基于改进BP神经网络的SRM间接位置检测技术

摘要

This paper presented a new approach of position sensorless control for switched reluctance motors (SRM) based on an improved BP neural network. This method constructed BP neural network,which of weight function became a function with adjustable parameters. And it was trained to form a single hidden layer network, then realize the non-linear mapping between motor current,flux and rotor position. Through analyzing the unique structural nature of the switched reluctance motor,the paper proposed a method of reducing the sample data, it achieves computing time. Simulation results show that this approach simplifies the complexity of the system to improve detection accuracy,it achieves indirect detection of the switched reluctance motor position.%文中提出了一种基于改进BP神经网络的开关磁阻电机间接位置检测的新方法.该方法构造了一个将权值变为参数可调函数的BP神经网络,经过训练,形成只有一个隐藏层的网络,实现电机电流、磁链与转子位置之间的非线性映射.通过对开关磁阻电机特有结构性质的分析,提出了一种减少样本数据的方法,从而节约计算时间.仿真结果表明,此方法简化了系统的复杂性,提高了检测精度,从而实现了开关磁阻电机位置的间接检测.