Predicting the equilibrium phase behavior of carbon dioxide/hydrocarbon mixtures using a continuous thermodynamics approach

摘要

Large-scale compositional reservoir simulations of enhanced oil recovery methods such as the carbon dioxide miscible process must compromise between the number of numerical gridblocks and hydrocarbon components that are utilized in the thermodynamic computations. One solution to this problem is to characterize the heptanes plus (C$sb7$+) fraction as a continuous statistical (e.g. Gaussian, Gamma) distribution in one or more physiochemical variables. The distribution is therefore defined by its mean and variance only, and can be used in Continuous Thermodynamics (CT) models where normal discrete summation calculations are replaced by integral operations.;The CT approach was investigated with respect to phase equilibria modeling by cubic equations of state (EOS) (Peng-Robinson, Soave-Redlich-Kwong). Existing EOS a,b parameter CT expressions correlated to molecular weight (MW) were expanded to bivariate (MW, boiling point (T$sb{rm B}$)) functions for generalized use with paraffin/napthene/aromatic (PNA) mixtures.;An experimental P-V-T apparatus was designed and constructed to generate equilibrium data on "synthetic" oils (C$sb4$-C$sb{13}$) and CO$sb2$ at temperatures of 120$spcirc$F and 170$spcirc$F. Experimental pressures ranged from about 800 to 1800 psia. Three different hydrocarbon systems were tested, ranging from an n-paraffins only mixture to a 23 component PNA fluid.;Both discrete component/critical property and semicontinuous/MW-T$sb{rm B}$ parameter correlation equilibrium flash models were written to predict the experimental data results using the Peng-Robinson EOS. The bivariate CT parameter correlations were reduced to simpler univariate expressions by using a mixture-specific second order polynomial relating T$sb{rm B}$ as a function of MW in the parameter equations. A gamma distribution was used to describe the continuous fractions.;The two models were found to do a satisfactory job of reproducing the experimental equilibrium K values and/or distributions of the System 1 data. They also accurately predict the liquid phase compositions of the System 2 and 3 PNA fluids, but overestimate the vapor phase hydrocarbon concentrations by as much as an order of magnitude. Comparison of the discrete and continuous model results show good cross-agreement of both liquid and vapor phase compositions and properties.