Académique Documents
Professionnel Documents
Culture Documents
3, SEPTEMBER 2009
725
I. INTRODUCTION
N RECENT years, wind energy has shown a rapid growth
as a clean and inexhaustible energy source all around the
world [1], [2]. However, as the penetration levels increase, it
is of considerable concern that a fluctuating power output of
wind farms will affect operation of interconnected grids [3],
especially weak power systems. Such cases may require some
measures to smooth the output fluctuation to have a reliable
power system [4].
Recent advances in electric energy storage technologies provide an opportunity of using energy storage to address the wind
energy intermittency [5]. Electric energy can be stored electromagnetically, electrochemically, kinetically, or as potential
energy. Two factors characterize the application of an energy
storage technology. One is the amount of energy that can be
stored in the device and the other is the rate at which energy can
be transferred into or out of the storage device. These factors depend mainly on the characteristic of the storage device itself [6].
A variety of storage technologies are available, which are
capable of smoothing out the unpredictable fluctuating power
output of the wind farms. Some of these storage technologies
are supercapacitors, superconducting magnetic energy storage,
flywheels, batteries, compressed air energy storage, and hydropumped storage. The basic issues with these storage technolo-
Manuscript received June 13, 2008; revised September 30, 2008. First published June 10, 2009; current version published August 21, 2009. Paper no.
TEC-00217-2008.
S. Teleke, M. E. Baran, A. Q. Huang, and S. Bhattacharya are with the Department of Electrical and Computer Engineering (ECE), North Carolina State
University in Raleigh, Raleigh, NC 27695 USA (e-mail: steleke@ncsu.edu;
baran@ncsu.edu).
L. Anderson is with Bonneville Power Administration, Portland, OR 97232
USA.
Digital Object Identifier 10.1109/TEC.2009.2016000
726
Fig. 1.
Fig. 2.
Fig. 3. Power and energy ratings for ideal BESS. (a) Power variations
dP = P se t P w in d (in megawatts). (b) Energy required for dispatch (in
megawatthours).
TELEKE et al.: CONTROL STRATEGIES FOR BATTERY ENERGY STORAGE FOR WIND FARM DISPATCHING
Fig. 4.
Fig. 5.
727
Fig. 6.
Fig. 7.
728
Fig. 8. Smoothing of wind farm power with BESS. P re f : desired set point; P w in d : wind power; P to ta l : net injected power; P to ta l with ideal BESS: net injected
power with ideal BESS model (in megawatts).
The control scheme we have developed is based on the controller design of [2], which uses the SOC as a feedback signal in order to keep the SOC of the battery within proper
limits.
This basic control scheme, however, needs to be enhanced in
order to address the other constraints pointed out before. Fig. 6
shows the proposed control scheme. This controller is designed
to charge/discharge the BESS through a STATCOM. Hence, the
controller serves as the outer loop controller, and provides the
set point, which is the current set point ib ess in Fig. 7, to
the inner controller of STATCOM [23].
Note that we want the controller to act as a regulator and
control the BESS to compensate for the fluctuations of the wind
farm power Pwind during the dispatching period. Hence, the
main input to the controller, Pset , is the desired set point, i.e.,
desired power dispatch level for the given hour. A good choice
for Pset is the hourly average of the estimated wind farm output
for the next hour, as recent advances in wind foresting can
estimate Pset reasonably well, with an rms error of 10% [24],
[25].
As Fig. 6 shows, a rate limiter is put after the input to prevent
overshooting when Pset is changed and also to limit the ramp
rate of the total power output. The filtered desired set point Pref
is then subtracted from the actual wind farm power output Pwind
to get Po , which indicates the amount of power the BESS should
compensate.
Po is then combined with the other feedback signal that indicates the SOC of the BESS. We used the BESS model in Fig. 5
to estimate the SOC by using the procedure proposed in [17].
This feedback loop ensures that the battery SOC remains within
its desired limits. In this feedback loop, offset = M C (where
C is the BESS capacity in MWh and M is SOC margin rate),
and = (C 2 C M )/(T PW F ) (where PW F is the rated
output of the wind farm in megawatts). Hence, the design parameters for the controller are smoothing time constant T and
SOC margin rate M .
Finally, an upper and lower limit block is utilized for the
power converter ratings and the combined signal Pb ess is then
TELEKE et al.: CONTROL STRATEGIES FOR BATTERY ENERGY STORAGE FOR WIND FARM DISPATCHING
729
Fig. 9. 10 Megawatthour BESS performance with SOC feedback. (a) SOC of the battery. (b) DC link voltage (per unit). (c) Power injected by the BESS (in
megawatts). (d) Current profile of the BESS (in kiloamperes).
730
Fig. 10. Power deviations in net power supplied P to ta l around the desired set point P re f with 10 MWh BESS. (a) Power deviations dP = P re f P to ta l (in
megawatts). (b) Histogram of power deviations.
average of Pwind and adding 10% noise to it. The controller then
charges/discharges the BESS based on this set point. Note that
the BESS is charged/discharged through a STATCOM, which
is much faster than the BESS; hence, the power output of the
BESS through STATCOM is approximately Pb ess in Fig. 6, as
the losses in STATCOM are quite small. Therefore, in the simulations, STATCOM is represented as a unity gain, and hence,
Pb ess obtained from the controller output is combined directly
with Pwind to get the total injected power to the grid, Ptotal .
Using this model, one-week-long simulations were made in
PSCAD/EMTDC with a time step of 1 s.
As indicated in the previous section, based on preliminary
studies assuming ideal BESS, two BESS sizes have been considered. Simulation results for 10 MWh BESS are given in
Figs. 8 and 9. The smoothing time constant T for the controller
for this case is set to 0.2 h, as described before.
Fig. 8 shows the Pref , Pwind , and the net power injected
Ptotal = Pwind + Pb ess , with ideal BESS model and actual
BESS model during a day when the wind farm power output
increases from 0 to 50 MW. It is seen that the total injected
power follows the desired set points in general; although large
deviations occur from time to time with actual BESS when wind
power has sudden severe drop or rise.
Fig. 8 also shows that the ideal BESS has better performance
than the practical onethe ideal BESS follows the set point
much closer. To help explain the difference, the SOC of the
actual and ideal BESSs are given in Fig. 9(a). As the figure
shows, when the wind power drops severely (for example, after
hour 135 in the figure), the energy needs to be provided by the
BESS increases considerably. With ideal BESS, this required
energy can be provided without severe drop in the SOC, as seen
in Fig. 9(a); however, due to the losses with the actual BESS, its
SOC drops more than the ideal one, and hence, it cannot provide
the required energy during the large energy demand.
Fig. 9(b) shows that SOC feedback control also helps to keep
the battery voltage within acceptable limits (+10/15% of rated
TELEKE et al.: CONTROL STRATEGIES FOR BATTERY ENERGY STORAGE FOR WIND FARM DISPATCHING
731
Fig. 11.
732
Subhashish Bhattacharya (M85) received the Ph.D. degree in electrical engineering from the University of Wisconsin, Madison, in 2003.
From 1994 to 1996, he was with York International Corporation. From 1996
to 1998, he was a Consultant to Soft Switching Technologies (SST). From 1998
to 2005, he was with the Flexible AC Transmission Systems (FACTS) and Power
Quality Division, Siemens Power Transmission and Distribution. Since August
2005, he has been an Assistant Professor in the Department of Electrical and
Computer Engineering, North Carolina State University, Raleigh. His current
research interests include FACTS, utility applications of power electronics such
as custom power and power quality issues, active filters, high-power converters,
and converter control techniques.
Loren Anderson received the B.S. degree from Oregon State University,
Corvallis, in 1980.
He is currently the Principal High Voltage DC Transmission (HVDC) and
Flexible AC Transmission Systems (FACTS) Engineer at Bonneville Power Administration (BPA), Vancouver, WA. He has vast experience working on HVDC
systems. His current research interests include HVDC control design, equipment maintenance, and failure analysis.