Power Transformer Protection Using Improved S-Transform

摘要

This article presents the detection and classification of a fault and abnormal conditions that occur in power transformers based on improved S-transform. The feature items are selected from improved S-transform, which is a very powerful time-frequency analysis tool. In order to avoid specificity in classifier strength determination, three different approaches are applied for feature selection, namely sequential forward selection, sequential backward selection, and the genetic algorithm. To further specify the feature for a given classifier, two classifiers-the probabilistic neural network and the support vector machine-are applied. The suitable performance of this method is demonstrated by the simulation of different faults and switching conditions in a power transformer using PSCAD/EMTDC software (Manitoba HVDC Research Center Inc., Manitoba, Canada). Results indicate that the proposed technique is suitable, reliable, and fast during the detection of a fault current and classification of different events.