Machine Learning framework for Reservoir Simulation

Take a look at this exciting development that brings the end-to-end neural network to 3D Reservoir Simulation and History Matching. Therefore, it promises several orders of magnitude simulation seed-up. But it's the claim that needs to be investigated. Here is the publication that describes the development: arxiv.2102.10304

It should not be treated as a substitution for finite-difference simulations that provide high accuracy but demand massive computational resources. Nevertheless, it's an excellent way for fast approximation with some caveats.

DeepField is licensed under the Apache License 2.0

Link for the paper
Link for the GitHub repo
Link for the LinkedIn post