The first application of compositional upscaling to the routine modeling of a major reservoir is described. The reservoirCupiagua, in Colombia-is a rich gas condensate field. Given that Cupiagua falls rapidly below dewpoint and produced gas is recycled, the main recovery process is the vaporization of liquid components into the gas phase, in which they are transported to producers. Therefore, the process is compositional, but flow is dominantly in one phase.Cupiagua is a heterogeneous reservoir dominated by natural fracture corridors that provide more than 80% of the permeability in some areas. It is not possible to represent these features explicitly within the full-field model (FFM), nor do they fit a conventional dual-porosity representation. Therefore, an upscaling process is required.The process described is the use of alpha-factor compositional upscaling functions, which modify the velocities of individual pseudocomponents. We show how fine-grid cross-section modeling, with a range of sensitivities, may be used to generate an appropriate set of alpha-factor functions, which are validated against detailed sector models and may then be used as the principal history-match parameter in the FFM.
IntroductionCompositional upscaling is a technique that has been developed at a theoretical level for a number of years. A set of tools to generate compositional upscaling functions for realistic reservoir and fluid descriptions now exists, and some commercial simulators have been adapted to apply these functions in full-field compositional simulation.There arguably has been some delay in appreciating the importance of the technique, given that many of the potential applications (such as miscible injection processes) involve a whole set of complexities (such as three-phase flow description and upscaling) in which compositional upscaling is just one more factor.Cupiagua, however, is a good candidate for compositional upscaling, with a production rate that cannot otherwise be simulated even approximately in an FFM.It is a good candidate because almost all reservoir flow occurs in the gas phase, including the transport of the heavier components. The liquid phase is relatively immobile. (An exception is the near-well region, in which the condensate-banking phenomenon is represented by techniques involving two-phase pseudopressure functions. 1 However, this region is smaller than a single FFM gridblock and can be decoupled from the reservoir flow problem.) The use of pseudorelative permeability upscaling functions is therefore not needed. Cupiagua is also a good candidate because it has strong heterogeneity, which is too small-scale to be represented explicitly in FFM gridblocks and therefore requires upscaling.Because the early gas/oil ratio (GOR) increase in Cupiagua is caused by injected gas movement through natural fractures, the only reasonable way to achieve an early FFM match without com-