Enhanced Fault Diagnosis in Transformer Oil using Duval Pentagon Method

Conventional transformer fault diagnosis employing Dissolved Gas Analysis (DGA) and the Duval Pentagon method is limited in resolving uncertain fault classifications. This research improves fault detection accuracy through the integration of the Duval Pentagon method and a Decision Tree classifier. The model proposed operates on raw gas concentrations and gas ratios, sharpening fault categorization through the identification of ambiguous cases with overlapping features. Implementation employed real-world and benchmark DGA datasets in MATLAB. Results show that the Decision Tree classifier enhances fault discrimination, attaining an accuracy of 91.84%, specifically in discriminating thermal and electrical discharge faults, to improve transformer diagnostics.