Accurate model of switched reluctance motor based on indirect measurement method and least square support vector machine. Issue 9 (1st November 2016)

• Record Type: Journal Article
• Authors: Zhong, Rui; Xu, Yuzhe; Cao, Yanping; Guo, Xiaoqiang; Hua, Wei; Xu, Shen; Sun, Weifeng
• Is Part Of: IET electric power applications. Volume 10:Issue 9(2016)
• Page Start: 916
• Page End: 922
• Publication Date: 2016-11-01
• Subjects: reluctance motors -- electric machine analysis computing -- least squares approximations -- support vector machines -- rotors -- machine theory; switched reluctance motor -- indirect measurement method -- least square support vector machine -- SRM flux linkage prediction -- performance prediction -- doubly salient structure -- rotor position -- phase current -- nonlinear function -- LSSVM -- static electromagnetic characteristics -- artificial neural network -- Matlab‐Simulink; Electric power -- Periodicals; Electric power systems -- Periodicals; 621.305
• DOI: 10.1049/iet-epa.2016.0112
• Languages: English
• ISSNs: 1751-8660

Abstract

Abstract : The accurate model of switched reluctance motor (SRM) is critical for performance prediction. However, due to the doubly salient structure, the flux linkage of SRM is a non‐linear function with phase current and rotor position, which makes the determination of the characteristic of SRM very difficult. In this study, indirect measurement method is used to obtain the sample data and least square support vector machine (LSSVM) is used for the prediction of flux linkage. First, the basic principles of indirect measurement and the exact methods of realisation are presented. Second, the errors in the indirect measurements are analysed and post‐processing is given to reduce them. Third, the entire static electromagnetic characteristics are obtained by the prediction of improved LSSVM. The results are compared with those of using artificial neural network and those of using less training data by LSSVM, which shows both strong learning ability and generalisation ability. Finally, a simulation model is built up using MATLAB/Simulink and the results are compared between simulation and measurement. The good agreement shows that the proposed model has good accuracy. In addition, the LSSVM does not need any prior knowledge which is much easier for modelling than the existing literatures.