关键词: 永磁同步电机, 强化学习, 近端策略优化, 转速波动

Abstract: A reinforcement learning-based proximal policy optimization ( PPO) feedforward compensation algorithm is proposed to address the speed fluctuation issue caused by unbalanced loads in permanent magnet synchronous motors. The causes and suppression principles of speed fluctuations in permanent magnet synchronous motors are analyzed. Taking advantage of the PPO algorithm’ s ability to achieve low steady-state error without the need for precise system modeling and its adaptive optimization training, a motor feedforward compensation control framework based on PPO is developed. Batch normalization is introduced to enhance the performance of the PPO algorithm by handling the general advantage estimation. Simulation and experimental results demonstrate that the proposed PPO feedforward control algorithm effectively suppresses speed fluctuations in the presence of unbalanced load conditions.

Key words: permanent magnet synchronous motor, reinforcement learning, proximal policy optimization, speed fluctuation