Application of Dempster-Shafer's theory of evidence for transformer incipient fault diagnosis

• Published in: 8th International Conference on Advances in Power System Control, Operation and Management (APSCOM 2009) p. 324
• Main Authors: Lee Hui Min, Chang, C.S
• Format: Conference Proceeding
• Language: English
• Published: Stevenage IET 2009
• Subjects: Knowledge engineering techniques; Neural computing techniques; Transformers and reactors; robust power transformer; fuzzy-neural-network; IEC 60599 gas-ratio method; fuzzy neural nets; fault location; Duval Triangle method; power transformer testing; inference mechanisms; Dempster-Shafer theory of evidence; transformer incipient fault diagnosis; dissolved gas analysis
• ISBN: 1849192146; 9781849192149
• DOI: 10.1049/cp.2009.1844

This paper focuses upon the development of a robust power transformer incipient fault diagnosis system using data collected from the dissolved gas analysis (DGA). Results from two established and two fuzzy-neural-network (FNN) based methods are fused together with a modified combination rule of Dempster-Shafer's Theory of Evidence (DST). The two established methods are the IEC 60599 gas-ratio method and Duval Triangle method. A platform called the Transformer Fault Clinic is designed and created for guiding the user to compare and combine the diagnosis results from the four methods. The DST combination approach estimates the belief levels of all the faults being diagnosed. Using this, the user will obtain a concrete overview of the insulation condition of a power transformer, and the nature and severity of the faults. This will also enable the appropriate maintenance scheduling to be carried out. (6 pages)