Application of Gene Expression Programming (GEP) in Power Transformers Fault Diagnosis Using DGA

Abstract

Accurate diagnosis of incipient faults in oil-filled power transformers is important in preventive maintenance of transformers. Dissolved gas analysis (DGA) is an effective tool to diagnose incipient transformer faults. The majority of the methods reported in literature to analyze DGA results lay more emphasis on user experience rather than mathematical formulation/justification. Furthermore, sometimes DGA results for a certain fault do not belong to any of the IEC/IEEE standard and cannot be categorized/diagnosed. To address these issues, we propose a new approach for DGA interpretation using gene expression programming (GEP). The proposed approach is employed for analysis of 552 DGA samples collected from transformers of Himachal Pradesh State Electricity Board, India, in conjunction with samples extracted from reliable literature. We use the aforementioned dataset to test and validate our proposed GEP model. We also compare the performance of our approach against other artificial intelligence-based techniques such as artificial neural network, fuzzy-logic, and support vector machine. Results and comparison against other soft computing approaches show relative superiority of GEP-based DGA interpretation in terms of classification accuracy.