Wideband transient data acquired by high speed collecting device with sensitive starting contains a large number of interference noises. It is urgent to automatically filter the flashover fault traveling wave from the mass of transient data. The entropy

mean value and variance of wavelet energy are used to characterize the flashover fault traveling wave and interference noise data prior to forming the characteristic matrix of the traveling wave data. The principal component analysis(PCA)is used to reduce the dimensions of the characteristic matrix and the main information on the original characteristic matrix is retained by the dimension-reduced matrix as best it can. By taking the dimension-reduced characteristic matrix as a sample

the identification of the flashover fault traveling wave and the interference noise is made using the clustering algorithms of Mahalanobis distance and mixed Gaussian model. Then the flashover fault traveling wave is filtered by the computer. A large number of field measured data show that the clustering algorithms of Mahalanobis distance and mixture Gaussian model is able to achieve reliable filtering of flashover fault traveling wave

while the mixed Gaussian model has a better identification effect with a correct rate above 97%.