Machine-learning algorithms for the computation of upscaled permeabilities
Master thesis
Permanent lenke
http://hdl.handle.net/11250/2565104Utgivelsesdato
2018Metadata
Vis full innførselSamlinger
Sammendrag
In full-scale reservoir simulation models, the characteristics of the rock arenot fully resolved. Upscaled permeabilities are used to account for theeffective behaviour of a composite region. By averaging the fine properties ofthe flow, they provide a simple linear correlation between the flux and thepressure drop. The goal of this master project is to assess the performance ofmachine-learning algorithms for the computation of upscaled permeabilities. Themethodology is to use high resolution simulations on a randomly generated set offine scale models. This data set will be used as training set for a machinelearning algorithm. Ordinary least squares and Kernel Ridge regression algorithms will be the two different machine learning algorithms that will be tested andcompared. We will also investigate the robustness of the algorithms with respect to the choice ofthe statistical distributions used for the generation of the fine scale models. Our results show that both the ordinary least squares and Kernel Ridge is capable to capture the upscaled behaviour of the flow andencode it into a single coefficient, namely the upscaled permeability. Hence they will capture the underlyingphysics of the problem. However, Ordinary least squares does so with a high error that is likely to limit the usefulness of that particular algorithm.