Risk assessment for gas transmission station based on cloud model based multilevel Bayesian network from the perspective of multi-flow intersecting theory

Abstract

Accidents occurred in the natural gas transmission stations may result in casualties, equipment damage, property loss and even political and ecological impact. It is necessary to propose an approach for the gas transmission stations risk assessment so as to identify risk factors and prevent accidents. However, insufficient attention has been paid to the complex interactions and coupling patterns of these factors with current risk assessment methods. Hence, a novel risk analysis strategy is needed for industrial site risk assessment. Given industrial site risk factors, the multi-flow intersecting theory (MIT) is first defined to consider material, information and behavior risk analysis. Then the multilevel Bayesian network is constructed to represent these factors from different flows. To handle issues of insufficient statistical data and subjectivity associated with expert judgement, cloud model and fuzzy Bayesian are incorporated. For the defuzzification, existing cloud model is improved by introducing the similarity degree of standard cloud and evaluation cloud. Then the prior and conditional probabilities obtained from cloud model are input into Bayesian network for further reasoning and accident probability prediction. To validate the utility of the proposed method, the gas transmission startup process was chosen for risk assessment.