USP Electronic Research Repository

Enhancing Open Circuit Switch Fault Localization in Two-Level Voltage Source Inverters through Machine Learning with 3D Current Trajectory Analysis

Peter, Naithan and Kumar, Rahul R. and Chand, Shyamal S. and Cirrincione, Maurizio (2024) Enhancing Open Circuit Switch Fault Localization in Two-Level Voltage Source Inverters through Machine Learning with 3D Current Trajectory Analysis. [Conference Proceedings]

Full text not available from this repository.

Abstract

This paper proposes an enhanced approach for localizing open circuit faults (OCFs) in two-level voltage source inverters (VSIs). OCFs in VSIs can lead to critical system failures which can also be catastrophic if not localized or attended to within a small time-frame. While detecting the OCFs is important, accurate localization of OCFs is paramount as it helps in reducing the downtime of the system. Leveraging machine learning algorithms, including a neural-based classifier called the Wide Neural Network (NN) and a non-neural classifier known as the Subspace KNN, the proposed study demonstrates promising results. The Wide NN achieves an accuracy of 97.16%, while the Subspace KNN achieves an impressive accuracy of 99.66% in fault localization. These high accuracies contribute to improved VSI reliability in various industrial applications, advancing the field of power electronics by exploring innovative fault detection and localization techniques.

Item Type: Conference Proceedings
Uncontrolled Keywords: Location awareness;Uncertainty;Voltage source inverters;Fault detection;Artificial neural networks;Electrical fault detection;Integrated circuit reliability;Open Circuit Fault;3D Current Trajectory;Machine Learning;Voltage Source Inverter
Subjects: T Technology > TK Electrical engineering. Electronics Nuclear engineering
Divisions: School of Information Technology, Engineering, Mathematics and Physics (STEMP)
Depositing User: Rahul Kumar
Date Deposited: 07 Mar 2024 23:12
Last Modified: 14 Mar 2024 23:03
URI: https://repository.usp.ac.fj/id/eprint/14463

Actions (login required)

View Item View Item