Accurate estimation of molecular diffusion coefficient is necessary for the design and modeling of solvent-assisted recovery techniques for exploitation of heavy oil and bitumen resources. Several studies on molecular diffusion measurement have been conducted using various pure gaseous solvents but there are very few diffusivity measurements available for the propane-bitumen system. The scarcity of these data is due to the complexities arising in the modeling of dissolution of propane in bitumen. These include high solubility of propane in bitumen and subsequently, dramatic volume change of the diluted oil, drastic reduction of bitumen viscosity during dilution and the possibility of asphaltene precipitation.
In this work, a rigorous numerical model is developed to model the diffusivity of propane in bitumen based on real data. This model accounts for bitumen solution density change (swelling) as a result of dilution. A front-tracking moving boundary algorithm and numerical procedure is proposed which accounts for volume change of each grid at each time step. Those grids which experience diffusive flux, change sizes based on non-ideal mixing data available from measurement of solution density versus concentration. Finally, the Levenberg-Marquartd method was applied to estimate the propane diffusion coefficient based on laboratory measured gas-oil interface movement. A constant-pressure diffusion measurement experimental technique was used to measure mass of gas dissolved and solution height change in a propane-bitumen diffusion experiment at 413.7kPa (60 psia) and 827 kPa (120 psia) both in 24°C.
The experimental results show that the "no-swelling" simplifying assumption often used in diffusion measurement modeling studies, cannot be generalized to all gaseous solvents because based on our experiments, the bitumen solution volume can be increased by 50% due to propane dissolution. The estimated diffusion coefficient is compared with the results of other mathematical techniques available in the literature. It is shown that the no-swelling simplifying assumption used in unidirectional modeling of gas diffusion in bitumen leads to estimation of erroneous parameters when propane is the gas and oil density and oil volume change is neglected.