Abstract：To address the parameter dependency issues present in traditional model predictive control

algorithms，a novel model predictive control method that employed an ultra-local model and updated its parameters in real-time using the recursive least squares （RLS）method was proposed. This approach eliminated the dependence on parameters in model predictive control methods and enhanced their robustness. Additionally，

addressing the lack of universal rules and reliance on experience in selecting weighting factors in the predictive

control cost function，a sequential prediction-based control method that achieves desired performance without the

need to choose weighting factors was utilized. Finally，the effectiveness and robustness of the method were validated through simulations in Matlab/Simulink on a three level- active neutral point clamped（3L-ANPC）inverter.

      Key words：model-free predictive control（MFPC）；recursive least squares（RLS）method；three level- active

neutral point clamped（3L-ANPC）inverter；ultra-local model；sequential predictive control