Fault-generated high-frequency noise has been proven to be effective for fa
ulted phase selection. A combined method using HF noise, fast Fourier trans
form (FFT), and neural networks (NN) for phase selection has been proposed
previously; however, FFT and NN have some implicit disadvantages. This pape
r describes a HF noise based method for phase selection using wavelets base
d feature extraction. It is shown, that the features extracted by wavelets
transform (WT) have a more distinctive property than those extracted by FFT
due to the good time and frequency localization characteristics of WT. As
a result, the proposed method dispenses with the neural networks and hence
is more reliable and simpler than the previous FFT-based method. Extensive
simulation studies have been made to verify that the proposed approach is v
ery powerful and apropos to phase selection.