摘要：磁悬浮开关磁阻电机(BSRM)的电感矩阵是电机建模的基础，本文提出了基于最小二乘支持向量机(LSSVM)的电机电感辨识建模方法。首先通过对BSRM电感特性的有限元分析，获得各参数对电感的影响规律，然后结合LSSVM在有限样本数据下对高维非线性的逼近能力，离线建立BSRM各种运行工况下的电感模型。另外在建模中，针对LSSVM超参数选取问题，采用粒子群优化算法(PSO)对其进行自动寻优，以提高电感模型精度。最后通过对比仿真研究，表明PSOLSSVM模型能够准确反映电机磁饱和下的电感特性，这为BSRM磁饱和模型的构建奠定了基础。

关键词：磁悬浮开关磁阻电机；支持向量机；粒子群优化；建模
Abstract: The inductance matrix is very important for the model of bearingless switched reluctance motors (BSRM). A novel modeling method of the inductance for BSRM using least squaressupport vector machine (LSSVM) was presented. First, the inductance characteristic of BSRM was analyzed by the finite elements method (FEM). For the nonlinear character of the inductance, this approach takes advantage of LSSVM with better solution for smallsample learning problem and good generalization ability. Through the offline learning, a better LSSVM was built to form an efficient nonlinear mapping for the inductance mode of BSRM. Then, the particle swarm optimization (PSO) algorithm was used to optimize parameters of LSSVM to improve the accuracy of the inductance model. Finally, the comparative simulation research showed that the PSOLSSVM model could accurately reflect the inductance characteristics of BSRM under magnetic saturation. This makes a contribution to the model of BSRM considering the characteristic of magnetic saturation.

Key words:  bearingless switched reluctance motors; support vector machine; particle swarm optimization; modeling