A Robust Control Scheme for PMSM Based on Integral Reinforcement Learning
Qi Guan, Xuliang Yao, Zifan Lin, Jingfang Wang, Herbert Ho‐Ching Iu, Tyrone Fernando, Xinan Zhang
Abstract
This article proposes an integral reinforcement learning (IRL)-based <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$H_{\infty }$ </tex-math></inline-formula> control algorithm for permanent magnet synchronous motor (PMSM) drives with excellent performance and guaranteed stability. Owing to its model-free nature, this algorithm achieves superior current regulation without any prior knowledge of motor parameters. Unlike the traditional offline reinforcement learning (RL) algorithms, which rely heavily on the quality of presampled data for training, the proposed algorithm optimizes the control strategy online using real-time data. The convergence of the proposed algorithm is proved. Moreover, a simple actor-critic structure-based neural network (NN) is employed to iteratively update the control policy by a recursive least-square (RLS) approach with low computational burden. The effectiveness of the proposed algorithm is experimentally verified on a 2-kW PMSM prototype.