This article compares machine learning methods for three-dimensional localization of partial discharges in power transformers. It focuses on predicting PD location using single-sensor electric field measurements and finds that the CNN model performs the best.