The Effect of Source Data on Graphical Pentagons DGA Methods for Detecting Incipient Faults in Power Transformers

Abstract

Detecting power transformer faults at an early stage is crucial for ensuring the safety and uninterrupted operation of electrical systems. Dissolved Gas Analysis (DGA) serves as a valuable strategy for diagnosing faults in oil-filled transformers. Oil plays a vital role in transformer operation, serving as both a cooling and insulation medium. It also acts as an indicator for assessing thermal and electrical stresses within the transformer by analyzing the composition of dissolved gases in the oil. In this paper, the graphical techniques of the Pentagons for each of Duval Pentagon method and Mansour Pentagon method were discussed at different data sources. The dataset comprises 357 samples, with 117 samples sourced from IEC TC 10 data and 240 samples obtained from Egyptian laboratories. The Duval Pentagon method demonstrates superior performance, achieving a diagnostic accuracy of 69.23% when applied to the IEC TC 10 data. Conversely, the Mansour Pentagon method outperforms the Duval Pentagon method with a diagnostic accuracy of 68.75% when applied to the Egyptian laboratory data. This paper is a proposal to find a new graphical method that adapts to the variables of the source data. The proposed approach aims to address the challenges posed by diverse data sources and provide a more comprehensive and adaptable framework for transformer fault diagnosis.