Adaptive AI-Assisted Model Predictive Current Control for PMSM Drives in Electric Vehicles with Robustness Enhancement and Real-Time Feasibility Analysis

In this paper, we propose an adaptive AI-assisted model predictive current control (AI-MPC) strategy for permanent magnet synchronous motor (PMSM) drivthe es in electric vehicles (EVs). Whereas classic finite control set model predictive current control (FCS-MPCC) relies on a predetermined set of weighting factors, a neural network-based solution is proposed to enable adaptability of cost function weights online. The prediction is done with a discrete-time PMSM model combined with vehicle load dynamics, while Lyapunov-based analysis ensur conducted toonline adaptation of cost function weightstem in the presence ofsence of disturbance speed variation, the simulation results presented under various EV operating scenarios, including overshoot by 40%, torque ripple by 35%, and RMSE by 30% versus FCS-MPCC. This also achieves a much faster settling time (≈ 3 ms) and increased robustness. These findings demonstrate that the predictive control framework proposed using AI is a promising solution for real-time high-performance EV motor drives.