Identification and prediction of Fault types in Transformer using Decision Tree Model

Power transformers are among the most vital and expensive assets in the electric power industry. Dissolved Gas Analysis (DGA) is widely recognized as the leading diagnostic tool globally for detecting early insulation failures in transformers. However, interpreting DGA results presents significant challenges for transformer owners and utility engineers due to the complexity and ambiguity associated with fault gas analysis. While various computational techniques and offline methods have been employed for DGA fault classification, issues such as limited data availability, ambiguity in interpretation, and model accuracy remain critical hurdles. This research addresses these challenges by utilizing a large and diverse dataset of DGA samples from in-service transformers operated by the Transmission Company of Nigeria (TCN). A total of 580 DGA samples were collected, with 80% used for training and 20% for testing a Decision Tree model. The results demonstrate that the model is highly effective in identifying and predicting transformer faults, achieving a weighted average accuracy, recall, and F1-score of 0.97. The findings suggest that nonparametric and non-linear models like Decision Trees are particularly well-suited for DGA fault classification. The high accuracy achieved in this study indicates that the Decision Tree model provides a reliable approach to transformer fault prediction. Implementing this model in the power industry can facilitate predictive maintenance strategies, enhancing the overall reliability and efficiency of power systems.