Development of Generalized Correlations for Thermophysical Properties of Light Hydrocarbon Solvents (C1–C5)/Bitumen Systems Using Genetic Programming

Abstract

Accurate modeling of thermophysical properties of solvent/bitumen mixtures is critical for proper design and implementation of thermal- and solvent-based bitumen recovery processes. In this study, three generalized correlations were developed for prediction of solubility, density, and viscosity of light hydrocarbon/bitumen mixtures. The generalized correlations were developed using symbolic regression based on genetic programming and employing a 10-year set of comprehensive phase behavior experimental studies conducted under the SHARP research program on solvent-aided thermal recovery of bitumen. The data set comprised Surmont, JACOS, Mackay River, and Cold Lake bitumen samples and five light hydrocarbon solvents including methane, ethane, propane, n-butane, and n-pentane. The developed correlations are valid for gaseous solvents. Finally, the developed correlations for solubility, density, and viscosity were validated against a large data set of experimental measurements collected from the literature. The validation demonstrates that the developed correlations are able to accurately predict the available experimental data of solubility, density, and viscosity reported in the literature.