A Novel Discriminative Approach Based on Hidden Markov Models and Wavelet Transform to Transformer Protection

Abstract

In this paper we present a combinatorial scheme based on hidden Markov models (HMM) and wavelet transform (WT) to discriminate between magnetizing inrush currents and internal faults in power transformers. HMMs are powerful tools for transient classification which compute the maximum likelihood probability between training and testing data signals for identification. The WT is employed to extract certain features which reduce the computation burden of HMMs and enhance detection accuracy. The newly extracted feature efficiently discriminates between faults by different trends. The k-means clustering technique is applied to reduce the training procedure time investment. Since the discrimination method is based on the probabilistic characteristics of the signals without application of any deterministic index, more reliable and accurate classification is achieved. This method is independent of the selection thresholds. Based on the proposed algorithm a highspeed relay response (a quarter of a cycle) can be achieved. The suitable performance of this method is demonstrated by simulation of different faults and switching conditions on a power transformer using PSCAD/EMTDC software.