s) (5.4)
∆Q2
Where Z2 is can be found form parallel component theory:
ZC (s) · (ZL + ZR )(s)

Z2 (s) = (5.5)
ZC (s) + (ZL + ZR )(s)
This gives:
s · (2 · L · a2 ) + 2 · a2 · rh · L · g
Z2 (s) = (5.6)
s2 · (L2 Ag) + s · rh · A(g · L)2 + 2a2 · g · A
Inserting the respective system values from table 5.1 gives the following function used in
simulations:
Figure 5.6: Transfer function block representing the RLC penstock model
5.2.2 Head Race Tunnel and Draft Tube
When assuming incompressible fluid and a rigid pipe, the time-traveling effect is negligi-
ble, and elasticity is insignificant [12]. The equation 5.1 is reduced to:
H(S)
= −Tw · s − hf (5.7)
Q(S)
Which gives the transfer function used in the model, where the output is needed to be a
fluid flow:
34
Q(S) 1
=− (5.8)
H(S) Tw · s + hf
In figure 5.7 the block used in the simulations are found values inserted.
Figure 5.7: Transfer function block from scilab describing the head-race tunnel used in the simula-
tions.
The same model is used for the simplified draft tube. The transfer function block can be
seen in figure 5.8
Figure 5.8: Transfer function block from scilab describing the draft tube tunnel used in the simula-
tions.
35
5.3 Surge Shaft
The transfer function describing the surge shaft water level height is deduced in subsection
3.4.2. The surge shaft filling time is given by:
As · H0
Ts = (5.9)
Q0
And the surge shaft transfer function is:
dH 1
= (5.10)
dQ Ts · s
The block used in simulations is found in figure 5.9. The values are presented in Table 5.1
and are based on the calculations presented in Appendix A.
Figure 5.9: Transfer function block from scilab describing the surge shaft used in the simulations.
36
5.4 Surgeshaft validation model
5.4 Surgeshaft validation model
This model is including an input representing a reservoir providing a particular flow to the
system. It also includes a pipe between the tank and the surge shaft and the surge shaft
itself. The model can be seen in figure 5.10
Figure 5.10: Block diagram for simulation of the test rigg at the laboratory.
During the simulations the head race tunnel from section 5.2.2 is used. A step function is
set from one to zero [pu] in the reservoir. The pressure or level of head will then oscillate
due to the system dynamics from the head race tunnel and the surge shaft. The signal can
then be seen in the scope. Adding the head pressure gives the scope output the correct
pressure to total pressure. There is no need for multiplying inside the the lope with the
head base since the head is 1m. This means that the change in head also is given in [m], as
well as [pu]. In figure 5.11 the scilab model from the surge shaft validation can be seen:
Figure 5.11: The model for surge shaft validation with scilab
37
5.5 Turbine
The Francis turbine model used in the simulations is based on the linearized turbine model.
By linearizing the turbine characteristics around an operating point gives a detailed rep-
resentation of the turbine model, but its accuracy is constricted around that specific point.
Equation 5.13 and 5.14 is found and used in literature such as: [12], [4] and [7]. The
dynamic characteristics for a Francis hydro-turbine can be described as:
q = q(h, n, G) (5.11)
mt = mt (h, n, G) (5.12)
Where m is the mechanical torque and q is the water flow through the turbine. h,n and
G is respectively turbine head, turbine speed and the guide-vane opening. By utilizing
first order Taylor expansion of Eq: 5.11 and 5.12 the dynamic expressions of the Francis
turbine is obtained [12]:
∂q ∂q ∂q
∆q = ∆h + ∆n + ∆G (5.13)
∂h ∂n ∂G
∂mt ∂mt ∂mt

∆mt = ∆h + ∆n + ∆G (5.14)
∂h ∂n ∂G
Turbine transfer coefficients are described as partial deviates of flow and torque with re-
spect to head, speed and guide vane position, where all variables are in per unit. For
optimal performance the transfer coefficients should be deduced at the specific operating
point of the turbine. In this thesis the operating point is set to BEP, and it is deduced by
a program based on Euler’s turbine equations. The program is programmed in labVIEW
and is made by Bjørnar Svingen. The layout can be seen in Figure 5.12.
38
5.5 Turbine
It is programmed using the modified Euler turbine equations in a dimensionless form,

deduced in Torbjørn Nilsens doctor thesis: ”Transient Characteristics of high head Francis
turbines” [1]. The turbine equations includes internal losses, and is represented in the
equations below:
dq q
Twt · = h − ( )2 − σ(e
ω 2 − 1) − Rq (q − qc )2 (5.15)
dt κ
and:
de
ω ∆h
Ta e s − ψn ω
= q(m e ) · (1 − e 2 − ηg
) − R3 ω (5.16)
dt h
where:
∆h = R1 q 2 + R2 (q − qc )2 (5.17)
These equations are deduced in his thesis. The different parameters are: Twt = turbine
water constant, dq = change in flow, dt = change in time, h = head, κ = guidevane opening
degree, σ = self governing parameter (droop), ω
e = Base runner speed, Rq = self governing
parameter, me s = specific dimensionless torque, R1 , R2 , R3 are lossparameters, and ψ =
machine constant
Figure 5.12: Picture of interface of the programe used to linearize the turbine characteristics.
The program automatically calculates changes in all the turbine parameters as incremental
changes is applied to flow, speed and guidevane opening. This allows the user to calculate
39
all the turbine transfer coefficients in a simple manner by the use of output variable from
the program. The calculations and and outputs can be found in appendix C. The transfer
coefficients where calculated to be:
∂mt ∂mt ∂mt ∂q ∂q ∂q

= 1.01 = 0.503 = −0.97 = 1.005 = 0.5 = −0.966
∂G ∂h ∂n ∂G ∂h ∂n
In figure 5.13 the model is given in a block diagram from. Calculations done for the
particular turbine used in the thesis can be found in Appendix B, and table 5.2 summarise
up the different parameters of the turbine.
Figure 5.13: Transfer function block from scilab describing the surge shaft used in the simulations.
Table 5.2: Turbine spesifications
Unit Description Value

3
Qbep Best efficiency point flow [ ms ] 50
Hbep Best efficiency head [m] 200
Nbep Best efficiency speed [rpm] 333.33
Tbep Best efficiency torque [Nm] 2700694
ηbep Efficiency at best point [-] 0.96
a1bep Guide vane angle [◦ ] 15
b1bep Angle of the rotor blade, (absolutt angle)[◦ ] 67.7
r1 Inlet runner radius [m] 1.311
r2 Outlet runner radius [m] 1.146
Rq Constant for RPT [-] 0
Rm Angular velocity damping constant [-] 0.04
Rd Design constant koeffisient [-] 0.03
Rf - 0.015
40
5.6 PID - Regulator
5.6 PID - Regulator
The PID regulator in this thesis regulates the active power production in response to the
variation of power demand, hence frequency in the grid, see equation 5.18. The frequency
and load variation has a relationship as follow [8]
∆f ∆Pg
= −ρ · (5.18)
fn Pd
Where ∆f is the change in frequency, fn is the nominal frequency in the grid, ∆Pg is the
change in active generated power, while Pd is power demand, and ρ is the static droop of
the turbine governor characteristic.
By regulating the guiding unit of the Francis wheel, the power regulation is made possible.
This chapter details mathematical modelling of the electro-hydraulic turbine governing
system PID regulator (proportional, integrating derivating), that will function as a governor
for the hydropower system.
As far as the literature describes the controller there is two ways of modelling. The one
which is ideal, includes Kp into the derivative and integral part deducing new constants
called integral and derivation gains. The model used in this thesis, is generated directly
out from the Laplace transformed equation 5.23. This is done to ease the tuning, where
tuning methods often are based directly on adjusting the time constants Ti ans Td .
The I-block or ( Ti1·s ) is the part where stationary deviation the eliminated. D part (Td · s)
is reducing the dynamic deviation and counteracts on oscillations in the process value, an
important part when running in island mode. [16] This leads to less wear on the water
guiding device, and less time to reach stationary speed. Kp is simply a proportional gain,
making sure that the gain is proportional to the deviation.
The mathematical relationships between the input and output for the PID regulator is as
follow [9] :
Z t
1 d
u = Kp · + dt + Td · + u0 (5.19)
Ti 0 dt
Where =deviation from reference input of what there is to be regulated. U is the output
signal where
u = uP + uI + uD + u0 (5.20)
41
In other words u is a function of the different contributions from the proportional, integral,
derivative, and nominal states given as:
uP = KP (5.21)
Z t
Kp
uI = · dt (5.22)
Ti 0
d
uD = Kp · Td · (5.23)
dt
Introducing Laplace transformation to 5.19 gives

1 u0
u(s) = KP · (s) 1 + + Td · s + (5.24)
Ti · s s

u(s) 1
= KP 1+ + Td · s (5.25)
(s) Ti · s
Equation 5.25 gives the following block diagram, as seen in figure 5.14
Figure 5.14: PID regulator for speed regulation and power regulation, as modelled in scilab. [17]
42
5.7 Servo System
5.7 Servo System
The servo system in his thesis are only used in the water hammer simulations. It is a part
of the mechanical system, making the system damp the water hammer, forces that may
occur. By the use of a rate limiter closing time of the guide vanes can be limited and set to
greater than the reflection time. The closing time is set to 10s during LVTrans simulations,
so did the scilab model during the simulations. This should be enough as the reflection
time in the system is less than 2s. Based on earlier experiences, if a pipeline of 1000m
is filled up with water and the closing time is 10s, the pressure transients will appear as
mildly [12]. In figure 5.15 the block used as rate limiter can be seen.
Figure 5.15: The ratelimiter block used for waterhammer simulations.
43
5.8 PID Tuning
5.8.1 Method 1
PID tuning can be done in many ways based on the type of system and is criteria for system
stability. In this thesis two different methods are reviewed. The first one is called ”Method
1” in the guide book of LVTrans [21], and it is the one used in this thesis. The procedure
steps can be seen in the list below:
1. Set the PID in island mode running at 85% til 100% load.
2. Set P = 1, Ti = 10, Td =0
3. Wait until the system is stable.
4. Set Ti = 1010 (or as high as possible)

5. Apply steps in the reference value of frequency of 10%
6. Increase P until a overshoot of 15% is observed in the frequency process value (PV)
compared to the step in reference.
7. Decrease Ti until the overshoot in PV is 30%

8. At last decrease P until the overshoot is acceptable. Typical 5% til 20%
9. For PID, increase Td until the overshoot is acceptable. Typical 5% til 20%
This methods is applied to the system in LVTrans and has been the reference thought the
system simulations done in chapter 6. As the regulator only has been working as a PI
regulator, the D parameter has been set to 0. The values are found to be:
• P=3
• I = Ti = 7s
44
5.8 PID Tuning
5.8.2 PID tuning by frequency analysis
When tuning the system in the frequency plane from the start a frequency response of the
system without the regulator is done. This transfer function is called h∗0 . By following the
the five stages in the list below, PI settings for the controller can be found.
1. Choose the type of regulator PI, PD, or PID.
2. Achieve Bode plot for the open loop transfer function without the regulator. This is
given by h∗0 (s)
3. Decide the amplitude cross frequency ωφc
4. Decide the regulator parameters. A proposal for this is found in table 5.3
5. Control the gain margin and perhaps adjust Kp
Table 5.3: Proposal for PID tuning by frequency analysis [9]
- P PI PD PID PID
(product) (sum)
∠h∗0 (jωφc ) −180◦ −160◦ −220◦ −200◦ −210◦
= ang +∆Φ +∆Φ +∆Φ +∆Φ +∆Φ
2.8 2.8 2.8
Ti ωφc ωφc ωφc
1 1 1
Td ωφc ωφc ωφc
[Kp]dB − |ang|dB − |ang|dB − |ang|dB − |ang|dB − |ang|dB

−1dB −3dB −4dB −2dB
In table 5.3, the different variables are: ∠h∗0 (jωφc ) which is the phase angle that decides
the amplitude cross frequency. This is just called ang further in the table, and the ∆Φ is
the phase margin for the given system. Once the frequency is found both Ti and Td can
be found. At last the gain called Kp can be found. This is done by taking the gain at the
calculated amplitude cross frequency and subtracting the right amount of dB given for the
specified regulator (P, PI, PD, or PID). The PID on sum or product is only a mathematical
preference where most of the new micro controller based regulators are on the sum form.
The product form is defined as:
45
1 + Ti · s 1 + Td · s
hprod
pid (s) = Kp · ·
Ti · s 1 + Tnd · s
And the sum form of the PID regulator is given by:
1 Td · s
hsum
pid (s) = Kp (1 + + ) (5.26)
Ti · s 1 + Tnd · s
PI constants are than analyzed and can be modified with frequency analysis to the system.
This, is done in section 6.3. In order to perform the frequency analysis an extra package
in the scilab program is installed. The package is called CPGE and makes it possible to
generate both bode plots and Nyquist diagrams directly from the block diagram interface.
The margins form the response in figure 6.10 in the frequency analysis chapter 6.3 are
well within limits, when looking at the general criteria for a stable system where the phase
and gain margins both are positive. Figure 5.16 illustrates two examples of a stable and
unstable system. The stable system is marked with the red line, while the blue is unstable.
Figure 5.16: Examples of unstable and stable systems based on phase and gain margin [11].
and the system is found to be stable during any external influences such as load rejections
and frequency variations the system are not adjusted any more than it is from the original
settings. This is a subject that could be taken further, and a proper PID analysis in the
frequency plane could be performed as future fork for the system.
46
Chapter 6
Simulation and Analysis
In this chapter, the analysis and simulations results are shown. The content involved in
this chapter is:
• System setup overview, p48.

• Functional requirements for hydropower generation and control in Norway p56.
• Stability analysis for Scilab model, p57.
• Evaluation of the hydraulic system in Scilab, p60.

• Model validation with LVTrans, p64.
The laboratory work can be found in chapter 7 page 71. The chapter contains surge shaft
model verification with the u-pipe test rig.
47
Chapter 6. Simulation and Analysis
6.1 System Setup
In this section models from the block diagrams are shown and explained. There are four
block diagrams form the Scilab program that is shown:
• Frequency regulation in island model
• Power regulation in a stiff network

• Analytic island model
• Analytic Island model with variable time dependant friction settings, used in water
hammer and mass oscillation simulations.
Each model is saved in individual xcos files, and can be opened by Scilab.exe. The pro-
gram has a ”write to CSV” block which logs the given input signal to a given time interval.
The log is saved as a text file and can easily get imported to Excel, MATLAB or similar
programs.
LVTrans is programmed into a more compressed simulation tool which makes the amount
of work programming the system minimal. By an advanced PID regulator, both island
and stiff network simulations can be switched by a button. Also, simulations for the water
hammer is done by the regulator by turning a switch.
48
6.1 System Setup
6.1.1 Scilab Setup
Frequency regulation in island mode
The model in figure 6.1 is frequency regulated turbine in island mode. It is a single ma-
chine, single load environment and it is much used in stability analysis on the hydraulic
part as the stiffness from a normal stiff network is absent. PID parameters are also set
in this configuration, and for Norwegian power plants, the settings should also be fully
functional during turbine power regulation with stiff loads.
Figure 6.1: Layout of the block diagram for frequency regulation in island mode.
49
Power regulation in a stiff network
Next system is the block diagram for a power regulated turbine connected to a stiff network
or load seen in figure 6.2. The system is based on the first model but has some changes
made to it. With this model, grid frequency is set as an input, as well are power and speed
references to the turbine too. The simulations are done with a step in grid frequency.
Figure 6.2: Layout of the block diagram for load regulation connected to a stiff load.
It is a single machine stiff load environment, with an implemented droop function which
decides how much the guide apparatus shall act during a change in load. Figure 6.3 illus-
trates the effect of different droop settings:
Figure 6.3: The effect of three different droop settings
As the droop get lower, the more will the turbine react to a change in frequency. In the
example, a drop of 0.5Hz is done at t=500. For the drop of 0.02, almost half of the machine
power gets reduced, compared to the droop of 0.12, where only 10% of the power gets
reduced. During the simulations, a droop of 0.06 is set.
50
6.1 System Setup
Analytic Island model.
Figure 6.4: The hydraulic system used in the simulations.
51
The analytic model in figure 6.4 is made to perform stability analyses and system param-
eter analysis. By using the ”set context” function in Scilab makes it possible to give the
parameters names and values. By inserting the different equations for PID parameters,
turbine parameters and hydraulic equations into the blocks, the program calculates the
transfer functions after what is defined in the context file.
By using the block: ”PARAM VAR”, different inputs can be simulated in one go. This
provides an effective way to perform parameter analysis to the system. The small red
blocks with letters in sets the reference point to where the program shall calculate the
transfer function. For a closed loop, frequency analysis between point A and B is inserted
in the BODE function block as start and end points.
Figure 6.5: Analytic functions for scilab model
In figure 6.5 the context for bode plot is set in the box to the left. In the box to the right,
settings for multiple variable simulations are done. In the figure, Kp is simulated for the
values 1 to 9.
52
6.1 System Setup
Waterhammer Simulations
To perform the simulations correctly, the guide vane closing time must be limited to about
10s, and the power from the grid must fall to zero. The guide vane must in the same time
be kept constant at zero. With the basis of the island mode diagram, modifications to the
governor and block infrastructure is done, see figure 6.6:
Figure 6.6: Changes in the infrastructure due to water hammer simulations.
After performing simulations with the setup from figure 6.6 in combination with the sys-
tem in figure 6.1, some more modifications had to be done to make corrections for the
friction factors as the flow changes states and value.
The friction still stays static during the event, but new values in the form of friction factor
can be changed with the system setup in figure 6.8. By use of relays and new blocks
describing the tunnels with new friction factors, the changes can be done in the time of
the valve closure. New friction settings are defined in the ”set context” function under
simulation settings see figure 6.7.
Figure 6.7: set context tool to inset system variables.
53
Analytic Island model with variable time-dependent friction settings
Figure 6.8: Analytic model with water hammer sim and friction modification.
54
6.1 System Setup
6.1.2 LVTrans Setup
The model in LVTrans can be seen in figure 6.9. It is programmed through using premade
blocks in the programme. After starting the program, each block can be defined in the
system specifications. In Appendix E, The parameters utilised in the simulations can be
found.
Figure 6.9: System in LVTrans
The setup is modelled as closely to the scilab model as possible. There are some smaller
deviations in the model such as the end tunnel and the servo implemented in the PID
regulator. By choosing a small length of the end tunnel (1m), and implementing a rate
limiter in the most transient simulations the deviations got minimised. The program has
a good implementation of logging functions and generates plots via MATLAB is easily
done. The program comes with a guide book as well, which makes it easy to get started
and ready to use.
55
6.2 Functional Requirement
The Norwegian power system administrator, Statnett is responsible for the operation of the
main grid in Norway. Through their book of guidelines (FIKS) [20], information on how
to operate energy production facilities are set, and different criterion for system stability
are given. Some of them are listed down below, and the simulations are to some degree
evaluated towards them.
General functional requirements
• Frequency regulation with a positioning feedback loop, with separate load adjust-
ments.
• Power regulation mode, measuring the generator power production. A droop should
also be implemented in the controller.
• Resolution of frequency measurements should not be less than 0.01% when operat-
ing in the area of 90% to 110% of the nominal frequency.
• Time constant Ty in the servo system should not be more than 0.4s This regards
Francis turbines as well as Pelton. The time constant the based on an opening of
100%.
• The permanent droop should be adjustable between 2% and 8%.
• For island operation: During a 10% change in load while running at 85% power
production, no more 6% frequency transient should occur.
Power response during destabilisation
• Unbalance between power production and and power consumption results in a change
in frequency. The turbine regulator should then compensate by regulating depended
on the measured frequency.
• The correction in production will vary based on the systems permanent droop.
• Stability analysis for an existing power station is identified by frequency analysis.
• Criteria for system stability is that the phase margins is in the interval between 25◦
35◦ and the frequency margin is between 3dB til 5dB at full load and a constant
voltage level.
• One should aim the regulation parameters by these margins for the best system sta-
bility possible.
56
6.3 Stability Analysis
The stability analysis is based on frequency response and a step response of the system.
It is performed bode plot from the system in open loop in the figures below. To perform
the frequency analysis, an extra package in the Scilab program is installed. The package
is called CPGE and makes it possible to generate both Bode plots and Nyquist diagrams
directly from the block diagram interface. The PI parameters are set to Kp = 3 and Ti = 7
during all the plots. Due to the many derivative blocks in the system, a delay in the gain
signal is implemented. This is necessary for the program to include the dynamics of the
turbine and penstock.
Figure 6.10: Bode plot for the open loop transfer function. Kp = 3, Ti = 7 and time delay = 0.01s
In figure 6.10 bode plot for the open loop including the PI regulator is made. A time delay
of 0.01s has been included in front of the turbine model, and the Kp and Ti settings are 3
and 7. This leads to a gain and phase margins of 24dB and 91◦ .
57
The tuning can be done by adjusting Ti and Kp to change the gain and phase margins. See
figure 6.11 where the cross frequency changes as Kp changes from 1 to 9:
Figure 6.11: Gain and phase response due to changes in Kp
In figure 6.12 the response for changing Ti from 1 to 9 as Kp is kept constant at three can
be seen. Choosing a lower Ti results in a lower phase angle at a frequency of 0.02 rad s to
10 rad
s . Doing so also increases the cross frequency which may result in a higher phase
margins. So there is a balance that must be set for the gain and the integral time.
Figure 6.12: Gain and phase response due to changes in Ti
The results from the frequency analysis show promising stability characteristics for the
system. With both positive phase and gain margins, the general stability criteria for a
dynamic system are set. The criteria set by statkraft is not fulfilled, and should be aimed
when systems shall be used for simulating with new plant data, or used for dimensioning
a new full-scale power plant.
Stability analysis are also done with a step response in load demand, simulating grid fre-
quency over a period of 2000s. The step occurs at t = 1000s with a negative magnitude of
0.5pu. The simulations are done in island mode with a simple machine, single load.
58
Figure 6.13: Frequency response for a load drop of 50%
Figure 6.13 shows the governor action in the form of PI regulation. As the drop in load
demand occurs, the governor stabilises the energy produced by the turbine, by closing the
guide vanes from fully open with and process value of 1.962 to 1.462 as seen in figure
6.14. The governor then slowly stabilizes the frequency by decelerating the turbine by
applying less amount of inlet flow, which corresponds well according to the mechanical
torque equation where the shaft torque is a function of fluid flow: Mt = ρQ(rin · cu,in −
rout cu,out ) and where the turbine power is a function of torque and angular speed, the
torque must decrease in order to maintain the speed as the load decreases, seen in equation:
Pt = Mt · wt . This must be, in order to keep the general stability and production criteria
where the energy demand and energy production mus be equal : Pl = Pt .
Figure 6.14: Gate opening, responding to a drop in load demand of 50%
59
6.4 Evaluation of Hydraulic Systems in Scilab
Evaluation of the dynamic penstock is done by comparing it to the classical, static model.
This is done to confirm the impact, which the dynamic model brings to the system. The
draft tube and lower surge shaft are not included in the model in the first simulations. First
off a comparison between the static and the dynamic penstock are done to see the impact of
frequency dynamics on the grid under a load change of −10%. The second comparison is
in water head level a larger load change. Last penstock model comparison is simulations
done with penstock as the only pipe in the system. Simulations of pressure pulsations
with different penstock lengths are also done to see the impact of pressure transients when
extending the penstock. Til last, the effect of adding a draft tube and a lower surge shaft
are being investigated.
6.4.1 Comparing Dynamic and Static Penstock
In the figure, 6.15 a comparison of the static and dynamic penstock model with surge shaft
dynamics is made. The blue line is simplified static penstock model, and the red line is
dynamic penstock model.
Figure 6.15: Comparing the frequency dynamics of two penstock models, when simulating a drop
in load of 10 % in island-mode.
The differences between the two penstock models are not visible during the decrease in
load, as seen in the right figure inside figure 6.15. A Small deviation in the start-up dy-
namics can be viewed in the figure to the left. The little difference could be because of
the short penstock length. This makes the effect of elastic water string small and almost
neglectable. As seen in section 6.18 the penstock of 500m has more effect on the system
than the shorter one of 250m.
60
Simulations for the pressure in front of the turbine can be seen in figure 6.16. It is the same
setup as in the simulations from figure 6.15, but with a larger load drop of 50%.
Figure 6.16: Comparing the pressure dynamics in front of the turbine when simulating a drop in
load of 50 % in island-mode.
The deviation from the load drop can be seen in figure 6.16 with a pressure difference of
0.006pu which in the 200m head base system is 1.2m. Due to deviation in friction, the
head in the dynamic penstock is some higher than the other. The models seem to behave
very similarly, and the penstock effect is obviously small in the system. Due to its short
length, the dynamic effect is almost neglectable from the view of pressure pulsations.
New simulations with only the penstock connected to the system gave better results for
comparison, see figure 6.17
Figure 6.17: Simulation of pressure and frequency, comparing the penstock models isolated.
By isolating the penstock in the simulations, differences became clearer. In the figure,
6.17 head pressure and frequency are plotted as a function of time. The simulations are
performed in island mode, with a load drop from 85% to 75% at t=1000. As seen the
dynamic model has a much larger influence in the form of increased pressure in front of
61
the turbine. There are also some deviations in the frequency dynamics as well. See table
6.1 for comparing figure values:
Table 6.1: Values from comparing dynamic and static penstock
Model ptot [m] ∆p [m] ∆f [pu]

Dynamic 216.8 19 0.028
Static 203 5.2 0.021
6.4.2 The Effect of Extending Penstock Length
To investigate the effect of penstock length, the change in head pressure in front of the
turbine is measured. It is simulated during a drop in load from 0.85 pu to 0.75pu, seen in
figure 6.18. The simulations are done with the analytic model, running in island mode.
Figure 6.18: Pressure pulsation effect for different penstock lengths
As the length increases so do the head pressure as a result of flow change due to a change
in load demand. More oscillations can be seen in the longer penstock as the elasticity time
constant is increasing, where:
lp
Te =
a
The deviation in total friction loss and pressure at the first transient are measured and
can be seen in table 6.2 The pressure ∆pt is measured from steady state head to the first
transient. The head loss is measured directly from steady state head. Values from the
graph are in per unit, and converted in the table to m.
62
Table 6.2: Values inserted in the scilab model during the simulations
Penstock length l [m] Friction loss ∆htot

f [m] Change in pressure ∆pt [m]
250 -3.86 4.34
500 -4.25 7.48
1000 -5.38 12.85
6.4.3 The effect of adding a draft tube and lower surge shaft
The effect of implementing a draft tube and a lower surge shaft can be seen in figure 6.19.
The figure shows pressure changes in front of the turbine given in pu. The simulations are
performed in island mode with a drop in load from 1pu to 0.9. It is done with a draft tube
with the same area as the penstock and the lower surge shaft dimensioned after the size of
the draft tube.
Figure 6.19: Frequency response of system start up and during a load rejection of 10% during full
load.
The implementation provides more friction which leads to a lower static head. A more
oscillatory behaviour during changes in flow can be observed with a higher pressure tran-
sient. Friction loss with draft tube and lower surge shaft implemented are at 4.302m and
without them, 4.048m. Both models provide a pressure increase til around 200m head,
which makes the model with the implantation some higher.
63
6.5 Model Validation with LVTrans
In this section the block diagram made in scilab is validated up against LVTrans. There
are performed three types of simulation:
• Frequency regulation in island mode
• Power Regulation in Stiff Grid Mode.
• Water Hammer Simulation
By comparing the two programs, an idea of how the Scilab model performs compared to
the physical model can be achieved. Since LVTrans has proven to be very accurate and
used in the industry as a standard simulation tool this program is used a validation method.
By using step response in both frequency and electric load, the model characteristics are
compared up against each other. The simulations are done with a step in load demand
for island operation and with steps in frequency for the power-regulated model. Since the
turbine and piping models have such large impact on how the responses are to pressure
pulsations and turbine speeds, a good system evaluation is possible to achieve by the step
responses.
Water hammer simulations are performed at the end of the chapter. This shows more
clearly how the Scilab model performs in the form of pressure pulsations and mass oscil-
lation simulation.
6.5.1 Frequency regulation in island mode.
Changes in grid frequency occur constantly, but since the load system is so substantial,
the frequency variation often appears to be small. Larger deviations in frequency occur
typically during line faults, and large load connect and disconnections. To provide good
quality and high reliability in the electrical power supply, governing in the production units
are necessary. This due to keep frequency and voltage fluctuations as low as possible and
maintain stability in a complex dynamic system. [8] In the simulations the systems are
stressed with frequency drops of 10% and a worse case of 50% load rejection.
In the figure below 6.20 a drop of 50% in load demand is applied to the systems. The Scilab
simulation was far off the LVTrans simulation during the heavy drop. It is a frequency
deviation of 4Hz, and the Scilab simulation is slower to reach steady state.
64
Figure 6.20: Frequency regulation during a drop from 100% to 50% in load demand
Figure 6.21 shows that the stability criteria is proven to be within the limits when running
in island mode. Acceptable results from both simulations in LVTrans and Scilab can be
seen as it is under a 6% transient peak during a 10% drop in load from 85% of BEP. The
LVTrans model had a peak of 1.026 pu while the Scilab simulation peaked at 1.021 pu, as
seen in the figure, deviation in frequency between the models are at 0.25Hz.
Figure 6.21: Frequency regulation during a drop from 85% to 75% in load.
6.5.2 Power Regulation in Stiff Grid Mode.
Simulations of Turbine power regulation connected to a stiff network is shown in this

section. In both simulations figure 6.22 and 6.23 the droop is set to: ρ = 0.06. The first
simulation can be seen in figure 6.22 and it is done with a drop in grid frequency if 0.5Hz.
65
Figure 6.22: Power regulation in stiff grid mode with a drop of 0.01% in grid frequency
The result is an increase in power production of 0.167 pu, equivalent to 15.74MW with
the rated turbine power of 94.27MW. The value corresponds well to theory. Equation 6.1
found in, [8] shows the mathematical relationship between changes in generated power as
a function of change in frequency:
Pd · ∆f 1pu · −0.01pu
∆Pg = = = 0.167pu (6.1)
fn · −ρ 1pu · −0.06
In figure 6.23 simulations are done with an increase in the frequency of 0.5Hz.
Figure 6.23: Power regulation in stiff grid mode with a increase of 0.01% in grid frequency
Simulations show the same ∆Pt as in figure 6.22, but oppositely. This naturally results
in a lower power production from the aggregate, and the turbine power output decreases
from 1pu to 0.833, a drop of 15.27MW.
66
6.5.3 Water Hammer and Mass Oscillations
The water hammer simulations are performed during BEP conditions of H = 200m and
3
Q = 50 ms . By using the modifications in the Scilab program, full closure of the guide
vane is done over a period of 10s. Full load rejection from the grid is done at the same
time, which makes the flow stop completely. Due to the full stop, pressure pulsations and
mass oscillations are generated by the hydraulic dynamics that can be seen in the figures
6.24 and 6.25 below.
Figure 6.24: Water hammer simulations done in scilab and LVTrans
As seen in the simulations, some clearly deviations of the response are present. Higher
transients in the first 50s of the simulations for the scilab model. The mass wave will then
decrease fast in amplitude for the scilab model while the LVTrans model is slower and
keeps on oscillate for the rest of the simulation time over 1200s.
Figure 6.25: A close look at the first transient from figure 6.24.
67
The first transients have a deviation of 8m in pressure and get close to identical in the
second one. LVTrans model catches up the smaller pressure oscillations in the first 100s
after the closure, while the Scilab model doesn’t perceive the fast pulsations.
Much of the results deviations are assumed to be due to system friction. In LVTrans the
friction varies as the flow changes. Due to this, a modification in the Scilab block diagram
is implemented to investigate the effect of a change in friction as the flow changes. The
friction factors are not variable during the simulations, but are changed to a lower static
value after the flow goes close to zero. The new friction factors count after 78s, 40s after
the guide vane starts to close. This delay is set to ensure that the flow has decreased
adequately. The new friction factors are set to the following:
• fD1 = 0.02
• fD2 = 0.005
• fD3 = 0.02
In figure 6.26 , the results of this simulation can be seen:
Figure 6.26: Simulation of a full system stop, with changes in the internal friction factors in the
scilab model.
As seen in figure 6.26 the results are much closer, but the Scilab model is reaching steady
state after around 3000s, while the LVTrans model keeps on oscillating. Simulations to
around 40000s are done with LVTrans, and the oscillations are sill going but with a small
amplitude.
As the friction factor of each part is changed, they all had their characteristic influence on
the system. When decreasing the friction factor for the head race tunnel, a much higher
amplitude is achieved. This factor has the biggest influence for how the mass oscillations
behave. The penstock has a minimal impact. The outcome when changing the friction
68
factor in this region is close to zero. Changing friction factor for the draft tube has a much
larger influence again. Is seems to induce much more oscillations during the transient stage
of the simulations. As seen, the first 250s are in the last simulations much more oscillatory
than the first one in figure 6.24.
69
70
Chapter 7
Laboratory Work
Figure 7.1: Picture of the rig.
71
Chapter 7. Laboratory Work
In this section, information and results from the laboratory work are presented. The test rig
is made to investigate harmonic oscillating flow which involves creating pressure transients
in a u-pipe rig. The rig is made by a co-student named Isak Bergset and the rig-results
presented in this section are done in cooperation with him.
The purpose of the testing is to perform verification of the surge shaft model that is imple-
mented in the modelled hydro power plant, see section 5.4. By using the rigs facility data
and perform simulations with the surge shaft model in xcos, it is possible to achieve some
knowledge of how the model performs to real physical behaviour.
In figure 7.1 the end of the test rig can be seen. Pressure gauges are located on the left
side. To create the water hammer, the flow is stopped by manually turning the valve in the
middle of the picture. Up in the left corner, the tank can be seen. The pump is located on
the floor below.
7.1 Laboratory Setup
Figure 7.2: A scetch of the test ring at the laboratory
In figure 7.2 the cross section of the test rig at the laboratory can be seen. It is designed
with an overflow function so that the head level can not exceed 1m. The piping diameter
is 150mm and the pipe length between the tank and the surge shaft is 11m. After the surge
shaft an valve which is operated manually. The flow can then go in return to a pump which
keeps the steady level of head to the system. The roughness of the piping is assumed to be
0.02mm and can be found in table 3.1. Summarized, the system setup has the following
settings:
72
7.1 Laboratory Setup
• H0 = 1m
• dpipe = 150mm
• e = roughness = 0.02mm
• Lpipe = 11m
• ft = Steel thickness = 3.3mm
20◦C
• µH2o = 1.002 · 10−3
There is performed three different tests on the rig, each with different steady state flow
3
rates before the valve closes. The first test is with a flow of 0.016 ms , the second is with a
3 3
flow of 0.01389 ms and the 3rd with 0.007 ms .
Calculations of the friction factor will be performed with the Colebrook-White equation:
1 e 9.35
√ = 1.14 − 2log10 · ( + √ )
f d Re f
While the friction loss is calculated with Darcy-Weisbach equation.
L V2
hf = fD · ·
D 2g
The different losses in the pipe can be seen in table 7.1
Table 7.1: Losses in the test pipe under different flow scenarios.
3
Test Number [#] Q [ ms ] Re [-] Friction factor f [-] Friction loss [m]
1 0.0160 135539 0.0177 0.0521
2 0.0139 117751.5 0.0182 0.0420
3 0.0070 59296.4 0.0206 0.0121
73
7.2 Results From the Rig
The results are plotted with a time step of 0.0002 (5000 pointss ) and have been averaged
every 500 points. The test results can be seen in figures 7.3 to 7.6 below. Notice that the
y-axis is scaled in all the figure. For comparison see figure 7.6.
Figure 7.3: Rig test result #1
Figure 7.3 shows the results from test one where the transient reaches 2.228 m in pressure.
This gives dH = 1.136m. Due to some deviation in the level of head form 1m, the
transient is not precisely 1.228m.
In figure 7.4 show the results from test number two where the transient reaches 2.11m of
pressure with an dH = 1m. Some smaller then test1 which is expected due to less flow.
74
7.2 Results From the Rig
Figure 7.5 shows the results from test three where the transient reaches 1.8m of pressure,
giving a pressure differential of dH = 0.586m. Again some smaller transient then the
other tests above due to flow reduction.
Figure 7.6: Comparison of all three test results from the rig.
In figure 7.6 all three tests are compared to each other. Note that test 3 is some offset in
the time of the event. This results in an offset phase in the oscillations. The transients are
clearly decreasing as the flow rate decrease, which was expected from equation 3.11, as
seen below:
c∆v Tr ∆Q L
∆h = · =2· ·
g Tc Tc gA
75
7.3 Rig results V.S Scilab model
When comparing the rig tests to the model in xcos, some of the modelling constants
changes as the flow changes throughout the tests. The changes in flow results in changes to
the water starting time, the surge shaft time constant and the friction losses. The different
parameters can be seen in table 7.2
Table 7.2: Values inserted in the scilab model during the simulations
Test Number [#] Tw [s] Ts [s] Re [-] Friction loss [m]

1 1.0152 1.1045 135539 0.0521
2 0.8820 1.2713 117751.5 0.0420
3 0.4442 2.5245 59296.4 0.0121
Compering the simulations done in xcos with a static pipe model to the laboratory tests
are done as shown in the figures below.
Figure 7.7: Comparing test 1 with the scilab model
In the figure, 7.7 the deviation in pressure is 0.2m from the first pressure transient. The
pressure from the laboratory test will then increase over time due to less friction as it
reaches steady state. This was not possible to simulate, as the friction model for the scilab
simulations where to simple. The friction stays the same throughout the whole time and
does not variate with the changes in flow.
76
7.3 Rig results V.S Scilab model
In figure 7.8 the deviation is smaller with a deviation in pressure of 0.15m. The same
friction effect is visible in this simulation. The steady state pressure is not shown in the
figure here, but the pressure is levelling out at 300s, about the same as in figure 7.7
The last comparison is with the smallest flow, gives the best results in the comparison tests.
The pressure deviation between the simulated and real values are 0.08m. The oscillations
stay in phase for almost a minute, and due to the low friction, the steady state value is also
similar.
77
78
Chapter 8
Discussion
The discussion part is divided into three main parts.
• Model validation
• Scilab model performance compared to LVTrans
• Surge shaft verification from the laboratory test rig
An analytic description of the simulation results and modelling aspects will be presented
in this chapter. The surge shaft verification will also be discussed and evaluated up against
the rig results. The primary emphasis is the comparison of the Scilab and LVTrans results.
This has been the main aim for the analysis part and is weighted accordingly.
79
Chapter 8. Discussion
8.1 Model Validation
The goal of the model verification chapter is to analyse the effect on the different hy-
draulic components in the system. Simulations of the frequency and pressure were done
comparing two different penstock models seen in figure 6.15 to 6.16. This was done in
island operation, with a step in a local load. The results show small deviations between
the models and in this case with a penstock of 250m the simplified model would be suf-
ficient. Simulations with only a penstock in the hydraulic system show different results.
Figure 6.17 displays pressure and frequency simulations for both penstock models during
a decrease in load from 85% to 75%. The pressure transient is clearly much higher with
the dynamic model as seen in the figure and is over three times greater than the static one.
The dynamic model also has an impact on frequency regulation characteristics and shows
some higher transient frequency than the static model.
To show the impact of penstock lengths, simulations were done with three different lengths,
250, 500 and 1000 meters. The simulations show that the more length added to the pen-
stock the higher change in pressure and more oscillations occurs due to a change in flow.
The water elasticity effect and the water staring time is influenced by the changes in length,
resulting in a more unstable system. From table 6.2 the results from the different penstock
lengths are shown. It indicates that when doubling the penstock length, the pressure tran-
sients increases close to twice the value. The friction loss for the different penstock lengths
is visually shown, as the graphs move down the pressure axis for the longer penstocks. This
was expected behaviour, due to the increase in water masses added to the pipeline. More
masses increases the force as a result of flow deceleration (F = ma).
The effect of implementing draft tube and surge shaft are also investigated. The purpose
of the draft tube is to transform kinetic energy from the runner to pressure energy. By
decelerate the water from the turbine, a higher head on the backside of the turbine can be
achieved. This prevents less wear on the turbine runner during changes in speed and flows
[3].
The simulations seen in figure 6.19 shows a more unstable and oscillatory behaviour with
this extra tunnel and surge shaft. In theory, there should be fewer pulsations and a more
stable head due to flow changes. The diagram show some tendencies of back pressure as
the first slope for the blue line in figure 6.19 is some deeper than the red one, representing
the system without the lower surge shaft and draft tube. Reasons for this misbehaviour is
the dimensions for the lower hydraulic parts. The draft tube in most cases has a character-
istic cone shape and a variable area along its length. This is not taken account for, making
the area dimensions for the draft tube to small. A consequence for this is an undersized
lower surge shaft. This may cause a negative effect on the stability and provide small pres-
sure pulsations. Averaging a new draft tube area could improve the performance. Transfer
functions for geometry such as a special designed cone shaped pipe may be possible, but
it is not done or researched during the thesis. From the study of draft tube designs in
”Hydraulic Turbines” by Brekke [3] it is stated that: ”The dynamic behaviour of the flow
in a real draft tube with bend and a rectangular outlet is not fully understood today”. So
80
8.2 Model Stability
modelling a draft tube to behave like a real life model could be very difficult.
8.2 Model Stability
The stability is analysed through frequency analysis, and step responses. The frequency
analysis is based on bode plot of the open loop transfer function. The plot shows the
system gain and phase response over a span of different frequencies. By looking at the
margins of phase and gain the system stability is characterised. The step response shows
the system stability physically. By causing changes in the external signals of the system,
such as the connected load, system stability can be evaluated.
The Bode plot for the system in figure 6.10 shows a stable system with both positive phase
ad gain margins of 91◦ and 24dB. Running tests with changes in Ti and Kp shows normal
behaviour in the form of how the margins and cross-frequency moves. An increase in Kp
gives a higher cross frequency, lifting the gain at lower frequencies. Increasing Ti results
in a decrease in cross frequency, and lifting the phase up over a span of 0.02Rad/s to 10.
Testing the system stability from a change in load demand shows good stability character-
istics for the system. In figure 6.13 a significant drop is impacting the system. 50% of the
load demand is cut off at t=1000s, and the governor immediately responds to decrease gate
opening seen in figure 6.14. The drop results in an increase of approx 0.1pu in frequency.
Due to PI action, the system regulates the frequency down to a reference value, and the
frequency is stabilised at 1pu or 50Hz. The regulator in the system works good, and the
stability of the system is within the limits of what statkraft recommend which is a transient
frequency of no more than 6% during a 10% change in load while running at 85% of ag-
gregate performance. This simulation of performance can be seen in section 6.5.1, figure
6.21. It shows a transient in the frequency of 2.1% in Scilab and 2.6% in LVTrans.
The system is found to be stable during any external influences such as load rejections and
frequency variations, (see figure 6.22 and 6.23). The system is not adjusted anymore from
the original settings. This is a subject that could be taken further, and a proper PID tuning
analysis in the frequency plane could be performed as future work for the system.
81
8.3 Scilab VS LVTrans
8.3.1 Frequency regulation in Island mode
The models responded as expected from theory. When the load demand drops, an increase
in frequency and machine speed occurs. Both systems responded by changing the guide
vane opening (G), and reducing the hydraulic force to the turbine wheel. By doing so,
the system speed is getting regulated back to its point of reference, as system speed is
proportional to the power-system frequency: n = fZ·60
p
.
The response can be linked to system frequency as well as turbine and generator speed.
This tendency of regulation can be seen in all the simulations done in island mode with
frequency control.
Two simulations are done with drops in load from 1pu to 0.5, and 0.85 to 0.75pu as seen
in figure 6.20 and 6.21. The models behave as expected. They gave different results in
both of the events of frequency regulations during load drops. Since the Scilab model is
linearized around the point of 1pu, a much larger drop away from 1pu should give a larger
error in the simulation. This could explain the large error in transient frequency seen in
figure 6.20 with 0.5pu load drop.
The whole event is much more similar in shape and governing action in the simulation with
a drop from 0.85 to 0.75pu, and with a smaller deviation in frequency transient between
the models. During the first simulation with o.5pu load-drop gives a frequency difference
of 4Hz or 0.08pu in frequency. In the second simulation, only 0.25Hz deviates between
the models and the regulators respond synchronised after some time and bring the system
back to steady state simultaneously 50s after the event.
8.3.2 Power regulation in Stiff Grid Mode
Simulations performed in a stiff grid can be seen in the figures 6.22 and 6.23. Both simu-
lations show power regulation as the frequency drops and increase 0.01 pu. The regulator
performs as it supposed to, and increases the production as the frequency drops and de-
creases power output as the frequency increases.
Comparing the two models, they both behave very similar in the form of production rate
during governing action, and final steady state values. The models behave accordingly to
theory, and it seems that the scilab model gives a good representation of such simulation
scenario.
Some more oscillatory behaviour is observed in the scilab model during the guide vane
regulation, and this could be due to the linearization issues as the turbine operation point
is changed during the governing. Simplifications in the linearized model such as system
82
8.3 Scilab VS LVTrans
friction and governing servos are simplified in the scilab model which has an adverse
impact on the behaviour of the system, leading to deviations between the models.
8.3.3 Mass oscillation and water hammer simulations
Simulations of water hammer, pressure transients and mass oscillations can be seen in
figure 6.24 to 6.26. The first Scilab simulation is done with the simple water hammer
simulation model that only utilises the standard island operation model but with the extra
guide vane closure implementations seen in figure 6.6. The simulations differ in several
ways. The first pressure transient deviates approximately 8m. The Scilab model simulates
a pressure increase of a 30m while LVTrans simulations reach about 22m. The mass oscil-
lations from Scilab model leads with 15s during the first periods, and pressure pulsations
on top of the mass oscillations in LVTrans are much more visible.
There are reasons to believe that that due to simplification, especially the friction model
causes these deviations. In the case of a fully closed guide vane, the friction is no longer
constant, but it varies as the flow stops in and the pressure pulsations force water masses
to pulsate for and backwards in the conduits. By keeping a constant friction such as it is
done in the scilab model, the simulation accuracy decreases and the simulation results in a
situation with too much friction and a pressure behaviour that decreases to fast and reaches
steady state too early. The water elasticity time constant Te has been too low during the
simulations, and as a consequence the simulations shows a little lower pressure transients.
Not a big influence overall, but it should be mentioned and corrected for further work with
the model. In appendix A-6 a suggestion for a more accurate elastic water time constant is
given.
This was later corrected in the next simulation seen in figure 6.26. By the use of the
system seen in figure 6.8, the friction in the conduits can be automatically changed at a
given time, e.g., at the moment of guide vane closure. In the simulation, new values of
friction coefficients are chosen to give a better pressure behaviour. As seen the results are
much better and the mass oscillations are almost in phase and at the same amplitude for
the first 600s of simulation. After 600s the friction has changed even more in LVTrans,
and larger deviations between the model are observed.
The electrical equivalent model or RLC model could improve the simulation, which is
used in the paper by Mishra [18]. By dividing the penstock into many sections a more
accurate simulation of the pressure pulsations can be achieved. In doing so, the whole
conduit system must be modelled in the same type of modelling. The RLC model only
gave worse results when trying to simulate water hammer and comparing with the test rig.
It adds up to the following conclusion regarding scilab model performance after multiple
tests and simulations comparing the two models. Scilab model simulates frequency regula-
tion well around its operating point, but it is not thrust worthy when the external influences
get too large. Simulations with turbine power regulation in a stiff grid environment give
the best results. Almost no error compared to LVTrans simulations in both scenarios of
83
frequency increase and drop. The model is not suited for accurate dynamic pressure sim-
ulations. It gives indications on how the masses oscillates and the first pressure transients
during the first few second after a full plant shutdown. Scilab model is also showing some
higher results compared to the LVTrans model here, so least it’s not underrating the forces
induced by the gate closure, which is a good thing regarding using the model to do plant
designs.
8.4 Surgeshaft Verification With Test Rig
The verification of the surge shaft model is found in section 7. Results comparing the
tests against the simulations are found in the figures 7.7 to 7.9. As seen in the figures the
simulations show reasonably good results compared to the rig test, but the friction model
causes errors inn the scilab model. Figure 7.7 indicates that the steady state head decreases
as the mass oscillations are damped out by pipe friction while the scilab head pressure is
the same even tho the water has no velocity in the pipe.
Another reason for deviations is that there is no rate limiter implemented in the scilab
model during the simulations. This causes a lead in the mass oscillations compared to the
test rig. The manometer in the test rig flushes with the pipe wall, which means that the
velocity pressure in the system is not included in the total sum of pressure measurements.
This could lead to minor errors during the tests, but are almost neglectable as the dynamic
velocity pressure is calculated to 0.041pa = 4.26 · 10−6 mW C at a velocity of 0.9 m s
3
during the highest flow rate of 0.0160 ms . This amounts to approximately 0.0004%.
Overall the surge shaft model provides to some extent a precise representation of the dy-
namics during a sudden stop in the system flow. The comparison between the scilab model
and the rig gives the model itself more heft and quality, knowing that the surge shaft is ac-
curate regarding mass oscillations and pressure transients.
84
Chapter 9
Conclusion
The master thesis has seven points that should be completed within the deadline. The
seven points are:
1. Add a surge shaft and an inlet tunnel to the existing model

2. Verify the surge shaft with experiments in the laboratory
3. Research the opportunities to add an elastic penstock model to the system

4. Theoretical deduction of the partial derivates for use in the xcos model, based on
Torbjørn Nielsen’s turbine model
5. Simulate the system with both an isolated and stiff grid
6. Look for possibilities to do a frequency response analysis

7. A short paper shall be written and presented at the 7th International Symposium on
Current Research in Hydraulic Turbines (CRHT-VII) at Katmandu University, April
2017
8. Write a report
85
Chapter 9. Conclusion
1. The surge shaft and inlet tunnel where implemented in the existing model. In addi-
tion to this, a draft tube and a lower end surge shaft were also applied to improve the
hydraulic accuracy. These implementations resulted in a more dynamic water path, hence
a hydraulic system influencing the regulation of power and frequency as seen in the sim-
ulations. Improvements for the draft tube model should be made in future work with the
system.
2. The surge shaft model is verified with a test rig at the laboratory. The results show good
performance for the scilab model, it giving a good representation of mass and pressure
behaviour during transient states. A new way of implementing a dynamic friction model
which varies with the changes in flow automatically would improve the results.
3. An elastic penstock is implemented to the system. By the use of Maclaurin series

expansion of the hyperbolic function tanh(Te · s), the implementation could be done
without iteration issues. During the system evaluation, the dynamic effect proved to be
small and neglectable for small penstock lengths.
4. The theoretical deduction of the partial derivatives was performed in the form of using
a hill chart generator which is based on Torbjørn Nielsen’s turbine equations. By doing so,
the transfer coefficients could be calculated by the output values given by the program.
5. Simulations were done in both isolated operations with a single machine and single
load and Stiff grid with a single machine. The simulations were compared up against
simulations done in LVTrans. The frequency simulations in island mode indicate that
the scilab model has a restricted area of operation. Performing simulations way off its
operating point give larger errors in the simulation, as seen in the simulation with a drop
of 50% drop in load. Simulations around the operating point give much better results,
and the two models give a very similar response in governing action and transient values.
Power simulations done in stiff grid operation shows good results and the two models acts
very similar during events of smaller frequency variations. These simulations gave the
best results for the scilab model.
6. A frequency response of the system is made, and a bode plot is generated automatically
from the open loop system transfer function with the use of an ad-don to the scilab program
called CPGE. The results show system stability with both positive phase and gain margins.
System stability is also analysed in the form of step responses in the electrical load. Good
stability is shown as the frequency transient does not exceed 6% during a load drop from
85% to 75%.
7. A paper is delivered in the event of the 7th International Symposium on Current Re-
search in Hydraulic Turbines, and it can be found in Appendix G. During the symposium
a presentation of the thesis was performed.
8. A report has been written, accordingly to the university standards.
86
Chapter 10
Further Work
During the process of developing the system, several marks on how to improve the system
have been noted. The draft tube should be redesigned, and new values which are represent-
ing an average diameter must be calculated. By doing so, changes in the characteristics of
the lower surge shaft has to be done. The use of RLC equivalent hydraulic system model
during pressure transient simulations would be interesting. In theory, this should give
better results for such simulations. Also improving the friction model to get the friction
loss to variate simultaneously with the rate of flow would give the model, overall a better
performance.
In form of system layout, a better representation in the block schematics should be done.
Creating ”superblocks” to improve the design and to get a better overview of the system
could improve the usability.
87
Chapter 10. Further Work
88
Bibliography
[1] Hermod Brekke. Transient Characteristics of High Head Francis Turbines. Doktor
engineer thesis, Trondheim, NTH, 1990.
[2] Hermod Brekke. Konstruksjon av Pumper og Turbiner. NTNU, 1998.
[3] Hermod Brekke. Hydraulic Turbines, design, erection, and operation. 2001.
[4] Wang.C. Goa.H. Effect of detailed hydro turbine models on power system analysis.
Zhejiang University, Hangzhou IEEE 2006, 2006.
[5] Odd Guttormsen. Vassdragsteknikk II. Akademika forlag, 2014.
[6] Kartein J. Olsen Hans M. faanes, Arne T. Holten. Stabilitet for kraftsystemer og
motordrifter. pages 45–50, 1990.
[7] Nkosinathi Dlakavu Hongqing Fang, Long Chen and Zuyi Shen. Basic modeling
and simulation tool for analysis og hydraulic transients in hydroelectric power plants.
IEEE. Transactions on energy conversjon, Vol. 23(NO. 3):p.834–841, 2008.
[8] James R. Bumby Jan Machowski, Janusz W. Bialek. Power System Dynamics: Sta-
bility and Control, Sec. edition. Wiley, 2012.
[9] Per Hveem Kåre Bjørvik. Reguleringsteknikk. Kybernesis, 2012.
[10] L.F.Moody. Trans. asme. Volume 66, 1944.
[11] MIT. Stability criteria in bode plot. http://www.mit.edu/afs.new/
athena/course/2/2.010/www_f00/psets/hw3_dir/tutor3_dir/
tut3_g.html, 29 september 2000. [Online; accessed 21-May-2017].
[12] German Ardul Munoz-Hernandez. Modelling and Controlling Hydropower Plants.
Springer, 2013.
[13] Torbjørn Nielsen. Dynamic Dimensioning Of Hydrop Power Plants. NTNU,
Vannkraftlaboratoriet.
89
[14] J. Riera O. Quiroga and C. Batlle. Identification of partially known models of the
susqueda hydroelectric power plant. Lat. Am. appl. res., Volume 33(Number 4),
2003.
[15] The Engineering Page. Roughness of different materials. http://www.
the-engineering-page.com/forms/dp/typ_eps.html, 2017. [On-
line; accessed 10-May-2017].
[16] B Vahidi R.A. Naghizadeh, S. Jazebi. Modeling hydro power plants and tuning
hydro governors as an educational guideline. International Review on modelling and
simulations, Vol. 5(NO. 4), 2012.
[17] Erik Rognaldsen. Project Work: Comparison of Blockdiagrams for Hydropower

Plants. NTNU, 2016.
[18] D.K. Khatod Sachin Mishra, S.K Singal. Effect of variation of penstock parameter
on mechanical power. International Journal og Energy Science, Volume 2(Number
3):110–114, 2012.
[19] Scilab Enterprises S.A.S. https://www.scilab.org/scilab/about. 2017.

[20] Statnett. Funksjonskrav i kraftsystemet (FIKS). Statnett, 2012.
[21] Bjørnar Svingen. LVTrans Manual. 2016.
[22] Nielsen T. Dynamisk dimensjonering av vannkraftverk. Trondheim, Sintef, 1990.
90
Appendix
A
A
Dimensioning Conduit Parameters
Friction losses in a conduit is found by Dracy Weisbach’s equation:
L V2
hf = fD · · (10.1)
Dh 2g
For the headrace tunnel the following constants are set as:
Table 10.1: Parameter values for the head race tunnel
Parameter Description Value Unit

A1 Area 25 m2
v1 Velocity @ Qr 2 m2
L1 Length 2000 m
fD1 Friction factor 0.05 −
To calculate the hydraulic diameter, the tunnel geometry must be decided. Typical geom-
etry used in Nowegian hydropower plants are the one introduced in the theory chapter:
A-1
Figure 10.1: Cross section of the head race tunnel
Hydraulic diamter is defined as:
4·A
Dh = (10.2)
p
For the given geometry, this can be deduced down to:
r
8·A
Dh = (10.3)
4+π
This gives the tunnel a hydrulic diamter of:
r
8 · 25
Dh = = 5.292m (10.4)
4+π
The friction loss can now be calculated:
2000m (2 m
s )
2
hf 1 = 0.05 · · = 3.85m (10.5)
5.292m 2 · 9.81 sm2
For the penstock the following parameters are set as in table
A-2
Table 10.2: Parameter values for the penstock
Parameter Description Value Unit

A2 Area 12 m2
v2 Velocity @ Qr 4.167 m2
L2 Length 250 m
fD2 Friction factor 0.01 −
The penstock has a round geometry, as seen in figure 10.2.
Figure 10.2: Cross section of the penstock
For such geometry with the pipe completely filled with water, the hydraulic diameter is the
same as its diameter.
4·A 4 · π · r2
Dh = = =2·r =D (10.6)
p 2·π·r
This gives the following hydraulic radius:
4 · 12m2
Dh = q = 3.91m (10.7)
2
2 · 12mπ
This gives the head loss:
250m (4.167 ms )
2
hf 2 = 0.01 · · = 0.57m (10.8)
3.91m 2 · 9.81 sm2
And for the Draft tube:
A-3
25m (4.167 ms )
2
hf 3 = 0.05 · · = 0.053mm (10.9)
3.91m 2 · 9.81 sm2
Water mass starting time are given by:
Lp Qbase
Tw = · (10.10)
Ap · g Hbase
Timeconstant for headrace are:
3
2000m 50 ms
Tw1 = · = 2.04s (10.11)
25m2 · 9.81 sm2 200m
and for the penstock:
3
250m 50 ms
Tw2 = · = 0.53s (10.12)
12m2 · 9.81 sm2 200m
The draft tube is assumed to be the same dimensions and form as the penstock, but with a
length of 25m. The water starting time is therefore:
3
25m 50 ms
Tw3 = 2 m · = 0.053s (10.13)
12m · 9.81 s2 200m
When designing the surge shaft, the area in the canal is given by the Thoma area. Typical
approach is to set the surge shaft area As = 1.5 · Athoma = 1.5 · Ath . Where:
L · A1 · v12
Ath = (10.14)
2 · hf 1 · g · H0
This gives an value of:
2000m · 25m2 · (2 m s )
2
Ath1 = m = 26.5m2 (10.15)
2 · 3.85m · 9.81 s2 · 200m
So the surge shaft area becomes:
A-4
As1 = 1.5 · Ath = 1.5 · 26.5m2 = 39.7m2 ≈ 40m2 (10.16)
Surge shaft time constant can now be calculated:
As · Hr 40m2 · 200m
Ts1 = = 3 = 160s (10.17)
Qr 50 ms
Surge shaft swing limits are given by:
s
L · At
∆Z = v · (10.18)
g · As
s
m 2000m · 25m2
∆Z = 2 · = 22.58m (10.19)
s 9.81 sm2 · 40m2
Calculations for the lower surge shaft are shown below. It uses the same equations and
procedure as the one higher up:
25m · 12m2 · (4.16 m s )

2
Ath2 = m = 4.68m2 (10.20)
2 · 0.283m · 9.81 s2 · 200m
As2 = 1.5 · Ath = 1.5 · 4.68m2 = 7.02m2 ≈ 7m2 (10.21)
Surge shaft time constant can now be calculated:
As · Hr 7m2 · 200m
Ts2 = = 3 = 28s (10.22)
Qr 50 ms
The wave velocity calculations is based on rigid water conduits and the simplified velocity
equation is used:
s s
K 2.2 · 109 pa m
a= = kg
= 1483 (10.23)
ρ 3
1 · 10 m3 s
A-5
Suggestion of elasticity water time constant for future simulations:
A new value for the elasticity constant for future simulations are found here. At first the
wall thickness is calculated, then a new wave velocity for elastic water string is deduced.
N
The wall thickness is based on a general rule of a stress limit of σ = 100M pa = 100 mm 2
The maximum forcess acting on the penstock are at the verry end, with a head pressure of
100mwc. In Pa this is:
kg
P = ρ · g · h = 1000 = 1962000P a (10.24)
m3
Forces acting on half a pipe wall with diameter of 3.91m over a lenght of 1mm gives:
F = P · A = P · D · 1mm = 1962000P a · 3.91m · 0.001m = 7671.42N (10.25)
And the pipe thickness is given by:
F = 2 · σ · A = 2 · σ · t · 1mm (10.26)
This gives a thickness of:
F 7671.42N
t= = N
= 38.36mm (10.27)
2·σ 2 · 100 mm 2
New velocity based on the pipe thickness is then calculated:
s v
N
2.2 · 109 m
u
K u 2 m
a= D·K
= u
9 N
= 1043.7 (10.28)
ρ(1 + f ·E ) t kg
1000 m
3.91m·2.2·10 m2 s
3 (1 + )
0.038m·2.22·1011 N m2
The elasticity time constant Te for the penstock then becomes:
l 250m
Te = = = 0.239s ≈ 0.24s (10.29)
a 1043.7 m
s
A-6
B
Dimensioning Francis Turbine Wheel
Calculations done in this chapter is based on theory from the book ”pumper og turbiner”,
made by Hermod Brekke [2]. The picture below in figure 10.5 is from the same book, and
defines the velocity components at the entrance and the end of the turbine wheel.
Figure 10.3: Velocity triangles for a Francis turbine. [2]
Where:
----- = Performance over BEP (Best efficiency point)
* = BEP
B-1
General equations
NPSH (Net Positive Suction Head) is used as a starting point when designing Francis
turbine wheel. The value of NPSH gives the pressure that the turbine wheel must withstand
without starting to cavitate.
Table 10.3: Values for the parameters, a and b based on earlier experiences.
Parameter Tubines Pumps

a 1.05 <a <1.15 1.6 <a <2.0
b 0.05 <b <0.15 0.2 <a <0.25
The relationship between NPSH and the velocities for the outflow is descibed as:
c2m2 u2
N P SHt = a +b 2 (10.30)
2g 2g
Also the following criteria must be fulfilled.
c2
N P SHt < − j = −Hs + hb hvp (10.31)
2g
Where hb is barometer pressure ans hvp is water vaporizing pressure. Hs is negative for
turbines with an negative head due to the lower reservoir. And also:
a(u2 · tanβ2 )2 + bu22

N P SH = (10.32)
2g
Typical angles for β2 are based on empiric values of 13◦ and 19◦ . Also typical velocities
of u2 is between 35 and 43 ms .
Further the relationships between cm and the outflow diameter D2 is given by:
4 · Q∗
cm2 = (10.33)
πD22
where Q∗ is dimensional flow rate.
The relationship between u2 and the parameters of D2 and n:
B-2
πD2 n
u2 = (10.34)
60
where n is the synchronous speed of the wheel in relation of the grid frequency (50Hz in
Norway):
f · 60 3000
n= = (10.35)
Zp Zp
where Zp is number of pole pairs, and the value being an integer. So in order to keep the
chosen vane angle, an correction for the wheel diameter is done to maintain the form of
velocity triangle.
Based on that β remains constant and that cu = 0 gives the equation for correcting velocity
cm
4 · Q∗ 4 · Q∗
cm2 = 2 → cm2corr = 2 (10.36)
πD2 πD2corr
This gives the following conditions:
cm2corr D2 2 u2corr
=( ) = (10.37)
cm2 D2corr u2
Further when β2 = constant and cu2 = cu2corr = 0 gives:
4Q∗
cm2 πD22
tanβ2 = = π·nD2
(10.38)
u2 60
and as β = constant the same relationship counts for corrected speed and diameter.
4Q∗
cm2corr 2
πD2corr
tanβ2 = = π·nD2corr
(10.39)
u2corr 60
This gives:
3
ncorr D2corr = nD23 (10.40)
FInaly D2corr is found to be:
B-3
D23 n 1
D2corr = ( )3 (10.41)
ncorr
It is possible to choose number of pole pairs at the begining and directly find D2corr , but
then u2corr must be found.
D2corr
u2corr = ncorr π (10.42)
60
further
4 · Q∗
cm2 = (10.43)
πD2corr
The turbine diving head must be checked by the equation 10.30 in order to make sure that
cavitation does not occur. If NPSH becomes to large relative to the available head, a lower
peripheral speed or a lower β2 will decrease the value.
The inner diameter is then found. The speed of cm1 is here found by keeping the flow
constant through out a control volume:
where:
cm = ρm = constant (10.44)
Which gives
cm 1ρ1
cm2 = (10.45)
ρ2
And using control volume as a test:
Z r22
cm rdr = Q (10.46)
r21
There is also necesary to controll the full load flow where cu1 6= 0. This is done til last,
where margins of N P SH must be atleast 0.5[mH2O] v.s N P SHt . Introducing Euler’s
pump and turbine equation the following relationship is given:
B-4
(u1 cu1 − u2 cu2 )
ηh = = 2 · (u1 cu1 − u2 cu2 ) (10.47)
gH
Assuming ηh = 0.96 and BEP with vertical outlet gives cu2 = 0. This leads to:
ηn = 2 · (u1 cu1 ) = 0.96 (10.48)
Choosing the relationship between u1 and cu1 in such way that the losses are reduced due
to pressure variations as angle of the guide vanes vary. Typical values are, u1 = 0.7−0.76.
This gives an value for cu1 of:
0.96
cu1 = = 0.67 (10.49)
2 · 0.71
Knowing cu1 , D1 can be found by the following equation:
√
2 · u1 2 · u1 · 2gH
D1 = = nπ (10.50)
ω 30
Next is to get the dimensjons for the height of the wheer, or the with. From equation of
continuity we have:
cm1 A1 = cm2 A2 and cm2 = 1.1cm1 (10.51)
This gives:
1.1πD22 0.275D22
B1 D1 π = → B1 = (10.52)
4 D1
Knowing that,
cm2 p
cm1 = and u1 = 0.72 2gH (10.53)
1.1
the entry angle β1 is found as:
cm1
tanβ1 = (10.54)
(u1 − cu2 )
B-5
Dimensioning the turbine wheel
The main dimensions for the Francis turbine wheel is calculated here. The values may
differ a little from the ones used in ”test-hill” programme and ”LVTrans”. This is due
to some values being rounded off, but they are based on the same facility data and are
approximately the same.
3 3
Facility data: Q∗ = 50 ms , ηh = 0.96, H = 200m, A = 8 ms , Tw = 0.95
kg m m3
P = ηρgQH = 0.96 · 1000 · 9.81 · 50 · 200m ≈ 94, 176M W (10.55)
m3 s2 s
β2 and u2 are chosen to 16◦ amd 40 m

s
cm2 m
tanβ2 = → cm2 = 11.47 (10.56)
u2 s
r r
4 · Q∗ 4 · 50
D2 = = = 2.35m (10.57)
π · cm2 π · 11.47
u2 · 60s 40 m
s · 60s
n= = = 325rpm (10.58)
π · D2 π · 2.35m
Number of pole pair:
3000 3000
Zp = = = 9.228 (10.59)
n 325
Choosing Zp = 9 for this machine. This gives new dimensjons:
3000
nsynch = = 333.33 (10.60)
9
s s
3 D22 u2 60 3 2.352 m · 40 m
s · 60s
D2corr = = = 2.33m (10.61)
πnsynch π · 333.33rpm
nπD2corr 333.33rpm · π · 2.33m m

u2corr = = = 40.66 (10.62)
60s 60s s
B-6
Further, calculations for the inlet gives:
cm2 cm2 11.47 ms

cm1 = = √ = = 0.166 (10.63)
1.1 · 2 · 9.81 sm2 · 200m
p
1.1 1.1 · 2gH
Calculations are based on η = 0.96 and chosen value of cu2 = 0.67:
0.96
ηh = 2u1 cu1 → u1 = = 0.716 (10.64)
2 · 0.67
This gives the inlet diameter of the turbine:

√
60 · 2 · 9.81 sm2 · 200m · 0.716
p
60u1 60s · 2gHu1
D1 = = = = 2.57m (10.65)
πn πn π · 333.33rpm
Now finding the height of the turbine:
3
Q∗ Q∗ 50 ms
B1 = = √ = = 0.6m
π · 2.57m · 0.166 · 2 · 9.81 sm2 · 200m
p
πD1 cm1 πD1 cm1 2gH
(10.66)
At last the inlet angles can be found:
cm1 0.166
α1 = tan−1 ( ) = tan−1 ( ) = 13.9◦ (10.67)
cu1 0.67
cm1 0.166
β1 = tan−1 ( ) = tan−1 ( ) = 74.51◦ (10.68)
u1 − cu1 0.716 − 0.67
B-7
C
Calculations of Turbine Transfer Coefficients
Figure 10.4: Calculations of transfer coefficients in reference to induced torque by the turbine
C-1
Figure 10.5: Calculations of transfer coefficients in reference to flow through the turbine
C-2
D
Scilab Models
D-1
Figure 10.6: Power regulation in a stiff network.
D-2
Figure 10.7: Frequency regulation in island mode.
D-3
Figure 10.8: The hydraulic system used in the simulations.
D-4
E
Parameters used in the LVTrans simulations
E-1
Figure 10.9: Parameters used in the LVTrans simulations
E-2
F
Figure 10.10: Risk assessment
F-1
G
Paper Presented as the 7th International Symposium on Current Research

in Hydraulic
G-1
Proceedings of the International Symposium on Current Research in Hydraulic Turbines
CRHT – VII
April 04, 2016, Turbine Testing Lab, Kathmandu University, Dhulikhel, Nepal
Modeling and simulation of hydropower plant in Scilab

Erik Rognaldsen1*
1
Department of Energy and Process Engineering, Norwegian University of Science and Technology,
Alfred Getz vei 4, Norway
* Corresponding author ([email protected])
Abstract
A hydropower plant is modeled in a computer application called Scilab, an open source MATLAB clone with
possibilities to create block diagrams and do system simulations. The system is designed to do both frequency
regulation when running in island operation and power regulation in a stiff power network. The conduit system are
set up with a dynamic penstock model, surge shaft and a rigid head-race tunnel. It is used a linear turbine model
allowing the author to linearize turbine characteristics of a specified turbine to get a detailed description of the
turbine directly in the model. The linearization are performed with a computer program based on modified Euler’s
turbine equations. The system are until now only implemented with a simple governor with a PID-regulator. A
servo system may be implemented at a later point of the mater thesis process.
In both simulation scenarios (grid and island), peak and mean steady state values are similar, but there is clearly a
more oscillatory behavior in simulations performed in LVTrans, compared to Scilab. In grid mode the frequency
responded with an increase of 8.5Hz when performing a step in load of 50%. This happened in both programs, but
the time to develop steady state are different where LVTrans where faster and more aggressive. Simulating in grid
mode, performing a drop of 0.5Hz with a droop set to 0.06, an increase of 17% in power generation occurred in
both programs. Also in this event LVTrans behaved more oscillatory.
Keywords: System modeling, Hydropower Simulaiton, Scilab, LVTrans, Model verification

1. Introduction
1.1. Context
During the spring 2016 the project work assignment at NTNU gave the thesis a start, providing the writer
knowledge of system modeling and hydropower systems through a literature study. The models in this
paper are found in different articles and books and are cited during the theory section. What there is
provided to the field of subject are the use of computer software tools and the method of system verification.
During literature study, a hydropower plant has not been found modeled inn Scilab, xcos before and the
thesis will give the program an evaluation of hydropower simulation performance with block diagrams.
Later in the process, a model verification of the surge shaft will be performed with the u-pipe test rig at the
laboratory at NTNU.
1.2. Motivation
Load variation on the grid or different events of faults causes changes on the grid frequency. In order to
obtain the goals for stability, safety and supply, regulations of the aggregate units are necessary. Simulations
of a hydro power plant in the transient states are therefore crucial to get predictability needed to build and
maintain the system, but also to know the limits of plant design. [1]
The modeling and simulations are performed in two different contexts. The one being frequency regulation
in island operation, with only a local load and a single machine see Figure 1. The other one is power
regulation on a stiff power network, Figure 2
Figure 1: General block diagram for simulating frequency regulation in island operation.
Figure 2: General block diagram for simulating power regulation in a stiff power network
2. System Modeling and Theory
2.1. Conduit system
The conduit design consists of a penstock, a surge shaft and a headrace tunnel as seen in Figure 3
Figure 3: System layout.
The level of Hr are considered constant in the model. The lengths L1 (2000m) and L2 (250m) are the
headrace tunnel and the penstock, while Ls (25m) is the height of the surge shaft. Same subscripts accounts
for areas and velocities in the model where 𝐴1 = 25𝑚2 , 𝐴2 = 12𝑚2 , 𝐴𝑠 = 40𝑚2. System head base are
𝑚3
200m, rated fluid flow is set to 50 𝑠
which gives base power at the runner 94MW. The dynamic penstock
is modeled in articles such as: [2]
Basic mathematical models are derived from Newton’s second law of motion, and conservation of mass
law. For a fluid element inside a tube, transient flow is described as the differential equation 1.
𝜕𝐻 1 𝜕𝑄 𝑓𝑄 |𝑄|
𝜕𝑋
+ 𝑔𝐴 ⋅ 𝜕𝑡
+ 2𝑔𝐷𝑎2 = 0 (1)
See table 1.1 for parameter explanation:
Table 1: List of parameters
Symbol Description
H Water pressure [m]
X Displacement [m]
Q Flow rate [m^3/s]
g Gravity acceleration =9.81
A Cross section area [m^2]
t Time [s]
f Friction coefficient [-]
D Tube diameter [m]
Introducing conservation of mass law for a CV (control volume) gives the continuity equation. This takes
account for water compressibility effect during transient periods, and tube elasticity:
𝜕𝑄 𝑔𝐴 𝜕𝐻
+ ⋅ =0 (2)
𝜕𝑋 𝑎 2 𝜕𝑡
Where “a” is the velocity of the pressure wave in meters per second. After Laplace transformation, per
unitizing and algebraic manipulation introducing the characteristic impedance Zc, water starting time
constant Tw and the elastic time constant Te gives the following equation set in a matrix format:
hU (s) cosh(𝑧𝑇𝑒 ) −𝑍𝑐 sinh(𝑧𝑇𝑒 ) h (s)

[ ]=[ 1 ]⋅[ D ] (3)
qU (s) − 𝑧 sinh(𝑧𝑇𝑒 ) cosh(𝑧𝑇𝑒 ) qD (s)
𝑐
This gives the classical wave solution as a transfer function, describing pressure and flowrate at each point
of the conduit in a hyperbolic function:
ℎ(𝑠) 𝑇 𝑓𝑄 1/2 𝑓𝑄 1/2
𝑞(𝑠)
= − 𝑇𝑤 (1 + 2𝐷𝐴⋅𝑠
0
) 𝑡𝑎𝑛ℎ ((𝑠 2 + 𝑠 2𝐷𝐴0 ) 𝑇𝑒 ) (4)
𝑒
This leads to the final function describing the dynamic penstock used in the simulations:
ℎ(𝑠) 𝑇𝑤
=− 𝑡𝑎𝑛ℎ(𝑠𝑇𝑒 ) (5)
𝑞(𝑠) 𝑇𝑒
For a rigid conduit without the elasticity taken account for, the transfer function is reduced to:
ℎ(𝑠)
= −Tw − ℎ𝑓 (6)
𝑞(𝑠)
Where 𝑇𝑤 , is the water mass time constant given by:

𝑄𝑏𝑎𝑠𝑒 𝐿
Tw = ∑ (6)
𝑔 𝐻𝑏𝑎𝑠𝑒 𝐴
The surge shaft is described as a change in surge shaft water level head Z over a change in the system flow
Q, which are given by:
𝑍(𝑠) 1
𝑄(𝑠)
= 𝑇 ⋅𝑆 (7)
𝑠
Where 𝑇𝑠 is the surge shaft time constant given by:
𝐴𝑠 𝐻0
Ts = 𝑄0
(8)
Total layout of the conduit system can be seen in Figure 4
Figure 4: Layout of the conduit system.
2.2. Turbine model

The turbine model used in the simulations is referred as the linearized turbine model and can be found in
literature such as: [3]. The model consists of two equation sets describing the changes in flow through the
turbine, Δ𝑞, and the changes in momentum Δ𝑚.
𝜕𝑞 𝜕𝑞 𝜕𝑞
Δ𝑞 = 𝜕ℎ Δℎ + 𝜕𝑛 Δ𝑛 + 𝜕𝐺 Δ𝐺 (9)
𝜕𝑚 𝜕𝑚 𝜕𝑚
Δ𝑚 = Δℎ + Δ𝑛 + Δ𝐺 (10)
𝜕ℎ 𝜕𝑛 𝜕𝐺
Where “h” is head, “q” is flow and “G” is guide vane opening. The partial derives from now on called
turbine transfer coefficients are found by different methods. Typical method is by calculations of Euler’s
internal turbine equations. In the thesis modified Euler turbine equations are used, deduced by T. Nielsen
at NTNU. By linearizing the equation around an operating point turbine transfer coefficients can be found.
Turbine used in the model can be seen in table 2:
Table 2. Turbine parameters
Parameter Definition Value
𝑚3
Qbep Best efficiency point flow [𝑠] 50
Hbep Best efficiency head [m] 200

Nbep Best efficiency speed [rpm] 333.33
Tbep Best efficiency torque [Nm] 2700694
n_bep Efficiency at best point [-] 0.96
a1bep Guide vane angle [°] 15
b1bep Angle of the rotor blade [°] 67.7
r1 Inlet radius [m] 1.311
r2 Outlet radius [m] 1.146
Rq Constant for RPT [-] 0
Rm Angular velocity damping constant [-] 0.04
Rd Design constant coefficient [-] 0.03
Rf - 0.0015
This gives the following turbine transfer coefficients:

𝜕𝑞 𝜕𝑞 𝜕𝑞 𝜕𝑚 𝜕𝑚 𝜕𝑚
= 2.75, = −1.198, = 0.86, = 2.87, = −1.12, = 0.87
𝜕ℎ 𝜕𝑛 𝜕𝐺 𝜕ℎ 𝜕𝑛 𝜕𝐺
Block diagram describing the turbine model are given in figure 5. The conduit system are in this figure
referred as: “Hydraulic system”.
Figure 5: Turbine model used in the simulations
2.3. PID regulator
A classical PID regulator is used in the system. Equation 11 describes the regulator:
1 t de
u = K p ⋅ (e + ∫ e ⋅ dt + Td ⋅ )+ u0 (11)
Ti 0 dt
Where “e” is the deviation from reference input value, and “u” is the output signal. The regulator is modeled
as in figure 6.
Figure 6: PID regulator

3. Results
Results from both Scilab and LVTrans are presented in this chapter. The system are modeled as similar as
possible in both programs. All simulations are done with PID parameters set to: Kp= 2.3, Ti=8 and Td=0.
3.1. Scilab simulations
Figure 7: Running in island mode, simulating frequency response to a load drop of 50%.
Figure 8: System in grid mode, with a change in grid frquency of 0.5 hz at t=500s.
3.2. LVTrans simulations
Figure 9: Simulation in island mode with a load drop of 50%
Figur 10: Simulation in grid mode with a drop of 0.5Hz in the grid frequency
4. Analysis
4.1. Discussion
Looking at the frequency response when the systems are running in island mode the peak value are
comparable as the peak reaches 1.17 pu frequency giving a 𝚫𝒇 = 𝟎. 𝟏𝟕 𝒑𝒖 in both programs. This gives
an increase of 8.5hz in a 50Hz base system. The dynamics of the transient effect are different comparing
the two programs. Regulation in LVTrans performs more aggressive compared to Scilab, resulting in a
more oscillating response. It is known that LVTrans utilizes a more advanced governing system with
implemented hydraulics. This provides the system with a higher order transfer function which in this case
results in a more detailed response.
The same behavior is observed in simulations in grid mode. Simulations done in LVTrans are more
oscillatory but the mean values of energy production are similar. When simulating a drop in 0.5 Hz with a
static droop at 0.06, gives an increase in power generation of approximately 17% which in a 94MW base
system gives increase of 16MW in production.
4.2. Conclusion
The system modelled in Scilab gives to some degree, accuracy when performing simulations for dynamic
response of frequency and power generation. The main differences in the two systems are the oscillatory
effect, which can be explained by the differences in governing system. By implementing a more accurate
governing and prime mover model for Scilab, this can be tested and evaluated once more, later in the master
thesis process.
5. Future work
During the thesis, a more thorough analysis shall be performed. A modal and Bode plot analysis will be
done during the spring. A more detailed model of the penstock shall also be made. By using an electrical
RLC component equivalent model makes it is possible to perform realistic simulations of pressure
propagation during water hammer.
Laboratory work with a u-pipe rig will be performed in near future, which will give an verification of the
surge shaft model. This will be done hopefully the last days of March and hopefully some results are ready
within 4.th of April .
References
[1] Statnett “FIKS”, 2012
[2] Working Group on prime mover and energy supply, ”Hydrualicturbine and turbine control models for
system dynamicstudies”, Transactions on power systems, Vol 7, No1, February 1992 p(167-179)
[3] Huimin. Gao, Chao. Wang, “Effect of detailed Hydro Turbine Model on Power System Analysis” PSCE
2006

Fulltext

Uploaded by

Copyright:

Available Formats

Fulltext

Uploaded by

Document Information

Original Description:

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

Fulltext

Uploaded by

Copyright:

Available Formats

Modeling and Simulation of Hydropower

Master of Energy Use and Energy Planning

Norwegian University of Science and Technology

Keywords: Hydropower modeling, block diagram, linearized Francis turbine, dynamic

Systemets ytese for trykktransienter og massesvinginger er også testet. Scilab modellen

Trondheim, June 15, 2017

List of Figures xvi

1.2 Thesis Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.3 System Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2.2 Paper Presented at the 7th International Symposium on Current Research

3.1 General Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

3.2 Transient Fudamentals . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

3.2.1 Pressure wave velocity . . . . . . . . . . . . . . . . . . . . . . . 9

3.2.2 Wave reflection time . . . . . . . . . . . . . . . . . . . . . . . . 10

3.2.3 Mass Forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

3.3 Friction Losses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

3.4 Surge shaft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

3.4.1 Surge shaft Oscillations . . . . . . . . . . . . . . . . . . . . . . 17

3.4.2 Surge Shaft Transfer Function . . . . . . . . . . . . . . . . . . . 18

3.5 Penstock and Tunnels . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

3.5.1 Dynamic penstock . . . . . . . . . . . . . . . . . . . . . . . . . 20

3.5.2 Approximation of Tanh . . . . . . . . . . . . . . . . . . . . . . . 20

3.5.3 RLC Penstock Model . . . . . . . . . . . . . . . . . . . . . . . 21

3.5.4 Static Tunnel Model . . . . . . . . . . . . . . . . . . . . . . . . 23

4 Modelling and Simulation Tools 25

5.1 Modeling scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

5.2 Hydraulic System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

5.2.2 Head Race Tunnel and Draft Tube . . . . . . . . . . . . . . . . . 34

5.3 Surge Shaft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

5.4 Surgeshaft validation model . . . . . . . . . . . . . . . . . . . . . . . . 37

5.6 PID - Regulator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

5.7 Servo System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

5.8 PID Tuning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

5.8.2 PID tuning by frequency analysis . . . . . . . . . . . . . . . . . 45

6 Simulation and Analysis 47

6.1 System Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

6.1.1 Scilab Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

6.1.2 LVTrans Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

6.2 Functional Requirement . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

6.3 Stability Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

6.4 Evaluation of Hydraulic Systems in Scilab . . . . . . . . . . . . . . . . . 60

6.4.1 Comparing Dynamic and Static Penstock . . . . . . . . . . . . . 60

6.4.2 The Effect of Extending Penstock Length . . . . . . . . . . . . . 62

6.5 Model Validation with LVTrans . . . . . . . . . . . . . . . . . . . . . . 64

6.5.1 Frequency regulation in island mode. . . . . . . . . . . . . . . . 64

6.5.2 Power Regulation in Stiff Grid Mode. . . . . . . . . . . . . . . . 65

6.5.3 Water Hammer and Mass Oscillations . . . . . . . . . . . . . . . 67

7.1 Laboratory Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

7.2 Results From the Rig . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

7.3 Rig results V.S Scilab model . . . . . . . . . . . . . . . . . . . . . . . . 76

8.1 Model Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

8.2 Model Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

8.3 Scilab VS LVTrans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

8.3.1 Frequency regulation in Island mode . . . . . . . . . . . . . . . . 82

8.3.2 Power regulation in Stiff Grid Mode . . . . . . . . . . . . . . . . 82

8.3.3 Mass oscillation and water hammer simulations . . . . . . . . . . 83

8.4 Surgeshaft Verification With Test Rig . . . . . . . . . . . . . . . . . . . 84

3.1 Typical surface roughness on different materials. [15] . . . . . . . . . . . 14

= Signal deviation [pu]