中文摘要：

针对开关磁阻电机磁场的强非线性和高饱和性，通过有限元法得到了磁化以及转矩特性曲线簇。在此基础之上，利用神经网络的非线性映射能力，分别采取反向传播神经网络以及径向基函数神经网络对曲线簇进行了学习，进而在Matlab中建立了开关磁阻电机驱动系统的非线性动态仿真模型。对神经网络的结构及训练参数进行了敏感性分析，优化了取值，提高了逼近及泛化能力。在不同控制方法下，通过与有限元分析所得相电流以及合成转矩曲线的比较，验证了所建立开关磁阻电机驱动系统神经网络动态仿真模型的有效性。

英文摘要：

According to strong nonlinearity and high saturation of magnetic field in switched reluctance machine （SRM）, the magnetization and torque characteristic curve family were obtained by finite element method （FEM）. On this basis, the nonlinear dynamic simulation model of SRM was developed, which made use of the nonlinear mapping ability of back propagation neural network （BPNN） and radial basis function neural network （RBFNN）. The sensitivity of structural and training parameters was analyzed, their values are optimized, and the approximation and generalization ability are im- proved. The validity of the developed neural network dynamic model was verified by comparing with the phase current and total torque curves obtained by FEM.