This repository maintain codes and logs for "Partial Discharge Localization in Power Transformer Tanks Using Machine Learning Methods" paper.
This paper detail an exploration of several machine learning (ML) methods used for three-dimensional partial discharge (PD) localization in a power transformer tank, which are examined from linear regression as the simplest model to convolutional neural networks (CNN) as a sophisticated deep learning (DL) model respectively based on their structural complexity. Moreover, other case studies are considered focussing on different attributes including the sensor position, size and shape of the transformer tank and finally effect of using three sensors. The PD location predicts in three-dimensional through single-sensor electric field measurement except for the triple sensor case study. Features of each method, such as input signal, core methodology, the correlation coefficient of predicted location with the actual location, and root mean square error (RMSE) analysis are summarized and compared. Among the considered methods, the CNN model founded to have the best performance in terms of location accuracy.
