IEEE Transactions on Energy Conversion, Vol. 10, No.

3, September 1995 555

EFFECT OF BATTERY ENERGY STORAGE SYSTEM ON LOAD

FREQUENCY CONTROL CONSIDERING GOVERNOR DEADBAND AND
GENERATION RATE CONSTRAINT

Chun-Feng Lu Chun-Chang Liu Chi-Jui Wu, Member

Department of Electrical Engineering Department of Electrical Engineering
National Taiwan University National Taiwan Institute of Technology
Taipei, Taiwan, ROC Taipei, Taiwan, ROC

Abstract- Since a battery energy storage system (BES) can One alternative to improve the reliability of supply
provide fast active power compensation, it also can be used to during peak load period is the use of storage facilities and
improve the performance of load-frequency control. In this especially a battery energy storage system (BES) [l, 21.
paper a new incremental model of a BES is presented and Energy is charged into the BES during off-peak load period
merged into the load-frequency control of a power system. A (at night), and released from the BES during peak load
comprehensive digital computer model of a two-area period (at day). The energy eficiency is expected to be
interconnected power system including governor deadband and greater than 0.8 from night to day [3]. A BES not only has
generation rate constraint is employed for a realistic response.benefits in peak-shaving and load leveling, but also in
Computer simulations show that the BES is very effective in dynamic improvement, such as voltage and frequency
damping the oscillations caused by load disturbances. The regulation, area regulation, and long line stabilization. The
BES model is suitable for charging mode and discharging mode applications of a BES have received great attention since the
operations. Optimization of controller gains is obtained by the successful demonstration at a 17 MW BES facility in Berlin
second method of Lyapunov. [4] and a 10 MW/40 MWh facility in California [5].
Since a BES can provide fast active power compensation,
Keywords: battery energy storage, modeling, stability, power it also can be used to improve the performance of load-
generation control frequency control (LFC). The LFC problems fundamentally
are that of sudden small load perturbations which
INTRODUCTION continuously disturb the normal operation of a power system
[6]. To balance power supply and load demand at dynamic
The Taiwan Power Company (TPC) is the only power periods, the automatic generation control (AGC) in some
utility in Taiwan, which serves all of residential, commercial, hydro and/or thermal power plants is synonymous with the
and industrial customers. In 1993 the system had a peak load LFC. However, the BES can give fast energy storage action
of 17670 MW and an average load of 11620 MW. The peak to help the LFC to repress the system oscillations after load
load of TPC occurs at day in summer, where air- perturbations. In this paper investigation on the effect of a
conditioning load is about 30%. The system reserve capacity BES on the LFC has been presented and the method to use a
is reduced to 8% in recent years. The improved GNP and small size is suggested.
standard of living create a continuously growing system peak Since linearized models are used in the investigations of
load. How to ensure affluent and reliable electric power load-frequency control [7], this paper derives an incremental
supply during the day time in summer is a challenging model for the BES operating at charging mode and
mission of TPC. discharging mode. A two-area interconnected power system
is considered for the computer simulation studies. The power
95 WM 086-9 EC A paper recommended and approved system model considers governor deadband and generation
by the IEEE Energy Development and Power Generation rate constraint (GRC) to obtain realistic responses. The
Committee of the IEEE Power Engineering Society f o r results show that the dynamic performance of the system can
presentation a t the 1995 IEEE/PES Winter Meeting,
January 29, t o February 2, 1995, New Pork, !iY. be greatly improved and overshoots of the dynamic
Manuscript submitted August 1 , 1994; made available frequency deviations can be reduced by the BES.
f o r printing January 3, 1995.
INCREMENTAL BES MODEL

The fundamental configuration of an equivalent BES

system is shown in Fig. 1, that containing an equivalent
battery composed of parallel / series connected battery cells,

0885-8%9/95/$04.00 8 1995 IEEE

556
a 12-pulse cascaded bridge circuit connected to a Y 1A - Y RBI = overvoltage resistance
transformer, and a control scheme. The ideal no-load
C,, = overvoltage capacitance
maximum DC voltage of the 12-pulse converter is expressed
as RBp = self discharge resistance
CBp = battery capacitance
We can obtain the expression of the DC current flowing into
where Vt is the line to neutral r.m.s. voltage. The terminal the battery from the equivalent circuit analysis as
voltage of the equivalent battery is obtainedfrom E
'T - B' OC - vBI
IBES = ( 3 )
RBT +RBS
where
-- 6
- 3& Vt(COSal +COSa2)--XcoIBES ( 2) V,, = RBP
( 4 )
IC IC IBES
where ~+SRBPCBP
Xco = commutating reactance
IBES = DCcurrent flowing into equivalent battery
ai = firing delay angle of converter i According to the converter circuit analysis, active power and
The equivalent circuit of the BES can be represented as a reactive power absorbed by the BES are [ 101
converter connected to an equivalent battery as shown in Fig. 3&
2 with a same cosine value of the firing delay angles in equ. PBEs =-Vt IBEs(COSal + C O W 2 ) ( 6 )
a
(2). In the battery equivalent circuit [8, 91
VBoc= battery open-circuit voltage Q B E ~= -Vt
3& IBES(SINa, + SINa2) (7)
a
VBl = battery overvoltage There are two control strategies
RB,= connecting resistance (1) P-Q modulation : a,= a2 = a
RBS = internal resistance 6&
PBES= -Vt IBES (COSa)
a
3 0POWER
SYSTEM

II
12-PULSE - (2) P modulation : a , = -a2 = a
BRIDGE BATTERY
CONVERTER
PTICT
= EmIBESCOSa = VCOI,,~ (10)
- CONTROL
QBES = O
where Vco=EmCOSa is the DC voltage without overlap.
SCHEME
Since only incremental active power is considered in load-
frequency control, we select P modulation in this paper.
Fig. 1. Fundamental Configuration of BES in power system. Linearization of equ. (10) gets the incremental power of the
BES :
uBES = v&A , IBE~ +I;ESAVCO (12)
The AIBES will be negative when the BES is under
continuous charging mode due to the increment of battery
- I BES
'BI
RBP E:
+ voltage AVm and AVB1. Although constant current
operating mode is the most efficient for the BES, we should
CBP ~ ~ v ~ o c
adjust the firing angle a, that is AVco in equ. (12), to keep
the BES in constant power mode for the sake of load-
frequency control. We decompose AVco into two
components: (a) E m AV, to compensate the power
Fig. 2. Equivalent circuit of BES
 557

deviation caused by AI,, , and (b) EmAVs to respond the value, that is, the direction of initial current within the
system disturbance. We can obtain BES. Since there are DC breakers to prevent too high
hPBEs = v&)AIBES (Avf currents which would endanger battery service life, the
deviation of battery current is limited. There is also a limiter
= &E,AV~ ( 13 1 upon AVs due to Vco 5 E.,
if we let
STUDIED SYSTEM
'BES
In order to study the effect of the BES, a digital computer
Then the use of the BES in LFC is obtained by a damping
model for LFC of a two-area system with governor deadband,
signal AVs. generation rate constraint (GRC), and the BES in area 1 is
AVs = KBP ASignal shown in Fig. 4. The computer programs are developed
1+ STBp with the standard FORTRAN language. Time domain
where KBP and TBP are the control loop gain and the simulations are conducted with a fourth-order Runge-Kutta
measurement device time constant, respectively. The method.
A Signal is a useful feedback from the power system in order
to provide damping effect. Combination of above equations,
the incremental model of the BES is shown in Fig. 3.
The discharging mode operation of the BES also can be
...............................................
expressed by Fig. 3. We can use the ignition angle p for the AIBES
converter in discharging mode [lo]. The power consumption vt i +
. -
of the BES is

!
i
6&
PBEs =-vtIB,Scosp, p= n - a APBES
4

n 4

= -E,IBEsCOSa = -VC~IBES (16)

The similar result is obtained.
DBES = -I;ESE,AVS (17)
The incremental BES model as shown in Fig. 3 can identify BATTERY
charging mode or discharging mode according to the sign of Fig. 3. Block Diagram of BES incremental model

L 1 I

@ @ Dead Band

Governor j :-
GRC
.........................................................................................
Reheat Steam Turbine

-w
I
2pJl ............................................................... I
Dead Band

EH-4: 1 i
1
Il+SKr2TR1
l+STr2 I
Govemor :
: ......................................................................................... ApD2
Reheat Steam Turbine

Fig. 4. Block diagram of LFC with governor deadband, generation rate constraint, and a BES
558

Governor Deadband According to the analysis in [13], if the maximal load

perturbation is 0.005 p.u.MW, a BES with power rating
Describing function approach is used to incorporate the 2000 MWx0.005 = 10 MW is needed. In this paper, a 10
governor deadband nonlinearity. It has been found that the MW/40 MWh BES is applied. The behavior of the power
backlash nonlinearity tends to produce a continuous system with the BES in charging mode and frequency
sinusoidal oscillation with a natural period of about 2 deviation feedback under the same disturbance is shown in
seconds. An approximate solution of Fourier series [ 111 has Fig. 6. It can be seen from Fig. 6 and the third column of
been developed as follows: Table 1 that the damping is greatly improved with the
DB, =NIX+-% N2 addition of the BES. The maximum overshoots of Af, , Af, ,
00 and AFj2 are reduced approximately 94%, 96%, and 90%,
where DB is the deadband. respectively. Fig. 6 also shows the deviations of open-circuit
The Fourier coefficients of a backlash of 0.05 % are voltage, DC current, and power in the BES. Since the BES
obtain as: may be operated in discharging mode during peak-load
N, = 0 . 8 and N, =-0.2 period, Fig. 7 shows the responses with the same disturbance
and the BES in discharging mode. The discharging mode
Generation Rate Constraint (GRC) BES also gives good damping effect.
In multiarea power systems it is desirable that each area
In the real situation, there is a maximum limit to the rate could carry its own load, least participating in the
of change in generation power of a steam plant. It has been
shown in [ 121 that the dynamic responses of the system with
the presence of GRC have larger overshoots and longer
settling times, compared to the system without considering
GRC. Furthermore, if the parameters of the controller are
not chosen properly, the system may become unstable. A
1 - lO
4.02
20
n w : SEC
40
4.02:k-! 4.00:F-l
0 20
TIME:SEC
40 0 20
TIME:sEc
40

generation rate limitation of 0.1 p.u. per minute [12] is

considered here, i.e.,
hpp I O.lp.u./min= 0.0017 p.u./sec A P s l (P.U/SEC)
4.002 4.002
20 40 0 20 40 20 40
TIME : SEC TIME : SEC nuE:sEc
EFFECT O F BES ON LFC LOOP 0.001 I I 0.001 , ,

Fig. 5 shows the deviations of system variables without

the BES in response to a step load change of 0.005 p.u.MW A I BES( P.U.) A P B E S ( P.U. )

in area 1. The integrator gains (K,, and K,,) setting are

determined by least square error criterion as illustrated in Fig. 6Responsesof the power system with charging modeBES and Ni
Appendix A. The curve for the performance index versus feedback
K, is also shown in Fig. 5 . It can be seen that the optimal

4,00:p
integrator gain is 0.48. The generation rates of both
generators reach lower and upper bounds such that system
oscillations sustain for a long period. The maximal
overshoots are large as shown in Table 1.
1 - bO
4.02
20
TIME : SEC
40
4.02;pzJ TIME
20: SEC 40 TIME:sEc
20 40

A P g l (P.U/SEC)
4.002 4o 4.002 4.002
0 20 20 40 20 40
TIME : SEC TIME:SEc TIME:SEC
TIME : SEC TIME :SEC TIME:sEC
0.002 0.002

PERFORMANCE
0 0 INDEX
A I BES(P.U. ) APBES (P.U.)
0.001 4.003
0 20 40 0 20 40
-0.002 4.002 0.0005 TIME :SEC TIME. SEC
20 40 0.1 0,s 0.9
TLME : SEC TlME : SEC UTEGUTOR GAIN : KI
Fig. 7 Responses of the power system with dischargingmode BES and
Fig. 5 Respomes of the power system without BES feedback
 559

Table 1 Overshoot with different control schemes

Hz'Hz 1 0.01804
100% I 0.00068
3.76% 1 0.0006835
3.79% I 1
0.0001996 0.0001996
1.106% 1.106% I
I I I 1 I I
.o.Or--
Apt2 ,P.U. 0.002378 0.000236 0.000234 0.0001562 0.0001562
100% 9.924% 9.837% 6.567% 6.568% 0.001 0.001 1

disturbances of other areas. This reduces the inadvertent

APBES(P.U.)
interchange of power between areas during frequency 4.003
m U) m
Tu(E:sEc
oscillations [ 6 ] . Since the deviation of tie-line powers and 'Iu(E:sEc
Fig. 8 Responses of the power system with charging mode BES and ACE
area frequencies are available, we can use the area control feedback
error (ACE) as the feedback signal for the BES control
scheme.

Fig. 8 and Fig. 9 show the power system behaviors with

the BES and ACE feedback in charging mode and
discharging mode, respectively. The maximal overshoots
are given in Table 1. The damping of power system is very
good. By comparing Figs. 6 and 7 with Figs. 8 and 9, the APd(P.UISEC) APgl (P.U/SEC) A V Boc (P.U.)
use of ACE feedback greatly reduces the deviation of Af2 -0.002
0 m 40 -0.002
m 40 -0.002
m 40
=:SEC 5IME:sEc 'IIME:sEc
and AP,, . Area 2 participates slightly since the disturbance
is in area 1. Fig. 10 shows the responses when there is a
0.005 p.u.MW load disturbance in area 2. Only the BES in
discharging mode is examined. Although the BES with Afl -0.003
m m
feedback provides good damping effect, it expands the tie- -:SEC TuIE:sEc
Fig.9 Responses of the power system with discharging mode BES and
line power deviation. This can be avoided by the ACE ACE feedback
feedback as shown in Fig. IO(c). It can be concluded that
the use of ACE feedback overcomes the energy storage unit
expanding the tie-line power deviation in case of load
disturbance in the other areas. The use of ACE feedback
introduces the best dynamic behavior and lowest overshoots.
Although with the BES, the optimal setting of the
integrator gain is altered to a higher value, that is, the
optimal value of K, is changed from 0.48 to 1.09, it is
recommended to keep the already setting without change.
The effect of optimal K, is reduced due to insensitivity of
the system to K,. It has been approved from the flatter of
performance index versus the integrator gain for the system
with the BES. Unstable oscillations caused by GRC effects
due to overaction while the BES is out of service should be
avoided. It has been examined, but is not revealed here, that
the system with a higher integrator gain but without the BES
will be unstable under the same disturbance.
4.025 o~wom AP12 (P.UMW )
560
REDUCTION IN BES SIZE ACKNOWLEDGMENT

The consideration of BES size includes BES MWh The authors would like to thank the National Council of
capacity and converter MW rating. With same converter ROC in financial support, project number NSC 83-0404-E-
MW rating, a BES of different MWh rating has different 011-002.
charging or discharging time. But the damping effect of the
BES is mainly determined by the converter MW rating. It REFERENCES
has been examined that a 10 MW120 MWh BES can provide
nearly the same damping force with that in Figs. 6-10. 1. J. L DelMonaco, P. k Lewis, H. T. Roman, and J. zemkodd, "Advanced
Fig. 11 shows the system responses of a 0.005 p.u.MW Batteries for Load-Leveling; the Utility Perspective on System
load disturbance in area 1 and a 5 MWl20 MWh discharging Integration", IEEE Trans. on Power Apparatus and Systems, Vol. 101, No.
mode BES with ACE feedback. Although a smaller size 9, pp. 3315-3321, 1982.
BES is used, the damping effect is still very good. It is 2. C. E. Lin, Y. S. Shiao, C. L. Huang, and P. S. Sung, " A Real and Reactive
concluded that for an area of capacity 2000 MW, a 5 MW120 Power Control Approach for Battery Energy Storage System", IEEE
MWh BES is sufficient if the maximal load disturbance is Trans. on Power Systems, Vol. 7, No. 3, pp. 1132-1140, 1992.
0.005 p.u.MW 3. A Pivec, B. M. Radimer, and E. k Hyman, "Utility Operation of
Battery Energy Storage at the BEST Facility", IEEE Trans. on Energy
CONCLUSIONS Conversion, Vol. 1, No. 3, pp. 47-54, 1986.
4. H. J. Kunisch, K. G. Kramer, and H. Dominik, "Battery Energy Storage,
An incremental model of BES is proposed for Another Option for Load-Frequency-Control and Instantaneous Reserve",
investigating its application in load-frequency control. It has IEEE Trans. on Energy Conversion, Vol. 1, No. 3, pp. 4146,1986,
been observed in the present work that a BES with simple 5. L. H. Walker, "10 MW GTO Converter for Battery Peaking Service",
control can effectively reduce frequency and tie-line power IEEE Trans. on Industry Applications, Vol. 26, No. 1, pp. 63-72, 1990.
oscillations following sudden small load disturbances. Both 6. S. Banerjee, J. K. Chatterjee, and S. C. Tripathy, "Application of Magnetic
frequency deviation and area control error are taken as the Energy Storage Unit as Load-Frequency Stabilizer", IEEE Trans. on
feedback signals. Simulation results show that the area Energy Conversion, Vol. 5 , No. 1, pp. 46-5 1, 1990.
control error is suitable for this purpose. It can be concluded -
7. 0. I. Elgerd, Electric Energy Systems Theory An Introduction,
that applications of a BES to LFC of interconnected power McGraw-Hill Publishing Co. Ltd., 2e, 1982.
systems will provide great improvement in system 8. Z. M. Salameh, M. k Casacca, and W. A. Lynch, "A Mathematical Model
performance. for Lead-Acid Batteries", IEEE Trans. on Energy Conversion, Vol. 7, No.
Further research effort is necessary in the addition of 1, pp. 93-97, 1992.
experimental data to compare the digital computer 9. M. A Casacca and 2. M. Salameh, "Determination of Lead-Acid Battery
simulation results. The issues about mobile battery energy Capacity Via Mathematical Modeling Techniques", IEEE Trans. on
storage systems used in electric vehicles are also suggested Energy Conversion, Vol. 7, No. 3, pp. 442-446, 1992.
for future study. Their effects on load leveling, voltage and 10. E. W. Kimbark, Direct Current Transmission, Volume 1, The John Wiley
frequency regulation, and peak-shaving are also important. & Sons,Inc. 197 1.
1 1 . S. C. Tripathy, G. S. Hope, and 0. P. Malik, "Optimization of Load-
Frequency Control Parameters for Power System with Reheat Steam

1 - IO
-0.02
20
TIME SEC
40
o
.o
. z50p-l

20
TIME : SEC
4o
op7zJ
-0.005
20
TIME. SEC
40
Turbines and Govemor Deadband Nonlinearity", IEE Proc., Vol. 129, h.
C, NO. 1, pp. 10-16, 1982.
12. W. C. Chan and Y. Y. Hsu, "Automatic Generation Control of
Interconnected Power System Using Variable-Structure Controllers", IEE
0 002 0 002
Proc., Vol. 128, Pt.C, No. 5, pp. 269-279, 1981
0 0 13. D. Kottick, M. Blau, and D. Edelstein, "Battery Energy Storage for
-0 002 4 002
A& (P.U/SEC) Frequency Regulation in an Island Power System", IEEE Trans. on
0 20 40 0 20 40
TIME. SEC TIME : SEC
Energy Conversion, Vol. 8, No. 3, pp. 455459, 1993.
0.001 0.001 T I
14. G. E. Gareis, D. P. Carroll, C. M. Ong, and P. Wood, "The Interaction of
Battery and Fuel Cells with Electrical Distribution System : Force
Con~lutatedConverter Interface", IEEE Trans. on Power Apparatus and
Systems, Vol. 96, NO. 4, pp. 1242-1250, 1977.
-0 003
0 20 40 0 20 40
TIME SEC nm SEC
15. J. W.Beck, D.P. Carroll, G. E. Gareis, P. C. Krause, and C. M. Ong, "A
Computer Study of Battery Energy Storage and Power Conversion
Fig. 11 Responses of the power System when the BES size is reduced to Equipment Operation", IEEE Trans. on Power Apparatus and Systems,
5 MW/2O MWh.
Vol. 95, NO. 4, pp. 1064-1072, 1976.
 561
APPENDICES In this paper, the undetermined parameters a and P
Appendix A : Least Square Error Algorithm are the integrator gains (Kl1 and KI2) of control areas. The
elements of the weighting matrix Q corresponding to Afl,
A linear system can be written in the vector-matrix
differential equation form A f 2 , and AP12 are 1.0, 1.0, and 2.0, respectively. The

X = A X + D Z , X(O)=O (A-1 1 others are set to 10".

where X is the state vector, 2 is the disturbance vector, and Appendix B : System Data
A and D are matrices with undetermined parameters a and
P. (A) Power System
Equation (A-1) can be written in the compact form Area capacity, PR1= PR2 = 2000 M w
Y = A Y , Y ( 0 )= -X,s, Y ( - ) = 0 ( A-2 ) Kp1 = Kp2 = 120 Hz/P. uMW TpI = Tp2 = 20 S ~ C
where Y = X - Xs, , and Xss = -A-'DZ is the steady state Krl = Kr, = 0.5 Trl = Tr2 = 10 sec
values of the state variables in equ. (A-1). TGl = TG2= 0.08 sec T T ~=Tn =0.3 sec
The performance index for the system in equ. (A-2) is NI = 0 . 8 N2 = -0.2
given by RI = R2 =2.4 Hz/p.u.MW A12 =-1.O
B1 = B2 = 0.425 P.U.MW/HZ
Na,P) = &Y
' TQY dt ( 24-3 1
The matrix Q is a symmetrical weighting matrix and a D, = D2 = 8 . 3 3 lod3
~ p.u.MW/Hz
positive semidefinite matrix also. Notice that minimization T12= 0.08674 p. u. MW/radian
of 77 leads to a minimum value of the sum of the weighted APD, = 0.005 P.U.MW
square error of the state variables. The dynamic responses of
( B ) BES ( 10MW/40 MWh ) [5, 14, 151
the system can, therefore, be improved through the
minimization of the performance index ?I. Battery voltage = 1755-2925 V dc
CBp = 52597 F RBp = 1OwZ
A method based on Lyapunov technique is proposed to
evaluate the integral in equ. (A-3). Because A is Hunvitz, a CBI = 1 F RBI = 0.001R
RBT = 0.0167 R R,, = 0.013 R
positive definite Lyapunov function yTPY can be defined
which have a negative semidefinite derivative to satisfy equ. X,, = 0.0274 R I:Es = 4.426 kA
(A-3). Accordingly, K BP = 100 kV/Hz (Af feedback ) a i = 15"
d
-(YTPY) =-YTQY
dt
( A-4 ) KBp = 117kV/Hz (ACE feedback) fly = 25"
where P is a positive definite symmetric matrix. Equation TBp = 0.026 sec
(A-?) can be written as
BIOGRAPHY
YTPY +YTPY = -YTQY ( A-5 )
Taking the transpose of equ. (A-2) and substituting it in equ. Chun-Feng Lu was born in Taichung, Taiwan on July 9, 1963. lie
(A-5), gives graduated from the National Taipei Institute of Technology in 1983, and
received his B.Sc. and M.Sc. degree in electrical engineering from the National
Y~A~+ P YY~ P A Y= - Y ~ Q Y (Ad) Taiwan Institute of Technology in 1990 and 1992, respectively. He is currently
Therefore, we obtain from equ. (Ad) studying toward the Ph.D. degree in the National Taiwan University. His
A ~ +PA
P = -Q ( A-7 ) research interests include power system dynamics, stability, and energy storage
Comparing equ. (A-3) with equ. (A-4), the performance systerns.
index can be written as Chun-Chnng Liu was born in Taiwan on October 4, 1925. He received
q(a,p)= J,"--(YTPY)dt
d =-YTPYI, his B.Sc. degree in electrical engineering from the National Taiwan University
dt in 1951. Since 1951, he has been with the National Taiwan University, where
he is now a professor. From 1987 to 1988, he was a visiting professor at the
=YT(0)PY(O) ( A-8 1
Kyushu University in Japan. At present, his mearch interests include power
because Y (CO)= 0 ,
system analysis and electrical machines.
With particular values of a and P , P is obtained from ChiJui \Vu ( M'89 ) received the B.Sc., M.Sc., and Ph.D. degree in
equ. (A-7). If the state vector Y has n components, then the electrical engineering from fhe National Taiwan University in 1983, 1985,
computation of the matrix P in equ. (A-7) is equivalent to and 1988 respectively. Now he is a professor in the Department of Electrical
the solution of a set of n(n+1)/2 linear equations. The Engineering, National Taiwan Institute of Technology. His research
matrix P is substituted in equ. (A-8) to evaluate 7 7 . This interests include stability, electric power quality, load management, and
process of evaluating 71 is repeated for various values of a power engineering education.
and P until a minimum value of 17 is found.