Study On Intelligent Prediction Model Of Suction Profile In Hei46 Area Of Jilin Oilfield

With the deepening of the development of old oilfields in China,CO₂ flooding has become an important means to improve the recovery of reservoirs in high water cut stage.However,affected by the static geological conditions and dynamic development mode of the reservoir,the suction capacity of different reservoir types is quite different.In order to improve the development effect of CO₂ flooding,it is urgent to study the characteristics of suction profiles of different sand bodies in underground reservoir.There are mainly two kinds of methods for the study of reservoir suction capacity:suction profile test and reservoir numerical simulation.The test of suction profile takes a long time and has high economic cost;The reservoir numerical simulation method mainly depends on the fine geological model and development performance data,and is greatly affected by the professional level of researchers.In order to improve the research efficiency and prediction accuracy of suction profile and solve the problems of many influencing factors of reservoir suction capacity and complex correlation between multiple factors in the process of CO₂ displacement,under the guidance of big data and artificial intelligence,oil and gas geology,reservoir engineering and other theories,the main controlling factors analysis of suction capacity and intelligent prediction of suction profile are carried out for the Qingyi Member system in Hei46 block of Daqingzijing oilfield in Songliao Basin.Based on the drilling coring and analysis test data,logging data and suction profile test data,the characteristics are optimized by using Pearson coefficient method,univariate feature selection method,random forest algorithm and recursive feature elimination method,and the main controlling factors of CO₂ drive reservoir suction profile such as reservoir thickness,permeability,permeability level difference,reservoir rhythm and water drive CO₂ drive alternate injection mode are revealed.Aiming at the problem of insufficient test data samples of reservoir suction profile,several sets of typical geological models are constructed,and the sample data set of suction profile is expanded by numerical simulation method.Guided by the theories of machine learning and reservoir engineering,an intelligent prediction model of suction profile based on LSTM neural network is constructed through data driving.According to the test of actual reservoir suction profile,the prediction accuracy of this method for suction profile is more than90%,which provides theoretical basis and data support for improving the development effect of CO₂ drive reservoir.