Online Multiparameter Estimation for Robust Adaptive Decoupling PI Controllers of an IPMSM Drive: Variable Regularized APAs

Abstract

This paper proposes the variable regularized affine projection algorithms (VR-APAs) for the online multiparameter estimation of the interior permanent magnet synchronous motors (IPMSMs). Unlike the conventional APAs with a fixed regularization factor, the normalized gradient of the mean-square error is introduced in the proposed VR-APA to update the variable regularization which ensures a fast convergence rate, accurate estimation, and low steady-state error. Moreover, the proposed VR-APA does not require any accurate priori information of the motor parameters, making it highly feasible for the IPMSM. In order to accurately estimate the stator d-q axis inductances, stator resistance, flux linkage, and load torque, the two-time scale approach in the proposed VR-APAs is used due to the difference in the IPMSM dynamics. Of various applications of the proposed VR-APAs, such as condition monitoring, fault analysis, and controller design, these estimated multiparameters are updated online to the adaptive decoupling PI controllers to achieve the robustness against the parameter variations due to the temperature increase and load disturbances under various operating conditions (i.e., speed change, load change, and speed reversal). Finally, the comparative experimental verifications via a prototype IPMSM with TMS320F28335 DSP programmed by Code-Composer-Studio are conducted to confirm the effectiveness of the proposed VR-APAs.