In power grids

operating and maintaining engineers have recorded plenty of texts or logs during maintaining and inspecting activities. These textual data contain abundant asset health information. So far

however

few researches

if any

have studied text mining techniques in the power grid. We take the circuit breaker(CB)as a case in point to establish a framework of text mining-based lifecycle condition assessment. Firstly

the key issues of text mining and lifecycle condition assessment models are listed based on reviewing the research of CB condition assessment. Then

use is made of the framework including a hidden Markov model(HMM)-based text preprocessing and vectorizion

self-interval searching k-nearest neighbor(KNN)-based text classification

and a proportional health-index fusion model(PHFM). Finally

we have collected real textual data from a certain power company to demonstrate an example that shows the text mining technique couldlearn similar defects from other assets by itself

and PHFM shows historical data stream and lifecycle health index much more rigorously.