Optimisation Algorithms in Multi-Step Prediction Horizon MPC for Two-Level Inverters

This paper presents a detailed comparison of three algorithms used in Finite Control Set Model Predictive Control (FCS-MPC) for two-level inverters: the standard Sphere Decoding Algorithm (SDA), the K-best SDA, and the Exhaustive Search Algorithm (ESA). ESA was implemented in two frameworks: Firstly, by listing all control sequences and secondly, by exploring all nodes in a search tree. The study focuses on a traditional three-phase two-level Voltage Source Inverter (VSI), using Matlab/Simulink for benchmarking the different simulation models. Our results demonstrate a reduction in the number of iterations and computational complexity with the SDA and K-best SDA, while maintaining similar control performance with ESA. The 3-phase load has a significant influence on the standard SDA, with the number of iterations varying for linear and non-linear loads. These findings suggest that the K-best SDA is a practical and efficient alternative to ESA for FCS- MPC applications, especially in scenarios of longer prediction horizons.