Static high resolution three dimensional geological models are routinely constructed to provide an integrated description of a reservoir which includes seismic, well log, and core data, and which characterize the reservoir heterogeneity at multiple scales. These models also represent the structure and stratigraphy of the reservoir within the design of the modeling grid, which may include fault blocks, faults, pinch-outs, layering and cross-bedding. The growth of computational resources has remained rapid, and both geologic models and flow simulation models have increased in size. 50 million cell geologic models are routine, while simulation models are typically one or two orders of magnitude coarser. Hence upscaling of the geologic models for flow simulation remains part of the subsurface workflows.
The industry also faces new reservoir engineering challenges. Unconventional reservoirs (tight gas / shale oil / shale gas) have sufficiently low permeabilities that the time for pressure transients are no longer measured in hours or days, but instead are measured in decades or longer. The separation between transient testing and steady state reservoir management is no longer applicable. Similarly, our upscaling algorithms have relied upon steady state concepts of flow, which may no longer be applicable.
In the current study, a novel diffuse source transmissibility upscaling approach is described. It applies pressure transient concepts to the calculation of the effective transmissibility between reservoir simulation coarse cell pairs. Unlike the usual steady state upscaling algorithms, it is a completely local calculation and is not dependent upon knowledge of, or assumptions about, global reservoir flow patterns. It is well suited to performance prediction within unconventional reservoirs as it utilizes drainage volume concepts to the calculation of coarse cell average pressures, although its use is not restricted to unconventional reservoirs. The approach is tested using the conventional reservoir SPE10 waterflood data set. It is then validated at field scale using an onshore US tight gas reservoir model. The approach is shown to reduce simulation run times by up to two orders of magnitude without significant loss of accuracy in performance prediction.