Thesis

Improvements to standards-driven DGA condition monitoring methodologies of oil-immersed transformers

Downloadable Content

Download PDF
Creator
Rights statement
Awarding institution
  • University of Strathclyde
Date of award
  • 2025
Thesis identifier
  • T17228
Person Identifier (Local)
  • 201554449
Qualification Level
Qualification Name
Department, School or Faculty
Abstract
  • This thesis focusses on advancing a Standard-driven methodology for interpreting Dissolved Gas Analysis (DGA) for condition monitoring of oil-immersed Transformers (TXs). The focus is on analysing and evaluating the methodology outlined in IEEE C57.104-2019, which differs substantially from both its predecessor and its closest peer: IEC 60599:2022. The implications of these changes are difficult to intuit. This thesis details a comparison of their relative behaviours via case studies using real TX DGA data. More generally, it can be unclear how to proceed when facing issues attempting a practical deployment. Modifications to a prescribed methodology can undermine the basis of its original validity, however, capturing the original intent can be a time-consuming and nebulous task. This thesis highlights potential barriers and presents relevant quantifiable and analytical advances to facilitate the deployment of the IEEE C57.104-2019 methodology in an automated setting. The rationale is based on a holistic overview of the topic area and on findings from real TX DGA case studies. The presented improvements to the methodology target the default limits as well as the derivations to both the input and output metrics. These improve the methodology’s noise tolerance, metric consistency, and output granularity, respectively. The proposals are intended to preserve the methodology’s perceived original intent whilst improving upon its provided decision-support for the screening of TXs using DGA in practical deployment. Additionally, the thesis explores extending the methodology to incorporate a measure of uncertainty. Though IEEE C57.104-2019 emphasises the importance of uncertainty, it provides limited guidance on its practical application. A novel methodology, grounded in a Standards-based literature review, is proposed to address this gap by quantitatively assessing measurement uncertainty via the propagation of probability distributions. The methodology uses a Monte-Carlo technique, which is validated and demonstrated as a scalable, simple-to-apply solution for larger datasets.
Advisor / supervisor
  • McArthur, Stephen, 1971-
Resource Type
DOI

Relations

Items

Thumbnail Title Date Uploaded Visibility Actions
file details PDF of Thesis T17228 2025-03-27 Public Download