Model-Free Predictive Control With Super-Twisting Observer of Single-Phase Neutral-Clamped H-Bridge Cascade Inverters With Fixed Switching Frequency

A continuous control set model-free predictive control (CCS-MFPC) algorithm study based on single-phase cascaded H-bridge neutral-point clamped inverter is proposed to address the problems of unfixed switching frequency and parameter mismatch in finite control set model predictive control (FCS-MPC). First, in order to achieve a fixed switching frequency, three voltage vectors are employed in each control cycle, with the number of applications inversely proportional to their respective cost functions. The 27 switching vectors are combined into a sequence of 32 switching vectors from the consideration of both the reduction of switching action as well as the balance of the neutral point potential. Second, the effect of neutral point potential on system performance is investigated, and multiobjective prediction is used to balance the neutral point potential effectively by adding the neutral point potential difference to the cost function. Finally, in order to solve the parameter dependence problem of model prediction, a ultralocal model of single-phase CHB-NPC inverter is established, and a super-twisting sliding mode observer (ST-SMO) is designed to estimate the centralized perturbation, which makes the controller more parametrically robust. The corresponding experimental platform is built to demonstrate the correctness and superiority of the proposed method.