Model of an artificial Enhanced Geothermal System (EGS)

This repository contains a stochastic model of an EGS, developed for the Wolrd Geothermal Congress 2020+1. The main model is in the directory wgc2020_model. The model uses Monte Carlo method for random generation of fracture network. For each fracture network, the simulation of EGS is run and the quantities of interest (temperature, power) are collected. The result is then in form of distribution characteristics of the quantities of interest: mean, standard deviation, probability density function.

This provides a much higher quality of information in comparison to single simulation, since the fracture network and hydrogeological parameters are mostly estimated with high level of uncertainty.

The model uses two main softwares:

• GeoMop http://geomop.github.io/, two of its libraries in particular:
  • MLMC https://github.com/GeoMop/MLMC
  • BGEM https://github.com/GeoMop/bgem
• Flow123d http://flow123d.github.io/

Single sample model

Each sample takes several computation steps:

1. random fracture network sample generation (by BGEM)
2. creation of compatible 1D-2D-3D computational mesh (by BGEM)
3. computational mesh optimization by HealMesh algorithm (by BGEM)
4. hydraulic stimulation model (poroelastic model in Flow123d)
5. modification of hydraulic parameters based on mechanical changes (by BGEM)
6. heat transfer model with modified parameters (thermo-hydraulic model in Flow123d)
7. heat transfer model with original parameters as a reference (thermo-hydraulic model in Flow123d)

Installation

Local

For running the simulation locally one has have Docker (https://www.docker.com/) or Singularity (https://apptainer.org/) available. The simulation is then simply run by command

./run_process_local.sh run <output_dir>

with Docker, or

./run_process_local.sh run <output_dir> sing

with Singularity.

The Docker image flow123d/geomop-gnu:2.0.0 will be pulled automatically, or it can be downloaded manually from https://hub.docker.com/r/flow123d/geomop-gnu, with current tag 2.0.0 (see also project webpage http://geomop.github.io/).

The configuration of the model can be changes in config.yaml

Cluster (with PBS)

For running the simulation one has to install Singularity, Python 3, and MLMC package. The MLMC library controls the sampling and scheduling. Sample computation requires running BGEM and Flow123d inside flow123d/geomop-gnu:2.0.0 container as mentioned above.

For installing MLMC in virtual environment one can use the enclosed script setup_python_envinroment.sh.

The simulation is then run through PBS script

qsub run_process_pbs.sh

which must be accommodated by users needs and cluster specification.

Running the simulation

Model can be run locally by script run_process_local.sh or on cluster (Metacentrum in CZ tested) by script run_process_pbs.sh. In case of cluster run, one has to specify the computational resources in file config_pbs.yaml, the PBS script itself is then created automatically. In both cases, the main configuration is loaded from file config.yaml.