首页>
外文会议>Offshore Technology Conference
>Applied Geostatistical Reservoir Characterization in R: Review and Implementation of Permeability Estimation Modeling and Prediction Algorithms - Part Ⅱ
【24h】
Applied Geostatistical Reservoir Characterization in R: Review and Implementation of Permeability Estimation Modeling and Prediction Algorithms - Part Ⅱ
Modeling and prediction the formation permeability is a decisive step in the reservoir characterization as it concerns the sparseness of the row data with different scales from different sources. Efficiently combining the different sources of rock characteristics, especially core and well logs data, should lead to accurate estimation of permeability for other wells that have no core analysis. That results in accurate reservoir characterization, precise geospatial modeling and solid reservoir modeling. The most conventional approach for combining the core measurements, well log data into permeability modeling is the Multiple Linear Regression, which considers the least-square equation to estimate the coefficient of parameters in the linear modeling. However, there are many other algorithms that can be used for more accurate modeling and prediction of formation permeability. The other methods include: Multivariate Multiple Linear Regression, Generalized Additive Modeling, Multivariate Adaptive Regression Splines, Least-Angel Regression, Bayesian Generalized Linear Modeling, and Robust Linear Modeling. In this paper, review of all the aforementioned algorithms were provided along with full implementation on the permeability modeling given the well log and core data in a well from Rumaila dataset, from South Rumaila oil field in Iraq. The comparison between these agorithms were performed based on the root mean square prediction error of the predicted permeability after conducting data sampling and cross-validation. These algorithms can be carried out through several commercial software packages with limitations in terms of availability and computation time consuming. In this paper, we introduce a simplified implementation of all these algorithm through R, the most powerful open-source statistical computing language. Detailed R-codes were prepared for all algorithms.
展开▼