Advanced Upscaling for Kashagan Reservoir Modeling

Abstract

Summary The Kashagan field is a huge carbonate formation located 4.5 km below the bottom of the North Caspian sea. The reservoir is saturated by overpressured light oil, and the development is based on first-contact-miscible gas injection. The reservoir is highly stratified, with a fine sequence of depositional cycles and long-range lateral correlations. Three porosity systems (matrix, karst, and fractures) can be organized in two main environments: a massive, low-permeability, matrix-like inner platform and a highly fractured/karstified rim. The reservoir geology is modeled by means of detailed geological grids consisting of tens of millions of cells, with vertical spacing of 1 m or even less to account for high-order depositional cycles. Geological grid cannot be used to run compositional simulations, and much-coarser grids, in which hundreds of geological layers are lumped in few tens of dynamic layers, are used by reservoir engineers. To minimize coarse-scale errors, an average lateral spacing of 250×250 m is used for both simulation and geological grid; nonetheless, upscaling remains a challenge. Traditional permeability (k*) upscaling methods, including flow-based methods, overestimate Kashagan field/wells production and injection potentials. We implemented a method in which the outcome of the upscaling is effective transmissibility (T*) instead of k*. T* upscaling has been proposed in the past as an alternative to k* upscaling, but it is neither part of commercial workflows nor widely accepted in the reservoir-modeling community. In our T* upscaling, the solution of local flow problems around coarse-cell interfaces is used to compute coarse transmissibility. T* and k* upscaling were compared by simulating both single-phase and gas-injection problems, including platform and rim, using the results of fine-scale simulation as a reference. We considered (1) single-porosity simulations with geological grid populated by only matrix (first medium) and karst+fracture (second medium) properties and (2) dual-porosity/ dual-permeability simulations encompassing both media. Contrary to k* upscaling, T*-based coarse simulations perfectly replicate fine-scale field and well injection/production potentials. Using T* upscaling as a cornerstone for company activities on Kashagan, we can run coarse-scale full-field simulations in a few hours without loss of consistency with the results provided by weeks-long, often unpractical, fine-scale simulations. On the contrary, the inaccuracy of k* upscaling would have required much finer and more computationally-expensive simulation grids together with the implementation of ad hoc multiphase upscaling.