With the increase of voltage grade in power transmission system and the aggravation of industrial pollution

flashover of contaminated insulators in HV and UHV system is more and more serious

threatening the security of power transmission severely.The key of improving the security of system is exploring the methods of monitoring states of insulator’s surface by some character values.In this paper

a method based on clustering and discriminating of statistics is introduced for distinguishing the flashover stages by leakage current and humidity.Through training with the experiment data

a judge equation is established

and the cross-validate results show that Fisher method is more accurate than Bayes in discriminating the falshover stage

which is important for deciding the insulator’s state and forecasting the risk of flashover.