遗传优化的最小二乘支持向量机在开关磁阻电机建模中的应用

Considering the nonlinear flux-linkage characteristic of switched reluctance motor (SRM), least square support vector machine (LSSVM) optimized by adaptive genetic algorithm (AGA) and implemented it for modeling nonlinear characteristic of SRM. When the LSSVM is trained with sufficient sample data, AGA is applied to optimize super parameters of LSSVM via minimizing fitting errors between forecasted data and measured data. With the trained LSSVM, the forecasted data of the model are compared with measured data, and error analyses are given to evaluate performances of the proposed model. The experimental results demonstrate that LSSVM optimized by AGA performs better forecast accuracy and successful modeling of SRM. © 2009 Chin. Soc. for Elec. Eng.