Open-Circuit Fault Detection in T-Type MLI Using XGBoost: A Machine Learning-Based Approach

Abstract

Prediction of open circuit (OC) faults in a T-Type multilevel inverter (MLI) with high accuracy using machine learning (ML) models is a critical issue for the reliability of power electronic systems. In this study, OC faults that may occur in any of the nine switches in the inverter circuit are detected and classified using a data-driven approach. K-Nearest Neighbors (KNN) and XGBoost algorithms are applied for machine learning based fault detection and their performances are compared. The data set for fault detection is obtained from randomly generated OC faults in the T-Type MLI circuit modeled in MATLAB/Simulink environment. For each fault class, 255 signal samples are obtained for two cycles (4 ms) at a sampling frequency of 5 kHz and used as input data to the ML models. The XGBoost model shows the best performance with an accuracy of 98.37%. This shows that it provides fast and reliable fault detection in large data sets. The confusion matrix, accuracy and loss plots are used to analyze the performance of the model and the study demonstrates the applicability of ML-based fault detection in inverter systems.