Corrosion Risk Prediction Of Gathering And Transportation Pipelines Based On Gaussian Process Regression

Gathering and transportation pipeline is the pipeline connecting wellhead and transportation station,which plays an important role in oil and gas transportation.Compared with other pipelines,the oil and gas transported by the gathering pipeline is not purified and contains more corrosive media,which is more prone to internal corrosion.With the increase of the pipeline operation time,the corrosion risk in the pipeline is increasing,once the accident will cause huge loss of life and property and environmental pollution,it is necessary for enterprises to timely control the corrosion risk in the gathering and transportation pipeline.Pipe corrosion risk prediction is one of the technology to control the pipeline internal corrosion risk,the risk prediction model is based on the mechanism or experience driving more,inside is used to predict pipeline corrosion risk factor is less and most of the existing models did not internal corrosion occurrence probability into consideration,this led to the model results in risk prediction accuracy is low,Risk description is incomplete.Therefore,it is of great significance for the risk control of pipeline to study the corrosion risk prediction method of gathering and transportation pipeline,to predict the current situation and future trend of pipeline corrosion,and then to find out the pipe section that needs to take measures.Taking a gathering and transportation pipeline in Bohai Sea as an example,the internal corrosion risk prediction of gathering and transportation pipeline is studied by Gaussian process regression.The main work of this paper includes:(1)The commonly used methods and theories of internal corrosion prediction of gathering and transportation pipelines are introduced.The characteristics of internal corrosion prediction models and the advantages and disadvantages of each machine learning method are analyzed and discussed.The Gaussian process regression method is selected as the model training method in this paper.(2)the output characteristics based on gaussian process regression method and transcend the concept of probability,build the pipeline internal corrosion prediction model,the model contains the cause analysis,data acquisition,data fitting,the risk prediction four link,and the above purpose,methods,processes,and the four parts of the link between the relevance are expounded.(3)The model of a gathering and transportation pipeline in Bohai Sea was verified by an example.Through the fault tree analysis of the target pipeline,15 basic events leading to the internal corrosion of the gathering and transportation pipeline were obtained,and various operating data along the pipeline were obtained through the flow pattern simulation analysis,and then the Gaussian process regression model of internal corrosion was established.Finally,the internal corrosion risk of the target pipeline was predicted and evaluated.The results show that in the aspect of Gaussian process regression,SE + RQ as the kernel function has the best prediction effect,and the prediction results have smaller error and stronger correlation compared with De Waard95 model.In terms of risk prediction,build the model of risk of internal corrosion of the pipeline to target characterization,severity and probability prediction results showed that the pipeline corrosion risk over the next five years in the overall low risk,risk increased 5 years later,in the next 10 ～ 15 years to repair work of high-risk pipeline section is carried out.