Machine Learning-Based Prediction of Asset Health Index for Distribution Transformers

The Asset Health Index (AHI) is an essential tool for assessing the condition and performance of transformers in electrical distribution networks. AHI enables predictive evaluations that assist in decision making related to asset maintenance and replacement, ultimately enhancing the reliability and efficiency of the electrical system. Their failure can lead to significant operational and financial losses. In this work, the application of three machine learning (ML) algorithms K-Nearest Neighbors (KNN), Artificial Neural Networks (ANN), and Random Forest was emphasized in predicting the health condition of the distribution transformer. Using a data set of transformer operational parameters, the models are trained and evaluated based on accuracy, precision, recall, and F1 score. Hyperparameter optimization combined with cross-validation was highly efficient in determining the optimal set of hyperparameters for the model, ensuring enhanced performance and greater robustness across datasets. In addition, the transformer health index serves as a valuable reference for planning maintenance schedules for distribution transformers and prioritizing maintenance tasks, ultimately optimizing both operational and asset maintenance costs. The Random Forest model achieved an accuracy of 95 percent, substantially surpassing the performance of KNN, which achieved 72 percent accuracy, and ANN, which achieved 88 percent accuracy. The study highlights the potential of machine learning in the monitoring of transformer health and provides insight into selecting the most suitable algorithm for the prediction of AHI of distribution transformers in PT. PLN XYZ.