Internal fault detection in three phase transformer using machine learning methods

Abstract

The study of ABC (artificial bee colony algorithm) and four different machine learning methods has been explored for internal fault detection in three phase transformer using differential protection scheme. The half cycle window of differential current has been sampled at 1 kHz sampling frequency for classification of five operating conditions i.e. normal, magnetizing inrush, over-excitation, internal and external fault condition. 420 samples have been generated by modeling the differential protection scheme of Y-Y transformer and simulating different operating conditions usingsimpowersys of MATLAB/SIMULINK. The training and testing result shows that random forest method gives best result as compared to decision tree, linear model and support vector method.The k-fold cross- validation has been used for measuring the accuracy and sensitivity of random forest machine learning method. This gives the best result for classification of internal fault and other operating conditions.