Abstract:A statistical method by using distance-discrimination analysis is presented to utilize the information obtained from wavelet packet decomposition.The method employs a random energy vector to express the energy of each frequency band of the transient energy component and collects the statistic features of normal lines.The fault line then can be selected by getting the line with the maximum distance between its energy vectors with the normal line energy distribution.The information of each frequent bind and the relative informations between different frequent binds are made full use of and then can increase the accuracy of fault line detection.The method also combines the demographic characteristics of transient direction components and steady-state fifth harmonic components to perform fault line selection.