Multi-parameter estimation of non-salient pole permanent magnet synchronous machines by using evolutionary algorithms

Abstract

This paper describes how to apply evolutionary algorithms (EA) for multi-parameter estimation of non-salient pole permanent magnet synchronous machines (PMSM). The encoding of estimated parameters is firstly described and the design of a penalty function associated with a proposed error analysis for PMSM multi-parameter estimation is then introduced. The PMSM stator winding resistance, dq-axis inductances and rotor flux linkage can be estimated by maximizing the proposed penalty function through evolutionary algorithms such as immune clonal algorithm (ICA), quantum genetic algorithm (QGA) and genetic algorithm (GA). The experimental results show that the proposed strategy has good convergence in simultaneously estimating winding resistance, dq-axis inductances and rotor flux linkage. In addition, the convergence speed of ICA in estimation is compared with GA and QGA, which verifies that the ICA has better performances in global searching. The ability of proposed method for tracking the parameter variation is verified by winding resistance step change and temperature variation experiments at last.