Fault Diagosis of CHB Inverter Using Machine Learning

Abstract

Renewable energy has gained popularity due to depleting natural resources and escalating fossil fuel and nuclear pollution. Power electronic engineers design grid-connected power conversion systems. MLIs provide more power and solutions. Cascaded H-Bridge (CHB) MLIs start with two or more 3L single-phase H-bridge inverters. Each H-bridge may produce three separate voltage levels. Combining the separated dc voltage sources produces a stepped output voltage with a step size equal to the magnitude of the connected sources. The present work develops a method for detecting and resolving switch failures in a three-phase CHB inverter, ensuring system dependability and allowing for system redundancy. The recommended approach uses Wavelet transform to extract features, then Decision Tree classifier to detect and characterise defects. Increased classification accuracy shows the DT-based fault diagnosis system's efficiency in identifying inverter switch faults.