Upscaling of Compositional Models for Single-Porosity and Discrete-Fracture-Matrix Systems Lou Durlofsky Stanford University Compositional flow simulation is essential for modeling enhanced oil recovery and carbon storage processes, but it can be very expensive computationally. This is of particular concern when many simulations must be performed, as is the case for uncertainty quantification or in the context of optimization. Computational demands can be mitigated by applying upscaling procedures, though it is essential that the coarse-scale models provide sufficient accuracy relative to fine-scale results. In this talk I will describe a set of accurate upscaling techniques applicable for compositional reservoir simulation. These methods include aggregation-based upscaling for complex discrete-fracture-matrix (DFM) models and global compositional upscaling for single-porosity (Gaussian and channelized permeability) models. For DFM models, the aggregation-based upscaling computations are not very expensive, though the degree of coarsening achievable may be limited in some cases. Global compositional upscaling, by contrast, is expensive, so it is important that the upscaled functions be “reuseable.” Numerical results will be presented to demonstrate the performance of the various upscaling methods for both DFM and single-porosity systems. The effective reuse of globally computed upscaled functions, both for optimization and uncertainty quantification, will also be demonstrated.