FCS-MPC Based on Geometric State Preselection for High Computational Reduction for Multilevel Converters

Abstract

Finite control-set model predictive control (FCS-MPC) has been extensively applied to multilevel converters (MCs) due to its simplicity, fast dynamics, and ability to handle multiple objectives. However, its main drawback is the exponential increase in computational load as the number of switching states grows, which limits its real-time applicability in MCs with many levels. This article proposes a novel strategy, FCS-MPC with geometrical preselection states (FCS-MPC-GPS), which significantly reduces the number of evaluated states while preserving current tracking and capacitor-voltage regulation performance. The method introduces a geometric redefinition of the $\alpha \beta$ voltage-vector frame, replacing the conventional hexagonal representation with concentric circles whose radii correspond to the voltage magnitudes. This transformation enables a hierarchical reduced-search selection process that separates current tracking and capacitor-voltage balancing into sequential stages. The proposed approach is experimentally validated on a three-cell four-level flying capacitor converter with 512 switching states, demonstrating a computational load reduction of nearly 98.55% while maintaining accurate current tracking, capacitor-voltage regulation, and fast dynamic response.