Simulation of Unstable Displacement with Focus on Gas-Injection Methods

Many operational scenarios require simulation of unstable displacements;  examples include gas injection for increased oil recovery and injection of CO2 into saline aquifers for long-term storage. A particular problem with such problems is that they require bridging very different scales: Whereas displacement fronts may flow kilometers in the lateral direction, crucial effects like segregation, reaction, and dissolution typically occur on a sub-meter scale in the vertical direction. The main time-scale of interest for a specific flow scenario is often months, years, or decades. In contrast, the thermodynamical behavior, which is a key determining factor for the flow, occurs on much shorter time-scales.

Numerical simulation is an essential tool to understand, predict, and optimize these scenarios. To describe the complex flow behavior of unstable displacements one must include a large number of physical effects, which increases simulation times significantly and makes it difficult to obtain sufficient accuracy in a reasonable time-frame. In particular, the simulator will spend much of the time solving thermodynamic subproblems that, fundamentally, are very similar to each other.

The purpose of this project is to develop robust and efficient numerical techniques for simulating unstable displacements. Our focus is on various divide-and-conquer strategies, that decompose the simulation into subproblems that can be solved separately using efficient and robust methods.  For instance, local thermodynamic behavior and flow features may be resolved by way of controlled numerical experiments in a small subsection of the reservoir that gives precise, upscaled input to the full-field simulator. By systematically giving specialized treatment tailored to individual physical effects, it is possible to significantly improve both efficiency and accuracy, ensuring that the computational budget can be spent on the effects that matter. At the same time, one must assure that the coupling between the subproblems is correctly resolved.

Collaborators for this project include researchers at Stanford University in the US, TU Delft in the Netherlands, and University of Manchester in the UK.