A Machine Learning Approach for Distribution Transformer Fault Detection

A distribution transformer fault refers to a malfunction or abnormal condition that occurs in a distribution transformer, which is a vital component in electrical power distribution systems. Distribution transformers lower high transmission voltage for use in homes and businesses. In this study, we address the critical challenge of transformer fault detection by employing a comprehensive approach that harnesses the power of advanced machine-learning techniques. We conducted extensive experiments utilizing a diverse dataset to predict and classify four distinct types of transformer faults. Leveraging the capabilities of twelve different classification algorithms, such as - SVM, DecisionTree(DT), AdaBoost, RandomForest(RF), ExtraTrees(ET), GradientBoosting(GB), MultipleLayerPerceptron(MLP), KNeighbors, LogisticRegression(LR), LinearDiscriminantAnalysis, XGBoost, and LGBM, we implemented a robust methodology that outperformed prior results, achieving a significantly higher accuracy rate. Key to our success is the integration of cross-validation, ensuring the reliability and generalization of our predictive models. This research advances the field of transformer fault prediction, offering a more accurate and dependable framework for identifying and mitigating critical issues in power distribution systems. The results hold substantial promise for enhancing the reliability and safety of transformer operations, with far-reaching implications for the energy industry.