Analysis of urban hydrogen-blended natural gas pipeline leak failure and accident evolution based on the combination of causal inference and probabilistic machine learning

Abstract

Integrating hydrogen into urban gas pipeline networks is a pivotal technology for energy transition yet poses critical safety threats, thus necessitating comprehensive risk assessment of hydrogen-blended natural gas pipelines. This study performs full quantitative risk assessment of leakage failure and accident evolution by proposing a novel framework that integrates causal inference (Bow-Tie analysis) with probabilistic machine learning (Bayesian networks), enabling systematic failure factor identification and dynamic accident progression simulation. Key findings indicate human factors and pipeline material degradation as primary triggers. The studied pipeline exhibits a low baseline failure probability, with dispersion emerging as the most likely consequence of leakage. Higher hydrogen blending ratios significantly elevate jet fire risk due to hydrogen’s low ignition energy, while hydrogen’s inherent buoyancy and high diffusivity notably mitigate the likelihood of flash fire and vapor cloud explosion. The case study verifies the model’s practicability, and macro-micro analyses provide holistic insights, offering a reliable method to guide pipeline safety and reliability improvement amid energy transition.