基于DNN神经网络的PMSM控制策略研究

Research on PMSM Control Strategy Based on DNN Neural Network

PMSM系统在电动汽车控制系统中应用广泛。为提高PMSM系统的控制效果,本次研究根据DNN神经网络的激活函数优化特点,将DNN神经网络应用到PMSM控制系统之中。结果显示DNN神经网络与其他算法相比性能更加优越,预测误差均值更低,在50与150个样本状态下误差均值分别为0.74N·m与0.58N·m。且本次研究改进后的PMSM控制系统输出的转矩在正常与突变状态下的波动范围较传统PI控制系统更小,分别为位于2.00N·m到2.50N·m之间与2.52N·m到2.63N·m之间。可见基于DNN神经网络的PMSM控制系统性能更加优越,控制效果更佳。

PMSM system is widely used in electric vehicle control system.In order to improve the control effect of PMSM system,DNN neural network is applied to PMSM control system according to the optimization characteristics of activation function of DNN neural network.The results show that compared with other algorithms,DNN neural network has better performance and lower mean prediction error,which are 0.74N·m and 0.58N·m respectively in 50 and 150 samples.The torque fluctuation range of the improved PMSM control system in the normal and sudden state is smaller than that of the traditional PI control system,which is between 2.00N·m and 2.50N·m and 2.52N·m and 2.63N·m respectively.It can be seen that the PMSM control system based on DNN neural network has better performance and better control effect.