PVT modeling of complex heavy oil mixtures

Abstract

This paper describes a systematic approach to model the phase and volumetric behavior of extra heavy oil and light solvent mixtures for a wide range of temperatures. A cubic equation of state (EOS) is first developed using crude assay data. We use a modified Jacoby correlation to describe the relationship of specific gravity and molecular weight for the heavy oil sample. A gamma molar distribution model is used to fit the crude assay data, then single-carbon-number (SCN) fractions are defined up to C90+. The Twu correlation is used for estimating SCN critical properties, including C90+, resulting in an EOS with 89 components. Pure solvent-crude oil mixture PVT experimental data are analyzed and used to tune the EOS model. The resulting model reproduces accurately all phase and volumetric behavior of pure-solvent-crude mixtures. The prediction of the final EOS model provides accurate prediction against measured PVT data for mixtures of extra-heavy oil sample and synthetic combustion gas solvents made up of CO2, N2, C1, and C2. For viscosity modeling, we propose a novel technique to model large variation of viscosity for a wide range of pressures, temperatures and solvent compositions. In our proposed approach, the LBC (Lorenz-Bray-Clark) correlation is used, with SCN critical volumes modified individually to ensure the LBC correlation estimates SCN viscosities as given by a modified Twu correlation. The viscosity predictions of the pure-solvent-saturated heavy oil at varying temperatures, the EOS/LBC model results in poor agreement against measured saturated viscosity data. Our solution was to split the heaviest fraction C90+ into two sub-fractions, where only critical volumes differ, resulting in “lower-viscous” and “higher-viscous” C90+ fractions (C90+L into C90+H). We find that the fraction of C90+L (fL) correlates well against pure solvent solubility and temperature, resulting in an accurate overall viscosity fit. The final EOS/LBC model is successfully employed to predict measured PVT and viscosity data that was not used during the training process. Our new approach can provide accurate phase properties and volumetric behavior calculations and as well as good predictions of large variation of liquid viscosity. Such developed EOS technique is vital for providing an accurate estimation of phase behavior and volumetric properties, and the viscosity variations required for modeling thermal-EOR techniques.