基于强化学习的刚性联接双电机系统无模型最优协调控制

Model-free Optimal Coordinated Control for Rigidly Connected Dual-motor Systems Using Reinforcement Learning

双永磁同步电机系统具有模型参数不确定,外部负载变化以及快慢动态共存的特性,这为其最优协调控制带来挑战。该文提出一种基于强化学习的无模型最优协调控制方法。首先,采用传统的主从控制结构和PI控制器构建双永磁同步电机系统数学模型;其次,利用输出调节理论和最优控制理论,设计最优协调控制器,解决系统外部负载变化问题;然后,针对系统的模型参数不确定以及快慢动态共存特性,提出一种独立于模型参数的强化学习算法,学习控制器增益;最后,仿真和实验结果验证所提出的控制策略可以有效改善双电机系统速度跟踪性能和转矩同步性能。该文所提控制方法能够提高双电机系统的跟踪性能和同步性能,抑制未知时变负载的干扰,并避免参数不确定所带来的影响。

Dual permanent magnet synchronous motor(PMSM)systems have the characteristics of uncertain model parameters,external load changes and coexistence of fast and slow dynamics,which brings challenges to their optimal coordinated control.This paper proposes a model-free optimal coordinated control method based on reinforcement learning(RL).First,the mathematical model of the dual-PMSM system under traditional master-slave control and PI controllers is formulated.Next,by output regulation and optimal control theories,an optimal coordinated controller is designed to solve the problem of external load change of the system.Then,a RL algorithm independent of model parameters is proposed for uncertainty of model parameters and the coexistence of fast and slow dynamic to learn the controller gain.The proposed control method can improve the tracking performance and synchroni-zation performance of the dual-PMSM system,suppress the interference of unknown time-varying loads,and avoid the influence of parameter uncertainty.Finally,the simulation and experimental results verify that the proposed control strategy can effectively improve the speed tracking performance and torque synchronization performance of the dual-PMSM system.