Boosting Barlow Twins Reduced Order Modeling for Machine Learning‐Based Surrogate Models in Multiphase Flow Problems

• Published in: Water resources research Vol. 60; no. 10
• Main Authors: Kadeethum, T., Silva, V. L. S., Salinas, P., Pain, C. C., Yoon, H.
• Format: Journal Article
• Language: English
• Published: Washington John Wiley & Sons, Inc 01-10-2024; American Geophysical Union (AGU)
• Subjects: Algorithms; Barlow Twin; boosting; Carbon capture and storage; Carbon dioxide; Carbon sequestration; Ensemble learning; Fluid pressure; geologic carbon storage; Injection wells; Learning algorithms; Machine learning; Model reduction; Modelling; Multiphase flow; Parameter robustness; Predictions; Reduced order models; Supervised learning
• ISSN: 0043-1397; 1944-7973
• DOI: 10.1029/2023WR035778

We present an innovative approach called boosting Barlow Twins reduced order modeling (BBT‐ROM) to enhance the reliability of machine learning surrogate models for multiphase flow problems. BBT‐ROM builds upon Barlow Twins reduced order modeling that leverages self‐supervised learning to effectively handle linear and nonlinear manifolds by constructing well‐structured latent spaces of input parameters and output quantities. To address the challenge of high contrast data in multiphase flow problems due to injection wells and faults, we employ a boosting algorithm within BBT‐ROM. This algorithm sequentially trains a set of weak models (i.e., inaccurate models), improving prediction accuracy through ensemble learning. To evaluate the performance of BBT‐ROM, we conduct three three‐dimensional multiphase flow problems, including waterflooding and geologic carbon storage (GCS), with varying numbers of input parameter cases and model domain features. The results demonstrate that BBT‐ROM excels at predicting non‐wetting phase saturation (e.g., oil or CO2 $\mathrm{C}{\mathrm{O}}_{\mathrm{2}}$ saturation) and fluid pressure, with average relative errors ranging from 0.5% to 3%. Importantly, BBT‐ROM showcases robustness when faced with limited input parameter space during GCS testing.
Key Points
Boosting Barlow Twins reduced order modeling (BBT‐ROM) to account for high gradient data around wells and faults in multiphase flow problems
BBT‐ROM achieves accuracy with a maximum relative error below 12% at any given time, even in complex fault geometries
Accurate pressure and CO2 saturation prediction in multiphase flow and geologic carbon storage examples