建立了基于开关磁阻发电机的波浪发电系统以及SRG的磁链特性和转矩特性模型,分别用Levenberg-Mar-quardt算法和弹性BP算法进行训练,比较了BP神经网络的几种常用训练算法的性能;模型的泛化性能用数据插值方法进行了检验。建模效果及模型泛化能力试验表明,BP神经网络建立的开关磁阻发电机模型性能优越。 The model of flux linkage and torque characteristic of a switched reluctance generator based on wave power generation system and SRG is established. The Levenberg-Marquardt algorithm and the elastic BP algorithm are respectively used for training. Several common training algorithms of the BP neural network are compared The generalized performance of the model was tested by the data interpolation method. Modeling effects and model generalization tests show that the switched reluctance generator model built by BP neural network has superior performance.