Physics-informed machine learning with differentiable programming for heterogeneous underground reservoir pressure management.

Avoiding over-pressurization in subsurface reservoirs is critical for applications like CO[Formula: see text] sequestration and wastewater injection. Managing the pressures by controlling injection/extraction are challenging because of complex heterogeneity in the subsurface. The heterogeneity typically requires high-fidelity physics-based models to make predictions on CO[Formula: see text] fate. Furthermore, characterizing the heterogeneity accurately is fraught with parametric uncertainty. Accounting for both, heterogeneity and uncertainty, makes this a computationally-intensive problem challenging for current reservoir simulators. To tackle this, we use differentiable programming with a full-physics model and machine learning to determine the fluid extraction rates that prevent over-pressurization at critical reservoir locations. We use DPFEHM framework, which has trustworthy physics based on the standard two-point flux finite volume discretization and is also automatically differentiable like machine learning models. Our physics-informed machine learning framework uses convolutional neural networks to learn an appropriate extraction rate based on the permeability field. We also perform a hyperparameter search to improve the model's accuracy. Training and testing scenarios are executed to evaluate the feasibility of using physics-informed machine learning to manage reservoir pressures. We constructed and tested a sufficiently accurate simulator that is 400 000 times faster than the underlying physics-based simulator, allowing for near real-time analysis and robust uncertainty quantification.

Improving deep learning performance for predicting large-scale geological [Formula: see text] sequestration modeling through feature coarsening.

Physics-based reservoir simulation for fluid flow in porous media is a numerical simulation method to predict the temporal-spatial patterns of state variables (e.g. pressure p) in porous media, and usually requires prohibitively high computational expense due to its non-linearity and the large number of degrees of freedom (DoF). This work describes a deep learning (DL) workflow to predict the pressure evolution as fluid flows in large-scale 3-dimensional(3D) heterogeneous porous media. In particular, we develop an efficient feature coarsening technique to extract the most representative information and perform the training and prediction of DL at the coarse scale, and further recover the resolution at the fine scale by spatial interpolation. We validate the DL approach to predict pressure field against physics-based simulation data for a field-scale 3D geologic [Formula: see text] sequestration reservoir model. We evaluate the impact of feature coarsening on DL performance, and observe that the feature coarsening not only decreases the training time by [Formula: see text] and reduces the memory consumption by [Formula: see text], but also maintains temporal error [Formula: see text] on average. Besides, the DL workflow provides predictive efficiency with 1406 times speedup compared to physics-based numerical simulation. The key findings from this research significantly improve the training and prediction efficiency of deep learning model to deal with large-scale heterogeneous reservoir models, and thus it can also be further applied to accelerate workflows of history matching and reservoir optimization for close-loop reservoir management.

Gas diffusion through variably-water-saturated zeolitic tuff: Implications for transport following a subsurface nuclear event.

Noble gas transport through geologic media has important applications in the characterization of underground nuclear explosions (UNEs). Without accurate transport models, it is nearly impossible to distinguish between xenon signatures originating from civilian nuclear facilities and UNEs. Understanding xenon transport time through the earth is a key parameter for interpreting measured xenon isotopic ratios. One of the most challenging aspects of modeling gas transport time is accounting for the effect of variable water saturation of geological media. In this study, we utilize bench-scale laboratory experiments to characterize the diffusion of krypton, xenon, and sulfur hexafluoride (SF6) through intact zeolitic tuff under different saturations. We demonstrate that the water in rock cores with low partial saturation dramatically affects xenon transport time compared to that of krypton and SF6 by blocking sites in zeolitic tuff that preferentially adsorb xenon. This leads to breakthrough trends that are strongly influenced by the degree of the rock saturation. Xenon is especially susceptible to this phenomenon, a finding that is crucial to incorporate in subsurface gas transport models used for nuclear event identification. We also find that the breakthrough of SF6 diverges significantly from that of noble gases within our system. When developing field scale models, it is important to understand how the behavior of xenon deviates from chemical tracers used in the field, such as SF6 (Carrigan et al., 1996). These new insights demonstrate the critical need to consider the interplay between rock saturation and fission product sorption during transport modeling, and the importance of evaluating specific interactions between geomedia and gases of interest, which may differ from geomedia interactions with chemical tracers.

Beyond Barnwell: Applying lessons learned from the Barnwell site to other historic underground nuclear tests at Pahute Mesa to understand radioactive gas-seepage observations.

An underground nuclear explosion (UNE) generates radioactive gases that can be transported through fractures to the ground surface over timescales of hours to months. If detected, the presence of particular short-lived radionuclides in the gas can provide strong evidence that a recent UNE has occurred. By drawing comparisons between sixteen similar historical U.S. UNEs where radioactive gas was or was not detected, we identified factors that control the occurrence and timing of breakthrough at the ground surface. The factors that we evaluated include the post-test atmospheric conditions, local geology, and surface geology at the UNE sites. The UNEs, all located on Pahute Mesa on the Nevada National Security Site (NNSS), had the same announced yield range (20-150 kt), similar burial depths in the unsaturated zone, and were designed and performed by the same organization during the mid-to-late 1980s. Results of the analysis indicate that breakthrough at the ground surface is largely controlled by a combination of the post-UNE barometric pressure changes in the months following the UNE, and the volume of air-filled pore space above the UNE. Conceptually simplified numerical models of each of the 16 historical UNEs that include these factors successfully predict the occurrence (5 of the UNEs) or lack of occurrence (remaining 11 UNEs) of post-UNE gas seepage to the ground surface. However, the data analysis and modeling indicates that estimates of the meteorological conditions and of the post-UNE, site-specific subsurface environment including air-filled porosity, in combination, may be necessary to successfully predict late-time detectable gas breakthrough for a suspected UNE site.

Identification of dominant gas transport frequencies during barometric pumping of fractured rock.

We demonstrate that although barometric pressures are complicated signals comprised of numerous frequencies, it is a subset of these frequencies that drive the overwhelming majority of gas transport in fractured rock. Using an inverse numerical analysis, we demonstrate that a single barometric component with seasonally modulated amplitude approximates gas transport due to a measured barometric signal. If past barometric tendencies are expected to continue at a location, the identification of this frequency can facilitate accurate long term predictions of barometrically induced gas transport negating the need to consider stochastic realizations of future barometric variations. Additionally, we perform an analytical analysis that indicates that there is a set of barometric frequencies, consistent with the inverse numerical analysis, with high production efficiency. Based on the corroborating inverse numerical and analytical analyses, we conclude that there is a set of dominant gas transport frequencies in barometric records.

Identification of pumping influences in long-term water level fluctuations.

Identification of the pumping influences at monitoring wells caused by spatially and temporally variable water supply pumping can be a challenging, yet an important hydrogeological task. The information that can be obtained can be critical for conceptualization of the hydrogeological conditions and indications of the zone of influence of the individual pumping wells. However, the pumping influences are often intermittent and small in magnitude with variable production rates from multiple pumping wells. While these difficulties may support an inclination to abandon the existing dataset and conduct a dedicated cross-hole pumping test, that option can be challenging and expensive to coordinate and execute. This paper presents a method that utilizes a simple analytical modeling approach for analysis of a long-term water level record utilizing an inverse modeling approach. The methodology allows the identification of pumping wells influencing the water level fluctuations. Thus, the analysis provides an efficient and cost-effective alternative to designed and coordinated cross-hole pumping tests. We apply this method on a dataset from the Los Alamos National Laboratory site. Our analysis also provides (1) an evaluation of the information content of the transient water level data; (2) indications of potential structures of the aquifer heterogeneity inhibiting or promoting pressure propagation; and (3) guidance for the development of more complicated models requiring detailed specification of the aquifer heterogeneity.