TL;DR
This paper introduces an extension to the multi-scale finite-volume (MSFV) method that allows for the accurate simulation of complex well configurations in subsurface flow problems. By integrating well models directly into the MSFV framework, the method effectively captures the intricate interactions between wells and the surrounding porous media, even with highly heterogeneous properties.
What Problem Does This Paper Address?
Accurately simulating fluid flow and transport in geological formations like aquifers and hydrocarbon reservoirs is crucial for various applications, including water management, pollution control, and resource extraction. These simulations often involve complex, heterogeneous domains with highly variable permeability. A significant challenge arises from the presence of wells, which are typically much smaller than the grid cells used in simulations. This scale difference makes it impossible to model wells explicitly. Traditional methods struggle to accurately represent the pressure differences between the wellbore and the surrounding grid block, and the complex interplay between multiple wells. Existing multi-scale methods, while effective for reservoir flow, have limitations in handling these complex well scenarios.
Methods
The authors extend the existing Multi-Scale Finite-Volume (MSFV) method by introducing "well functions." These functions are analogous to the basis and correction functions used in the standard MSFV method but are specifically designed to capture the behavior of wells. Unlike previous approaches that required separate sub-domains for wells, this new method integrates well functions directly within the existing dual coarse grid cells that intersect with the wells. This approach allows for the direct modeling of wellbore pressures and flow rates without requiring a finer grid resolution around the wells. The method handles both pressure-constraint and rate-constraint wells, introducing additional degrees of freedom for the latter to solve for the unknown wellbore pressure. The core of the MSFV method involves solving a coarse-scale pressure problem and then reconstructing a conservative fine-scale velocity field, which is then used to solve the transport equation.
Datasets
Not discussed in the paper. The paper focuses on developing and validating a numerical method using synthetic test cases.
Key Insights and Ideas
- Seamless Integration: The new well model fits naturally into the existing MSFV data structure, avoiding the need for complex modifications or additional sub-domains. This is a significant advantage for computational efficiency and implementation.
- No Upscaling of Well Productivity: The method directly uses fine-scale well productivity indices, eliminating the need for potentially error-prone upscaling of these parameters.
- Accurate Interference Capture: The approach effectively captures the interference between multiple complex wells, even when they are in close proximity or intersect within the same grid cells. This is a critical improvement over methods that struggle with such scenarios.
- Handling Heterogeneity: The MSFV method, with its new well model, demonstrates robustness in simulating flow through highly heterogeneous permeability fields, maintaining good accuracy.
- Line Sources are the Real Challenge: The paper emphasizes that the true difficulty in well modeling lies in accurately representing line sources (realistic wells) rather than simple point sources, especially for rate-constraint wells where an additional unknown (well pressure) needs to be solved.
Key Results
The MSFV method with the new well model demonstrated excellent agreement with fine-scale reference solutions across various scenarios, including:
- Single-phase flow: Relative errors in injection-well pressures were consistently low (e.g., 1.8% and 0.4% for two scenarios). Tracer transport simulations also showed very good agreement in representing sharp tracer fronts.
- Heterogeneous permeability: Even with high variances in permeability (up to $\sigma^2_{ln k} = 15.9$), the MSFV method maintained reasonable accuracy, with errors in well-bore pressure generally below 10% for moderate heterogeneity and up to 30% for extreme cases.
- Multi-phase flow with gravity: Simulations of multi-phase flow, including gravity effects, showed good agreement in both saturation and pressure fields for both homogeneous and heterogeneous permeability fields.
Limitations and Future Work
- Localization Assumption: The paper acknowledges that deviations from fine-scale solutions can occur in regions of strong interference, such as where wells intersect within a few fine grid cells. This is attributed to the localization assumption inherent in the MSFV method.
- Computational Cost for Extreme Heterogeneity: While the MSFV method significantly reduces computational cost compared to fine-scale simulations, the accuracy can be affected by extreme permeability heterogeneity, leading to higher errors in well-bore pressures.
- Future Work: The authors suggest that the method is well-suited for massive parallel computations due to the independent nature of local problems. Further exploration of its application to more complex multi-phase flow and transport problems, especially those involving strong gravity effects, is implied. The paper also notes that the framework can be extended to more general well-bore pressure relationships.
Summary updated