Direct Control of Multilevel Power Converters for Microgrid Applications

Multilevel power converters (MLPCs) are essential for the grid integration of renewable energy sources and energy storage systems in microgrids, yet their operation is challenged by the diversity of power electronic topologies, varying functional requirements, and conflicting performance objectives. While finite control‐set model predictive control (FCS‐MPC) offers superior flexibility for managing the nonlinear switching behavior of power converters, the existing implementations suffer from topology‐specific modeling requirements, high computational burden due to exhaustive search, and limited adaptability across applications, leading to fragmented control solutions that must be redesigned for each scenario. This paper proposes a generalized FCS‐MPC framework that addresses the existing limitations through three major contributions: (1) a universal switched state‐space representation (SSSR) model that represents any MLPC topology through a parametric formulation, removing the need for topology‐specific controller redesign; (2) a flexible control architecture that supports diverse microgrid applications by adapting only the cost function while retaining the same predictive framework; and (3) a graph‐based optimization approach leveraging Dijkstra’s algorithm, enabling computationally efficient real‐time implementation with execution times of only 10 μs per control cycle for three‐level topologies. The framework’s universality is validated through three comprehensive case studies spanning grid‐forming, hybrid energy storage, and power‐sharing applications across three major MLPC topologies. Simulation results demonstrate fast dynamic responses (7–20 ms for voltage transients), scalability from 27 to 256 switching states, and seamless adaptability to diverse control objectives. The proposed framework establishes a systematic approach for direct control of MLPCs in complex microgrid scenarios, reducing development time and enhancing control performance across applications.