Controller Design for PMSM Based on Multi-step Predictive Neural Network

摘要

Based on multi-step predictive control algorithm, a novel speed controller is proposed for the high performance drives of permanent-magnet synchronous-motor (PMSM) with artificial neural networks(ANN).The weights of ANN are trained online by using the extended Kalman filter(EKF) algorithm, and the training starts simultaneously along with the PMSM. The usefulness and validity of the proposed method are verified in a PMSM drive system in Matlab. The simulation results indicate that the controller has good performance at both transient and steady states, and also at variable-speed operation or load variation.