Multiphase Simulations of Alternative Production Scenarios in Fractured Reservoirs

Wang, Huabing¹, Craig Forster², Yao Fu², Chung-Kan Huang², Yi-Kun Yang², Milind Deo² (1) University of Utah, Salt Lake City, UT (2) University of Utah,

A fully-implicit, three-dimensional, three-phase, discrete fault/fracture, black-oil simulator provides new insight regarding oil production from fractured reservoirs. Results obtained with a controlled volume finite element (CVFE) method compare favorably to those obtained using both single- and dual-porosity finite difference methods (e.g., ECLIPSE). A regularized network of 30 orthogonal faults within a 1000 by 1000 by 200 feet, impermeable, model domain is used to test simulator performance. In this simple reservoir, cumulative oil recoveries over 900 days of primary and water-flood production are similar for CVFE, single-porosity and dual-porosity approaches. CVFE simulations of a complex discrete feature network illustrate the consequences of uncertainty in knowing fracture/fault properties (e.g., porosity, permeability, thickness, dip orientation, connectivity and flow transmissibility. For example, spatial variability in permeability values within the fault planes yields enhanced oil production due to the high-permeability pathways developed within the faults. Two different, equally plausible, intra-fault permeability structures yield simulated cumulative production 25% and 14% greater than that obtained for the uniform permeability case. CVFE simulations of production from reservoirs containing both faults/fractures and significant formation permeability (e.g., clastic reservoirs) reveal the combined effects of formation permeability heterogeneity, fracture/fault network geometries and spatial variability in fault/fracture properties – captured in a multiphase context – that cannot be evaluated with conventional simulators. The inherent variability in fault/fracture properties, combined with uncertainty in knowing the detailed geometric connections between different regions of the discrete feature network, impacts the pattern of sweep in ways that are best understood using a multiphase simulator.