Model-Free Predictive Control Using Sinusoidal Generalized Universal Model for PMSM Drives

To address the issue of weak robustness caused by parameter mismatches in model predictive control (MPC), model-free predictive control is widely adopted in motor driving systems. This approach replaces the a priori model with a data-driven model, eliminating the influence of physical parameters. However, the rotating coordinate system has various impacts on model accuracy, especially the inaccurate position signals and asymmetric controller that can cause errors. This article proposes a model-free predictive control strategy using a sinusoidal generalized universal model to directly control sinusoidal signals and improve model accuracy in reflecting the motion characteristics of the plant considering the past data. The model is designed based on signal variations from the plant, with online estimation of the coefficients using the recursive least square (RLS) algorithm to ensure model accuracy. To implement the control in a continuous-control-set (CCS) controller, the predictive signal is designed considering time-shift and Lagrange algorithms, and the basic performances and stability are analyzed in terms of pole/zero locations and differential equations. The proposed strategy is applied to a permanent magnet synchronous motor (PMSM) driving system as a predictive current control (PCC) strategy, and the experimental results under different operating conditions are provided to demonstrate the effectiveness of the proposed method, as well as its advantages in terms of model accuracy, current quality, and enhanced robustness compared to conventional control strategies. IEEE