Optimal SVM switching for a multilevel multi-phase machine using modified discrete PSO

Abstract

This paper searches for the best possible switching sequence in a multilevel multi-phase inverter that gives the lowest amount of voltage harmonics. A modified discrete particle swarm (MDPSO) algorithm is used in an attempt to find the optimal space vector modulation switching sequence that results in the lowest voltage THD. As with typical PSO cognitive and social parameters are used to guide the search, but an additional mutation term is added to broaden the amount of area searched. The search space is the feasible solutions for the predetermined vectors at a given modulation index. Comparison of the MDPSO algorithm to an integer particle swarm optimization (IPSO) is presented for all three modulation indices tested. The resulting switching sequences found show that the MDPSO algorithm is capable of finding a minimal THD solution for all modulations indices tested. The MDPSO algorithm performed better overall than the IPSO in terms of converging to the best solution with significantly lower iterations.