Extended state observer-actor–critic architecture based output-feedback optimized backstepping control for permanent magnet synchronous motors

초록

In this paper, we propose a serial extended state observer-based optimized backstepping control approach for permanent magnet synchronous motors (PMSMs). We obtain a new PMSM model, where an additional acceleration state variable is proposed to lump the uncertainties and the load torque into one disturbance. We first design the optimized backstepping controller, where reinforcement learning (RL) is used to solve the associated Hamilton–Jacobi–Bellman equation. Our RL is an actor–critic algorithm in which we use neural network (NN) approximators to design optimal backstepping virtual control, actual control, and system performance evaluation. Furthermore, an extended state observer in a first-order cascade structure is applied to estimate external disturbances and uncertainties; thus, no additional identifiers are required to implement output-feedback control. Therefore, the number of NN approximators is reduced, and substantial computational power problems are solved. The stability of the closed-loop system is proven by the input-to-state stability (ISS) property. Finally, various simulation and experimental results are presented to evaluate the effectiveness of the proposed method. Indeed, the experimental results for the PMSM model show that the proposed strategy improves computational performance by an average of 17.2% under external disturbances.

키워드