TL;DR
This paper introduces a novel upscaled model for simulating fluid flow and geomechanics in fractured porous media. By combining an embedded fracture model with a constrained energy minimization approach, they achieve a computationally efficient and accurate coarse-grid simulation method that significantly reduces computational cost while maintaining high fidelity.
What Problem Does This Paper Address?
Simulating fluid flow and geomechanics in fractured porous media is crucial for understanding processes in reservoirs, but it's computationally very expensive. This is due to the complex geometry of fracture networks, which often involve multiple scales and require very fine grids for accurate representation. Traditional methods struggle to capture these fine-scale details efficiently. The paper addresses this by developing a method to create a coarser, yet accurate, model that captures the essential physics without the prohibitive computational cost of fine-grid simulations.
Methods
The authors propose a two-pronged approach:
- Embedded Fracture Model (EFM): They first develop a fine-grid model that treats fractures as lower-dimensional entities embedded within a porous matrix. This "dual continuum" approach uses a finite volume method for fluid flow and a finite element method for mechanics, capturing the coupled interactions between fluid pressure and solid deformation.
- Nonlocal Multicontinua (NLMC) Upscaling: The core innovation lies in their upscaling technique. They use a constrained energy minimization principle to construct "multiscale basis functions" on local, oversampled domains. These functions are designed to capture the essential physics of both flow and mechanics across different continua (matrix and fractures) within a coarse grid cell. The key is that these basis functions have a spatial decay property, meaning their influence is localized, and they are constructed such that their degrees of freedom have direct physical meaning on the coarse grid. This allows for the creation of an efficient coarse-grid model that accurately represents the fine-scale behavior.
Datasets
Not discussed in the paper. The paper focuses on developing and validating a numerical method using synthetic test cases.
Key Insights and Ideas
- Physical Meaning on Coarse Grid: The NLMC upscaling method is designed so that the degrees of freedom on the coarse grid directly correspond to physically meaningful quantities (e.g., average pressure, average displacement in a coarse cell), which is a significant advantage over some other upscaling techniques.
- Decoupled Basis Function Construction: For simplicity, the authors construct multiscale basis functions for flow and mechanics separately. While they note that coupled construction could yield better results, their decoupled approach still demonstrates strong accuracy.
- Spatial Decay Property: The multiscale basis functions exhibit a spatial decay property, meaning their influence diminishes away from the region they are designed to represent. This is crucial for ensuring that the upscaled model remains computationally tractable.
- Fracture Representation: The embedded fracture model effectively handles the complex geometry of fractures without explicitly meshing every fine detail, which is a major bottleneck in traditional methods.
Key Results
The paper presents numerical results for two-dimensional test cases with 30 and 60 fractures.
- Accuracy: The upscaled coarse-grid model achieves high accuracy, with relative errors for pressure typically less than 1% and for displacement around 2.5% at the final simulation time, even with significantly reduced degrees of freedom.
- Computational Speedup: The upscaled model reduces the system size dramatically. For a test case with 30 fractures, the fine grid system had 59,124 degrees of freedom (DOF), while the upscaled coarse grid system had only 1,393 DOF. This translates to a significant computational gain, with the coarse grid solver being orders of magnitude faster than the fine grid solver.
- Convergence: The results show good convergence behavior with respect to the number of oversampling layers used in the construction of the multiscale basis functions.
Limitations and Future Work
- Decoupled Basis Functions: The authors acknowledge that constructing decoupled basis functions for flow and mechanics is a simplification. They suggest that a coupled construction of multiscale basis functions could provide even better results and is a direction for future work.
- General Heterogeneity: While the method is shown to be robust, the paper mentions that for highly heterogeneous cases, more basis functions might be needed, and the constrained energy minimization (CEM) approach can identify important modes.
- 3D Extension: The current work focuses on 2D problems. Extending the method to three dimensions would be a natural next step.
- Coupled Construction: As mentioned, a fully coupled construction of multiscale basis functions is a promising avenue for future research.
Summary updated