An effective combined method for symmetrical faults identification during power swing

This paper presents a new protection scheme based on combination of S-Transform (ST) and Probabilistic Neural Network (PNN) methods to identify symmetrical faults during power swing conditions. The ST, an effective time-frequency decomposition transform, is used for extraction of beneficial features from a half cycle of the current signal. The constructed features vector is fed to a PNN classifier as an input pattern. No need to set the initial weights is the key attribute of PNN. The discrimination capability of extracted features which is the main contribution of the proposed method is investigated under different circumstances. The simulation results show that the protection scheme identifies a symmetrical fault during power swing correctly. Moreover, the proposed intelligent scheme has a fast performance due to the low computational burden of the combined methodology. The efficacy of the proposed scheme is confirmed by comparing with some of existent algorithms.