Predictive Model for Incipient Faults in Oil-Filled Transformers

The power transformer is an invaluable piece of device in the power system. To prevent catastrophic failures and the ensuing power outages, the status of a transformer linked to a system must be examined for any possible faults. Despite using DGA as a global tool for detecting faults, it is limited...

Full description

Saved in:
Bibliographic Details
Main Authors: Michael Osajeh, Linda Usiosefe, Efosa Igodan
Format: Article
Language:English
Published: Sakarya University 2024-08-01
Series:Sakarya University Journal of Computer and Information Sciences
Subjects:
Online Access:https://dergipark.org.tr/en/download/article-file/3637223
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850146156670091264
author Michael Osajeh
Linda Usiosefe
Efosa Igodan
author_facet Michael Osajeh
Linda Usiosefe
Efosa Igodan
author_sort Michael Osajeh
collection DOAJ
description The power transformer is an invaluable piece of device in the power system. To prevent catastrophic failures and the ensuing power outages, the status of a transformer linked to a system must be examined for any possible faults. Despite using DGA as a global tool for detecting faults, it is limited by the inability to accurately solve the problem associated with results variability due to the intrinsic nature of the IEC TC 10 database. This study proposed a data-driven fault/defect diagnostic model using four ensemble models with three base classifiers respectively. The base classifiers are comprised of SVM, C4.5 decision tree, and naive Bayes while the ensemble methods are comprised of stacking, voting, boosting and bagging respectively. The DGA dataset used comprises seven features and 168 instances split into training (i.e. 56%) and test (i.e. 44%) datasets respectively. The results indicate that C4.5 obtained a 98.33% accuracy while stacking obtained a 99.89% accuracy as the best-performing base and ensemble models respectively. The high classification performance accuracy achieved by our proposed models indicates its capacity for real-world applications. It can be applied to advance automation in mobile-based technology.
format Article
id doaj-art-e696f0ca54eb4d48af8d95b695dd035e
institution OA Journals
issn 2636-8129
language English
publishDate 2024-08-01
publisher Sakarya University
record_format Article
series Sakarya University Journal of Computer and Information Sciences
spelling doaj-art-e696f0ca54eb4d48af8d95b695dd035e2025-08-20T02:27:54ZengSakarya UniversitySakarya University Journal of Computer and Information Sciences2636-81292024-08-017230231310.35377/saucis...141411528Predictive Model for Incipient Faults in Oil-Filled TransformersMichael Osajeh0https://orcid.org/0009-0000-8917-9446Linda Usiosefe1https://orcid.org/0009-0000-7425-4204Efosa Igodan2https://orcid.org/0000-0003-2102-3597University BeninUniversity of BeniUnviersity of BeninThe power transformer is an invaluable piece of device in the power system. To prevent catastrophic failures and the ensuing power outages, the status of a transformer linked to a system must be examined for any possible faults. Despite using DGA as a global tool for detecting faults, it is limited by the inability to accurately solve the problem associated with results variability due to the intrinsic nature of the IEC TC 10 database. This study proposed a data-driven fault/defect diagnostic model using four ensemble models with three base classifiers respectively. The base classifiers are comprised of SVM, C4.5 decision tree, and naive Bayes while the ensemble methods are comprised of stacking, voting, boosting and bagging respectively. The DGA dataset used comprises seven features and 168 instances split into training (i.e. 56%) and test (i.e. 44%) datasets respectively. The results indicate that C4.5 obtained a 98.33% accuracy while stacking obtained a 99.89% accuracy as the best-performing base and ensemble models respectively. The high classification performance accuracy achieved by our proposed models indicates its capacity for real-world applications. It can be applied to advance automation in mobile-based technology.https://dergipark.org.tr/en/download/article-file/3637223dissolve gas analysispower transformernaive bayessupport vector machinec4.5ensemble learning
spellingShingle Michael Osajeh
Linda Usiosefe
Efosa Igodan
Predictive Model for Incipient Faults in Oil-Filled Transformers
Sakarya University Journal of Computer and Information Sciences
dissolve gas analysis
power transformer
naive bayes
support vector machine
c4.5
ensemble learning
title Predictive Model for Incipient Faults in Oil-Filled Transformers
title_full Predictive Model for Incipient Faults in Oil-Filled Transformers
title_fullStr Predictive Model for Incipient Faults in Oil-Filled Transformers
title_full_unstemmed Predictive Model for Incipient Faults in Oil-Filled Transformers
title_short Predictive Model for Incipient Faults in Oil-Filled Transformers
title_sort predictive model for incipient faults in oil filled transformers
topic dissolve gas analysis
power transformer
naive bayes
support vector machine
c4.5
ensemble learning
url https://dergipark.org.tr/en/download/article-file/3637223
work_keys_str_mv AT michaelosajeh predictivemodelforincipientfaultsinoilfilledtransformers
AT lindausiosefe predictivemodelforincipientfaultsinoilfilledtransformers
AT efosaigodan predictivemodelforincipientfaultsinoilfilledtransformers