Reliability Assessment and Fault Prediction in a 13-Level Multilevel Inverter Through Machine Learning with SVM

Multilevel inverters appear to be a potential substitute for traditional inverters in medium-power applications. Real-time applications now heavily rely on modern power converters from renewable energy sources. This study examines the factors that affect the failure rate of power semiconductor devices, including temperature and current rating. The bathtub curve determines the lifespan of the gadget. This article makes a thirteen-level asymmetric multilevel inverter by using fewer switches and has undergone a thorough analysis to determine switching loss, conduction loss, and failure rate in terms of reliability. This study investigates the prediction of defects in switches within a 13-level multilevel inverter using four machine learning models. Our investigation demonstrates that the Support Vector Machine (SVM) model surpasses other models with a remarkable accuracy rate of 96.56%. The abstract outlines the creation of a confusion matrix specifically for Support Vector Machines (SVM), providing a comprehensive analysis of key parameters including Accuracy, Precision, Recall, and F1 score. The study emphasizes the SVM model’s robustness, providing insights into its training and validation accuracy for fault detection in switches.