关键词: 无轴承永磁薄片电机, 深度确定性策略梯度, 比例-积分-微分控制

Abstract: Due to characteristics such as dynamic eccentric magnetic disturbances and parameter variations in bearingless permanent magnet slice motor ( BPMSM) , traditional fixed-parameter proportional-integral-derivative (PID) magnetic levitation control performs poorly. This paper proposes a BPMSM magnetic levitation PID control method combining the deep deterministic policy gradient ( DDPG) algorithm. This method utilizes a DDPG agent to dynamically adjust the PID controller parameters,thereby improving the accuracy and response speed of magnetic levitation control. The mathematical model of BPMSM levitation is analyzed and based on this, a levitation PID control strategy is constructed. A PID parameter adaptive strategy based on the DDPG algorithm is proposed and a reward function consisting of error penalty,precision reward, and overshoot penalty is designed. The verification is conducted through MATLAB / Simulink simulation. The results show that compared with fixed-parameter PID control, the DDPG-PID controller exhibits smaller overshoot,shorter settling time,and stronger anti-interference and adaptive capabilities in radial displacement tracking and radial load disturbance experiments.

Key words: bearingless permanent magnet slice motor, deep deterministic policy gradient, proportional-integralderivative control