Empirical mode decomposition based probabilistic neural network for faults classification

Abstract

This paper presents a novel method of detecting and classifying the power system faults of voltage sags based on Empirical Mode Decomposition (EMD). A technique employed for analyzing power system fault data in terms of voltage sags is required. Also, provides information about the underlying event i.e. the fault type. EMD is to method which decomposes a non stationary signal into mono component and symmetric component signals called Intrinsic Mode Functions (IMFs). Further the Hilbert Transform (HT) of IMF provides magnitude and phase angle information. The characteristic features of the first three IMFs of each phase are used as inputs to the classifier Probabilistic Neural Network (PNN) for identification of fault type. Four types of shunt faults are taken for classification. A comparison is also made with wavelet Transform (WT). Simulation results show that the classification accuracy is better for EMD, which proves that the method is efficient in classifying the faults.