A Hierarchical Finite-Control-Set Model Predictive Control Framework for Permanent Magnet Synchronous Motor Drives via PINN-RLS and Virtual-Vector Extension

To address the degraded prediction accuracy, increased torque ripple, and weakened dynamic response of conventional finite-control-set model predictive control (FCS-MPC) under magnetic saturation, parameter mismatch, and load disturbances in permanent magnet synchronous motors (PMSMs), this paper proposes a hierarchical FCS-MPC framework based on PINN-RLS and virtual-voltage-vector extension, termed HRPV-MPC. Built upon a unified nonlinear motor model, the proposed method integrates PINN-RLS-based online parameter correction, virtual-voltage-vector extension, disturbance-observer-based feedforward compensation, maximum-torque-per-ampere (MTPA) and quadratic-programming (QP) reference reconstruction, and deep-neural-network (DNN)-based torque-ripple compensation into the same closed-loop control framework. Unlike existing studies that usually optimize parameter identification, disturbance compensation, or ripple suppression separately, the proposed method emphasizes their coordinated interaction within the predictive control chain so as to simultaneously improve steady-state precision, disturbance rejection, and dynamic recovery performance. Simulation results show that the proposed HRPV-MPC achieves coordinated improvements in steady-state precision, dynamic response, and disturbance rejection under various operating conditions; compared with baseline FCS-MPC, it exhibits clear advantages in torque-ripple suppression, torque-error reduction, load-disturbance recovery, and speed-tracking performance, thereby validating the effectiveness and superiority of the constructed hierarchical collaborative framework.