基于蝴蝶结-贝叶斯网络的站场储罐动态风险评价

Dynamic Risk Assessment of Storage Tank in Station Based on Bow-Bayesian Network Shi Haotong

为实现站场储罐区风险的实时评估和事故预防,在建立蝴蝶结模型的基础上,通过映射规则引入时间维度,归纳了储罐泄漏的主要危险源和不同安全屏障下的事故后果类型,通过遗漏概率模型确定了不同节点的条件概率,以此形成站场储罐的动态贝叶斯网络模型,推理出储罐事故发生概率随时间的变化规律,并进行实例分析和验证。结果表明:储罐泄漏概率先增大后减小,并在第3个时间片处达到峰值;后验概率较大的基本事件均与储罐腐蚀的中间事件相关,随着时间片的延长,严重腐蚀等级呈线性上升,中等腐蚀等级基本保持不变,轻微腐蚀等级呈指数下降;在第10个时间片上,油气聚集的事故后果发生概率最大,蒸气云爆炸的事故后果发生概率最小,说明强化紧急停车系统和避免形成受限空间是减轻事故后果的有效途径。

To realize the real-time risk assessment and accident prevention in the storage tank area of the station,based on the establishment of the bow tie model,the time dimension is introduced through mapping rule,the main hazard sources of storage tank leakage and the types of accident consequences under different safety barriers are summarized,and the conditional probabilities of different nodes are determined through the omission probability model,so as to form the dynamic Bayesian network model of the storage tank of the station.The variation law of the probability of tank accident with time is deduced,and is analyzed and verified with the case.The results show that the leakage probability of the tank first increases and then decreases,and reaches the peak value at the third time slice.The basic events with higher posterior probability are related to the intermediate events of tank corrosion.With the extension of time slice,the severe corrosion grade increases linearly,the medium corrosion grade keeps basically unchanged,and the minor corrosion grade decreases exponentially.In the 10th time slice,the accident consequence probability of oil and gas accumulation is the highest,and the accident consequence probability of steam cloud explosion is the least,indicating that strengthening the emergency shutdown system and avoiding the formation of confined space are effective ways to reduce the accident consequences.