Optimizing fault diagnosis in variable load conditions: A machine and deep learning approach for voltage source inverters

Abstract

This study investigates industrial three-phase voltage source inverters (VSI) diagnostics. The system assesses and suggests a novel approach. The research finishes with developing a novel problem-detection tool that identifies individual and multiple open switches and adjusts to various loads. This approach utilizes feature selection and machine learning algorithms to classify issues based on load conditions, aiming to enhance intelligence and performance by leveraging historical data and real-time operational conditions. An in-depth analysis of VSI fault diagnostics is conducted, encompassing preventive measures and sophisticated control techniques to enhance reliability. The utilization of DWT analysis for feature extraction and machine learning classifiers is recommended to identify challenges encountered in three-phase induction motors. The thesis concludes with implementing the Track and Hunt optimization-based Deep Neural Network. The methodology demonstrates a greater ability to anticipate faults than existing diagnostic approaches under various load circumstances. The simulation findings confirm the efficiency of the suggested system in precisely identifying faults, even in dynamic operational circumstances.