SVM Based Methodology for Classification of Internal Faults from Other Disturbances in Power Transformer

This paper introduces a differential protection scheme for power transformers based on a Support Vector Machine (SVM) approach, designed to effectively classify internal faults while distinguishing them from other disturbances, such as magnetizing inrush currents and overexcitation conditions. The feature vector used for classification is extracted from the differential signal by computing the energy of the detail 2 coefficients obtained through wavelet transform analysis. This feature vector serves as input to the SVM classifier, ensuring accurate discrimination between fault conditions and non-fault disturbances. Comprehensive simulations are performed to evaluate the scheme under various scenarios, including internal transformer faults, diverse magnetizing inrush conditions, overexcitation states, and normal operating conditions with varying load levels. The power transformer is modeled and analysed using the MATLAB Simulink software environment.