Go With the Flow Agent Based Modelling tool for hydro carbon migration

The “Go With the Flow” tool originated from a Xeek data science challenge. Dr Bastian Steffens and Dr Quentin Coraly – the winners of the competition –  were enhanced and transformed into a product that is now available to the market in beta.

The tool allows users to quickly iterate and explore fluid movement models in the subsurface through a user-friendly interface. Users simply upload seismic volume as a SEGY file, then select parameters for how they think fluids would behave kilometers below the Earth’s surface. In minutes the tool displays a GIF of how the agents accumulate within the seismic, allowing geoscientists to imagine how fluids such as hydrocarbons, water, and CO₂ would migrate and be trapped in the subsurface.

The tool description and imprementation is presented in Can Agents Model Hydrocarbon Migration for Petroleum System Analysis? A Fast Screening Tool to De-Risk Hydrocarbon Prospects  by Steffens, B., Corlay, Q., Suurmeyer, N., Noglows, J., Arnold, D. & Demyanov, V.1 Feb 2022In: Energies. 153, 902.

The Fluvial GAN for 2D simulation of fluvial depositional systems

FluvialGAN code was developed and tested by Dr Chao Sun  for the publication Geological realism in Fluvial facies modelling with GAN under variable depositional conditions
by Chao Sun,  Vasily Demyanov,   Daniel Arnold, Computational Geosciences, 10.1007/s10596-023-10190-w

GitHub repository
The dataset is available at https://geodatascience.hw.ac.uk/gan-river-i/

FluvialGAN3D for conditional GAN simulation of fluvial depositional systems

FluvialGAN3D code by Dr Chao Sun is a further development of Fluvial GAN for 3D conditioning case is described in our manuscript ‘A Conditional GAN-based Approach to Build 3D Facies Models Sequentially Upwards’ submitted to Computers & Geosciences

GitHub repository
The dataset is available at https://geodatascience.hw.ac.uk/gan-river-i/

pySeismic: seismic segmentation and geobody detection in 3D seismic

A workflow to automatically segment 3D seismic reflection data into objects and detect and extract geobodies most similar to the given template object (manually/automatically interpreted examples).

This work if the outcome from the PhD thesis of Dr Quentin Corlay “Detection of Geobodies in 3D Seismic using Unsupervised Machine Learning

An implementation with a benchmark F3 data set is provided with the code.
The code outputs the point cloud object in an ASCII file that can then be then opened in Petrel. 

GitHub repository https://github.com/GeoDataScienceUQ/pyseismic/ 

More details in the presentation