Three-Dimensional Coarse-Scale Simulation of Enhanced Oil Recovery Processes in Highly Heterogeneous Reservoirs

Upscaling of the geological model is an important stage in a reservoir simulation work-flow that introduces error due to averaging of sub-grid heterogeneity and numerical diffusion. Upscaling errors are generally more severe in compositional flow simulation than in black oil modeling due to the inherent equilibrium assumption that increases numerical diffusion significantly. When coarse grid blocks involve high and low permeability channels or layers, current single-porosity upscaling techniques are very sensitive to the level of upscaling and generally fail in representing the fine scale response adequately. In this work, the effect of heterogeneity on displacement efficiency in coarse-scale modeling is studied and a dual-porosity upscaling approach is proposed that significantly reduces the upscaling error due to averaging of heterogeneity. Pore space is divided, based on flow contribution, into two distinct continua and a dual-porosity dual-permeability model is used for coarse-scale flow simulation. A streamline index (SI), defined as the ratio of streamline density to time-of-flight, is used for division of storage into primary pore volume that covers the high flow pathways in the porous system and secondary pore volume that acts mainly as source/sink and feed into major flow pathways. Global single-phase upscaling is applied to calculate inter-continuum and inter-block transmissibility. The proposed technique is used to simulate miscible and immiscible displacements on 3D models. Displacement calculations are performed on the original fine grid and on a uniform coarse grid with single-porosity (classical approach) and dual-porosity upscaling. Several simulation results demonstrate the superiority of the proposed approach in effectively capturing the sub-grid heterogeneity effects on multi-phase flow. The sensitivity of the proposed upscaling technique is then investigated for settings where gravity forces are significant. We show that the proposed technique is accurate at high upscaling ratios and that it is significantly less sensitive to the upscaling level relative to single-porosity upscaling techniques.