controlling ramp rates of wind power output

CONTROLLING RAMP RATES OF WIND POWER OUTPUT
USING A STORAGE SYSTEM AND PATTERN MATCHING FORECASTING
Duehee Lee, Graduate Student, Phone +1 512 350 4704, University of Texas at Austin, [email protected]
Joonhyun Kim, Graduate Student, Phone +1 512 534 9110, University of Texas at Austin, [email protected]
Ross Baldick, Professor, Phone +1 512 471 5879, University of Texas at Austin, baldick @ece.utexas.edu
Overview
In order to be compatible with peak-load generators, ramp rates of wind power output should be limited to the ramping
capabilities of peak-load generators. Under the assumption that there is a penalty for deviating outside the ramp rate
limits (RRL)s, storage operation policies are decided optimally by designing a linear program. Furthermore, by
forecasting the trend of short-term wind power, a storage system allocates the state of the charge (SOC) level wisely in
preparation for severe ramp down or up events while minimzing the total penalty costs. Wind power output for six
hours are recursively forecasted through the Gaussian Processes (GP).
Methods
A storage system includes the round-trip efficiency, maximum SOC level, minimum SOC level, and power rating
controller with respect to the SOC level. The storage system either charges or discharges, so it cannot charge and
discharge at the same time. The power rating controler reduces the charging rate when the SOC is close to the maximum
level, and it reduces the discharging rate when the SOC is close to the minimum level, in order to reduce the ramp
shocks to the power system. The operation time line for a storage system is show in Fig. 1.
Fig. 1. Time line for a storage system operation
An objective function of an optimization framework is designed to impose penalty costs when net production, which is
the sum of wind power and storage output, violates the RRLs. It is modeled as a piecewise convex function. The penalty
is assumed to be linearly proportional to the number of MW deviated from the RRLs. An objective function is decided
among four functions regarding the relationships between the previous net production and current wind power. In
addition to the objective function, many constraints are added to the objective functions: the SOC management
constraints, power rating constraints, and output power control constraints.
Sum of objective functions for forecasted wind power becomes the final objective function. In this case, the objective
function is decided based on the relationship among forecasted wind powers. The discount rate is used to differentiate
the penalty incurred by the current operation from penalties to be incurred by future operations. Since the penalty of the
current operation is more expensive than other penalties, the current operation would not be biased by wrong forecasts.
Even though storage operations are forecasted for six hours, only the current operation is performed and the decision
process is repeatedly performed. Finally, the storage system limits the ramp rate of the net production that is the sum of
wind power output and storage output.
Wind power outputs for six hours are forecasted through the Gaussian Processes (GP). First, the GP forecasts the one
hour ahead wind power output. Based on the forecasted value, the GP forecasts the wind power two hours ahead. In this
way, the GP forecasts the wind power for six hours recursively.
Results
Wind power output and its conditional distribution is forecasted through the GP in Fig. 2. As shown in Fig. 3., storage
operations limit ramp rates of wind power succesfully. Total wind power output measured in 2008 from Electric
Reliability Council of Texas (ERCOT) is used. In Fig. 3. the optimal storage operation is compared to the ideal case of
limiting ramp rate. The capacity of wind farm was 7000 MW. The power rating and storage size of a storage system is
272MW and 1761 MWh. Decisions made by solving the optimization problem performed well under the long-term
ramp-up or ramp-down events.
Fig. 2. Forecasted mean and variance of wind power
Fig. 3. Storage operation results
Conclusions
An optimization framework of a storage system with wind power forecasting is designed to limit ramp rates of the net
production. A storage system limits ramp rates until it reaches its maximum power rating and storage size. Furthermore,
a storage system could manage its SOC level through forecasting the future wind power. Furthermore, a storage system
decides an optimal storage operation and manage the SOC level when wind power does not violate the RRL.
References
D. Lee, and R. Baldick, “Limiting ramp rates of wind power output using a battery based on the variance gamma
process,” in International Conference on Renewable Energies and Power Quality, no. 771, Mar. 2012.
J. Eyer, and G. Corey, “Energy storage for the electricity grid: Benefits and market potential assessment guide,” Sandia
National Laboratories, Albuquerque, New Mexico, Feb. 2010.
S. Teleke, M. Baran, A. Huang, S. Bhattacharya, and L. Anderson, “Control strategies for battery energy storage for
wind farm dispatching,” Energy Conversion, IEEE Transactions on, vol 24, no. 3, pp. 725-732, Sept. 2009.