The results of modeling injection into and long-term storage of CO2 in a deep saline aquifer with a commercial black-oil reservoir simulator are presented. Realistic CO2/water pVT properties covering all pressure, temperature and compositional conditions accounted for during the simulations have been used. The pressure and temperature in the aquifer is above the CO2 critical pressure and temperature giving rise to the existence of a two-phase fluid system of CO2 as a fluid and an aqueous phase containing dissolved CO . The impact of CO2 injection rate, reservoir layering, capillary pressure and residual CO2 saturation for water re-entering CO2 filled volumes (hysteresis in fluid saturations) on CO2 distribution and storage in the deposit have been studied. The re-distribution of water occurs after stop of CO2 injection due to gravity segregation of denser CO2 saturated water and CO2-free water.
For upscaling studies a small section of the reservoir model (typically a reservoir grid block) was finely gridded and studied separately for various reservoir parameters and injection rates. The fine model showed that the sweep efficiency of CO2 displacing saline water and both position and quantity of CO2 in each grid block of the reservoir model are sensitive to the CO2 injection rate and the effective vertical permeability. By increasing the viscous to gravity force ratio, the sweep efficiency increases resulting in more CO2 trapped as residual gas. More CO2 is also trapped in the case of increasing CO2-water capillary pressure. Estimating the amount of trapped CO2 from the average CO2 saturation is possible when the sweep efficiency is good (evenly distributed CO2) but may give bad estimates when CO2 only enters part of each large grid block. This shows that to get reliable estimates of the amount of trapped CO2 the distribution of CO2 in each grid block is important and the traditional upscaling methods have to be extended to incorporate the effect of CO2 distribution on trapped gas. Especially the effect of fine-scale layering must be upscaled properly.