Multi-factor risk analysis and early warning system of oil and gas accidents based on N–K coupling model

Current research on oil and gas accidents generally has the problems of scattered risk factor identification, linear and static methods, and difficulty in modeling multi-factor nonlinear coupling and its dynamic evolution path. This paper introduces the N–K coupling model to construct a multi-factor dynamic risk coupling analysis framework. Based on typical cases and literature on oil and gas accidents, this paper extracts multi-category core risk factors and constructs a hierarchical indicator system. Then, based on the N–K coupling model, each factor is treated as a system node, and its coupling relationships with other nodes are established to form a multi-factor coupling network. The state of each factor is encoded and assigned a coupling weight. The overall risk of the system under different state combinations is calculated, and high-risk coupling paths and key factors are identified. Finally, the dynamic analysis of risk evolution is simulated to predict risk accumulation and potential loss of control, thereby achieving quantitative early warning of multi-factor coupling risks. Experiments show that the average F1 score of the proposed model across six accident scenarios is 84.8%, which is significantly better than that of the BP Neural Network, Random Forest, and other models. The average recall rate of multiple accident cases is 89.8% (false alarm rate is 14.3%). The system’s accident time prediction deviation is only 2.7 days, enabling early warning, responsiveness, and ultimately quantitative risk analysis and dynamic warning.