基于动态贝叶斯网络的非常规突发灾害事故情景推演

针对非常规突发灾害事故演变路径不清晰,演变过程及发展态势复杂,应急决策主体在应急救援过程中很难做出正确决策的现状,在分析非常规突发灾害事故情景演变规律的基础上,基于情景状态(S)、处置目标(T)、处置措施(M)和自身演变(E)四个要素分析了非常规突发灾害事故情景演变的路径,基于动态贝叶斯构建了非常规突发灾害事故动态情景网络,并利用联合概率公式进行相应节点变量的状态概率计算,实现了非常规突发灾害事故的关键情景推演.以大连"7·16"油库爆炸火灾事故为例,演示了非常规突发灾害事故的情景推演流程及关键技术,并对情景推演的结果进行了分析.推演结果表明:事故按输油管线爆炸→油罐爆炸起火→原油泄漏、污染海域的路径演变;其中,输油管线爆炸出现的概率为90.2%,T103罐爆炸起火出现的概率为84.1%,原油泄漏、污染海域出现的概率为80.3%,推演结果与实际灾害事故的情景发展状态基本一致,证明了该方法的合理性和可行性.

Scenario Deduction Model of Unconventional Emergency Based on Dynamic Bayesian Network

The unclear evolution path and complex development of unconventional emergency could make it difficult for decision-makers to make right decisions. A model based on the dynamic Bayesian network was proposed to solve the key scenario deduction problems of unconventional emergency. In this model， the scenario evolution law of unconventional emergency was first analyzed to formulate the four factors including scenario situation(S)， disposal target (T)， disposal measure (M) and evolution (E). Then the scenario evolution path was performed based on the four factors. Finally， the state probabilities of corresponding node variables were calculated by using the joint probability formula. For the purpose of illustration and verification， the case of Dalian “7·16” oil depot fire and explosion accident was presented. The results showed that the evolution path follows oil pipeline explosion， oil tank explosion and fire， and oil spill and offshore pollution， whose probabilities are respectively 90.2%， 84.1% and 80.3%. Thus， it could be concluded that the proposed dynamic Bayesian network is both reasonable and feasible.