Abstract: Remaining useful life prediction can not only improve the maintenance efficiency of relay protection equipment of the power system but also enrich the full-lifecycle management of power equipment in the state grid. Based on the relay protection equipment of JC city power supply company and its characteristics of data, this paper defines the design life course and actual life course according to the variable installation date, damage date, current date, and design life related to attributes. Support vector regression(SVR), regression tree(RT), and random forest(RF) methods are used to predict the remaining useful life of unmonitored equipment. Case analysis shows that the prediction method of random forest performs best.

Key words: support vector regression, regression tree, random forest, remaining useful life, relay protection