Data-Driven Receding Horizon Predictive Current Control for PMSM Drives Using Ultralocal Model

Traditional model-free predictive current control (MFPCC) improves system robustness by eliminating parameter dependency. However, it still exhibits limitations in terms of disturbance rejection performance and dynamic response characteristics. This paper proposes a new data-driven receding horizon predictive current control (DDRHPCC) method that integrates a multi-step ultra-local model with a generalized receding horizon estimator (GRHE) to further optimize the control performance of permanent magnet synchronous motor (PMSM) drives. The multi-step ultra-local model replaces the traditional motor model to establish a multi-step prediction framework relying solely on system input-output data. This approach avoids parameter sensitivity and strengthens long-term prediction capability. Furthermore, GRHE is designed to estimate unmodeled dynamics, external disturbances, and parameter variations in real time. By compensating for these factors, GRHE solves the problem of insufficient anti-interference ability caused by model simplification in traditional MFPCC. Utilizing pre-estimation disturbance information, the GRHE optimizes error estimation, significantly improving disturbance rejection capability and state estimation accuracy. Simulation results validated the effectiveness of the proposed method.