Real-Time Parameter Identification for Model Predictive Control of PMSM Using Incremental LSTM

Permanent Magnet Synchronous Motors (PMSMs) are widely used in high-precision applications due to their efficiency, but traditional Field-Oriented Control (FOC) is limited by slow dynamic responses, complex parameter tuning, and sensitivity to system variations. This paper proposes a Finite Control Set Model Predictive Current Control (FCS-MPCC) strategy with real-time motor parameter identification using an Incremental Long Short-Term Memory (LSTM) neural network. The approach effectively addresses parameter variations, reduces reliance on Pulse Width Modulation (PWM), and selects optimal voltage vectors, leading to improved dynamic response, reduced harmonics, minimized overshoot, and simplified control. MATLAB/Simulink simulations confirm that FCS-MPCC outperforms traditional FOC in complex operating conditions. Additionally, the Incremental LSTM-based parameter identification achieves higher accuracy than conventional methods, offering enhanced adaptability and stability under dynamic conditions, providing a reliable solution for high-precision PMSM control.