Improved Fluid Property Predictors for Reservoir Compositional Simulation

Lawal, Akanni Segun L
Journal Title
Journal ISSN
Volume Title

Binary interaction parameter kij was correlated as functions of molecular weight and boiling point for hydrocarbon, non-hydrocarbon, and mixed binaries. The kij-functions yield the value of zero for kij when i=j and are extended to binaries involving heptanes-plus pseudo-component. The functions were tested against 1399 data points in 246 systems and an Average Ab so 1 ute Percent Deviation (AAPD) of 1.1% resulted. These functions provide new techniques for establishing the kij-coefficient in the equation-of-state mixing rules without resorting to arbitrary values, as is commonly done. Critical pressures, critical temperatures, and acentric factor correlations of heavy fractions developed from petroleum fractions and pure components are based on a new characterization scheme which uses molecular weight, boiling point, specific gravity, and refractive vii index. The use of this non-iterative scheme for correlating paraffinic, naphthenic, and aromatic categories of hydrocarbons has never before been achieved. An AAPD of 2% resulted from the prediction of 1254 values of critical properties and acentric factor. A van der Waa 1 s mode 1 based so 1 e ly upon pure component critical data is developed for viscosity. An analysis of the prediction results from 5728 pub 1 i shed experi menta 1 data points of dilute to dense gas viscosities of fourteen components resulted in an AAPD of 4.7%, Mixing rules are proposed and comparisons of 9414 experimental data points of several binary, a ternary, and multicomponent systems resulted in an AAPD of 2.5%. The extension of these rules to reservoir oils viscosity prediction was generally within± 6% of the experimental values. But, with the improved mixing parameters for heptanes-plus, the prediction was within ± 0.5% of the experimental values. This prediction of phase viscosity from a single equation has heretofore been unattainable. A generalized cubic equation of state (EOS) is developed as functions of coefficients tt and ~ in the attractive pressure term and the limiting critical· volume b/Vc. From this generalized model, all the patterns of the previously published cubic equations of state are unfolded. The versatility of this model was demonstrated by constructing two- and four- parameters EOS' denoted by LLSF and LLSV respectively. The LLSF EOS was designed for a fixed critical compressibility factor Z = 0.273 and b/V = 0.267 while a and ~ of the LLSV c c viii EDS were selected to match experimental Zc and b/Vc of individual component and a technique of extending the procedure to heptanes-plus pseudo-component was provided. This has never before been considered by previous investigators. These Lawal-Lake-Silberberg equations (LLS) predict better liquid densities of pure components than any cubic equations of state presently in use. Extension of LLS equat{ons to mixtures was demonstrated by using the established form of mixing rules for the attractive and repulsive terms while new mixing rules were proposed for the a m and ~ m of the LLSV EDS. These mixing rules were tested with binary VLE data and was found to be acceptable in general. Liquid densities of reservoir fluids were predicted with the LLS and three published equations. The predictions with the LLS were generally better than those published equations which more often than not predict lower liquid densities.