Selective Forgetting-Based Unbiased Estimation of PMSM Parameters Under Complex Operating Conditions and Measurement Noise

Accurate parameter estimation for permanent magnet synchronous motors (PMSMs) under complex operating conditions is challenging, as not all frequency components of the input signal contribute equally to parameter estimation, and not all sampled data are equally informative. This article presents a theoretical analysis of how specific frequency components influence the estimation of individual PMSM parameters. Based on these insights, a selective forgetting mechanism is proposed to enhance estimation accuracy by retaining informative data while discarding outdated information in targeted parameter directions. In parallel, an instrumental variable method is employed to inject new information while mitigating the adverse effects of measurement noise. To ensure real-time feasibility, a computationally efficient inversion scheme for the information matrix is designed, significantly reducing the algorithm’s complexity. The resulting method achieves high-accuracy, unbiased parameter estimation with strong robustness across diverse operating conditions. Comparative simulations validate the effectiveness and superiority of the proposed approach.