Paper Title
Fault Detection in Three Phase Transformer using Ensemble Model

Abstract
Three phase transformers are an important part of electrical power distribution systems. This type of transformer is composed of primary and secondary windings and can be connected in either star or delta formations. Detection of faults in such high power transformers is crucial and can help in reducing such faults up to a great extent. The ensemble model is used for the detection of five faults in three phase transformer. Seven feature selection techniques and seven machine learning models have been used, out of which five best possible combinations are selected for ensemble model. Five different operating conditions are discussed namely normal condition, inrush, over excitation, internal fault and external fault. These faults are classified by 30 features i.e. 10 samples of three phase differential current have been used. 420 simulated samples have been generated using sim power systems of MATLAB/SIMULINK under different conditions of Y-Y transformer. Cross Validation is done to show the robustness and consistency of the best predictive models on which ensemble model is applied. The results show that on applying ensemble techniques, the accuracy improves significantly to about 85.652% in fault prediction based on the dataset. Index Terms - Fault Detection, Three Phase Transformer, ARC, Random Forest, Gini, Internal Fault Simulations.