Robust Predictive Control of PMSM via Dual-Loop Parameter-Disturbance Adaptation

To address the control performance degradation caused by parameter mismatches and external disturbances in permanent magnet synchronous motor (PMSM) model predictive control, this study proposes a composite predictive control strategy integrating online parameter identification and disturbance observation. By constructing an incremental motor model, a forgetting factor-based recursive least squares (FF-RLS) method is designed to dynamically identify stator resistance, inductance, and permanent magnet flux linkage parameters in real time, enabling adaptive correction of the predictive model to suppress parameter drift effects. Innovatively, a third-order extended state observer (ESO) is introduced to uniformly model unmodeled dynamics, load torque disturbances, and residual parameter errors as a total disturbance term for estimation, which is actively compensated through a feedforward voltage reference injection mechanism. The core innovation lies in the synergistic optimization framework of parameter identification and disturbance compensation: A dynamic coupling framework for parameter updating and observer bandwidth adjustment is established based on Lyapunov stability theory, ensuring uniform boundedness of parameter estimation errors and disturbance observation errors. Simultaneously, voltage feedforward compensation achieves decoupled control of predictive model errors. Results demonstrate that the proposed strategy significantly enhances system robustness against parameter uncertainties and external disturbances by dynamically updating model parameters and proactively suppressing composite disturbances. While preserving the fast dynamic response of model predictive control, it effectively resolves steady-state errors and harmonic distortions induced by parameter mismatches, offering a theoretically rigorous and practically viable solution for high-precision motor drive systems.