Study On Location Of Leaking Points Of Crude Oil Pipeline And Determination Of The Amount Of Leakage

Leakages from pipelines occur frequently because of pipelines corrosions,construction damages and man-made stolen behaviors,which will not only damage to national interests,but also bring a huge threat to the lives and property of the residents.When leak happens,locating the leaking points is the base to reduce the accident level and handle the accident.Calculating pipeline leakage is a crucial premise for estimating economic losses and assessing leakage accidents.Therefore,it is of great significance to study locating technique of the leaking points and calculating method to leakage from crude oil pipelines.Crude oil pipelines are taken as the main research objects in this paper.SPS software is used to build the model of the Dongying-Huangdao crude oil pipeline belonging to Sinopec.Based on this model,the‘Qingdao 11·22'leak and explosion accident of the pipeline occurred in November 22^nd,2013 is simulated.The simulation results through use of SPS,which are verified by the real accident data reported,are analyzed in this paper.To solve the problem of location of leaking points and leaked amount of crude oil pipeline due to the accident,the pipeline leakage predicting model on the section of the Dongying-Huangdao crude oil pipeline has been established by use of SPS and different leakages have been simulated,from which large amount of leakage data of the crude oil pipelines has generated,and been collected,and to be used to train leak predicting models through some machine learning algorithms,including Ridge regression method,Lasso regression method,Decision Tree method and Random Forest method.Furthermore,learning abilities of these different algorithm are also compared.Lasso regression algorithm and Ridge regression algorithm are selected to train the leak locating model,while Random Forest algorithm are selected to train the leak predicting model to predict diameters'of leaking holes.Adaboost algorithm is used to optimize the models.In this dissertation,leakage model under the non-steady through crude oil pipelines is established based on computational fluid dynamics to determine leakage amount.Considering changing of the crude oil's friction coefficient and specific heat capacity with the temperature variation,the non-steady-state pipe flow regularity in pipelines after leakages,the outflow regularity through the small hole are combined to accomplish the coupling calculation of“pipelines flow-flow through leaking holes”.The results of SPS model are used to verify reliability of this kind of coupling calculation.The mean error resulting from the proposed leak locating model is less than 35m,and the mean error of diameters of leaking holes from the predicting model of leaking holes is less than 0.8mm.The error from the calculation model for leakage amount is not more than 6.5%.In summary,the proposed models in this paper provide high accuracy.