A New Approach Based on S-transform for Discrimination and Classification of Inrush Current from Internal Fault Currents Using Probabilistic Neural Network

Abstract This article presents a new approach for differential protection of power transformers. The proposed method uses S-transform and a probabilistic neural network to discriminate internal faults from inrush current. S-transform is utilized to extract some useful features of non-stationary signal analysis, giving the information of transient currents both in time and frequency domains. The features extracted using S-transform are applied to train probabilistic neural network classifiers. This approach has been realized through two different stages. In the first stage, discrimination of inrush current and fault current has been done; in the second stage, different types of fault current will be recognized in four steps. The performance of this algorithm is demonstrated by simulation of different faults and switching conditions on a power transformer using PSCAD/EMTDC software (Manitoba HVDC Research Center, Winnipeg, Manitoba, Canada). The simulation results show that the combination of S-transform and a probabilistic neural network can effectively detect inrush current from fault currents and that it can also classify the fault currents with high accuracy and speed, even in a noisy environment.