Reinforcement Learning-Based Feedforward Compensation Control in PMSM

This paper presents a Proximal Policy optimization(PPO) feedforward compensation algorithm based on reinforcement learning for Permanent Magnet Synchronous Motor (PMSM). The PPO algorithm eliminates the need for precise system modeling and achieves lower steady-state errors and improved generalization capabilities through adaptive optimization training. Additionally, we introduce modules to enhance the performance of the PPO algorithm. Simulation results demonstrate the effectiveness of the PPO feedforward control in reducing speed fluctuations and stabilizing the motor’s operation near the target speed.