Abstract
When neglecting capillarity, two-phase incompressible flow in porous media is modeled as a scalar nonlinear hyperbolic conservation law. A change in the rock type results in a change of the flux function. Discretizing in one dimension with a finite volume method, we investigate two numerical fluxes: an extension of the Godunov flux and the upstream mobility flux, the latter being widely used in hydrogeology and petroleum engineering. In the case of a changing rock type, we demonstrate that the upstream mobility flux often fails to provide the physically correct solution.
Introduction
Original Research Report Details Source, p. 1
In reservoir simulation, modeling the movement of fluids like oil and water through porous rock is a critical task. This process is typically described by nonlinear hyperbolic conservation laws. A cornerstone of this modeling in the petroleum industry is the upstream mobility (UM) scheme. Valued for its physical intuitiveness and numerical stability, it has become a standard tool for engineers. However, research by Siddhartha Mishra and Jérôme Jaffré (2008) reveals that this industry staple possesses a fundamental mathematical flaw when applied to heterogeneous environments.
The Challenge of Heterogeneity
Real-world reservoirs are rarely uniform; they consist of various rock types such as sandstone and limestone. At the interface where one rock type meets another, the "flux function"—which dictates fluid velocity—changes abruptly. This creates a "discontinuous flux" problem.
While the UM scheme is known to be stable and monotone, the authors investigate whether it consistently converges to the physically correct solution. They compare it against the Exact Riemann Solver (ERS), an extension of the Godunov flux that rigorously solves the jump discontinuity at the interface.
Methodology and Convergence
The study employs a 1D finite volume framework to compare three approaches:
- Upstream Mobility (UM) Flux: The industry standard, calculating flow based on upstream fluid dominance.
- Exact Riemann Solver (ERS): A mathematically rigorous approach solving the Riemann problem exactly at the interface.
- Averaged Flux (AV): A staggered mesh Godunov-type scheme.
Using a singular mapping technique, the authors provide a formal proof that the UM scheme converges to a weak solution. However, the critical question is whether this solution satisfies the "interface entropy condition"—the mathematical requirement ensuring uniqueness and physical realism.
Key Insights: Stability vs. Correctness
The research highlights several critical issues with the UM scheme:
- Entropy Inconsistency: Monotonicity ensures stability but not "entropy consistency." A stable simulation can converge to an unphysical solution.
- Undercompressive Waves: In heterogeneous media, waves can become "undercompressive." The ERS scheme is designed to exclude these unphysical states, whereas the UM scheme is not.
- Spurious Artifacts: The UM scheme frequently generates "numerical boundary layers" at rock interfaces. Unlike typical numerical errors, these artifacts do not vanish as the grid is refined; their magnitude remains significant even as their spatial extent shrinks.
- Failure of Intuition: In specific scenarios where mobility values align in certain ways, the UM scheme can completely miss discontinuities, producing physically impossible constant-state solutions.
Numerical Results
Through five distinct numerical experiments using synthetic benchmarks (based on the Buckley-Leverett model), the authors demonstrated the UM scheme's qualitative failures. For instance, in experiments involving changes in absolute permeability, the UM scheme's trace at the interface significantly deviated from the physically correct entropy solution provided by the ERS. Similarly, the staggered "AV" scheme often failed to resolve the interface correctly, mirroring the UM scheme's shortcomings.
Conclusion and Future Work
The primary limitation of this study is its focus on 1D incompressible flow without capillarity. In 3D reservoirs, capillary pressure might smooth some of these discontinuities. However, since 1D logic forms the core of most commercial 3D simulators, these errors likely persist in large-scale models.
The authors conclude that the industry should transition from standard upstream mobility fluxes to Godunov-type (ERS) fluxes. This shift is essential to ensure that simulations of complex, heterogeneous reservoirs are not just stable, but physically accurate.
Summary updated