Uncertainty and sensitivity analysis of relative permeability curves for the numerical simulation of CO2 core flooding

Various error sources exist in the specified relative permeability (kr) curves for numerical modeling of CO2/water core flooding process, including model selection, parameter estimation, and heterogeneity setting. In this study, we quantitatively investigate their effects on model predictions by a numerical sensitivity analysis. Numerical simulation results indicate: (1) The errors in residual gas saturation (Sgr) and residual liquid saturation (Slr) could significantly influence the estimation of the total volume (TPV) of injected CO2 when CO2/water displacement reaches a steady state. A change of Sgr or Slr by 0.1 can induce deviation in TPV up to 0.5 pore volume. (2) The change in Sgr could cause a bigger predictive deviation in local CO2 saturation (SCO2) than that in Slr. In the case of the same degree of change, the deviation of the former is almost twice as much as that of the latter. (3) The model predictions remain fairly consistent when utilizing either Corey's curve or Brooks–Corey curve, but become significantly different if changing to van Genuchten curve. In this case, the difference in TPV is greater than that caused by the errors in Sgr or Slr. The maximum deviation in the predicted local SCO2 exceeds 0.3, influencing the analysis of CO2 local trapping behavior. Comparatively, the local SCO2 buildup is only slight if van Genuchten curve is used, but obvious using Corey's curve. (4) kr heterogeneity only presents a small effect on the local SCO2 and could be reasonably ignored compared to the effects of the heterogeneity in capillary pressure and permeability. © 2020 Society of Chemical Industry and John Wiley & Sons, Ltd.