Abstract
The mechanical properties of hydrocarbon reservoirs significantly depend on the elastic properties of the fluids occupying the pore space in the rock frame. Accurate data and models for the mechanical properties of fluid mixtures in a petroleum reservoir containing supercritical CO2 should be available at the same reservoir conditions for reliable design of well-completion, maximizing reservoir productivity, and minimizing risk in drilling operations. This work investigates the change in the bulk modulus of the higher hydrocarbon fluid (decane C10H22) after the injection with supercritical CO2 at reservoir conditions. The isothermal bulk modulus βT of liquids under pressure, simply defined as the first-order derivative of pressure with respect to volume, is determined in this study from the derivative of pressure with respect to density. The density data were obtained from experimental measurements of mixtures of supercritical CO2 + C10H22 for a range of CO2 mole fractions from 0 to 0.73, at temperatures from 40 to 137 °C and pressures up to 12000 psi. The isothermal derivative coefficients of the pressure as a function of density are reported for each CO2 concentration measured in this work. Common fluid-substitution models, including the Gassmann model, which is only valid for the isothermal regime, have limited predictive power because most fluids are treated as simple fluids, with their mechanical properties only characterized by their densities. However, under different environments, such as when supercritical CO2 is injected into the geological formation, the fluid phase and its mechanical properties can vary dramatically. At high pressure, the density of CO2 can equal to that of the hydrocarbon phase ρ(CO2)/ρ(C10H22) ≈ 1, while the bulk modulus of CO2 remains as low as only βT(CO2)/βT(C10H22) ≈ 7 %. Excessive decrease in the bulk modulus can easily cause subsidence, although the pore pressure and the fluid mixture density remain unchanged, even at pressures up to 4000 psi.
Probabilistic Flight Delay Predictions Using Machine Learning and Applications to the Flight-to-Gate Assignment Problem
