Design And Development Of Intelligent Analysis System For Development Dynamic Index Of Polymer Injection Flooding Block In Oil Reservoir

With regard to the production dynamic analysis of oil reservoir polymer flooding block,the change law of development dynamic index is very important reference information and its change law is very important for the evaluation of polymer flooding effect,the further planning of later production and the calibration of recoverable reserves.Therefore,this project studies the prediction method and system development of development dynamic indicators based on actual production data.Taking the actual production data of accumulation and flooding block in Daqing X reservoir as the research object,the production factors with high influence importance are extracted from it,and the development dynamic index are predicted based on the hybrid prediction model（CBi GRU-Attention）.The main research contents are as follows:1.An iterative interpolation method based on machine learning is proposed to solve the problem of missing values and outliers in the data records in the production process of reservoir accumulation and flooding.Firstly,the distribution rules of missing values and outliers in the data were summarized.Then,according to its distribution regularity and numerical characteristics,an iterative interpolation method based on random forest（miss Forest）is used to deal with the missing data and abnormal data in the data.The results show that compared with the commonly used imputation methods such as mean imputation and zero imputation,the imputation method based on miss Forest improves the standardization of the sample database and the accuracy of prediction.2.Aiming at the problem that a single production factor cannot accurately express the actual production law,and the influence of multiple production factors should be comprehensively considered,a Random Forest Regression Algorithm（RFRA）based on embedded feature selection is proposed.Firstly,the actual production data of A₁,A₂ and A₃blocks of X reservoir were collected and collated.On the basis of sufficient literature and data research,through communication with field experts,a variety of production indicators such as water injection,production time and the number of Wells opened were initially selected as sample databases for feature selection after considering a variety of influencing factors.Then,the production index is analyzed using RFRA model.Finally,by comparing the weight ratio of each production index,10 factors,such as production time and number of Wells opened,are selected as the input characteristics for the prediction of polymer flooding development dynamic index of X reservoir.3.In view of the particularity of the oil displacement principle of the poly flooding block and the problems of large amount of calculation,complex numerical formula and limitations of traditional prediction methods,a hybrid prediction model of convolutional neural network based on attention mechanism and bidirectional recurrent gated neural network is proposed.Firstly,the structure of the neural network was preliminarily set and its change law was learned through modeling.The parameter Settings of the model were determined by stepwise trial and error method after multiple trials and modifications,and a hybrid prediction model of convolutional neural network and gated recurrent unit based on attention mechanism was established.Then,the normative sample database is used to predict the important development dynamic indexes（oil production,water production）in the dynamic analysis of accumulation and flooding production,and the average value of the determination coefficient（R²）is more than 90%.Finally,a comparison model based on random forest,BP neural network,and Convolutional Neural Network（GRU）was constructed.Through multiple comparative experiments,the high accuracy and superiority of the hybrid prediction model were proved.