Transformer Winding Fault Classification Based on Pattern Recognition Technique

Recognition of transformer winding fault conditions by analyzing the current waveform disturbances is very important task for transformer health monitoring. This paper presents a novel approach for classification of transformer winding fault conditions viz. ageing, arcing and inter-turn short circuit at different loads, which is based on discrete wavelet transform (DWT), Parseval's theorem and neural network. In the proposed technique DWT is used to decompose signals into different resolution levels and then energies of these levels are calculated to determine powerful features. Further, useful feature information is used to develop neural network classifier. The effectiveness of neural network classifier is verified by giving unknown winding fault inputs and comparing the results with known outputs. Results thus obtained show the versatility of the classifier ill classifying the winding faults of transformers. Copyright (C) 2012 Praise Worthy Prize S.r.l. - All rights reserved.