Development of pore network modeling for prediction of two-phase flows in rocks

Abstract

Geological Carbon Storage is one of the promising methods for reducing the atmospheric carbon dioxide concentration, which is injecting the carbon dioxide to subsurface. As the interest on GCS increases, it becomes more important to predict the behavior of injected $CO_2$. This research focused on predicting two-phase flow in pore-scale, when capillary force was dominant. For some limitations in lab-scale experiments, pore network simulation was used extracted by maximal ball algorithm. In this study, the flow simulation system was developed based on Hagen-Poiseuille equation and Young-Laplace equation. Using the developed pore network simulation, primary drainage was simulated. The results of pore network simulation were compared to those of the experiment including oil invasion experiment into the Berea sandstone and $scCO_2$ invasion experiment into the Domengine sandstone. The absolute permeability of the Berea sandstone was used in the determination of throat length coefficient ($\alpha$) and throat length coefficient was determined to 1. The oil invasion simulation results confirm the fact that as the pressure difference increases, residual oil saturation increases. The relative permeability with respect to degree of oil saturation shows similar tendency of general results of Berea sandstone with exception in low oil saturation. The oil relative permeability in low oil saturation was underestimated. In conclusion, the comparison results figure out that flow simulation gives reasonable results except when oil saturation is low. The $scCO_2$ invasion to the brine-saturated Domengine sample was also simulated. For the computational convenience, one subvolume of nine subvolumes were selected to flow simulation. Representative elementary volume test was conducted for porosity and absolute permeability to determine if subvolume 3 was sufficient to characterize the whole Domengine sample. Both porosity and absolute permeability converged to 22% and 2000 mD respectively, as elementary volume increased. The distribution of $scCO_2$ from flow simulation was compared to that of X-ray microCT images. As a result, it was confirmed that the developed flow simulation gives the reliable results in hydraulic properties and the pattern of displacement.