基于强化学习自适应自耦PID的可控永磁悬浮系统控制研究OA

The controllable permanent magnet levitation belt conveyor is a novel type of magnetic levitation belt conveyor that uses controllable permanent magnet levitation rollers as its supporting structure.Aiming at the nonlinearity and susceptibility to interference of its controllable permanent magnet levitation system,a control method integrating the reinforcement-learning deep deterministic policy gradient(DDPG)algorithm and self-coupling PID(SC-PID)was proposed.The DDPG algorithm serves as the upper-layer control strategy,which learns the nonlinear characteristics of the controllable permanent magnet levitation system to adaptively adjust the speed factor parameters of the lower-layer SC-PID control.Moreover,to address the low exploration efficiency in the early training phase of the DDPG algorithm,a double replay buffer mechanism was introduced to enhance the algorithm's training efficiency in the initial stage.The simulation and experimental results show that compared with the PID and SC-PID control methods,the proposed control method can quickly adjust control parameters according to the system tracking error,demonstrating excellent tracking and anti-interference performance.Under external load impact disturbances,the change in the system's levitation gap is 0.29 mm,which is a reduction of 32.5%and 19.4%compared to the PID and SC-PID control methods,respectively.This improvement in the dynamic performance of the controllable permanent magnet levitation system proves the superiority brought about by the intelligent adjustment of controller parameters.