基于模糊神经网络预测转角的无传感器无刷直流电动机换相控制研究

针对位置检测和准确换相是无刷直流电机运行的关键，本文提出一种新的方法来实现无传感器无刷电机的换相控制。即构造一个基于遗传算法训练结构和参数的模糊神经网络，通过检测电机的磁通和电流来预测电机实际转角实现电机的准确换相。仿真结果表明，利用模糊神经网络预测转角来控制电机换相能取得很好的效果。而文章利用遗传算法作为模糊神经网络的训练算法，此算法具有收敛速度快，不易陷入局部极小的特点。

The Commutation Control Research of BLDCM Based-on Fuzzy Neural Network for Estimating the Rotor Position

In allusion to the key problem of BLDCM position measuring and precise commutation, a new method of controlling the commutation of BLDCM is presented in this paper. Namely, a fuzzy neural network is proposed whose structure and parameters are trained based on improved genetic algorithm. Through measurement of the phase flux linkages and phase currents the network is able to estimate the rotor position, and then the commutation is realized. Simulation results demonstrate that the algorithm is feasible. Furthermore this paper uses GA (Genetic Algorithm) as the training algorithm of fuzzy neural network, which has the advantage of fast convergence and will not converge at local nadir.