Power Transformer Protection Using Fast Discrete S-Transform and Optimized Support Vector Machine Classifier with Bee Algorithm

This study presents a Fast Discrete S-Transform based method to discriminate internal fault currents of power transformer from other disturbances. A criterion function is proposed based on some extracted features from the obtained S-Matrix and frequency contours. First, the Support Vector Machine (SVM) is extended for feature classification. Then, the Bee optimization algorithm is implemented to select optimal parameters of SVM classifier. To do this, several conditions of external and internal faults, inrush current and different levels of current transformer saturation are simulated using PSCAD/EMTDC software. In addition, differential currents are contaminated by noise for modeling real conditions. To evaluate the performance of proposed scheme, the obtained results are compared with results of other methods. Comparing the results shows that the proposed method remains stable with high accuracy during transformer excitation and external faults. Also, the proposed approach is effective, fast and not affected by noise during classification of different events.