Dynamic Economic Dispatch of Thermal-Wind-Storage Systems Based on Reinforcement Learning

Abstract

This paper studies a dynamic economic dispatch (DED) problem which includes thermal and wind-storage hybrid units, aiming at minimizing the total generation cost and penalty costs involving generation regulation, load shedding, and wind curtailment. Each unit is assigned with a fixed, discrete, constrained virtual action set, and its cost function is unknown. Based on the developed model, a reinforcement learning algorithm is applied to solve the DED problem under the wind uncertainty. Simulation results illustrate the effectiveness of the algorithm.