Partial Discharge Detection in Transformer Using Machine Learning

Partial discharge (PD) is a significant problem in electrical insulation systems, frequently resulting in substantial harm and malfunction of electrical equipment, such as transformers. Timely and accurate locating PD is essential to prevent potential risks and enhance the reliability of power networks. With the use of machine learning (ML) methods and a two-dimensional (2D) model, this study offers an approach for localizing PD sources inside a transformer tank. This method examines the time of arrival (ToA) of acoustic waves at three sensors positioned on the outside of the tank. The machine learning algorithms utilized comprise XGBoost, Random Forest, and Decision Tree models. The models are trained and tested using synthetic data derived from ToA computations. XGBoost has superior performance compared to the other models, with the lowest error.