A black-oil, pseudo-miscible simulator, a compositional simulator, and a compositional/ incomplete mixing simulator were used to model first-contact miscible!, carbon dioxide, enhanced-oil recovery experiments. The experiments were conducted by the University of Wyoming in a three-dimensional, high-pressure, physical model to investigate the effect of injection well geometry on oil recovery. These laboratory scale experiments showed that oil recovery could be significantly increased using a horizontal injection well instead of a vertical injection well. The objectives of this study were to evaluate the capability of each simulator to accurately model these displacements and to use the most appropriate simulator to investigate the mechanisms that led to improved oil recovery.

Both the black-oil, pseudo-miscible simulator and the compositional simulator were unable to provide a match of the experimental results. The pseudo-miscible simulator overpredicted the recovery rate and the ultimate recovery of oil because it could not accurately account for the large volume change when mixing carbon dioxide and heptane (laboratory oil). Only limited success in providing better matches was attained when the effects of viscous fingering were incorporated by lowering the value of the mixing parameter. The overproduction of oil resulting from the poor phase behavior description was too large to be compensated for by altering the mixing parameter. The compositional simulator provided a more realistic description of the phase behavior and, therefore, yielded good matches of oil production before breakthrough. However, the recovery of oil after breakthrough was overpredicted. It was not possible to improve these simulation results by accounting for viscous-fingering effects since the compositional simulator assumes complete mixing in all grid blocks.

A compositional/incomplete mixing simulator that empirically accounted for the macroscopic effects of viscous fingering by assuming incomplete mixing within grid blocks was then employed. An additional parameter was required to define the rate of mass transfer of oil from the ‘bypassed’ phase of a gridblock into the oil phase where mixing with the carbon dioxide was allowed to occur. A high value of the parameter (100 or more) corresponded to no viscous fingering as commonly assumed in conventional simulators, while a low value corresponded to the maximum degree of fingering effects. Good matches between simulation and experimental results for both the vertical and horizontal injection schemes were obtained using this model with parameter values of 0.1 and 2.5, respectively. These results indicated that viscous fingering probably occurred in all experiments, with the effect being much less pronounced when the horizontal injector was used.

You can access this article if you purchase or spend a download.