GRAPH NEURAL NETWORKS applied to reservoir engineering

Proposing predictions of the oil production and pressure evolution and in a field is usually done using dynamic models and, thus, requires a proper model to be built and match. This process takes several months to years depending on the complexity of the field fluids, geology, the number of wells and available data.

Our objectives are to evaluate if the pressure and the production of wells in an oil field can be estimated with graph neural networks in situations of fast decision-making or when the model is not available or under construction.

For this demonstration project, python libraries (keras and tensorflow) are used to build graph neural models. The training process is done using a Jupyter notebook. Different types of graph neural networks are available. In addition, different types of classical deep neural networks can be used as encoders and decoders. The developed Python script enables easy training, saving and loading of the graph neural network models.

Results are obtained using the Volve field (North Sea) open-source dataset and show that neural networks are able to capture the evolution of pressure depending on the production/injection patterns. The prediction of the oil production rate is also possible but with more difficulties.

A comparison with the open-source Volve dynamic model must still be carried on, and a comparison with material balance could also be interesting. Testing the methodology on other fields would also be very useful to further validate the approach.