A novel hybrid model for multiphase flow in complex multi-scale fractured systems
L. Li (China University of Petroleum (East China), TU Delft - Reservoir Engineering, Khalifa University)
Denis Voskov (TU Delft - Reservoir Engineering, Stanford University)
More Info
expand_more
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Abstract
We present a multi-level discrete fracture model (MLDFM) to guarantee a robust and efficient solution for naturally fractured reservoir simulation. In MLDFM, we apply a triple continuum model using structured grid for forward simulation where large-scale fractures are represented with numerical embedded discrete fracture model (EDFM) and the secondary fractures are upscaled as third continuum. What makes the triple continuum model different from the previous work is that both the numerical EDFM and the third continuum are treated in a dynamic approach by considering the effect of flow direction on the complex local-scale flow response. For that purpose, we construct a finer unstructured discrete fracture matrix (DFM) grid which represents all fractures explicitly and is conformal to the boundary of coarse structured grid. During a simulation run, we apply a basis function to generate the local boundary conditions at fine scale using the global solution. Benefit from that, we can use a more accurate flow-based approach in the extended local upscaling to re-compute the transmissibility in triple continuum model. Moreover, we apply a local-global upscaling formalism to guarantee dynamically updated local boundary conditions for upscaling. Besides, we present several cases using synthetic and realistic fractured networks to demonstrate the performance of MLDFM. The results prove that the proposed MLDFM approach more accurately captures the flow in complex fractured systems than EDFM solutions by comparing against fine-scale DFM. At the same time, MLDFM is more computationally efficient in comparison with fine scale DFM.