Research On Gas Condensate Flow Algorithm Based On Multi-model Data Fusion

With the development of oil and gas resources gradually moving towards the deep sea,the application of underwater production system is more and more.Under the condition of deep sea production,if we continue to use the traditional multiphase flow meter for production measurement,the development cost will be greatly increased,what’s more,the daily calibration and maintenance work will encounter many problems.Therefore,it is of great significance to study the new measurement method of underwater oil and gas fields.In recent years,virtual flow meter has been widely used in deep-water oil and gas fields because of its low cost,high precision and convenient maintenance.The first domestic virtual flow meter system for underwater condensate gas field has been put into use in the South China Sea,but it presents the problem of low precision and low reliability in the process of operation.In order to solve this problem,the author firstly revises the heat transfer algorithm of original wellbore flow model,and designs 3 new single flow model algorithms,including one new nozzle model,one cylinder-nozzle model,and one jumper model.Then,based on the calculation results of the 4 single flow model,according to the idea of data fusion,this paper designs three kinds of multi model data fusion gas condensate flow algorithms,including data fusion algorithm based on the D-S theory,the uncertainty theory and BP neural network theory respectively.Finally,use the error criterion to evaluate the results of single flow model and multi-model fusion algorithm.Practical engineering application shows that the mean absolute percentage error of single flow models maintain within 8%,while the mean absolute percentage error of multi model fusion algorithms maintain less than 5%,even when one or two single flow models present a large error,fusion algorithm can still give the optimal results,and ensure the accuracy and reliability of the system of virtual meter.