Application of nonlinear models in predicting storage-permeation body
Storage-permeation body are independent reservoirs formed by the superposition of pores and fractures with certain permeability in the formation.The distribution of storage-permeation body is mostly directly related to the development of pores,fractures and caves.The biggest difference between storage-permeation body prediction and conventional reser-voir prediction is that storage-permeation body prediction is often distributed in areas with developed fractures,so it is necessary to predict fractures.The pore structure of the dense sandstone in the second member of Xujiahe formation in Penglai area is complex,and the correlation between porosity and permeability is poor.Conventional logging data and methods are basically unable to identify fractures,which affects the classification and prediction of storage-permeation body.In this paper,the random forest classification algorithm with few parameters,fast modeling and suitable for multivariate parameter characteristics of well log-ging data in the study area is selected to establish the classification model.The random for-est classification algorithm,which can independently assess the importance of variables and can balance the error,realizes the prediction of vertical distribution of single well storage-permeation body.Based on the analysis of typical wells and the correlation between the gas content,testing,and production testing of existing wells and the distribution range of porosity and permeability,the evaluation criteria for reservoir classification were summarized.By us-ing some typical wells for modeling,a reservoir prediction accuracy of 96％was achieved.The reservoir type prediction interpretation diagram made by this model can intuitively see the location and effective reservoir thickness of different types of storage-permeation body in the well section,solving the problem of classification and prediction of tight sandstone stor-age-permeation body in the second member of Xujiahe formation in Penglai area.This pro-vides a new solution for the classification and prediction of tight sandstone reservoirs with extremely strong heterogeneity,and has certain reference value.
the second member of Xujiahe formationlogging datarandom foreststorage-permeation body prediction
万方数据
维普
