Modeling of flux linkage for the BPMSM based on LS-SVM within the Bayesian evidence framework

Abstract

According to the modeling of the flux linkage model for bearingless permanent magnet synchronous motors(BPMSMs) by using the conventional analytic method,a novel modeling method based on least squares support vector machine(LS-SVM) within the Bayesian evidence framework was proposed.Fistly,the nonlinear modeling of the flux linkage model was analyzed briefly.Secondly,the regression theory of the LS-SVM and the basic concept of the Bayesian evidence framework(namely,the three levels inference) were introduced in detail.And then the level 1 inference was used to determine the weight vector wof the LS-SVM.The level 2 inference was used to ascertain the model regularization parameter γ of the LS-SVM.The level 3 inference was used to obtain the kernel parameter σ of the LS-SVM,and thus the flux linkage model of the BPMSM based on LS-SVM within the Bayesian evidence framework was established.Finally,the simulation studies were carried out with the Matlab7.0 software to illustrate the performance of the proposed method.Simulation results show that this model has high accurate precision,good generalization ability,flexible structure,and rapid calculation speed.