Abstract:With dramatic increase of power equipment, how to manage and handle defect records effectively has become an urgent problem. The conventional mode of manual processing is so inefficient that it can hardly respond to challenges in text mining. To solve this problem, an accurate retrieval method of defect text for power equipment based on knowledge graph and entropy weight is proposed in this paper. According to defect codes and standards, the knowledge graph of power equipment defects is built, and the path database of standard defects is obtained. Considering characteristics of the electric-power industry, pre-learning strategy for professional terms and phrases is adopted in the word segmentation process. With the large-scale knowledge graph available, the tough problem of coreference resolution is solved. Based on the built knowledge graph and entropy weigh model, the standard path retrieving and similarity ranking are realized. Through case study, results show that this method retrieves defect text and locates standard path accurately, and provides significant reference for localization of codes and standards. Importantly, this research is indeed of benefit to the operation and maintenance management of power equipment in both quality and efficiency, and the safe grid with reliable power supply besides.