TX 75083-3836, U.S.A., fax 01-972-952-9435.
AbstractPredicting primary depletion recovery below the bubble point is a very difficult challenge, especially for carbonate reservoirs. Accurate predictions require sufficient laboratory data to cover the variability in response from the different rock types and laboratory procedures, and proper upscaling of laboratory data to simulation grid scales.Many recent studies have shown that laboratory-derived gas-oil relative permeability curves often depend on whether the tests were conducted using depletion drive or steady state methods. Though it may seem preferable to use laboratory data from depletion drive experiments for modeling solution gas drive recovery, these tests are complex and there is concern using data from tests conducted at pressure depletion rates much higher than that in the field. More significantly, such tests are expensive and it is not practical to conduct many depletion tests.We used mercury injection, porosity, permeability, CT scans, and thin section data to classify the giant carbonate reservoir into different rock types. Steady state gas-oil relative permeability tests on plug samples were then conducted to span these rock types, and P10-50-90 curves developed. We also carried out steady state and reservoir condition multi-rate depletion on select whole cores. This data was used to apply a correction to the steady state plug data, and this typically lowered the gas relative permeability curves.We next conducted numerous upscaling studies using sector models extracted from the full field simulation model. These studies showed that areal upscaling cause the core based critical gas saturation to be increased and the oil relative permeability to be decreased. Vertical upscaling, including the presence of high permeability streaks to represent fractured zones, has not yet shown much impact on the laboratory derived curves.Finally we developed a template to systematically modify the steady state relative permeability plug data to account for process and scale effects. The new P10-50-90 curves were then used in simulation predictions and results compared to that from the plug based curves.