Intelligent Condition Monitoring of Power Transformers via Machine Learning

Power transformers are critical infrastructure in the energy distribution sector. Thus, constant monitoring of the condition and health of the system is crucial. Digital technologies such as machine learning have shown promising applications in fault diagnosis and condition monitoring of smart power systems. In this study, seven machine learning models were developed and trained on dissolved gas analysis (DGA) data to determine the condition of the power transformers. Techniques explored include logistic regression (LR), decision tree (DT), random forest (RF), gradient boosting (GB), k-nearest neighbors (KNN), support vector machines (SVM), and multilayer perceptron (MLP). The performance of the models was studied with and without correction for data imbalance via oversampling, as well as feature scaling techniques. For hyperparameter tuning, the GridSearch technique with 5-fold cross-validation was applied.