DEVELOPMENT OF TRANSFORMER PROTECTION ALGORITHM USING WAVELET AND ARTIFICIAL NEURAL NETWORK

Abstract

The dissertation aim at classiFying and discriminating the healthy and faulted conditions that arise in a three-phase power transformer system. To deal with the problem of detection and classification of fault, two different algorithms have been proposed and are presented in this work. The first fault discrimination method uses the wavelet energy pattern obtained up to fifth level after the decomposition of difièrential current signals using Discrete Wavelet Transfbrm (DWT) with 'sym 2' as mother wavelet. In the second proposed method, a combined DWT and ANN based technique is used for discriminating the faulted condition and inrush behavior of power transformer. Further is proposed a complete digital transformer protection algorithm. The different fault situations are simulated using PSCAD/EMTDC software and the perfonmrnce of the proposed schemes are evaluated in MATLAB software. The results obtained have turned out that the proposed methods offer good accuracy in discriminating between the faulted and inrush condition of the power transformer when compared to the classical protection techniques