A Comparative Study of Power Transformer Winding Fault Diagnosis Using Machine Learning Algorithms

G. A.Z. Dlamini, B. A. Thango, P. N. Bokoro

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

In today's distribution and transmission infrastructure, power transformers have proven to be one of the maximum crucial components. Assessing the integrity of power transformer winding is imperative as winding failure is the most dominant failure location. Based on the reported literature a common trend with the research work is that they only consider one machine learning method and do not use other existing methods as benchmarks to corroborate proposed approaches. In this work, a comparative analysis of two trained well known, and broadly applied Machine Learning Algorithms (MLAs): Artificial Neural Network (ANN) and Support Vector Machine (SVM) to assess the power transformer winding fault is carried out. This work also introduces a machine learning algorithm that has the highest performance in terms of accuracy (validation and testing) with measured data from 300 power transformers that are independently inspected by industry connoisseurs. The obtained results show that the ANN method yields the best results when compared to the other algorithm.

Original languageEnglish
Title of host publicationProceedings of the 32nd Southern African Universities Power Engineering Conference, SAUPEC 2024
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9798350371345
DOIs
Publication statusPublished - 2024
Event32nd Southern African Universities Power Engineering Conference, SAUPEC 2024 - Stellenbosch, South Africa
Duration: 24 Jan 202425 Jan 2024

Publication series

NameProceedings of the 32nd Southern African Universities Power Engineering Conference, SAUPEC 2024

Conference

Conference32nd Southern African Universities Power Engineering Conference, SAUPEC 2024
Country/TerritorySouth Africa
CityStellenbosch
Period24/01/2425/01/24

Keywords

  • Machine Learning Algorithms
  • Neural Networks
  • Support Vector Machine
  • frequency response analysis
  • power transformers

ASJC Scopus subject areas

  • Energy Engineering and Power Technology
  • Renewable Energy, Sustainability and the Environment
  • Electrical and Electronic Engineering
  • Mechanical Engineering
  • Safety, Risk, Reliability and Quality

Fingerprint

Dive into the research topics of 'A Comparative Study of Power Transformer Winding Fault Diagnosis Using Machine Learning Algorithms'. Together they form a unique fingerprint.

Cite this