High-frequency cycles and rock-fabric units are the two important scales for modeling shallow-water carbonate reservoirs. A reservoir model of the Seminole San Andres Unit (SSAU) was constructed that incorporated the high-frequency sedimentary cycles that compose the geologic framework and identified the rock fabric types of the principal flow units. Three rock-fabric units are used: dolograinstone, grain-dominated dolopackstone and medium crystalline mud-dominated dolostones, and fine crystalline mud-dominated dolostones. Rock properties such as capillary pressure and relative permeability are grouped according to rock fabric. Sensitivity studies and history matching in a two-section area of the SSAU suggest that rock-fabric-averaged models provide fluid saturation distribution more accurately than do cycle-averaged models.

The use of rock-fabric-dependent properties (initial water saturation, residual oil saturation, and relative permeability) in simulation predicts lower waterflood recovery. For carbonates with separate-vug porosity, waterflood recovery decreases and residual oil saturation increases with increasing vuggy porosity ratio (VPR). Residual oil saturation determined by the steady-state method is significantly lower than that determined by the unsteady-state method.

Upscaling permeability values into simulation block sizes was investigated using outcrop and core data. Using detailed minipermeameter data from outcrop it can be shown that the variogram range increases and the sill decreases with increasing horizontal block size. Effective vertical permeability of the simulation block is normally estimated using the ratio of harmonic mean to arithmetic mean of permeability. An analytical equation shows that this ratio is a function of variance rather than thickness.

History matching of primary production in a two-section area of the SSAU is dominated by vertical permeability, which controls the fractions of gas flowing vertically. In order to match production history, the vertical permeabilities used in the simulation are much lower than those measured from cores.

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