The FluidFlower International Benchmark Study
Numerical benchmark studies of multi-phase flow in porous media generally lack a clearly defined ground truth. In this benchmark study, we consider a problem inspired by CO2 storage, where the dominant processes are associated with multiphase flows, capillarity, dissolution, and convective mixing. Supporting the numerical benchmark study, we conduct a series of physical realizations in the FluidFlower experimental setup.
Benchmark study
To leverage the experimental capabilities of the FluidFlower, we are conducting an international benchmarking study for modeling, simulation and forecasting of multiphase, multicomponent flows in porous media. Several multiphase flow phenomena are present in the experimental setup, including both heterogeneous and hysteretic two-phase flow properties (leading to both structural and residual gas trapping) and two-component mixing with concomitant development of gravity fingers, as illustrated in a sample geometry in Figure 1.
Figure 1 – Sample experiment in a downscaled FluidFlower rig. Visible here are multiple fluid-filled gas layers, free-phase CO2, and variable concentrations of dissolved CO2 in the water phase. Important processes for CO2 storage, such as capillary entry effects, hysteresis, gravity fingering, are all apparent.
Benchmark process
The main goal of the benchmark is to provide a full-physics validation of the state-of-the-art simulation capabilities within the international porous media community. To this aim we envision the following benchmarking process:
1) A full geometric, operational, and petrophysical description is provided to all participants, including commonly measured multiphase flow parameters for porous media simulation based on ex situ measurements (grain size distributions, porosities, permeabilities, end-point relative permeabilities, entry pressures). Single-phase flow data (well tests and tracer tests) are provided to the participants for model calibration.
2) On the exact geometry provided to the participants there will be conducted at least 5 identical multiphase flow experiments.
3) Participants must themselves provide equation-of-state data, based on the liquid and gas used.
4) All participants are asked to provide simulation output in terms of target quantities, together with both quantitative and qualitative confidence intervals, according to a common protocol.
Benchmark outcome and vision
The combination of A), B), C) and D) within the same benchmarking study is to our knowledge unprecedented within the porous media community. This will allow us, within the context of this setup, to address for the first time the correlation between the numerical uncertainty quantification and a measure of real system uncertainty.
Together, the physical FluidFlower experiments and numerical simulations will allow us to address the confidence with which the main physical processes associated with subsurface CO2 storage can be modeled:
A) The physical experiments will provide both a “ground truth”, and will also allow us to quantify to some extent the inherent uncertainties associated with the porous media flow itself by means of repeated experiments on the same geometry.
B) The in-house parameter estimation study will allow us to address any discrepancies between ex situ measured parameters and in situ properties, as well as to provide a sense of the extent the standard model equations for porous media can capture the observed flows.
C) The numerical simulations reported by each research group will allow us to quantify the precision and intra-group uncertainty as assessed by each of the research groups individually.
D) The total panel of research groups will allow us to quantify the inter-group variability in the results (and thus to some extent the reproducibility of numerical modeling), and the relationship between the ensemble precision and uncertainty.
Participating academic institutions
