A Quantitative and Qualitative Comparison of Coarse-Grid-Generation Techniques for Modeling Fluid Displacement in Heterogeneous Porous Media

Abstract

Summary To apply upscaling techniques is an undeniable demand in reservoir simulation, when one considers the difference between the level of detail in a geological model and the level of details that can be handled by a reservoir simulator. Upscaling the reservoir model involves first constructing a coarse grid by using gridding algorithms and then computing average properties for coarse gridblocks. Although various techniques have been proposed for each of these steps, one needs to be aware of strengths and weaknesses of each technique before attempting to apply them. In this paper, we focus on different gridding methods and evaluate their performances. Three main grid-generation techniques are considered: permeability-based (PB), flow-based (FB), and vorticity-based (VB) methods. We apply all three methods to a number of 2D heterogeneous models and simulate two-phase flow on the constructed grids. Then we compare their obtained global and local results. Fluid cuts at the producer is employed as the global performance indicator and saturation-distribution error as the local indicator. We show that FB and VB gridding, which are dynamic methods, are superior to PB gridding, which is a static method. On the basis of this analysis, we then concentrate on FB and VB gridding and investigate their performance in greater detail. While FB gridding uses fluid velocity as gridblock density indicator, VB gridding combines velocity and permeability variation in gridding according to its definition and takes advantage of both. Therefore, although performance of FB and VB gridding is comparable in many cases, VB has the benefit of producing coarse gridblocks with more-uniform permeability and fluid-properties distribution. This in turn yields more-accurate global and local results and reduces application of sophisticated upscaling techniques and fulltensor permeability upscaling.