The use of computational reservoir simulation plays a key role in understanding fluid flow in geological formations, enabling the development of advanced models and accurate predictions. These simulations rely heavily on parametric constitutive relations to represent essential properties such as relative permeability (Krel), which is required for modeling continuous-scale multiphase flow in porous media. However, the majority of models described in the existing literature contain numerous empirical parameters that must be determined. Regardless of how complex their parameterization is, these parameters have no direct relationship to the problem's underlying physics. As a result, there is a need to estimate these parameters and regularly assess the associated uncertainties. Nevertheless, achieving satisfactory uncertainty quantification requires a preliminary investigation into sensitivity and linear dependence, a step that is often overlooked, especially in the context of core flood experiments. Given this issue, this study aims to analyze the reduced sensitivity coefficient of relative permeability parameters, parameterized using the LET model, in unsteady-state core flooding experiments, considering both bump and no bump conditions, while also taking into account the capillary pressure parameterized via LET. In these experiments, a plug saturated with oil undergoes axial water injection at one end, resulting in oil (and water after breakthrough) being produced at the opposite end. Experimental data includes the pressure difference between the water inlet and the oil (and water) outlet, along with the accumulated volume of produced oil. Based on experimental data, computational simulations were conducted using the Cydar® software to optimize the relative permeability and capillary pressure. The dynamic behavior of the relative permeability parameters' local sensitivity is showcased, alongside their dimensional comparison.
Read full abstract