Research Of Switched Reluctance Motors' Torque Ripple Suppression Based On Neural Network

With the rapid development of modern power electronic technology,switched reluctance motor has been widely used,which has simple structure,low production cost,wide speed range and high output efficiency.Due to the doubly salient structure,nonlinear core magnetic circuit and the phase current of pulse works,switched reluctance motor has serious torque ripple,especially in low speed will lead to speed oscillation and noise.All of that affected the SRM's widely used in the transmission system.Therefore,reducing the switched reluctance motor torque ripple can broaden its application field and improve performance.To solve this problem,this paper presents instantaneous torque control based on back-propagation and radial basis function neural network,then this two methods are compared with traditional current chopping control in the control system.The torque observer is designed based on the strong generalization and approximation ability of the neural network.Combing with sample data of switched reluctance motor and control requirement,torque observer is designed which can realize the nonlinear relationship between motor phase current,rotor position angle and torque.The online calculation of three-phase instantaneous torque of switched reluctance motor was realized by torque observer,then the three-phase instantaneous torque is synthesized to total torque,which torque inner loop is constituted.Next,the motor reference torque is tracked by transient torque perfect.Finally,this switched reluctance drive is accomplished based on torque hysteresis-controller and neural network.A dynamic simulation model of switched reluctance drive is established in the Matlab environment.The digital simulation results demonstrate that the control strategy can effectively minimize the SRM's torque ripple and has the advantages of fast response,high control accuracy,adapt to speed variety,and the effect of torque ripple minimization based on RBF neural network is better than BP neural network.Finally,hardware circuit is designed based on DSP TMS320F28335 as the CPU,a three-phase 6/4 switched reluctance motor as control object,using C language to complete the writing of BP and RBF neural network program.By debugging the software program of DSP,the instantaneous torque ripple of switched reluctance motor is suppressed.Through the torque waveform comparison under traditional current chopping control and the control strategy,which based on back-propagation and radial basis function neural network torque observer,the validity and reliability of the proposed control strategy in this paper are verified.