基于遗传优化支持向量机的开关磁阻电机非线性建模方法

Nonlinear Modeling Method of Switched Reluctance Motor Based on GA Optimized Support Vector Machine

将有限元法(FEM)与非线性映射技术相结合,得到了开关磁阻电机(SRM)动态仿真模型。利用FEM获取了SRM的磁化特性和转矩特性数据,并依此对支持向量机进行了训练,进而在MATLAB中建立了仿真模型。采用改进型遗传算法对支持向量机的超参数进行全局寻优,提高了其逼近和泛化能力。基于对磁化特性数据的分析,引入了分段训练的思想,进一步提高了模型在小电流下的仿真精度。将所建模型的动态仿真结果与FEM分析结果相比较,验证了建模方法的有效性。

The finite element method （FEM） and nonlinear mapping technique were combined to get switched reluctance motor （SRM） dynamic simulation model. The magnetization and torque characteristics of SRM were obtained by FEM, then support vector machine （SVM） was trained based on these data, finally the simulation model was built in MATLAB. Improved genetic algorithm was adopted to find the global optimal hyper parameter value of SVM, and the approximation and generalization capability were improved. Segmented training strategy was introduced based on analysis of magnetization characteristic, and the simulation accuracy of the model Under small current was enhanced. The validity of the modeling method was proved by comparing its dynamic simulation results with that of the FEM.