Backstepping Control of PMSM Based on RBF Neural Network

Abstract

Based on PMSM dynamics and nonlinear load characteristics, a new nonlinear speed controller is designed with vector control scheme. The proposed controller was composed of backstepping speed controller and error regulator based on RBF neural network. The former was designed to ensure a desired speed tracking control, and the later was derived to realize the robust adaptive control against load torque variations. A second-order filter is adopted to reduce the speed overshoot in the starting course of PMSM. Backstepping control system of PMSM based on RBF neural network was established in Simulink. With dSPACE system and the external drive circuit, the completed control system hardware-in-loop real-time simulation was achieved successfully. Simulation and experimental results show the backstepping control system of PMSM based on RBF neural network Given in this paper can remain its good speed dynamic tracking performance and strong robustness when load torque disturbances appeared.