Leakage current is an on-line monitoring parameter of contaminated insulator

which can dynamically reflect the discharge state of insulator surface. In this paper

the artificial contamination discharge test of porcelain insulator is carried out. The intrinsic mode function components

Hilbert marginal spectrum and time-frequency entropy of leakage current in different pollution flashover stages are analyzed by using-Hilbert Huang transform. 15 characteristic quantities are extracted from time-frequency domain and waveform details. The contamination discharge state is identified by using principal component analysis and least squares support vector machine classifier. The results show that there are many components of leakage current intrinsic mode function in the initial discharge stage and flashover stage. The upper frequency of leakage current in Hilbert marginal spectrum is mainly distributed in the range of 0 ~ 150 Hz and 200 ~ 250 Hz. The time-frequency entropy of leakage current before flashover is always greater than that after flashover. When the number of training samples is 5 times or more than that of test samples

the comprehensive criteria accuracy of the classifier can reach 99%

and the classification and recognition of contamination discharge state can be realized accurately. The results of this paper can provide a basis for the set up of insulator contamination flashover early warning system.