Filter Design for Single-Phase Grid-Connected Inverter Based on Reinforcement Learning

As an important part of the grid-connected inverter, the filtering characteristics directly determine the gridconnected performance of the inverter. In this paper, the two fifth-order filter topologies are used as templates for Reinforcement learning selection based on the performance of the filters. The Reinforcement learning algorithm rewards the performance of the filters under different templates accordingly, and then continuously trains the neural network to obtain the optimal filter templates and component parameters. The simulation results validate the rationality of the proposed reinforcement learning-based filter design method.