GA优化的BP神经网络在SRM间接位置检测的应用

Application of BP neural network optimized by genetic algorithm for Indirect Position Detection of SRM

针对传统BP神经网络系统对开关磁阻电机间接位置检测过程中训练时间长、收敛速度慢、易陷入局部极小值等问题,提出一种基于遗传算法优化的BP神经网络检测方法,该法先利用遗传算法全局寻优能力修正BP网络的初始权值与阈值,进行网络训练,再利用训练好的BP网络实现电机电流、磁链与转子位置之间的非线性映射;仿真结果验证了遗传算法在提高间接位置检测精度方面的显著作用,从而实现了开关磁阻电机位置的间接检测。

As indirect position detection of SRM based on traditional BP neural network have shortcomings of long training time, slow convergence and easy to fall into local minimum, this paper presented a method of indirect position detection based on BP neural network opti- mized by genetic algorithm. The method used the global optimization ability of genetic algorithm （GA） to correct weights and thresholds of BP network, then used the trained BP network to achieve the non--linear mapping between the current, flux and rotor position of motor. Simulation results demonstrate that the genetic algorithm has a significant effect to improve detection accuracy of indirect position detection, then achieves indirect position detection of switched reluctance motor.