Skip to content
University of El OuedResearch Portal
Other

Overcoming Issues with Source Data for Early Fault Detection in Power Transformers: A Comparison with Combined Triangle Methods

Abstract

Early identification of power transformer disorders is essential to guarantee the safety and continuous functioning of electrical systems. The technique of Dissolved Gas Analysis (DGA) is highly useful in diagnosing problems in oil-filled transformers. In these transformers, the oil functions as both a cooling and insulating medium, effectively reflecting internal stresses. This study presents a comprehensive evaluation of the various defect diagnostic algorithms belonging to either Duval or Gouda Triangle approaches set across different datasets. The samples that were examined were a total of 287, representing 117 samples from IEC TC 10 data, and 170 other samples that derived from Egyptian scientific laboratories. In the IEC TC 10 data set, the Gouda Triangle method (88.89% accuracy) outperformed the Duval Triangle method (85.47%). On the other hand, for Egyptian data, specificity was 76.47% and PPV could be obtained using the Gouda Triangle method and respectively were 65.88% and 94.44% with Duval Triangle method; while in either of these methods. Notably, the use of the Combined Triangle methods, which combine the advantages of both the Duval and Gouda Triangle approaches, led to remarkable improvements in performance, with accuracies of 93.16% for Data 1 and 88.24% for Data 2. This novel methodology provides a more resilient and flexible diagnostic instrument, yielding consistent and dependable outcomes across diverse data sources.

Details

Original languageEnglish
EventInternational Conference of Advanced Technology in Electronic and Electrical Engineering (ICATEEE)
Article number11406463
Pages1-6
DOI10.1109/icateee68170.2025.11406463
Publication statusPublished — 10 December 2025

Keywords

UN Sustainable Development Goals

  • 9 — Industry, innovation and infrastructure

Fingerprint