Waterflood simulations in two-dimensional cross-sections are used to investigate the effects of fine-scale relative permeabilities and permeability heterogeneity on displacement performance. Heterogeneity is found to be most significant when deviations from the mean are highly correlated in the direction of flow, i.e., a layered reservoir. As spatial correlation decreases, the production behavior approaches that of a homogeneous medium. The critical features of the fine-scale relative permeabilities are the end-point relative permeability to water at residual oil saturation and the approach of the oil relative permeability curve to residual oil saturation. Relative permeability variations are found to be less important, but not necessarily negligible, when permeability heterogeneity is large and highly correlated in the displacement direction.
Under most circumstances, the minimum relative permeability data required to avoid gross errors in simulation predictions are the end-point effective permeabilities to each phase and the fractional flow curve approaching residual oil saturation. Waterflood simulations using exact relative permeability data are compared with simulations where complete relative permeability information is not available. Cases are considered where relative permeabilities are estimated by straight lines between known end points and relative permeabilities derived from end point measurements and fractional flow data for high water cuts. An approximate method is proposed to derive relative permeability information from limited production history data. The importance of production behavior at low fractional flow of oil also suggests the use of centrifuge measurements to help define the oil relative permeability curve approaching residual oil saturation.