Torque Ripple Control of Switched Reluctance Motor Based on RBF Neural Network

Mengqiu LI,Ling FAN,Xiuyong REN,Ping LIU,Wenhao YUAN

DOI: https://doi.org/10.3969/j.issn.1003-8930.2017.12.005

2017-01-01

Abstract:Torque ripple is a primary disadvantage of switched reluctance motor(SRM)owing to its doubly salient struc-ture and obvious nonlinear electromagnetic characteristics.To solve this problem,a scheme of instantaneous torque con-trol method based on radial basis function(RBF)neural network is presented.With the combination of requirement for sample data and control,a torque observer is designed based on the advantages of RBF neural network including its gen-eralization and approximation ability,which realized a nonlinear mapping of current and angle to torque.Then,torque inner loop is constituted by a torque hysteresis-controller,and torque is used as feedback directly. The proposed ap-proach accomplished the torque control of the motor by constructing transient torque to track the reference torque, which is the output of speed outer loop.Simulation and experimental results demonstrated that the proposed control strat-egy can effectively reduce the torque ripple of SRM,and have the advantages of fast response,high control accuracy and adaptability to speed variation.