Electrical Power and Energy Systems 77 (2016) 337344

Contents lists available at ScienceDirect

Electrical Power and Energy Systems

journal homepage: www.elsevier.com/locate/ijepes

Energy management at the distribution grid using a Battery Energy

Storage System (BESS)
Ehsan Reihani a, Saeed Sepasi b,, Leon R. Roose b, Marc Matsuura b
a
Renewable Energy Design Laboratory, Department of Mechanical Engineering, University of Hawaii at Manoa, 2540 Dole St., Holmes Hall 347, Honolulu, HI 96822, USA
b
Hawaii Natural Energy Institute, School of Ocean and Earth Science and Technology, University of Hawaii at Manoa, 1680 East-West Road, POST 110, Honolulu, HI 96822, USA

a r t i c l e i n f o a b s t r a c t

Article history: In 2008, the State of Hawaii initiated a clean energy initiative that set an ultimate goal of 70% clean
Received 19 December 2014 energy by 2030 (40% from renewable energy and 30% from energy efficiency). A controllable Battery
Received in revised form 16 September 2015 Energy Storage Systems (BESSs) can be used to manage intermittent renewable resources on a power
Accepted 17 November 2015
system to address both circuit and system level issues. Simulation and experimental results of applying
a novel algorithm for the charging and discharging of a BESS are presented, using actual grid data for con-
trolling a BESS for the purpose of peak load shaving, power curve smoothing, and voltage regulation of a
Keywords:
distribution transformer. Two optimization objectives for peak shaving are presented in which proposed
Battery energy storage system
Peak shaving
load forecasting methods are used. The application of a BESS for voltage regulation is examined and ana-
Power smoothing lyzed with different tests, and the observed results are discussed.
Voltage regulation 2015 Elsevier Ltd. All rights reserved.
Grid optimization

Introduction be avoided. Energy storage can also be used for peak demand shav-
ing on a particular distribution feeder transformer, with the objec-
The addition of renewable energy resources to power grids in tive to reduce the peak power demand on the transformer and
the U.S. has grown rapidly in recent years. Photovoltaic (PV) devices extend its useful life. The Battery Energy Storage System (BESS) is
are the fastest growing renewable category with a 60% growth rate, a battery equipped with bidirectional converters which can absorb
followed by wind power at 27% and biofuels at 18% [1]. The inherent or inject active and reactive power at the designated set points. In
intermittent nature of renewables poses some challenges to the this paper, an algorithm is developed to manage stored energy
continued expansion of their use due to limitations of existing con- and storage capacity effectively for peak shaving and load leveling
ventional generation facilities that are designed more for efficiency purposes and which considers estimates of future hourly pricing
than flexibility and existing transmission and distribution systems and renewable generation output.
that are designed for one-way power flows and load connection There is a growing number of research works which employ dif-
rather than generation interconnections. ferent storage technologies for dealing with the intermittency of
Energy storage is one of the ways to deal with the variability of renewables. In [2], different technologies used in battery energy
renewable resources. Energy storage devices can harvest excess storage systems deployed at the grid level are introduced. The opti-
energy during periods of low demand and inject the stored energy mal power and size of a hybrid energy storage system consisting of
when needed during peak usage periods. The storage devices can BESS and a high-speed superconducting flywheel energy storage
also play the role of reserve power plants, providing extra energy system are investigated in [3] for the purpose of stabilizing the
in case of power system contingencies or a rapid change in demand. power system. In [4], a real-time State of Charge (SOC) based con-
A popular use of energy storage is for system peak demand shaving, trol method is proposed to reduce the fluctuations in the power
which involves absorbing energy when there is excess energy, gen- system in response to a high level of integration of variable energy
erated either by renewables or base power plants, during off-peak sources such as PV and wind. The sizing of energy storage for
times and injecting the stored energy back into the distribution sys- micro-grids is examined in [5], where a neural network is used
tem during system peak load times. As a result, renewable genera- to forecast the PV and wind power generation levels, and the opti-
tion curtailment is reduced, and expensive fast generating units can mal size of BESS is determined with and without connection to the
main grid. In [6,7], a scheme consisting of wind generation in com-
Corresponding author. Tel.: +1 808 956 0767; fax: +1 808 956 2373. bination with a BESS is proposed for scheduling short-term power
E-mail address: [email protected] (S. Sepasi). dispatch to maximize the energy harvested from wind generation.

http://dx.doi.org/10.1016/j.ijepes.2015.11.035
0142-0615/ 2015 Elsevier Ltd. All rights reserved.
338 E. Reihani et al. / Electrical Power and Energy Systems 77 (2016) 337344

Different methods have been proposed for battery operation opti- nomial. A general fitting for a straight line to a first degree
mization and leveling the load profile. polynomial statement is as follows [15]:
In [8,9], dynamic programming techniques are used to find the
y a0 a1 x 1
optimal battery energy storage and power levels for peak load
shaving applications. Battery storage is examined in [10] for reduc- The residual is the difference of forecast and actual values:
ing transmission and distribution losses, and a set of normalized X
n
charts are provided to quantify the benefit of BESS for leveling R2 yi  a0 a1 xi 2 2
the utility load. Finally, in [11], BESS is used to regulate active i1
and reactive power according to SOC limits, and the control signals
Taking the partial derivative with respect to each coefficient ai
are fed into the switches using a current control loop.
and arranging in matrix form we get the Vandermonde matrix as
follows:
BESS 2 3
Xn
" #  6 yi 7
Here, a grid scale BESS (1 MW, 1 MWH) is connected to a distri- 1 x1    xk1 a0 6 i1 7
6
6X
7
7 3
bution feeder via a 1 MVA step-up transformer and is used for peak 1 x2    xk2 a1 4
n
5
shaving of the distribution grid circuit shown in Fig. 1. xi yi
i1
A 69 kV transmission grid provides the energy balancing needs
of the distribution circuit and BESS collectively via a 69/12.47 kV Rearranging Eq. (3) for yi gives the following:
distribution transformer. The goal of peak shaving is to optimally
2 3 2 3
control the BESS to reduce the peak load of the circuit. y1 1 x1
The BESS consists of twelve Li-ion battery racks and a master 6y 7 61 x2 7  
6 27 6 7 a0
6 76. .. 7 4
control rack. A single battery rack contains 22 trays (2 columns 6 .. 7 6 . 7 a
of 11) each populated with 38 prismatic flat pack cells and one Bat- 4 . 5 4. . 5 1
tery Management System (BMS) tray at the top. Together, these yn 1 xn
components form a 1 MW, 1 MW h energy storage system. The
The matrix shown in Eq. (4) can be written as follows:
BESS is connected to a 1 MW bidirectional three phase inverter
with 12,470 V AC output. The battery management system has a y Xa 5
SOC estimation algorithm, which estimates the amount of usable
electrical energy stored in the battery pack [12]. The SOC is limited Then the a coefficients can be calculated with a simple
to an operating range of 0.20.8 in which the battery is neither manipulation:
fully depleted nor fully charged [13,14], in order to avoid adversely  1
impacting the battery life. Control modes, set points, and active a XT X XT y 6
and reactive power commands are sent from the dispatch room
to the BESS controller using the Maui Electric supervisory control The paper discusses a parallel load forecasting method which
and data acquisition (SCADA) system utilizing the DNP3 protocol. is required for peak shaving of the load curve. The advantage of
In the context of a deregulated energy market system, a Distri- the proposed load forecasting method is its light computation
bution System Company (DISCO) can offer peak load shaving and burden. Two BESS control use cases are then evaluated and pre-
load smoothing services with optimal operation of a BESS under sented. The first use case focuses on a peak shaving method
its control at a market based price to the Independent System which presents fairly accurate performance since the magnitude
Operator (ISO). The ISO can in turn then utilize this DISCO provided of load uncertainty is low during the primary periods of BESS
resource to meet its system operational objectives, such as peak charging and discharging in the early morning and early eve-
demand shaving and operational reserves. ning hours. The method, however, may not perform as well in
time periods when PV generation variability is high. The second
BESS use case builds upon the first case by adding a power
Peak shaving
smoothing algorithm that utilizes an improved reference power
curve to address periods when PV production and power output
Peak shaving is used to reduce the peak demand on a power
variability is high, while maintaining the capability of peak
system, either at the balancing area as a whole or on a sub-
shaving.
system such as a distribution feeder. This can be accomplished
in several ways depending on the needs of the system and the
objectives of the strategy used. An example of this is to shift cur- BESS experiments
tailed renewable energy or lower priced energy generated during
times of low demand to periods of high demand to increase the In order to develop a good understanding of BESS operation on
utilization of renewable energy or reduce the use of more expen- the power grid, several charge/discharge experiments are per-
sive peak generating units. BESS are one of the emerging grid formed, and the electrical measurements from SCADA equipment
level options for shifting generation to when it is needed and at the distribution transformer are plotted.
smoothing the power fluctuations. In order to schedule the bat-
tery operation for the next 24 h, a forecast of the circuit power Active power flow
profile is needed. A linear regression method is used for obtaining
the power profile needed in the forecasting optimization In this experiment, BESS is charged with 50 kW steps. This
algorithm. test is done to figure out the impact of charging on the voltage
level and transformer Load Tap Changer (LTC) operation. The
Linear regression method step changes are kept small for safety purposes and also to see
the effect of incremental power changes on the grid. BESS and
In this method, the predicted value for each time step for n col- circuit active power and transformer voltage level and BESS
lected samples is calculated based on the least square fitting poly- SOC graphs are depicted in Figs. 2 and 3, respectively.
 E. Reihani et al. / Electrical Power and Energy Systems 77 (2016) 337344 339

1MVA
BESS
DC Transmission
1MW AC Grid
1.1MWHR
7.2 / 12.47KV 12.47 / 69KV
480V
Distribuon Grid

Fig. 1. BESS connection to distribution grid system.

In Figs. 2 and 3, 1blue and red graphs in the power measurements Combining Eqs. (8) and (9) into a single objective function, we
represent active and reactive power, respectively. In the voltage arrive at the following equation:
measurement plot, phases A, B and C are indicated by blue, red,
and green colors, respectively. Charging BESS draws current from P k Dt
J k1 SOC k  SOC max Lk Pk W k 10
the main grid, and there is a gradual voltage drop sensed by the Etot
LTC. As a result, the transformer compensates for the voltage drop Minimizing the objective function leads to taking a partial
by increasing its tap position, as observed in the abrupt changes in derivative with respect to Pk and summing over the planning hori-
the voltage graph. zon gives the following result:
!
Reactive power flow XN
Dt
Min 3Lk Pk 2 11
E
k1 tot
In order to analyze the impact of reactive power flow from the
BESS on the voltage level, reactive power tests are performed in Subject to SOC constraints:
which reactive power is injected and absorbed in 200 KVAR incre-
ments. Results from reactive power injection tests and the corre- 1 1
SOC min Etot  EK  < Pk < SOC max Etot  EK  12
sponding measurements are plotted in Figs. 4 and 5. Dt Dt
Voltage level changes measured at the lower side of the
69/12.47 kV distribution transformer of approximately 0.015 kV Ek1 Ek Pk Dt 13
for each incremental 200 KVAR injection and 0.06 kV for the total
800 KVAR test are recorded. Based on the observed test results, where Ek denotes BESS stored energy level at time step k. The
reactive power flow by the BESS does not have a significant impact objective function is nonlinear with a sequential quadratic pro-
on voltage regulation in the presence of the transformer LTC which gramming method which is used to obtain the BESS optimal
serves as the primary voltage regulation equipment for the distri- points. If there is an effective forecast of renewable energy gen-
bution circuit. Thus, it is preferred to utilize the BESS capacity for eration available, such as a wind forecast, the objective function
active power management. Remaining BESS capacity can, however, can be revised to make use of renewable generation for charging
be used to regulate the distribution circuit power factor. the battery rather than curtailing it in the off-peak time. This
can be done by defining a reference power curve, Pref, at the dis-
tribution transformer with the electricity price and renewable
Optimization algorithm
energy generation assumptions embedded in the reference
power curve. The revised objective function minimizes the error
The optimization algorithm finds the optimum active power
of active power flowing from the transformer and the reference
flow of the BESS at each time step. The objective function for peak
power curve using the battery capacity subject to SOC
shaving consists of two components, SOC cost and load cost which
constraints:
are to be minimized. The SOC cost at time step k + 1 is defined as
follows: " #
XN
 2
P k Dt Min Lk Pk  Pref k 14
J SOCk1 SOC k  SOC max 7 k1
Etot
It is up to the utility to define the reference power curve. A typ-
In the above equation, Pk denotes discharged power and SOCk is
ical reference power curve consists of two parts, peak time and off-
the value of the SOC of the BESS with capacity Etot at time step k.
peak time. The off-peak power profile makes use of a renewable
The cost of SOC, represented by JSOC, includes the updated SOC
energy forecast for charging the BESS. The peak time profile follows
value minus the maximum SOC value, SOCmax. The SOC cost is
setpoints obtained from optimal power flow. The overall flowchart
added to keep the battery full for as long as possible and incur a
of the algorithm is shown in Fig. 6.
cost for battery operation. The total cost at step k + 1 can be
The algorithm presented in Fig. 6 describes the peak shaving
described as a weighted load flowing through the distribution
method that is implemented in the BESS in Fig. 1. In the above
transformer Lk plus the SOC cost:
flowchart, Qres denotes the remaining reactive power injection
J k1 J SOCk1 Lk Pk W k 8 capability of the BESS after it has dispatched active power deter-
mined by the optimization objective function. If this value is
The weight component Wk is a quadratic function of load, which greater than the reactive power demand of distribution grid QL,
penalizes the high load passing through the distribution available reactive power capacity from the BESS is used to raise
transformer: the power factor of the circuit to the desired level. Although the
W k Lk Pk 2 9 optimization algorithm takes into account SOC changes on the
horizon, the real value of SOC is read from SCADA to account for
losses and inaccurate dispatch of the BESS. By following this tech-
1
For interpretation of color in Figs. 2 and 3, the reader is referred to the web nique, we make sure that the BESS is maintained within its desired
version of this article. SOC range.
340 E. Reihani et al. / Electrical Power and Energy Systems 77 (2016) 337344

Fig. 2. Circuit and BESS measurements in 50 kW active power flow test.

Fig. 3. Circuit voltage and BESS SOC measurements in 50 kW active power flow test.

Fig. 4. Circuit and BESS measurements in 200 kVAR reactive power flow test.

A potential concern of load shifting performance by a BESS charging and discharging cycles are easily managed by the
is the impact of the power flow changes on the regulation of system Load Frequency Control (LFC) function through dis-
grid frequency. In the case of the BESS evaluated here, the patch of the conventional generators. Moreover, a frequency
amount of power injected or absorbed is insignificant from a error signal can be embedded in the optimization algorithm,
perspective of overall system generation load balance. Thus, by which part of the deviation can be corrected by the BESS
any minor frequency deviations that result from the BESS power flow.
 E. Reihani et al. / Electrical Power and Energy Systems 77 (2016) 337344 341

Fig. 5. Circuit voltage and BESS SOC measurements in 200 kVAR reactive power flow test.

Simulation and results The SOC trajectories also have some discrepancies due to some
nonlinearities of the BESS and also some errors from the SOC esti-
Load forecasting simulation results mation subsystem in the BESS. In order to prolong the battery life,
SOCmin and SOCmax parameters are set to 0.2 and 0.8, respectively.
Linear regression method for next day load forecasting is The objective function defined in Eq. (11) tries to flatten the overall
applied to 108 days of historical data with 1 min resolution. The power curve by finding the BESS power setpoints considering the
forecast data point (y) for each time step (X) is obtained by insert- forecasted load. As a result, the BESS is charged when the load is
ing the given time step in Eq. (5). If the load curve for 24 h is given low (early morning) and discharged when the load is high (early
with 1 min resolution, the above equation should be executed evening). In the case of conducting our test, the BESS SOC is near
1440 times to obtain a forecast value for the next day. The load 80% at the start of the test, there is not a significant power flow into
forecasting algorithm is performed on 14 weeks of data and the the BESS until 6:00 AM. Then, from approximately 6:00 AM to 9:00
predicted weekday in the 15th week is compared with the actual AM, the peak shaving algorithm called on the BESS to discharge a
value. The load forecast and the corresponding actual load for a little to reduce an early morning peak demand on the feeder and
weekday at the distribution transformer are shown in Fig. 7. approach the optimized profile. PV fluctuations change the load
It can be seen in Fig. 7 that the linear regression approach does profile between the hours of 9:00 AM and 6:00 PM when the fore-
not perform well in predicting the fluctuation caused by PV cast and actual loads vary quite significantly. The evening circuit
resources in the circuit around 12:00 PM to 3:00 PM when PV pro- peak demand is then shaved well from 6:00 PM until about
duction is highest. However, we are not concerned about the devi- 11:00 PM.
ation of the forecast value from the actual value during this time Another scenario can also be considered where for some reason
period when PV fluctuations are high because the load shifting the BESS is unavailable to shave the peak circuit load. In this case,
function of the BESS does not occur during these hours. The fore- normal dispatch of thermal generation by the grid Energy Manage-
cast value is very close to the actual load during the BESS load ment System (EMS) operates to pick up the load. Other peak shav-
shifting charge and discharge cycles which coincide with times of ing methods implemented in an EMS such as load management
low to no PV generation on the grid. The Root Mean Square Error can also be dispatched to reduce circuit peak demand in coordina-
(RMSE) for this forecast is 49.36. tion with action of the BESS.
In order to effectively address the high variability of the load
profile during periods when there is high PV production and
Peak shaving simulation and infield test results fluctuation, a real time smoothing scheme can be used. Since
real-time measurements of active power in the transformer are
The optimization algorithm is applied to a 1 MW/1 MW h BESS available in the dispatch room, an active power setpoint is defined
in the circuit shown in Fig. 1. The time step is taken once every for the BESS. Any deviation from this setpoint is compensated for
15 min for a total of 96 steps in a 24 h planning horizon. The first by charging/discharging the difference in power to maintain the
optimization algorithm shown in Eqs. (11)(13) is applied to the defined level. The charge/discharge of the BESS away from this
forecast load for both smoothing and shaving the peak of the level should be almost equal to keep the SOC level needed in early
power curve. The shaved peak, load, BESS, optimized active power evening for peak shaving. However, to ensure that the BESS SOC is
profile, and BESS SOC are depicted in Figs. 8 and 9, respectively. The at its desired level based on the peak load shaving algorithm,
output of a PV inverter in the circuit is plotted in Fig. 10. power smoothing capability is suspended at 5:00 PM, one hour
The optimization simulation is done with an initial SOC of 70%. in advance of the start of anticipated peak shaving, to allow the
Due to lack of communication with SCADA until 12:00 PM, the BESS an opportunity to recharge. Simulations of this power
BESS is operated according to optimization points after this time, smoothing algorithm are performed. The load curve, accompanied
and thus SOC goes above 80% for a short time to comply with the by PV fluctuations, along with the BESS active power setpoint, is
obtained power points. As can be seen from Fig. 10, there is a small depicted for a sample day in Fig. 11.
amount of power generation from PV resources causing the load This feature is useful for reducing the system regulating reserve
curve to ramp up from 12:00 PM. The load forecast cannot predict to the degree that the BESS can flatten the fluctuation and thus
the stochastic variations caused by weather conditions (e.g. cloud minimize the operational and cost burden on thermal generation
movement) on the power profile. However, the optimized and regulating grid frequency. In order to find the optimum number
shaved power profile curves are very close after 5:00 PM, and the of smoothing levels, the maximum and average SOC error for 10
peak is shaved even better than the expected curve. smoothing levels is plotted in Fig. 12.
342 E. Reihani et al. / Electrical Power and Energy Systems 77 (2016) 337344

Fig. 6. Flowchart of peak shaving method.

The average value for the above 10 active power setpoints is the overall effectiveness of the peak shaving objective. The opti-
771.67 kW. If this value is applied across an entire week, the max- mization algorithm for the second method in Eq. (14) is applied
imum SOC error at 7:00 PM, one hour into the peak shaving period, to the circuit, for which a reference power curve is provided. The
is 13%. The BESS can effectively compensate for this 13% shortfall in BESS tries to follow the reference power curve considering SOC
SOC by charging for approximately 40 min with minimal impact to constraints. For example, it is preferred to charge the BESS at a
 E. Reihani et al. / Electrical Power and Energy Systems 77 (2016) 337344 343

Fig. 7. Actual and forecast load curves using linear regression method.

Fig. 10. Power output of an inverter in the circuit.

Fig. 8. Power curves for first optimization algorithm.

Fig. 11. Distribution circuit load curve with PV production and BESS active power
setpoint.

Fig. 9. Actual and forecast SOC curve for first optimization method.

constant rate from 2:00 AM to 5:00 AM and discharge it from 6:00

PM to 11:00 PM. The utility can define the reference power curve
based on the optimal power flow in the grid. The BESS power out-
put and SOC values are depicted in Figs. 13 and 14, respectively.
The SOC value rises up in the charging time interval based on
the duration defined and drops sharply in the peak time interval Fig. 12. Maximum and average SOC error for 10 smoothing levels.
to meet some of the demand. It remains constant in other time
intervals as the reference power curve is defined as the forecast
load.
This approach uses a reference load following algorithm in charging of the BESS can decrease its lifetime, which makes the
which operational and planning constraints can be embedded first approach more desirable [16]. Moreover, after getting real
and used for defining the reference power curve. Considerations data from SCADA, the planning can be updated for the next time
such as load forecast, demand response, and reserve scheduling horizon. The objective of the first method is to flatten the load
can be easily integrated into the reference power curve and thus curve with the BESS considering the imposed constraints. The
make it a better approach. On the other hand, charging and dis- disadvantage of the first method is its vulnerability to growing
344 E. Reihani et al. / Electrical Power and Energy Systems 77 (2016) 337344

smoothing. In this application (bulk storage at the substation end

of a feeder), the BESS does not have much impact on the feeder
voltage, but can be used to serve the VAr load on the circuit and
reduce the VAr load on the system. Two optimization methods
for peak shaving are introduced and the resulting power curves
are discussed.

Acknowledgement

The authors would like to gratefully and sincerely thank Chris

Reynolds for his assistance, patience, time, and, most importantly,
his friendship during BESS tests at the Maui Electric Company.

References
Fig. 13. Power curves for second optimization algorithm.
[1] Sioshansi FP. Smart grid: integrating renewable, distributed and efficient
energy. Academic Press; 2011.
[2] Sparacino A, Reed G, Kerestes R, Grainger B, Smith Z. Survey of battery energy
storage systems and modeling techniques; 2012.
[3] Lee H, Shin B, Han S, Jung S, Park B, Jangi G. Compensation for the power
fluctuation of the large scale wind farm using hybrid energy storage
applications. IEEE Trans Appl Supercond 2012;22(3).
[4] Xiangjun L, Hui D, Xiaokang L. Battery energy storage station (BESS)-based
smoothing control of photovoltaic (PV) and wind power generation
fluctuations. IEEE Trans Sustain Energy 2013;4(2):46473.
[5] Chen S, Gooi H, Wang M. Sizing of energy storage for microgrids. IEEE Trans
Smart Grid 2012;3(1):14251.
[6] Teleke S, Baran M, Bhattacharya S, Huang A. Optimal control of battery energy
storage for wind farm dispatching. IEEE Trans Energy Convers 2010;25
(3):78794.
[7] Yao D, Choi S, Tseng K, Lie T. Determination of short-term power dispatch
schedule for a wind farm incorporated with dual-battery energy storage
scheme. IEEE Trans Sustain Energy 2012;3(1):7484.
[8] Oudalov A, Cherkaoui R, Beguin A. Sizing and optimal operation of battery
energy storage system for peak shaving application; 2007.
[9] Riffonneau Y, Bacha S, Barruel F, Ploix S. Optimal power flow management for
grid connected PV systems with batteries. IEEE Trans Sustain Energy 2011;2
Fig. 14. Actual and forecast SOC curve for second optimization method. (3):30920.
[10] Nourai A, Kogan V, Schafer C. Load Leveling reduces T&D line losses. IEEE Trans
Power Deliv 2008;23(4):216873.
[11] Grid-connected lithium-ion battery energy storage system for load leveling
uncertainty in the grid especially with higher integration of and peak shaving; 2013.
distributed renewable generations. [12] Sepasi S, Ghorbani R, Bor Yann L. SOC estimation for aged lithium-ion batteries
using model adaptive extended Kalman filter; 2013.
[13] Sepasi S, Ghorbani R, Bor Yann L. Improved extended Kalman filter for state of
Conclusion charge estimation of battery pack. J Power Sources 2014;255(1):36876.
[14] Sepasi S, Ghorbani R, Bor Yann L. A novel on-board state of charge estimation
method for aged Li-ion batteries based on model adaptive extended Kalman
In this paper, BESS is investigated for use in peak shaving and
filter. J Power Sources 2014;245(1):33744.
voltage regulation of a distribution feeder. Several experiments [15] Ostertagov E. Modelling using polynomial regression. Procedia Eng 2012;48:
are carried out on the BESS and measurements obtained by SCADA 5006.
[16] Sepasi S, Roose LR, Matsuura MM. Extended kalman filter with a fuzzy method
are analyzed. Application of BESS for peak shaving, voltage regula-
for accurate battery pack state of charge estimation. Energies 2015;5
tion, and power smoothing is studied and it is shown that the BESS (6):521733.
capacity can be used effectively for peak shaving and power