Model-Free Predictive Control for Three-Level NPC Inverters Based on Adaptive Sliding-Mode Observer and RLS Algorithm

Model predictive control (MPC) has been widely used in power converter regulation due to its easy implementation, fast dynamic response, and superior performance. The traditional finite control set model predictive control (FCS-MPC) method is prone to parameter mismatch when relying on the exact system parameters, while the ultralocal model-based MFPC can reduce the dependence on the system model, but its control performance will still be affected by the observer parameter settings. To address the above problems, an enhanced sliding mode observer is designed in this paper to improve the antidisturbance capability of the system, and the RLS algorithm is introduced to adaptively adjust the observer parameters to improve the estimation accuracy of the system disturbances. In this approach, the prediction of the current is accomplished by the adaptive SMO observation ultralocal model, and the coefficients of the observer are updated by the RLS algorithm. In simulation experiments, the proposed method exhibits lower total harmonic distortion (THD) and more stable midpoint voltage control under both matched and unmatched parameter conditions, as well as better robustness and dynamic response performance compared with the conventional FCS-MPC and SMO-based MFPC methods.