Analysis of potential use of the SVM technique for transformer protection

Reliable and fast discrimination between internal faults and inrush conditions is still a challenging issue. In this paper an application of Support Vector Machine (SVM) for the transformer differential protection is discussed. To achieve the satisfactory classification strength various input vectors and training parameters were considered. Finally, 16 different versions of SVM classifiers are proposed. The developed SVM based power transformer protection units have been trained and tested with EMTP-ATP generated signals. The operation performance of designed SVM classifiers is compared to standard differential protection with traditional second harmonic stabilization approach. Moreover, potential hardware implementation of the presented SVM classifiers is analyzed.