An Open-circuit Fault Detection and Location Strategy for MMC with Feature Extraction and Random Forest

Abstract

With the rapid development of industrial applications based on modular multilevel converter(MMC), the operation reliability is receiving more and more attention. This paper concentrates on open-circuit fault detection and location(FDL) of sub-modules(SMs), and proposes a new strategy based on machine learning(ML) classifier. According to the analyzation the open-circuit fault characteristics, the proposed strategy selects the voltages of SMs as original features for FDL. In order to construct the dataset as well as reduce the amount of calculation, 9 time-domain features are extracted as the new inputs. A ML model based on random forest is then trained and tested with the dataset, considering FDL as classification. To prove the effectiveness of the proposed strategy, the MMC models are built both in the simulations and experiments.