Detection and Classification of Internal Faults in Power Transformers using Tree based Classifiers

Abstract

This article discriminates the internal faults from magnetizing inrush for a 3-phase transformer using a Decision Tree (DT). Afterwards, the internal faults are classified with DT, Random Forest (RF), and Gradient Boost (GB) classifiers. An array of time, frequency, and time-frequency features are extracted from the 3-phase differential currents. Sample entropy is chosen to distinguish the faults; and change quantile and absolute energy are used to classify the internal faults. The internal faults and magnetizing inrush cases are created by altering the system parameters in PSCAD/EMTDC software. The DT performs well with 100% accuracy for fault detection, and the GB based fault classifier performed the best among the three classifiers with an accuracy of 95.4%.