Effect of connectivity misrepresentation on accuracy of upscaled models in oil recovery by CO 2 injection

An upscaling method such as renormalization converts a detailed geological model to a coarse one. Although flow equations can be solved faster on a coarse model, its results have more errors. Numerical dispersion, heterogeneity loss, and connectivity misrepresentation are the factors responsible for errors. Connectivity has a great effect on the fluid distribution and leakage pathways in EOR processes or CO 2 storage. This paper deals with the description and quantification of connectivity misrepresentation in the upscaling process. For detection of high‐flow regions, the flow equations are solved under simplified single‐phase conditions. These regions are recognized as the cells whose fluxes are greater than a cut‐off. The connectivity is investigated by checking whether the high‐flux region forms a spanning cluster across the model. An indicator called the flux connectivity distortion is defined to measure the connectivity misrepresentation of coarse grids. This indicator is shown to have a strong correlation with the oil saturation error. For renormalization method, this relation helps to propose an upper limit of coarsening beyond which the coarse model results are not reliable. Once the proper sizes of coarse blocks are determined, flow behavior can be simulated accurately in a project of CO 2 ‐EOR or CO 2 sequestration. © 2015 Society of Chemical Industry and John Wiley & Sons, Ltd