Knowledge graph based early warning method for accident evolution in natural gas pipeline station abroad for harsh environmental conditions

In recent years, harsh environmental conditions (lightning, wind and rain) have posed significant threats to the safe operation of long oil and gas pipelines, especially in oil and gas pipelines in other countries. The operation and emergency responses for oil and gas pipelines abroad have faced various problems such as insufficient pipeline risk accident data and difficult cross-border coordination during harsh environmental conditions. Current pipeline accident early warning models rely too much on field data or accident-related data. Thus, this paper presents a station accident risk evolution early warning method based on a knowledge graph that uses a small amount of accident report data from natural gas pipeline stations for harsh environmental conditions. The method uses a bidirectional long short-term memory-conditional random field algorithm (Bi-LSTM-CRF) to extract the causal relationships from station accident reports, with a Neo4j graph database then used to establish the knowledge graph for accident risk evolution in natural gas pipeline stations abroad experiencing harsh environmental conditions. The results show that this knowledge graph based station accident risk early warning method not only provides earlier warnings of station accidents than traditional station accident early warning methods for long pipelines, but also predicts the accident path with recommended accident responses. The results show that this early warning method can effectively help safety management personnel in natural gas pipeline stations abroad provide better risk control and accident prevention.