Vasily Demyanov delivered an invited keynote talk at AAPG Decision Based Integrated Reservoir Modelling, Al Khobar 2025,

Explainable AI in Reservoir Modeling Workflows

also accompanied by a poster by Roman Baishev – Geosteering Under Geological Uncertainty  that presents geosteering model probability update workflow across ensemble of model realisation based on Watt UQ benchmark case .

The overview covers AI applications to integrated reservoir modelling workflows:

1. Discover patterns in geological and reservoir engineering data;
2. Describe variability and heterogeneity of subsurface properties;
3. Predict outcomes of subsurface resource development;
4. Decide how to develop the resource confident proactive response to monitoring.

This overview demonstrates a few examples of how unsupervised and supervised learning algorithms applied in reservoir modelling workflows can gain interpretability and explainability in describing geological realism and uncertainty for more accurate predictions and more efficient learning from data. AI applications covers several above listed tasks of reservoir modelling workflow (i, ii, iii, iv), specifically:

1. AI seismic interpretation and geobody detection (i) [1].
2. Constrain geological conceptual modelling with learning from outcrops (i) [2].
3. Generative facies modelling of meandering reservoirs based on learning from depositional process modelling (ii) [3].
4. Dynamic and static data integration with VAE and uncertainty representation via latent space to predict reservoir dynamic response (iii) [4].
5. Inform well operation decision-making based on PTA pattern recognition (iv) [5].

 

Abstract: Demyanov XAI AAPG

More details can be found in the Geostats Congress chapter:

Uncertainty in AI Based Reservoir Modelling Workflows by V. Demyanov, Q. Corlay, A. Nathanail, C. Sun, G. Shishaev, D. Arnold, Published in International Geostatistics Congress 2024, Springer Nature Switzerland, 2025