Intelligent Frequency Control Strategy Based on Reinforcement Learning of Multi-Objective Collaborative Reward Function

Abstract

Large scale wind power integration into the power grid will pose a serious threat to the frequency control of power system. If only Control Performance Standard (CPS) index is used as the evaluation standard of frequency quality, it will easily lead to short-term centralized frequency crossing, which will affect the effect of intelligent Automatic Generation Control (AGC) on frequency quality. In order to solve this problem, a multi-objective collaborative reward function is constructed by introducing a collaborative evaluation mechanism with multiple evaluation indexes. In addition, Negotiated W-Learning strategy is proposed to globally optimize the solution of the objective function from multi dimensions, it avoids the poor learning efficiency of the traditional Greedy strategy. The AGC control model simulation of standard two area interconnected power grid shows that the proposed intelligent strategy can effectively improve the frequency control performance and improve the frequency quality of the system in the whole-time scale.