Discrimination of Magnetizing Inrush and Internal Fault Currents Based on Stockwell Transform and ANN Approach for Transformer Protection

Abstract

In this study, Stockwell transform and artificial neural network were used in determining the inrush current and the internal current fault based on the power transformer protection. The S-transform is a robust transform that incorporates the time and frequency characteristics used in the analysis of non-stationary short term transient signals. It is used for pattern recognition for distinction between internal faults and inrush current. Time-frequency images were obtained by using S-transform, and the obtained images were observed to be different in internal faults and inrush current. The feature extraction is based on statistical methods, standard deviation and average value, the classification process was performed with the multilayer feed forward artificial neural network. The classification performance is calculated at a hundred percent accuracy. © 2019 Chamber of Turkish Electrical Engineers.