Solar Photovoltaic Output Smoothing:

Using Battery Energy Storage System

R P Sasmal1, Subir Sen2, Ankur Chakraborty3
Power Grid Corporation of India Ltd.
Gurgaon, Haryana, 122001
[email protected]

Abstract— Battery Energy Storage System (BESS) is widely authors [1]-[3] in the past have described the effect of
being implemented along with Solar PV to mitigate the inherent increasing Renewable energy penetration in the grid. Methods
intermittencies of solar power. Solar smoothing is one such such as use of Battery Energy Storage, use of dump loads and
application of BESS. In this paper, different techniques for solar curtailment of solar PV output power has been suggested to
power smoothing is compared. An energy compensation based
smoothing technique is proposed in this paper. The smoothing
reduce the fluctuations from solar power in the grid [4].
method not only ensures an optimal sizing of the battery but also In small grid the renewable intermittency brings in frequency
keeps the state of charge of the battery same at the beginning and changes. However, researchers have shown in case of
end of any random day. The different techniques are simulated on distribution networks with high R/X ratio increase of
a typical moving cloud day output of a 5 MWp solar power-plant. Renewable Energy penetration can lead to instability in the
network-voltage [2].
Keywords— Battery Energy Storage System (BESS), Smoothing, To handle the issues of renewable energy
State of Charge (SOC), Moving Window, Solar Photovoltaic, intermittencies Battery Energy Storage System (BESS) is being
Renewable Energy, Intermittency
implemented widely. Several studies dwell on design and
I. INTRODUCTION modelling of Battery Energy Management System for reduction
of power in solar PV [5]-[6]. Some researchers have also
With increasing penetration of renewable energy the suggested use of Hybrid Energy Storage System (HESS) along
grid is becoming increasingly weather dependent. The variable with Solar PV for power-fluctuation mitigation [7]. In this
cloud-cover or wind speeds on a typical day results in paper authors have proposed a super-capacitor bank along with
intermittency of renewable generations. This in turn leads to Vanadium Red-Oxide Battery as a part of the HESS.
permanent fluctuation of power at the terminals of conventional The different applications of BESS with Solar PV integration
generating units. The fast change in power requirement at the are energy time shift, frequency regulation and solar PV output
generator terminal causes higher stress of generator smoothing. In energy time shift application of BESS the solar
components and faster ageing and wear of components. The energy is stored in the battery during day time. The energy
sudden change in the power input to the grid has also stored in the battery is discharged during peak-load hours at a
detrimental effect on the connected load. later time of the day. The battery sizing is typically large for
In case of micro-grid systems, such as an electricity grid in an this application. The frequency regulation application of BESS
island the result of such variation may result in significant is to provide energy support to the grid by charging the battery
changes in frequency. In a grid with a high contribution of solar in over-frequency situations and discharging it while the grid
power, the fast changes in power output must be negated by frequency is less than the nominal frequency [8]. In this paper
changing the governor position of conventional generators. the authors describe a control strategy for BESS tailored for
This results into significant wear and tear of the governor Indian grid conditions. In smoothing application, the fast
system of the generators. Further, the quick variation of solar variations in solar PV output are negated using BESS. The
power through-out the day also means fast variation of loads on result is a smoother power output with a relatively small size of
the conventional generators. As a result, Generators are not in battery.
a steady thermal condition in the grid. Due to fluctuating From the above literature survey it is understood that,
thermal cycle and increased number of stop/start cycles the application of BESS along with solar PV is common for
thermal fatigue of the generators are increased which have frequency regulation, output smoothing and energy time shift.
adverse effect on the life of the conductors and insulation. In this paper the different strategies for solar PV output
When renewable energy penetration increases in a grid the smoothing is carried out. Further, the State of Charge (SOC) of
inherent intermittencies may result in voltage instability of the Battery at beginning and end of the day is also carefully
grid in case of weak distribution networks. considered. A smoothing technique which ensures an optimal
To mitigate the energy variation from solar power size of battery and almost same SOC at beginning and end of
output Battery Energy Storage System is being used. Several
978-1-4799-5141-3/14/$31.00 ©2016 IEEE

Authorized licensed use limited to: Universiti Teknikal Malaysia Melaka-UTEM. Downloaded on December 19,2021 at 18:27:51 UTC from IEEE Xplore. Restrictions apply.
 the day is proposed. The studies are carried out for a sample 5 The smoothing results are obtained for variable window size. It
MWp solar power-plant. may be noted that higher the window size smoother the curve
and lesser the variations in the output of the solar PV and
II. PROPOSED TECHNIQUES battery system. A curve is presented as follows for a 60 min
A. SIMPLE MOVING AVERAGE BASED SMOOTHING smoothing window period.
For solar-smoothing a moving window based
Figure 2 Smoothing Results with Simple Moving Average and
algorithm is proposed. At each point of time the algorithm
decides a target power. The difference between the target and
actual power output from the solar PV inverter is the battery

P (kW)
power.

P (kW) &
E (kWh)
Time (min)

60 min window period

The energy supply curve of the battery can be

inversely correlated with the State of Charge of the battery.
Higher the energy supplied lower the state of charge of battery.
It can be seen from the above curves that total energy supplied
by the BESS over a 24 hour period is zero.
It is further observed, with a higher averaging window size the
smoothing curve obtained is smoother. However, higher
window sizing results into a higher size of battery. The green
No curve in the second set of curves in the above figures the energy
requirement from the battery. With an increase in window size
from 30 min to 60 min battery sizing increases almost by 100%.
Hence the window size is to be judiciously chosen as a trade-
off between the smoothing desired and the battery size.
Yes
One major limitation of this method is the target power
always lags behind or leads ahead the actual power output
curve. As a result the difference in the power between the two
curves keeps on integrating over time and the energy sizing of
the BESS is much more. Higher the window size higher the
lag/lead and higher sizing of battery.
P (kW)

B. SINGLE AND HALF WINDOW BASED MOVING

AVERAGE
P (kW) &
E (kWh)

To tackle the problem of lag/lead in simple moving average

based algorithm a single and half window based moving average
is chosen. In this method moving average is calculated for the
Time (min)
whole window period and also for the second half of the window
The smoothing curves, obtained using simple moving period. Let, the average over the entire window period be x and
average from simulation, on a typical moving cloud day are average over the last half of the window period be y. Then the
presented in this section. estimated output at the end of window period will be y+(y-
x)=2y-x.
Figure 1 Smoothing Results with Simple Moving Average and The example of the following sequence can be taken for
30 min window period understanding purpose: {0,1,2,3,4,5,6,7,8,9}. In this case the
The figure presents two sets of curves. The first set of average over the entire period x=4.5 and over latter half of the
contain two curves (Red- Actual solar PV output & Blue-Total window period y=7. Hence the modified moving average is 2y-
System Output). The next set of curves represents Power x=9.5 which is closer to the real-time value of 9.
supplied by BESS (Red) and Energy supplied by BESS (green).

Authorized licensed use limited to: Universiti Teknikal Malaysia Melaka-UTEM. Downloaded on December 19,2021 at 18:27:51 UTC from IEEE Xplore. Restrictions apply.
The simulation result with this modified algorithm and 30 min minutes in the simulated problem). This energy is then
integration period is presented below. compensated over the next period of time.
The pseudo-code annexed to the algorithm can be written as
follows:
P (kW)

For (t=1 to end)

{
Revised Target Power (t) =Target Power (t) +Difference (t);
P (kW) &

If (t<=20)
E (kWh)

{Difference (t) =0 ;}
Time (min)
Else
{Difference (t) = [ ( − )−
Figure 3 Smoothing with Single and Half Window Algorithm ( − )];}

It can be observed the lag and lead of the previous method End if;
has been completely omitted in this method. Similar to the End For;
previous algorithm a smoother curve has been obtained with a
60 min window period. The results of the novel method are shown as follows:
P (kW)

P (kW)
Steady State
Difference in
Battery SOC
P (kW) &
E (kWh)

P (kW) &
E (kWh)

Time (min)
Time (min)

Figure 4 Smoothing with Single and Half Window and 60min Figure 5 Smoothing Results with Novel Method and with 30
Window Size. min Window Period

However, a closer scrutiny of the curves presents a hitherto

unprecedented problem. The single window and half based
P (kW)

moving average algorithm removes the lag and lead between the
target power and actual solar PV output. As the target power Smaller
closely follows the solar PV output the battery storage Steady State
requirement is also reduced. The problem with this method is Difference in
P (kW) &
E (kWh)

Battery SOC
that the SOC of the battery changes over the entire day. It is
observed from the curve that the net energy supplied over the
entire course of the day is non-zero. This means application of Time (min)
this technique makes the battery gain/lose some net charge over
the course of a day’s operation. Hence, the starting state of Figure 6 Smoothing Results with Novel Method and 60 min
charge (SOC) of the battery changes to a different level at the Window Period
end of the day. A change in the algorithm is hence proposed to
tackle this problem. III. RESULTS AND DISCUSSION
C. SINGLE AND HALF WINDOW BASED SMOOTHING The results of the different techniques used for solar PV
WITH ENERGY COMPENSATION output technique is compared in this section. The criteria on
which the fitness of algorithm are judged are Battery Sizing in
To negate the end of day change in SOC an integral control
terms of Energy and the final SOC of the battery.
is incorporated in algorithm. The change incorporated in the
algorithm monitors the difference in energy between target A. Simple Moving Average- 30 Min Window Period
energy and actual energy over a certain period of time (20
• Battery Sizing in terms of Energy: Maximum energy
exchange through BESS is 600 kWh.

Authorized licensed use limited to: Universiti Teknikal Malaysia Melaka-UTEM. Downloaded on December 19,2021 at 18:27:51 UTC from IEEE Xplore. Restrictions apply.
 • State of Charge: Unaltered at the end of the day. project, the benefits and suitability of two BESS technologies-
Lithium Ion and Lead Acid (250 kWh each) are being studied
B. Simple Moving Average- 60 Min Window Period extensively. Further, it is prudent to explore additional
• Battery Sizing in terms of Energy: Maximum energy applications of BESS, such as solar PV output smoothing which
exchange through BESS is 1200 kWh. will facilitate integration of Renewable Energy in the grid.
• State of Charge: Unaltered at the end of the day. The paper presents a comparative analysis of different solar
power smoothing techniques. The starting point is a simple
C. Single And Half Window-30 Min Window Period
moving window based algorithm using Battery Energy Storage
• Battery Sizing in terms of Energy: Maximum energy System. The simple Moving Window Average smooths the
exchange through BESS is 200 kWh. power output significantly but however presents with a high
• State of Charge: Changed at the end of the day sizing of battery. The reason for the high sizing of battery is that
D. Single And Half Window-60 Min Window Period the time difference between the actual power from PV and
target power. To reduce the lag a single window and half
• Battery Sizing in terms of Energy: Maximum energy moving average is proposed. This algorithm reduces the lag and
exchange through BESS is 300 kWh. consequently reduces the battery sizing. However, the SOC of
• State of Charge: Significant change at the end of the the battery changes significantly over a day’s operation. To
day. solve the issue an integral control based strategy is incorporated
E. Single and half window based with energy compensation- in the same algorithm. This has reduced the battery sizing as
30 min well as the change in SOC over a 24 period operation
The proposed methodology for smoothing is suitable for any
• Battery Sizing in terms of Energy: Maximum energy size of solar PV installation. Application of the novel-technique
exchange through BESS is 200 kWh. with similar window periods will result into similar smoothing
• State of Charge: No change in SOC. results and proportionate battery sizing.
F. Single and half window based with energy compensation-
V. ACKNOWLEDGEMENT
60 min
Authors are thankful to the management of
• Battery Sizing in terms of Energy: Maximum energy
POWERGRID for granting permission to present the paper.
exchange through BESS is 250 kWh.
Views expressed in this paper are of the authors only and need
• State of Charge: Small change in SOC.
not necessarily be that of the management of POWERGRID.
From the above discussion, it is clear that simple REFERENCES
moving average based smoothing results into a higher sizing of [1] C. Limsakul, R. Songprakorp, A. Sangswang and P. Parinya, "Impact of
battery. The battery sizing is typically 3-4 times the battery photovoltaic grid-connected power fluctuation on system frequency
sizing required with any other smoothing technique. deviation in contiguous power systems," Industrial Electronics Society,
With Single and half window based smoothing the battery IECON 2015 - 41st Annual Conference of the IEEE, Yokohama, 2015,
pp. 003236-003241.
sizing is reduced significantly. For a Solar power-plant of 5
[2] R. Yan, S. Roediger and T. K. Saha, "Impact of photovoltaic power
MWp the typical sizing for BESS is 200-300 kWh. This method fluctuations by moving clouds on network voltage: A case study of an
does not guarantee same state of charge at the end of the day. urban network," Universities Power Engineering Conference (AUPEC),
With the introduction of energy compensation in the above 2011 21st Australasian, Brisbane, QLD, 2011, pp. 1-6
problem the change of SOC is restricted. In the novel method, [3] R. Tonkoski, D. Turcotte and T. H. M. EL-Fouly, "Impact of High PV
Penetration on Voltage Profiles in Residential Neighborhoods," in IEEE
there is marginal change in Battery Sizing for 30 min and 60 Transactions on Sustainable Energy, vol. 3, no. 3, pp. 518-527, July 2012.
min integration period. However, there is small change in SOC [4] W. A. Omran, M. Kazerani and M. M. A. Salama, "Investigation of
for 60 min period using the novel technique, whereas there is Methods for Reduction of Power Fluctuations Generated From Large
no change in SOC for 30 min window period using the novel Grid-Connected Photovoltaic Systems," in IEEE Transactions on Energy
technique. Conversion, vol. 26, no. 1, pp. 318-327, March 2011.
Hence, it may be suggested that, the Single and a Half [5] K. Thirugnanam, S. G. Kerk, C. Yuen and B. Thirunavukarasu, "Battery
integrated solar photovoltaic energy management system for micro-grid,"
Window based Smoothing using Energy Compensation Smart Grid Technologies - Asia (ISGT ASIA), 2015 IEEE Innovative,
technique and a 30 min window period is most suitable for Bangkok, 2015, pp. 1-7.
Solar Smoothing application of a 5 MWp solar power plant. [6] N. Adhikari, B. Singh, A. L. Vyas, A. Chandra and Kamal-Al-Haddad,
"Analysis and design of isolated solar-PV energy generating system,"
IV. CONCLUSIONS Industry Applications Society Annual Meeting (IAS), 2011 IEEE,
Orlando, FL, 2011, pp. 1-6.
The Government of India has set an ambitious target [7] G. Wang, M. Ciobotaru and V. G. Agelidis, "Power Smoothing of Large
of generating 175 GW power by 2022 from Renewable Energy Solar PV Plant Using Hybrid Energy Storage," in IEEE Transactions on
Sources (RES), including 100 GW from solar. To accommodate Sustainable Energy, vol. 5, no. 3, pp. 834-842, July 2014.
the intermittencies associated with the RES, grid connected [8] I. S. Jha, S. Sen, M. Tiwari and M. K. Singh, "Control strategy for
storage is needed. As a proof of concept, POWERGRID is Frequency Regulation using Battery Energy Storage with optimal
utilization," 2014 IEEE 6th India International Conference on Power
implementing 500 kWh BESS at Puducherry with Frequency Electronics (IICPE), Kurukshetra, 2014, pp. 1-4
Regulation and Energy time shift applications. In the above

Authorized licensed use limited to: Universiti Teknikal Malaysia Melaka-UTEM. Downloaded on December 19,2021 at 18:27:51 UTC from IEEE Xplore. Restrictions apply.