Research On The Method Of Quantitative Identification Of Interbed In Lamadian Oilfield

After decades of development,the old oilfields in eastern China had basically entered the stage of high water cut and high recovery.However practice showed that interbed affected the vertical permeability and water drive efficiency because the interbed controled the fluid flow in the reservoir and there was considerable relative oil enrichment zone.Effective identification of the interbed and the identification of the distribution of the interbed was of great significance for the development of later reservoir development and the prediction of remaining oil.Taking the Saertu oil layer and the Putaohua oil layer in the middle west-2 area in south block of Lamadian Oilfield as the research object,the stratigraphic correlation and division were carried out,and every layer of interbed were divided to determine the cause and distribution of interbed.On the basis of the cause of the interbed,logging response and the coring data were used to identify types of interbed and select the identification parameters.Due to the shortage of logging data in the research area,this paper took full account of the combination relationship between various logging parameters,adopted methods of parameter polynomial processing and data dimensionality reduction,selected 9 feature parameters as new evaluation parameters,and determined the type of interbed.Thus,the known samples of the interbed quantitative evaluation were established.Support Vector Machine(SVM)was used to establish the interbed quantitative identification model,which provided a fast and accurate method for quantitative identification of reservoirs in the middle and later stages of reservoir development.Based on data preprocessing and model parameter optimization,a interbed identification model based on SVM was established by using 214 sets of learning samples.Then 68 sets of test samples were tested.The prediction accuracy was up to 80%,which was much higher than the prediction performance of the discrimination plate.In order to optimize the model,the feature set was optimized by improved F-score method and the redundant information in the parameter combination was removed.The optimal feature subset was established.Considering the influence of unbalanced data on the model,the learning sample of the optimal feature subset was studied.Data unbalanced processing was performed,and the optimized model prediction accuracy was up to 86.76%.The results showed that the established interbed quantitative identification model systematically reflected the complex mapping rules between various geological factors and interbed types.It was feasible to apply SVM to carry out interbed identification,and it was extended to other geological fields.