Kalman Filter Based Model-Free Double-Vector Predictive Control for Voltage Source Inverters

Conventional Model Predictive Control (MPC) methods are sensitive to parameter variations and exhibit limited robustness. Model-free predictive control (MFPC) serves as an effective approach to enhance the robustness of conventional MPC. This paper presents a novel model-free predictive control methodology for converters, which is based on the ultra-local (UL) model of the converter but overcomes it’s previous dependency on initial values of α inherent in algorithms based on UL models. Specifically, this paper develops an UL model for three-level Neutral Point Clamped (3L-NPC) converters and employs an enhanced Kalman filter (KF) approach to simultaneously estimate the parameters of the UL model. Subsequently, a two-vector method is utilized to improve the accuracy of the output voltage. Finally, the effectiveness of this method is validated through simulations in MATLAB/simulink and experimental evaluation.