Praks Pavel, Rasmussen Atgeirr, Lye Kjetil Olsen, Martinovič Jan, Praksová Renata, Watson Francesca, Brkić Dejan
IT4Innovations, VSB - Technical University of Ostrava, 708 00, Ostrava, Czech Republic.
SINTEF Digital, 0373, Oslo, Norway.
Heliyon. 2024 Nov 1;10(22):e40044. doi: 10.1016/j.heliyon.2024.e40044. eCollection 2024 Nov 30.
Open Porous Media (OPM) Flow is an open-source reservoir simulator used for solving subsurface porous media flow problems. Focus is placed here on carbon sequestration and the modeling of fluid flow within underground porous reservoirs. In this study, a sensitivity analysis of some input parameters for carbon sequestration is performed using six different uncertain parameters. An ensemble of model realizations is simulated using OPM Flow, and the model output is then calculated based on the values of the six input parameters mentioned above. CO injection is simulated for a period of 15 years, while the post-injection migration of CO in the saline storage aquifer is simulated for a subsequent period of 200 years, leading to a final analysis after 215 years. The input parameter values are generated using the quasi-Monte Carlo (QMC) method in the region of interest, following specified patterns suitable for analysis. The optimal convergence rate for quasi-Monte Carlo is observed. The aim of this study is to identify important input parameters contributing significantly to the model output, which is accomplished using sensitivity analysis and verified through symbolic regression modeling based on machine learning. Global sensitivity analysis using the Sobol sequence identifies input parameter 3, 'Permeability of shale between sand layers,' as having the most influence on the model output 'Secondary Trapped CO2.' All regression models, including the simplest and least accurate ones, incorporate parameter 3, confirming its significance. These approximations are valid within the designated area of interest for the input parameters and are easily interpretable for human experts. Sensitivity analysis of the developed time-dependent carbon sequestration model shows that the significance of each physical parameter changes over time: Sand porosity is more significant than shale permeability for roughly the first 120 years. Consequently, the presented results show that simulation timescales of at least 200 years are necessary for carbon sequestration evaluation.
开放多孔介质（OPM）流动是一种用于解决地下多孔介质流动问题的开源油藏模拟器。这里重点关注碳封存以及地下多孔油藏内流体流动的建模。在本研究中，使用六个不同的不确定参数对碳封存的一些输入参数进行了敏感性分析。使用OPM流动模拟了一组模型实现，然后根据上述六个输入参数的值计算模型输出。模拟了15年的CO注入过程，随后又模拟了200年盐水储存含水层中CO注入后的迁移情况，在215年后进行最终分析。输入参数值在感兴趣区域内使用准蒙特卡罗（QMC）方法生成，遵循适合分析的指定模式。观察到了准蒙特卡罗的最优收敛速度。本研究的目的是识别对模型输出有显著贡献的重要输入参数，这通过敏感性分析来完成，并通过基于机器学习的符号回归建模进行验证。使用索博尔序列的全局敏感性分析确定输入参数3，即“砂层间页岩的渗透率”，对模型输出“二次捕获的CO2”影响最大。所有回归模型，包括最简单和最不准确的模型，都包含参数3，证实了其重要性。这些近似值在输入参数的指定感兴趣区域内是有效的，并且易于人类专家解释。对所开发的时间相关碳封存模型的敏感性分析表明，每个物理参数随时间的重要性都会发生变化：在前120年左右，砂层孔隙度比页岩渗透率更重要。因此，所呈现的结果表明，碳封存评估至少需要200年的模拟时间尺度。
