Virtual Small Signal-Based Online Identification of PMSM Drive System for Model Predictive Control

This article presents a virtual small signal-based online parameter identification method for permanent magnet synchronous machine (PMSM) drive systems under model predictive control (MPC). By extracting inherent high-frequency features from the MPC-sampled data via convolution operations, the proposed method enables accurate parameter estimation without the need for external signal injection, thus avoiding additional torque ripple and power losses. The method decouples the parameter estimation problem into several single-input single-output (SISO) observers, substantially reducing computational complexity and improving real-time compatibility. Compared with conventional time-domain identification approaches, the proposed method achieves full-order parameter estimation with a magnetic parameter identification error of less than 5% across the entire operating range. It requires only 13.7 floating-point operations (FLOPs) per parameter and introduces a minimal 14% overhead in execution time, making it highly suitable for embedded controller implementation. Experimental validation confirms improved accuracy and fast convergence under dynamic conditions, demonstrating the method’s effectiveness for real-time MPC-based drive control applications.