Defect analysis method for improving performance of mapping knowledge domain embedded model of main power equipment

The invention provides a defect analysis method for improving the performance of a knowledge graph embedding model of power master equipment, which comprises the following steps of: predicting defects of the power equipment by adopting a fine tuning and knowledge graph embedding model KGEMs based on BERT; the method comprises the following steps: firstly, creating a defect data set by collecting and arranging defect information of power master equipment; secondly, defining an ontology and designing a knowledge graph; then, a pre-trained BERT model is used for processing a specific field corpus, and BERT is finely adjusted to better understand sentence semantics and extract more valuable features; and finally, more accurate and effective electrical equipment defect prediction is realized by combining training and testing of KGEMs. According to the method, the defects of direct use of BERT in professional field vocabularies are overcome, the performance of a prediction task is remarkably improved, efficient and