Model Predictive Current Control for Permanent Magnet Synchronous Motor based on Neural Network
Yaohua Li, Dongmei Liu, Ting-Xu Wu, Weichao Guo, Xinquan Zhang, Yizhi Deng
Abstract
In order to solve the problem of high computational burden of finite set multi-step model predictive current control, a neural network based predictive current control method (NMPCC) for permanent magnet synchronous motor (PMSM) model was proposed. Firstly, the traditional one-step model predicted current control model was established, and the nerve dissociation was trained to learn and approximate the voltage vector selection rule of the one-step model predicted current control, so as to replace the one-step model predicted current control algorithm. The simulation verifies the feasibility of using neural network instead of one step model to predict current control algorithm. For multi-step control, calculation burden will increase exponentially with the increase of step, but output voltage vectors will not change. Therefore, a neural-network-based two-step MPCC for PMSM established. Simulation results show neural network operates well instead to two-step MPCC controller with almost the same control performance. And two-step MPCC requires 56 predictive calculations, while the number of classification tasks for neural network control is still 7. Real-time experiments show the real-time performance of neural network is better than two-step MPCC and calculation time is decreased by 29.58%.