Multi-group Fault Positioning Model of Modular Multilevel Converter Based on FOA-LSSVM

Abstract

In order to solve the problem of large amount of calculation of IGBT open circuit fault location of modular multilevel converter (MMC), a multi-group fault location method of submodule (SM) based on Fruit Fly Optimization Algorithm (FOA) optimized LSSVM model is proposed. The fault characteristics of MMC are analyzed , the SMs of bridge arm are grouped, and the output voltage, terminal voltage of SMs group and capacitor voltage of SMs are selected as fault parameters respectively. The wavelet packet decomposition is used for data compression and noise reduction, and a model based on fruit fly algorithm is constructed to optimize the regular parameters and kernel function parameters of LS-SVM to achieve accurate fault location. Adopting NLM modulation strategy, a 31 level MMC is built based on SIMULINK, and a MMC semi-physical simulation platform is built based on STARSIM. The results show that, when an open-circuit fault occurs, the node energy of the three fault characteristic parameters will change obviously. The proposed fault positioning model is compared with the single group positioning FOA-LSSVM model, WPD-FOA-LSSVM model, SSAE-SOFTMAX model, 1-D CNNs model, WPD-PSO-SVM model and multi-group positioning FOA-LSSVM model, the test accuracy can reach 100%, and the anti-noise performance is excellent. The experimental results show that this method can accurately locate the faulty bridge arm, submodule and IGBT.