Research On Risk Assessment Of Oil Pipeline Based On Bayesian Network

Pipeline risk assessment is one of the core contents of pipeline integrity theory.In view of the risk factors such as corrosion,defects in pipe production,operation error by human and third-party damage in oilfield pipelines under complex environment,the pipeline risk model established by traditional risk assessment method is difficult,time-consuming,and the assessment process relies too much on expert knowledge and experience.These problems will lead to serious deviations in the evaluation results.In order to solve the above problems,this paper proposes a method of risk assessment of oil field pipeline based on Bayesian network,and tries to solve the accuracy problems faced by oil field pipeline risk assessment through the study of the Bayesian network structure and parameter determination method.Firstly,in view of the accuracy problem of traditional Bayesian network structure learning method,a Bayesian network structure determination method based on fault tree analysis is proposed in this paper.Fault tree analysis is used to identify the possible risk factors of the pipeline and establish the pipeline fault tree model.Mapping algorithm of the fault tree and Bayesian network is used to determine the structure of the pipeline risk Bayesian network model.Considering that there are redundant nodes in the model to increase the network complexity and affect the evaluation accuracy,a method of attribute reduction of network nodes based on genetic algorithm is proposed.In the fitness function,the dependence of the decision attribute on the condition attribute is introduced.Due to the algorithm can not only ensure the characteristics of global optimization,but also strengthen the ability of local search.The purpose is to reduce the complexity of the model structure to obtain a concise and accurate network structure.Secondly,for the subjective and static problems caused by expert knowledge and experience,and the efficiency problems faced by traditional parameter learning methods.A parameter learning method based on genetic algorithm is proposed.In the algorithm,the likelihood function of each network topology is constructed and used as fitness function to output the optimal network parameters and complete the construction of conditional probability table.Considering that there is a certain correlation between network parameter nodes,which does not meet the hypothesis of conditional independence required by Bayesian network.a method of parameter weighting based on gray correlation analysis is proposed in this paper.By calculating the correlation among the parameter nodes and determining the index weight.The Bayesian network probability inference formula is improved by weighting.The weight of the constructed conditional probability table is allocated to weaken the correlation between the parameter nodes.The purpose is to meet the conditional independence hypothesis.Finally,the pipeline risk assessment method proposed in this paper is applied to the actual oil field pipeline risk assessment problem.The failure probability of oil field pipeline is calculated by Ge NIe.The failure probability value is obtained.At the same time,the inversereasoning and evidence updating ability of Bayesian network are used to analyze each risk factor and obtain the cause chain that affects the pipeline failure.Experimental results show that the proposed pipeline risk assessment method has significantly improved in accuracy.