Fault Detection and Classification in Multilevel Inverter using Machine Learning

Multilevel inverters (MLIs) are predominantly employed in commercial applications that require high voltages and powers. As the switch count in a system increases, the probability of failure of a multilevel inverter increases. It is essential to identify switch faults in power converters. This study exclusively examined open-circuit switch faults in MLI. The proposed Machine Learning (ML) based open-switch fault diagnosis method employs only the output voltage data. Three features were extracted from the output voltage. The ML methods employed such as Decision Trees (DT), Support Vector Machines (SVM), K-Nearest Neighbors (KNN), and Random Forests (RF). The proposed diagnostic approach was implemented in the MATLAB/Simulink environment. KNN achieved a maximum accuracy of 99.52% with a training and testing data split of 70% and 30% respectively.