A Model Predictive Torque Control Method of Switched Reluctance Motor based on Deep Neural Network Torque Estimator and Torque Error Pulse Modulation

This paper proposes a model predictive torque control method based on a deep neural network torque estimator with a torque error pulse modulation module for switched reluctance motor(SRM). A deep neural network based on torque estimator is established and trained on data from different operation conditions and verified to have higher accuracy. In addition, a torque error pulse modulation module is also proposed in this method, which can obtain higher accuracy for FCS-MPTC and significantly reduce torque ripple. The simulation results of conventional MPTC method are compared to the proposed DNN-MPTC method in a three-phase 12/8 polses SRM.