Knowledge Graph Construction For Intelligent Maintenance Of Power Plants

The intelligent maintenance of power plants greatly relies upon previous experiences and historical information to make informed decisions. Previous research is predominantly focused on collecting data from the maintenance process while little work has been done on the automatic capture and mining of knowledge resources from the data accumulated. Focusing on the experience feedback issue in power plants maintenance, this work proposes a novel process of automatic construction and reasoning of knowledge graphs to support the intelligent maintenance of complex power equipment. In this process, the Bi-LSTM-CRF model and the attention-based Bi-LSTM are specifically used to identify and extract entities and relations from unstructured status reports. On this basis, the knowledge graph construction method based on the neo4j graph database is developed. This paper details the preliminary work towards implementing the proposed process.