Development of an Intelligent GUI with ChatGPT for Efficient and Accurate Transformer Fault Diagnosis Using DGA, Duval Triangle Method, and Machine Learning Automation

This study presents an intelligent graphical user interface (GUI) for transformer fault diagnosis, utilizing Dissolved Gas Analysis (DGA), the Duval Triangle method, and automated machine learning model selection. Key gases—Methane (CH<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">4</inf>), Ethylene (C<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> H<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">4</inf>), and Acetylene (C<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> H<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf>)—are analyzed to classify fault types, including partial discharges and thermal/electrical faults, through gas concentration ratios. The system leverages TPOT's model selection capabilities, with the Multi-Layer Perceptron (MLP) emerging as the optimal classifier, achieving over 95% accuracy. The GUI, developed using Gradio, enables CSV uploads for batch processing and real-time fault classification, complemented by ChatGPT-generated explanations and maintenance suggestions. A case study on mineral oil-based transformers demonstrates the tool's high diagnostic accuracy, user-friendly design, and potential for expanded applications in predictive maintenance.