30 May 2022 to 2 June 2022
Asia/Dubai timezone

Upscaling of Realistic Discrete Fracture Simulations Using Machine Learning

1 Jun 2022, 12:00
15m
Oral Presentation (MS15) Machine Learning and Big Data in Porous Media MS15

Speaker

Nikolai Andrianov (Geological Survey of Denmark and Greenland (GEUS))

Description

Fine-scale discrete fracture simulations provide a natural means to model fluid flows in fractured reservoirs. However, an application of discrete fracture modeling on the field scale is challenging due to uncertainties in fractures' properties, difficulties in creating conforming meshes, and the computational complexity of fluid flow simulations. Upscaling of flows in fracture networks has been traditionally used to cope with these challenges. One common approach is to use the dual porosity/dual permeability (DPDP) model (Hill and Thomas, 1985), where all fracture properties are encoded in effective coarse-scale fracture porosity, permeability, and a mass transfer term accounting for the crossflows between the fractures and the matrix.

Recent results (Andrianov and Nick, 2021; Andrianov, 2021) show that it is possible to map the pixelated fine-scale fracture geometry to the DPDP model closures via a convolutional neural network (CNN), and that the DPDP single-phase flow simulations with both reference and learned model closures agree reasonably well with each other. The goal of the present contribution is to extend these results to the case of two phases.

To this end, we simulate the fine-scale single-phase flow in each fractured coarse-scale grid block with a customized Discrete Fracture-Matrix (DFM) module of the open-source numerical simulator DuMuX (Flemisch et al., 2011). These simulation results are used to calculate the reference upscaled fracture permeabilities and to obtain the shape factors, which characterize the fracture-matrix mass transfer.

We demonstrate that the predicted model closures can be obtained by using the pre-trained CNNs from and implement the two-phase DPDP model on top of the dual porosity module of the open-source numerical simulator MRST (Lie, 2019). The DPDP results agree well with the fine-scale two-phase DFM solutions for a realistic fractured outcrop geometry.

References

Andrianov N. and Nick, H. M. 2021. Machine learning of dual porosity model closures from discrete fracture simulations. Advances in Water Resources, 147:103810. doi.org/10.1016/j.advwatres.2020.103810.

Andrianov, N. [2021] Upscaling of Realistic Discrete Fracture Simulations Using Machine Learning. Society of Petroleum Engineers (SPE). doi.org/10.2118/203962-MS

Flemisch, B., Darcis, M., Erbertseder, K., Faigle, B., Lauser, A., Mosthaf, K., Müthing, S., Nuske, P., Tatomir, A., Wolff, M. and R. Helmig. [2011] DuMux: DUNE for Multi-{Phase, Component, Scale, Physics, ...} Flow and Transport in Porous Media. Advances in Water Resources 34(9): 1102-1112.

Hill, A.C., Thomas, G.W. [1985] A new approach for simulating complex fractured reservoirs. Society of Petroleum Engineers. doi:10.2118/13537-MS

K.-A. Lie. [2019] An Introduction to Reservoir Simulation Using MATLAB/GNU Octave: User Guide for the MATLAB Reservoir Simulation Toolbox (MRST). Cambridge University Press.

Participation In person
Country Denmark
MDPI Energies Student Poster Award No, do not submit my presenation for the student posters award.
Time Block Preference Time Block A (09:00-12:00 CET)
Acceptance of the Terms & Conditions Click here to agree

Primary author

Nikolai Andrianov (Geological Survey of Denmark and Greenland (GEUS))

Presentation materials

There are no materials yet.