Abstract
Large-scale storage of CO2 in saline aquifers is considered an essential technology to mitigate carbon emissions. The injected CO2 will form a buoyant plume that can migrate long distances as a thin sheet under a sloping impermeable caprock. The plume migration can be described by mass conservation of CO2 and brine, which together with Darcy’s law form a mixed elliptic-hyperbolic system of PDEs. Solving this system with the accuracy required to track the plume movement over thousands of years is rarely computationally tractable. In the talk, I present suitable model simplifications.
The simplest possible model to estimate storage capacity assumes infinitesimal flow rates and uses geometrical analysis of the caprock topography to determine catchment areas, spill paths, and structural traps. As a mental picture, you can think of water trickling down a terrain to form ponds, rivers, and lakes. The next type of model assumes that vertical fluid equilibrium is reached instantaneously. Integrating the 3D model equations vertically, one obtains a 2D semi-analytical, elliptic-parabolic problem posed over a surface grid. This not only reduces the required number of grid cells, but also improves time constants significantly. If needed, one can form multilayered models and/or use a combination of depth-integrated and 3D models in different parts of the aquifer.
In the last part of the talk, I discuss how to mathematically optimize the placement of injection points and their fluid rates to maximize storage while minimizing the risk of CO2 leaking back to the surface. This includes use of adjoint formulations and simplified forecasts of migration paths for early termination of the forward simulations.
The simplest possible model to estimate storage capacity assumes infinitesimal flow rates and uses geometrical analysis of the caprock topography to determine catchment areas, spill paths, and structural traps. As a mental picture, you can think of water trickling down a terrain to form ponds, rivers, and lakes. The next type of model assumes that vertical fluid equilibrium is reached instantaneously. Integrating the 3D model equations vertically, one obtains a 2D semi-analytical, elliptic-parabolic problem posed over a surface grid. This not only reduces the required number of grid cells, but also improves time constants significantly. If needed, one can form multilayered models and/or use a combination of depth-integrated and 3D models in different parts of the aquifer.
In the last part of the talk, I discuss how to mathematically optimize the placement of injection points and their fluid rates to maximize storage while minimizing the risk of CO2 leaking back to the surface. This includes use of adjoint formulations and simplified forecasts of migration paths for early termination of the forward simulations.