基于改进的开关磁阻电机无位置传感器技术

由于开关磁阻电机（SRM）的双凸极结构与高饱和性，其磁链为相电流和转子位置的高度非线性函数。在简化磁链法中建立精确而小规模的SRM磁链模型，是实现SRM无位置传感器控制的关键。采用预处理方法优化神经网络（BPNN）模型可以减小尺度，提高精度；为了进一步优化该模型，在此基础上提出粒子群（PSO）优化BPNN的建模方法。利用PSO算法进行全局优化，避免了BP神经网络易陷入局部极小值的缺点。并通过实验样机实测磁链值，对两种磁链模型进行对比分析。仿真结果表明，此改进方法具有一定的准确性。最后通过Matlab仿真验证将精确度高的磁链模型应用于简化磁链法中，实现SRM无位置传感器控制具有可行性。

Based on the Improved Sensorless Technology of Switched Reluctance Motor

Because of the double salient pole structure and high saturation of SRM, its flux linkage is a highly nonlinear function of phase current and rotor position. The key to SRM sensorless control is to establish a precise and small-scale SRM flux model in the simplified flux method. Using the preprocessing method to optimize the neural network (BPNN) model can reduce the scale and improve the accuracy. To further optimize the model, a particle swarm optimization (PSO) modeling method for BPNN is proposed. PSO algorithm is used for global optimization to avoid the disadvantage of BP neural network falling into local minimum. The two models are compared and analyzed by the experimental prototype. Simulation results show that the improved method has a certain accuracy. Finally, Matlab simulation proves that it is feasible to apply the highly accurate flux linkage model to the simplified flux linkage method to realize SRM sensorless control.