Wavelet and Neural Network-Based Fault Location in Power Systems Using Statistical Analysis of Traveling Wave

Abstract

This paper describes a traveling wave-based fault location in power systems without using global positioning system (GPS) timing. To extract the transient wave from the recorded waves at the bus bars, wavelet denoising is used. The residual signal in this procedure has a large amount of information about the fault. The proposed algorithm uses the statistical analysis parameters of the traveling wave as the inputs of a defined artificial neural network. All the possible fault types are generated using the ATP-EMTP and results are discussed. Extensive simulation studies indicate that proposed network has a reliable response. 0.92 % Average error shows the power of this algorithm against the reactance-based techniques. In contrary, this is a higher error than the traveling wave-based fault location using GPS timing. Nevertheless, in proposed method because of omitting complicated utilities such as GPS receiver, the costs will be decreased and a higher degree of performance in an efficient manner is achieved.