基于BP神经网络的开关磁阻电机直接转矩控制建模与仿真

Modeling and Simulation of SRM Direct Torque Control Based on BP Neural Network

开关磁阻电机具有特殊的双凸极结构,导致其运行时会产生严重的转矩脉动,直接转矩控制可显著降低转矩脉动,但由于磁路的严重饱和,传统方法利用公式计算转矩过程非常复杂。提出一种基于BP神经网络来预测开关磁阻电机转矩,进而进行直接转矩控制的策略。通过有限元仿真得到训练数据经离线训练之后,即可得到输入量到转矩的非线性映射。该控制方法利用BP神经网络泛化、逼近能力强的优点,省去了复杂的转矩计算,同时可以对转矩脉动进行抑制。仿真和实验结果表明,该方法响应速度快,控制精度高。

Switched reluctance motor(SRM)has a special biconvex structure,which causes serious torque ripple when it runs.Direct torque control(DTC)can significantly reduce torque ripple,however,due to the serious magnetic saturation,the traditional method which uses the formula to calculate the torque process is very complicated.This paper presents a strategy of direct torque control based on BP neural network to predict SRM torque,then the strategy of direct torque control is carried out.After off-line training of the training data obtained by finite element simulation,the nonlinear mapping from input to torque can be obtained.This method takes the advantages of BP neural network,which will get rid of the complex torque calculation,at the same time can be suppress counter rotating torque pulsation.Simulation and experimental results show that this method has fast response speed and high control precision.