Abstract: There are a lot of equipment operation rules in power equipment defect text，which can be mined by

natural language processing technology to provide guidance for equipment maintenance. Aiming at this situation，a defect text mining method based on enhanced representation through knowledge integration（ERNIE）pre-trained was constructed. Firstly，the limitations of traditional word vector model were analyzed，and the characteristics of power equipment defect text were summarized. Aiming at the above problems，ERNIE was used for pre training to get the word vector containing the context information of the defect text. Then，bidirectional long short term memory（BiLSTM）was used to learn the defect text in two directions，and the semantic features of the defect text were further extracted and classified. Finally，an example shows that ERNIE-BiLSTM has better classification performance，which provides a new idea for future power equipment defect text management.

      Key words: power equipment；defect text mining；pre training model；knowledge enhancement；bidirectional long short term memory（BiLSTM）