Speaker
Description
In 2021, the FluidFlower validation benchmark study was initiated to assess reservoir simulation performance in a meter-scale, geologically complex setting [1]. The benchmark provided a unique dataset in which experimental observations were systematically compared against simulation results from multiple research groups. Among its key contributions were high-resolution imaging datasets of CO$_2$ storage, offering unprecedented detail for model validation. Building on this foundation, Landa-Marbán et al. (2025) [2] introduced a novel history-matching framework that leverages the Wasserstein distance as a quantitative metric for comparing simulated and observed images, using the OPM Flow simulator [3].
Our results achieve the lowest errors for both the sparse-data and Wasserstein-distance metrics when compared with previous benchmark submissions and with the study of Saló-Salgado et al. (2024) [4], in which parameters were manually calibrated using experimental data from a smaller-scale setup. The implemented workflow allows the five‑day FluidFlower experiment to be simulated in only about two minutes, highlighting its suitability for time‑critical applications, including digital twins. These successful outcomes further support the conclusions from the FluidFlower benchmark study [1], indicating that the system can be accurately represented using standard flow equations, conventional saturation functions, and typical PVT properties for CO$_2$-brine mixtures.
One of the main outcomes of this study is the pofff tool [2], an open-source framework that generates the necessary input files for OPM, including corner-point grids, saturation function tables, and injection schedules, through TOML configuration files. This workflow ensures reproducibility of the results and facilitates further studies of the history matching. The methodology is designed to align with the FAIR (Findable, Accessible, Interoperable, Reusable) principles [5], which have not been consistently adopted in recent years [6], yet remain essential for advancing reservoir simulation technology. Additional open-source tools related to OPM Flow are available at https://github.com/cssr-tools.
This presentation highlights the lessons learned from this challenging history matching study, including methodological advances and limitations encountered. Particular attention is given to the role of image segmentation bias, which remains a critical obstacle in achieving robust history matches. Finally, we outline future directions aimed at mitigating these biases and advancing the integration of image-based validation into reservoir simulation workflows.
References:
[1] Flemisch, B., et al. 2024. The FluidFlower validation benchmark study for the storage of CO2. Transp. Porous Med. 151, 865-912. https://doi.org/10.1007/s11242-023-01977-7.
[2] Landa-Marbán, D., Sandve, T. H., Both, J. W., Nordbotten, J.M., and Gasda, S. E., 2025. Performance of an open-source image-based history matching framework for CO2 storage. To appear in Transp. Porous Med. https://arxiv.org/abs/2510.20614.
[3] Rasmussen, A.F., et al., 2021. The open porous media flow reservoir simulator. Comput. Math. Appl. 81, 159–185. https://doi.org/10.1016/j.camwa.2020.05.014.
[4] Saló-Salgado et al., 2024. Direct comparison of numerical simulations and experiments of CO2 injection and migration in geologic media: Value of local data and forecasting capability. Transp. Porous Med. 151, 1199-1240. https://doi.org/10.1007/s11242-023-01972-y.
[5] Wilkinson, M., et al., 2016. The FAIR Guiding Principles for scientific data management and stewardship. Sci Data 3, 160018. https://doi.org/10.1038/sdata.2016.18
[6] Liu, N., et al. 2025. Trends in porous media laboratory imaging and open science practices. https://arxiv.org/abs/2510.05190.
| Country | Norway |
|---|---|
| Acceptance of the Terms & Conditions | Click here to agree |
