Machine Learning Based Fault Detection and Classification in Multilevel Inverter for Industrial Applications

Multilevel InvertersMultilevel Inverters (MLIs) are widely used in commercial applications that demand high voltage and power levels. An increase in switch count in the MLI leads to higher likelihood of component failures. Detecting switch faults in power converters is therefore crucial. This study focused specifically on diagnosing open-circuit (OC)Open-circuit (OC) faults switch faults in MLIs. The proposed fault diagnosisFault diagnosis (FD) technique is depending on machine learningMachine Learning and relies on output voltage data. From this voltage, three key features were extracted such as Total Harmonic Distortion (THD), Root Mean Square (RMS) value and Mean value. The Machine LearningMachine Learning (ML) algorithms used for diagnosis of OC switch faults were, Decision Trees (DT), Support Vector Machines (SVM), Random Forests (RF), and K-Nearest Neighbors (KNN). The diagnostic framework was developed and tested in the MATLAB/Simulink tool. Among the models, DT achieved maximum accuracy of 99.76% with a training and testing data split of 70% and 30% respectively.