Key points are not available for this paper at this time.
The ill-convergence of multiparameter estimation due to the rank-deficient state equations of permanent-magnet synchronous machines (PMSMs) is investigated. It is verified that the PMSM model for multiparameter estimation under i d = 0 control is rank deficient for simultaneously estimating winding resistance, rotor flux linkage, and winding inductance and cannot ensure them to converge to the correct parameter values. A new method is proposed based on injecting a short pulse of negative i d current and simultaneously solving two sets of simplified PMSM state equations corresponding to i d = 0 and i d ≠ 0 by using an Adaline neural network. The convergence of solutions is ensured, while the minimum | i d | is determined from the error analysis for nonsalient-pole PMSMs. The proposed method does not need the nominal value of any parameter and only needs to sample the winding terminal currents and voltages, and the rotor speed for simultaneously estimating the dq -axis inductances, the winding resistance, and the rotor flux linkage in nonsalient-pole PMSMs. Compared with existing methods, the proposed method can eliminate the estimation error caused by the variation of rotor flux linkage and inductance as a result of state change due to the injected d -axis current in the surface-mounted PMSM. The method is verified by experiments, and the results show that the proposed method has negligible influence on output torque and rotor speed and has good performance in tracking the variation of PMSM parameters due to temperature variation.
Liu et al. (Thu,) studied this question.