Abstract: Aiming at the difficulty of rapid decision support and plan formulation in the firefighting and emergency response process, this article proposes an emergency model for gas leakage and fire accidents at gas storage facility sites. This model uses a knowledge graph as a means of risk characterization, employing Bidirectional Encoder Representations from Transformers (BERT) and Bidirectional Long Short-Term Memory Model Conditional Random Field algorithm (Bi-LSTM-CRF) for entity recognition and relationship extraction from textual intelligence. The emergency knowledge graph for gas leakage and fire accidents at gas storage facility sites is constructed using the Neo4j graph database. The results show that compared to traditional emergency handling and firefighting strategy research methods, the emergency model proposed in this paper for gas leakage and fire accidents at gas storage facility sites not only enables early emergency handling but also identifies the risk propagation paths of accidents, providing support for firefighting emergency command and emergency decision-making.

Key words: firefighting and rescuing, emergency decision, knowledge graph, natural gas leakage, BERT-Bi-LSTM-CRF