加氢站泄漏燃爆事故的动态贝叶斯风险分析  

Risk analysis of hydrogen refueling station leakage and explosion accidents using dynamic Bayesian networks

作　　者：程方明 蒋萌 苏畅 屈姣 李卓 王郦楠 CHENG Fangming;JIANG Meng;SU Chang;QU Jiao;LI Zhuo;WANG Linan(College of Safety Science and Engineering,Xi'an University of Science and Technology,Xi'an 710054,China;Xi'an City Public Safety and Fire Rescue Key Laboratory,Xi'an 710054,China;Key Laboratory of Urban Safety and Emergency Rescue in Shaanxi Province Higher Education Institutions,Xi'an 710054,China)

机构地区：[1]西安科技大学安全科学与工程学院,西安710054 [2]西安市城市公共安全与消防救援重点实验室,西安710054 [3]城市安全与应急救援陕西省高等学校重点实验室,西安710054

出　　处：《安全与环境学报》2024年第9期3405-3417,共13页Journal of Safety and Environment

基　　金：国家重点研发计划课题(2021YFB4000905)。

摘　　要：针对加氢站氢泄漏燃爆事故风险分析中不确定性、复杂性和时间序列动态性的问题,提出了一种新的加氢站氢泄漏燃爆事故动态风险分析方法。该方法基于动态贝叶斯网络建立,利用GM(1,1)灰色预测和模糊集理论构建先验概率模型,能够在概率数据缺失的条件下,依据加氢站风险规避能力的时序变化规律,实现风险因素先验概率的预测。引入Leaky Noisy-or gate模型、Markov假设处理动态贝叶斯网络条件概率的不确定性,使模型更加客观。该方法适用于实际工况下事故后果概率的短期预测。以某加氢站为应用对象,验证了方法的合理性,通过定量分析、诊断分析、敏感性分析和影响强度分析分别得到了“喷射火”等事故后果概率及其时序变化曲线、“管线防腐失效”等关键风险因素、“高压氢气管路失效”等高敏感性风险因素、泄漏燃爆事故关键致因链,并依据风险因素影响模式提出了针对性管理对策。We propose a novel and dynamic risk analysis method for addressing the uncertainty,complexity,and time series dynamics associated with hydrogen leakage and explosion accidents in hydrogen refueling stations.This method aims to provide a more thorough and effective approach to assessing the risk factors involved in such accidents.This approach is grounded in dynamic Bayesian networks and employs a fusion technique that combines GM(1,1)grey prediction and fuzzy set theory to build a prior probability model.When there is a lack of risk factor probability data,it can forecast the prior probability of risk factors by analyzing the temporal variation pattern of the hydrogen refueling station's risk avoidance capability.Introducing the Leaky Noisy or gate model and Markov hypothesis can aid in addressing the impact of omitted risk factors and handling the uncertainty of conditional probabilities in dynamic Bayesian networks.This approach can also make the model more objective and suitable for short-term prediction of accident evolution under actual working conditions.The Dynamic Bayesian Network(DBN)model was developed to analyze the risk of leakage and explosion accidents at a specific hydrogen refueling station.The accuracy and reliability of this method were confirmed through the application of the three axioms theorem and risk factor sensitivity analysis.Through quantitative analysis,diagnostic analysis,sensitivity analysis,and influence strength analysis,we have determined the likelihood of accident consequences like“jet fire”and its time-sequence change curve.Additionally,we have assessed the probabilities of catastrophic accidents resulting from leaks in various work units and at different hydrogen leakage flow rates.Key risk factors,such as“pipeline corrosion prevention failure”and highly sensitive risk factors like“high-pressure hydrogen pipeline failure”,have been identified.Moreover,we have established the key causal chains of leakage and explosion accidents.Utilizing the risk factor impact model

关 键 词：安全工程 加氢站 泄漏燃爆事故 风险分析 动态贝叶斯网络(DBN) 

分 类 号：X937[环境科学与工程—安全科学]