Application of memetic algorithm for selective harmonic elimination in multi‐level inverters. Issue 9 (1st September 2015)

• Record Type: Journal Article
• Authors: Niknam Kumle, Alireza; Fathi, Seyed Hamid; Jabbarvaziri, Faramarz; Jamshidi, Mehdi; Heidari Yazdi, Seyed Saeed
• Is Part Of: IET power electronics. Volume 8:Issue 9(2015)
• Page Start: 1733
• Page End: 1739
• Publication Date: 2015-09-01
• Subjects: invertors -- harmonics suppression -- particle swarm optimisation -- search problems -- genetic algorithms; memetic algorithm -- selective harmonic elimination -- multilevel inverters -- MLI -- high‐quality output power -- low switching losses -- high conversion efficiency -- fundamental frequency switching strategy -- low order harmonics minimisation -- particle swarm optimisation -- PSO‐based MA -- mesh adaptive direct search -- harmonic optimisation problem -- modulation index -- evolutionary algorithms -- convergence probability -- genetic algorithm -- bee algorithm -- cascade seven‐level inverter; Power electronics -- Periodicals; 621.31705
• DOI: 10.1049/iet-pel.2014.0209
• Languages: English
• ISSNs: 1755-4535

Abstract

Abstract : Multi‐level inverter (MLI) is a promising technology, able to generate high‐quality power with lower switching frequency. Therefore it leads to high conversion efficiency and low switching losses. Selective harmonic elimination is a fundamental frequency switching strategy that theoretically provides desirable output waveform for MLIs by elimination of low order harmonics. Unfortunately, complexity, non‐linearity and solvability features attached to this method have limited its industrial application. In this study, a particle swarm optimisation (PSO)‐based memetic algorithm (MA) guided by mesh adaptive direct search is introduced as a suitable choice for the harmonic optimisation problem. This algorithm is precise and applicable to any MLI. The results show that MA converges to the exact solution for feasible modulation index. When the problem has no exact solution, the algorithm finds a relatively proper solution to regulate the fundamental component of the voltage. Furthermore, the comparison between MA and other methods shows that the probability of convergence of MA is higher than others. For 48% of the analysed data MA reaches to a fitness value lower than 10 −30 whereas this probability is 5% for PSO and almost zero for genetic algorithm and bee algorithm. The proposed method has been implemented on a cascade seven‐level inverter.