This paper proposes a strategy for constructing a knowledge graph for pre-control of power operation risks based on bidirectional encoder representations from transformers(BERT) technology

aiming to significantly enhance the safety and efficiency of power operations through intelligent means. Firstly

a training dataset specifically for the risk domain of power operation is constructed. Secondly

the BERT-BILSTM-CRF(where BILSTM stands for bidirectional long short-term Memory

and CRF stands for conditional random field) model is used to identify the entities related to transformer operation and maintenance. Furthermore

the BERT-BILSTM-Attention model is used to effectively recognize the relationships between entities. The performance of these two models are demonstrated through comparative experimental results. Finally

686 entities and 720 entity relationships are successfully imported into the Neo4j graph database

achieving intuitive visualization of the knowledge graph. On this basis

auxiliary decision-making functions are developed.