Optimal Distributed Coordinated Reinforcement Learning for Secondary Voltage Control in Time-Delayed Microgrid

Chanjuan Zhao, Chao Wang, Zhiyou Cheng, Wei Sun, Bao Zhao

Abstract: This article proposes an optimal distributed secondary voltage control method based on distributed coordinated reinforcement learning (DCRL) algorithm in the time-delayed microgrid. First, a distributed secondary voltage control protocol through a sparse communication network is designed using consensus-based algorithm and Lyapunov–Krasovskii theory. Then, by combining the consensus control method with reinforcement learning together, a novel DCRL algorithm is utilized to dynamically adjust the voltage controller gains. Since the DCRL method is an online intelligent algorithm, it can adaptively improve the voltage control performance in the presence of various external disturbances encountered in microgrid secondary control. The effectiveness of the proposed method is verified through simulations under different scenarios on a four inverter-based microgrid test system. Simulation results indicate that our method shows superior dynamic voltage regulation response over traditional methods.