Model Predictive Current Control With Adaptive Neural Network-Based Sliding Mode Observer for IPMSM

This study presents a novel model predictive current control (MPCC) strategy for an interior permanent magnet synchronous motor (IPMSM) drive. The proposed MPCC integrates a sliding mode observer (SMO) with a data-driven adaptive neural network (ANN) to enhance control performance. Traditional continuous control set model predictive current control (CCS-MPCC) is highly sensitive to motor parameter variations, necessitating improved robustness. To overcome this limitation, the proposed method aims to mitigate CCS-MPCC’s parameter sensitivity while strengthening disturbance rejection in current control. The study first formulates the modeling and control strategies for CCS-MPCC, incorporating the effects of time delay on the dq-axis of the IPMSM. Additionally, the lumped parameter disturbances in the dq-axis are characterized. A detailed analysis of ANN-based SMO is then presented, demonstrating its ability to estimate the lumped parameter disturbances in dq-axis current control. To further enhance performance, an ANN is integrated into the traditional SMO to estimate the dq-axis lumped parameters disturbance, thereby reducing the required switching gains. Finally, experimental results validate the effectiveness of the proposed MPCC approach, which integrates CCS-MPCC with ANN-based SMO, in improving the performance of IPMSM drives operating in the constant torque region.