A DDPG Algorithm Based Reinforcement Learning Controller for Three-Phase DC-AC Inverters

Abstract

This paper proposes a DC-AC inverter controller based on reinforcement learning (RL) algorithm. Compared with the traditional PID control method, the structure of the RL algorithm based controller is simpler. The deep deterministic policy gradient (DDPG) algorithm in RL algorithm is used to realize model-free control of inverters, so that the control algorithm has adaptive ability to different types of DC-AC inverters. It can avoid the dependence of the control strategy on the system model. Through simulation, the control strategy of three-phase two-level DC-AC inverter based on DDPG algorithm is compared with the traditional PID control. The simulation results show that the total harmonic distortion (THD) is reduced by 12% and the current tracking performance is improved by at least 75%.