Model-Free Predictive Current Control for PMSM Using Improved Adaptive Moment Estimation-Based Extended State Observer

To address the parameter uncertainty in permanent magnet synchronous motors, model-free predictive current control (MFPCC) has gained widespread attention, with extensive research on inductance parameter adaptation. However, study on disturbance adaptation remains limited and are mainly on complex data-driven approaches. To simplify the disturbance adaptation process, this article innovatively proposes an extended state observer (ESO)-based MFPCC that incorporates an improved adaptive momentum estimation (AMSGrad) algorithm, with enhanced dynamic response performance. Specifically, a nested optimization method is employed, that combines global optimization capability of the AMSGrad optimizer with local adjustment of the disturbance observer, thereby constructing an integrated framework for the momentum coordination and adaptive learning rate mechanism. Experimental comparison has been carried out between the proposed method, conventional ESO-based MFPCC and three advanced ESO methods, validating the advantages of the proposed method on dynamic response improvement and current disturbance suppression.