基于动态贝叶斯网络的高后果区油气管道风险评价

Risk assessment of oil and gas pipelines in high consequence areas based on dynamic Bayesian network

为实时动态反映油气管道在高后果区的风险水平,结合蝴蝶结(BT)和动态贝叶斯网络(DBN)模型对高后果区管道泄漏行为进行了分析,利用模糊概率和Leaky Noisy-or Gate模型分别确定DBN模型的先验概率和条件概率,并在人为因素、腐蚀情况等节点引入时间维度进行动态设置,得到管道泄漏概率和事故后果发生概率的时变特性。结果表明:管道泄漏概率呈先增大后减小再增大的趋势,在投产初期和运行后期的泄漏概率较大,应引起足够重视;管道占压和违章施工是引发高后果区管道泄漏的主控因素;随着时间的延长,不同事故的发生概率逐渐增大,事故发生概率从大到小依次为油气聚集、喷射火、闪火、油池火、火球和蒸气云爆炸。研究结果可为油气管道高后果区的动态系统分析和安全管理提供实际参考。

In order to dynamically reflect the risk level of oil and gas pipelines in the high consequence area in real time,the bow knot(BT)and the dynamic Bayesian network(DBN)model are combined to analyze the pipeline leakage behavior in the high consequence area.Fuzzy probability analysis and Leaky Noisy-or Gate model are used to determine the prior probability and conditional probability of the DBN model,respectively.In addition,the time dimension is introduced into the nodes of human factors and corrosion conditions for dynamic settings,and the time-varying characteristics of pipeline leakage probability and accident consequence probability are obtained.The results show that the probability of pipeline leakage increases first,then decreases,and then increases.The probability of pipeline leakage is larger in the early stage of production and later stage of operation,which should be paid enough attention to.Pipeline occupation and illegal construction are the main controlling factors of pipeline leakage in high consequence areas.With the extension of time,the occurrence probability of different accidents increases,and the occurrence probability of accidents ranked in descending order is oil and gas accumulation,jet fire,flash fire,oil pool fire,fireball,and steam cloud explosion.The research results can provide a practical reference for the dynamic system analysis and safety management of oil and gas pipelines in high consequence areas.