Abstract
Compositional flow simulation, which is required for modeling enhanced oil recovery (EOR) operations, can be very expensive computationally, particularly when the geological model is highly resolved. It is therefore difficult to apply computational procedures that require large numbers of flow simulations, such as optimization, for EOR processes. In this paper we develop an accurate and robust upscaling procedure for compositional flow simulation. The method requires a global fine-scale compositional simulation, from which we compute the required upscaled parameters and functions associated with each coarse-scale interface or well block. These include coarse-scale transmissibilities, upscaled relative permeability functions, and so-called α-factors, which act to capture component flow rates in the oil and gas phases. Specialized near-well treatments for both injection and production wells are introduced. An iterative procedure for optimizing the α-factors is incorporated to further improve coarse-model accuracy. The upscaling methodology is applied to two example cases, a two-dimensional model with eight components and a threedimensional model with four components, with flow in both cases driven by wells arranged in a five-spot pattern. Numerical results demonstrate that the global compositional upscaling procedure consistently provides very accurate coarse results for both phase and component production rates, at both the field and well level. The robustness of the compositionally upscaled models is assessed by simulating cases with time-varying well bottom-hole pressures that are significantly different from those used when the coarse model was constructed. The coarse models are shown to provide accurate predictions in these tests, indicating that the upscaled model is robust with respect to well settings. This suggests that upscaled models generated using our procedure can be used to mitigate computational demands in important applications such as well control optimization.
