Field scale model generation and upscaling toolkit

Repository: github.com/ImperialCollegeLondon/StrataTrapper

• The StrataTrapper codes
• Structure
• Running
• Output compression
• MEX acceleration
• Versions
• Contributing
• References

The StrataTrapper codes

This is the StrataTrapper upscaling toolkit. It can also generate heterogeneous fine-scale models with specific correlation lengths to re-upscale given coarse-scale two-phase flow models.

Another tool is the reduced-physics model CO2GraVISim.

In StrataTrapper-models repository, we publish field-scale models upscaled with StrataTrapper.

Structure

Top-level scripts and functions are in the repository root, and the rest is in src/ folder.

demo.m script is an implementation of the running guideline below.
Feel free to play with it and use as an example for your own scripts.

Running

1. Run startup.m to setup MATLAB Path.
   1. Optional: start a parallel pool to run computations there.
2. Read or generate target coarse grid dimensions and input fine-scale porosity and permeability for each coarse block.
3. Setup input rock-fluid properties and algorithm options represented by Params and Options classes. Pass an array of Params for multiple-region models.
4. Create integer mask to:
   • filter out impermeable cells and/or compute an arbitrary subset of cells by putting 0s
   • Put index n of the input array of Params to map coarse cells to the corresponding regions
5. Run strata_trapper function with arbitrary number of parallel workers optionally enabling a UI progress bar.
6. Visualise outputs with plot_result function
7. Export the outputs to PFLOTRAN-OGS or OPM Flow format using ogs_export or opm_export function, respectively.

Tips:

• Usually, MATLAB runs startup.m scripts automatically if they are in a startup folder.
• The heaviest part of the algorithm is essentially parallel with no synchronisation. So, using several parallel workers usually results in a proportional performance boost.

Output compression

By default, strata_trapper outputs one set of saturation tables per coarse cell per direction. Such a number might be found too big in some situations.

With quantize, toolkit can now compress the set of saturation tables if they are similar enough or the acceptable approximation error is high.

Read compression README for more details.

MEX acceleration

We provide the CodeGenMex class to automatically build a MEX-accelerated version of computationally-demanding functions such as upscale.

% 1. compile MEX functions
codegen_mex = CodeGenMex().config().build();

% 2. use strata_trapper normally
...

% 3. recompile MEX functions when the source code changes
codegen_mex.build();

% 4. delete MEX functions to use the original
codegen_mex.clear();

Requires MATLAB Coder.

Versions

The original version of the toolkit is v0.1.0.
It has its own structure and some unique functionality,
so it may worth attention as much as later versions.

CHANGELOG.md describes the version history and key changes.

Other versions can be accessed via tags and releases sections of the repository.

Contributing

Everyone is welcome to open issues and pull requests.

References

The StrataTrapper algorithm as well as motivation and theory behind it are in the paper:

Samuel J. Jackson, Samuel Krevor
Small-Scale Capillary Heterogeneity Linked to Rapid Plume Migration During CO2 Storage
Geophysical Research Letters | 2020
https://doi.org/10.1029/2020GL088616