Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Model reduction”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 109 records · Page 6
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Efficient Dimension Reduction of Complex Three-dimensional CO2 Saturation using Deep Learning Models

In the domain of deep learning (DL), dimension reduction is crucial for enhancing training efficiency and mitigating overfitting, particularly when managing complex data such as three-dimensional (3D) saturation data. The 3D saturation data in the context of geological carbon storage (GCS) presents unique challenges due to its inherent sparsity and the abrupt transitions at plume boundaries, known as shock fronts. To address the challenges, we proposed a novel DL framework that integrates dimension reduction with advanced 3D reconstruction techniques. Our model leveraged latent variables derived from 2D average saturation data, offering a robust and efficient solution tailored to the intricate dynamics of 3D saturation fields. The proposed framework can extract the critical features of the high-dimensional data while reducing the variable numbers, which is more tractable for DL models and enhances the model robustness and accuracy. Therefore, it provides a novel approach for modeling and analyses in complex geological scenarios, which finds great potential applications in environmental monitoring and energy storage.

Wang, Hongsheng↗
DOE OSTI2024

Assessment of Technoeconomic Opportunities in Automation for Nuclear Microreactors

Achieving full decarbonization of all economic sectors remains a challenge, especially in niche markets. For example, remote communities and industrial or mining activities detached from the main electric grid heavily rely on fossil fuels, similar to urban and industrial microgrids with combined heat and power needs. A combination of renewables and energy storage is often not suitable due to cost, reliability, intermittency, and large storage requirements. Small nuclear reactors with a flexible purpose could serve these applications. Microreactors (MR) are a class of reactors that are compact, factory manufactured, transportable, and self-regulating. Typically, they generate much less power than their large reactor counterparts. The main advantages of microreactors include the versatile nature of the energy produced, the reliability of supply, and freedom from having to transport and store large quantities of fuels on-site, coupled with the absence of dependence on an electrical grid. A strong business case is needed to move from the microreactor prototype to the commercialization phase. In fact, fossil fuels are still relatively inexpensive, and in the near term, carbon credits will be available to virtually compensate for emissions. For microreactors, one of the main costs in operation and maintenance (O&M) is their staffing levels. In this study, we investigate how to optimize the number (and thus the cost) of workers, moving from a traditional, fully manned, on-site personnel approach to an unmanned, remote personnel approach. We examine four different staffing models that can be implemented as the technology matures and evolves. We estimate the staffing needs of each model and build a business case to justify the substitution of on-site personnel with adequate technologies. To do so, we propose a cost model to quantify potential cost reductions from automating O&M activities. The model accounts for both the reduction in cost derived from the reduced number of full-time-equivalent (FTE) employees and the increase in cost derived from the need to buy new control hardware as needed. Applying the cost model that we created to different scenarios, an on-site O&M cost reduction exceeding 80% can be expected. Additionally, we found that it is more impactful to focus on automating routine O&M tasks rather than attempting to automate transient management (shutdowns, restarts, monitoring condition deviations). In fact, transients typically account for less than 1% of the total FTE time spent on the reactors.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗
DOE OSTI2026

Optimal Operation of Residential High Performance Water Heater for Reduction of Electricity Cost and Peak Demand Through Field Validation

Water heating accounts for about 18% of a typical US home’s energy use. Modern water heaters have enabled control options through APIs, offering customers the opportunity to reduce their energy cost and peak demand by dynamically adjusting settings. A water heater’s capacity to store energy using its storage tank makes it an asset for peak demand reduction and energy cost savings. For this reason, a mixed-integer linear programming model is proposed to minimize the energy cost of a high-performance water heater while also reducing the peak demand of the residential household under a time-of-use utility rate by dynamically changing the water heater’s running mode. Specifically, a multi-objective optimization model is formulated to determine the mode settings of the water heater considering hot water use, time-of-use rate, and peak demand limit of the residential household. The mode settings are associated with different dead bands of water temperature for triggering on/off action of the heat pump and heating element. A 66-gal hybrid electric high performance water heater was used for numerical simulation and practical experiments. The simulation results were well aligned with measurements of practical experiments, validating the soundness of the thermodynamic model. In addition, reductions of energy cost, enabling affordability, and reducing peak demand are demonstrated. The research team also developed a software framework with dashboards to automatically and continuously monitor and manage devices.

Liu, Guodong [ORNL] (ORCID:0000000213498608)↗
DOE OSTI2026

Multi-Scale Modeling and Prototype Development for Electrochemical CO2 Reduction (CRADA Final Report)

In this CRADA project, Lawrence Livermore National Laboratory, Stanford University, SLAC National Laboratory, and TotalEnergies collaboratively executed a multidisciplinary investigation of electrochemical reduction of CO2 to produce sustainable fuels and chemicals. Overall, the project led to an increased understanding of the fundamental processes involved in CO2 electrolysis, from the atomistic scale to the full electrolyzer device scale, ultimately leading to design guidelines for CO2 electrolyzers that will help in their future commercialization. As the model systems, Ag- and Cu-based catalysts were investigated in various forms depending on the electrochemical platform that was utilized to study the activity, selectivity, and durability towards electrochemical CO2 reduction. By employing experimental, theoretical, and computational techniques, the project team experimentally validated multi-physics models, evaluated the experimental levers that lead to increased electrolyzer reaction selectivity and energy efficiency, and used computational optimization to design higher performance electrodes. The learnings of this project were extensively documented in publicly available peer-reviewed journal publications and conference presentations, which serve as a foundation for further work to build from.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2025

Coupling flux balance analysis with reactive transport modeling through machine learning for rapid and stable simulation of microbial metabolic switching

Integrating genome-scale metabolic networks with reactive transport models (RTMs) provides a detailed description of the dynamic changes in microbial growth and metabolism. Despite promising demonstrations in the past, computational inefficiency has been pointed out as a critical issue to overcome because it requires repeated application of linear programming (LP) to obtain flux balance analysis (FBA) solutions in every time step and spatial grid. To address this challenge, we propose a new simulation method where we train and validate artificial neural networks (ANNs) using randomly sampled FBA solutions and incorporate the resulting surrogate FBA model (represented as algebraic equations) into RTMs as source/sink terms. We demonstrate the efficiency of our method via a case study of Shewanella oneidensis MR-1. During aerobic growth on lactate, S. oneidensis produces metabolic byproducts (such as pyruvate and acetate), which are subsequently consumed as alternative carbon sources when the preferred nutrients are depleted. To effectively simulate these complex dynamics, we used a cybernetic approach that models metabolic switches as the outcome of dynamic competition among multiple growth options. In both zero-dimensional batch and one-dimensional column configurations, the ANN-based surrogate models achieved substantial reduction of computational time by several orders of magnitude compared to the original LP-based FBA models. Moreover, the ANN models produced robust solutions without any special measures to prevent numerical instability. These developments significantly promote our ability to utilize genome-scale networks in complex, multi-physics, and multi-dimensional ecosystem modeling.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2024

ReEDS Performance Improvement

The Regional Energy Deployment System (ReEDS) is an open-source, spatially explicit, long-term capacity expansion model for the bulk electric power system of the contiguous United States, encompassing multiple scenarios with technological and political assumptions (see https://github.com/NREL/ReEDS-2.0). With the increased needs for capabilities, higher temporal and spatial resolutions to model the evolution of the power system with modern technologies and low-carbon pathways, ReEDS' model solution times have increased significantly from 4-6 hours in 2018 to 18-48+ hours in 2023 . Also, the model size for commonly-run ReEDS scenarios reached 22 and 28 million equations and variables, respectively. These runtimes can be especially challenging under certain scenario settings (e.g., very high temporal or spatial resolution) or with limited computational power. In this presentation, we will discuss several methods we used to improve model runtime, including data preparation, model modification, and solver tuning. The implementation of these methods shrank the model size to 7.2 and 7.3 million equations and variables, respectively. Furthermore, this led to a 77% reduction in the model's run time for commonly-run ReEDS scenarios. We will discuss the process of identifying areas for solve time improvements and how the specific enhancements for the ReEDS model might be applied to other similar large-scale models.

ENERGY PLANNING, POLICY, AND ECONOMY,MATHEMATICS A↗
DOE OSTI2024

Reductive Analysis with Compiler-Guided Large Language Models for Input-Centric Code Optimizations

Input-centric program optimization aims to optimize code by considering the relations between program inputs and program behaviors. Despite its promise, a long-standing barrier for its adoption is the difficulty of automatically identifying critical features of complex inputs. This paper introduces a novel technique, reductive analysis through compiler-guided Large Language Models (LLMs), to solve the problem through a synergy between compilers and LLMs. It uses a reductive approach to overcome the scalability and other limitations of LLMs in program code analysis. The solution, for the first time, automates the identification of critical input features without heavy instrumentation or profiling, cutting the time needed for input identification by 44× (or 450× for local LLMs), reduced from 9.6 hours to 13 minutes (with remote LLMs) or 77 seconds (with local LLMs) on average, making input characterization possible to be integrated into the workflow of program compilations. Optimizations on those identified input features show similar or even better results than those identified by previous profiling-based methods, leading to optimizations that yield 92.6% accuracy in selecting the appropriate adaptive OpenMP parallelization decisions, and 20-30% performance improvement of serverless computing while reducing resource usage by 50-60%.

Input-Centric Optimization↗
DOE OSTI2025

Mechanistic Multiscale Uncertainty Propagation in Support of Accelerated Fuel Qualification

Taking a nuclear fuel concept through the research, development, and qualification stages has historically taken on the order of 20 to 25 years because of extensive irradiation tests required for a variety of conditions. The concept of accelerated fuel qualification (AFQ) has been proposed to increase the innovation pace for nuclear fuels. The goal of AFQ is not to replace the traditional qualification approach but rather to reduce the total number of experiments required to ensure approval from the regulatory authority. Of the many AFQ approaches being explored, advanced modeling—and, in particular, mechanistic modeling—is in a uniquely cross-cutting position to reduce the number of required integral tests through the inclusion of separate-effects testing, while helping to extrapolate reactor performance during rare events. We make the case that propagation of uncertainty through various computational length scales helps contextualize mechanistic modeling. We will utilize UO 2 fission gas diffusion predictions from the atomistically informed cluster dynamics code Centipede to inform fuel performance rodlet simulations using the BISON finite element code as the metric for showing how multiscale mechanistic uncertainty quantification can help reduce uncertainty in fuel performance. In conclusion, by quantifying uncertainty and its reduction through multiscale modeling, the qualification process may be accelerated through the reduction of costly irradiation experiments.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2025

Low‐dimensional manifold learning for uncertainty quantification in complex multi‐scale stochastic systems

Broadly speaking, the goals of the project are to develop techniques to use manifold learning to develop reduced‐order and surrogate models for "hyper‐reduction" of very high‐dimensional complex multi‐scale systems. This is being achieved by employing a newly proposed form of manifold projection and learning that leverages recent advancements in computational geometry and data‐driven modeling. In particular, we are applying a manifold projection technique to project the solutions of very high‐dimensional systems onto the so‐called Grassmannmanifold, a Reimannian manifold comprised of orthonormal matrices. We then apply data‐driven machine learning techniques to classify the solutions on the manifold (e.g. clustering techniques) according to their proximity on the manifold and leverage a further nonlinear dimension reduction to organize the structured data on the manifold. Finally, we are developing novel techniques that enable us to directly interpolate the hyper‐reduced data such that we can predict the solution of the complex, high‐ dimensional system without need to call the full expensive computational model. Given their adherence to the underlying structure of the solution of the physical system, it is expected that these approximate solutions will be sufficiently constrained so as to (approximately) adhere to physical principles.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Redox Poise during Rhodospirillum rubrum Phototrophic Growth Drives Large-scale Changes in Macromolecular Synthesis Pathways

During photoheterotrophic growth on organic substrates, purple nonsulfur photosynthetic bacteria like Rhodospirillum rubrum can acquire electrons by multiple means, including oxidation of organic substrates, oxidation of inorganic electron donors (e.g., H2), and by reverse electron flow from the photosynthetic electron transport chain. These electrons are stored as reduced electron-carrying cofactors (e.g., NAD(P)H and ferredoxin). The overall ratio of oxidized to reduced cofactors (e.g., NAD(P)+:NAD(P)H), or ’redox poise’, is difficult to understand or predict, as are the cellular processes for dissipating these reducing equivalents. Using physics-based models that capture mass action kinetics consistent with the thermodynamics of reactions and pathways, a range of redox conditions for heterophototrophic growth are evaluated, from conditions in which the NADP+/NADPH levels approach thermodynamic equilibrium to conditions in which the NADP+/NADPH ratio is far above the typical physiological values. Modeling predictions together with experimental measurements indicate that the redox poise of the cell results in large-scale changes in the activity of biosynthetic pathways and, thus, changes in cell macromolecule levels (DNA, RNA, proteins, and fatty acids). Furthermore, modeling predictions indicate that during phototrophic growth, reverse electron flow from the quinone pool is a minor contributor to the production of reduced cofactors (e.g., NAD(P)H) compared to other oxidative processes (H2 and carbon substrate oxidation). Instead, the quinone pool primarily operates to aid ATP production. The high level of ATP, in turn, drives reduction processes even when NADPH levels are relatively low compared to NADP+ by coupling ATP hydrolysis to the reductive processes. The model, in agreement with experimental measurements of macromolecule ratios of cells growing on different carbon substrates, indicates that the dynamics of nucleotide versus lipid and protein production is likely a significant mechanism of balancing oxidation and reduction in the cell.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2025

Redox poise in R. rubrum phototrophic growth drives large-scale changes in macromolecular pathways

During photoheterotrophic growth on organic substrates, purple nonsulfur photosynthetic bacteria like Rhodospirillum rubrum can acquire electrons by multiple means, including oxidation of organic substrates, oxidation of inorganic electron donors (e.g., H 2 ), and by reverse electron flow from the photosynthetic electron transport chain. These electrons are stored as reduced electron-carrying cofactors (e.g., NAD(P)H and ferredoxin). The overall ratio of oxidized to reduced cofactors (e.g., NAD(P)+:NAD(P)H), or ’redox poise’, is difficult to understand or predict, as are the cellular processes for dissipating these reducing equivalents. Using physics-based models that capture mass action kinetics consistent with the thermodynamics of reactions and pathways, a range of redox conditions for heterophototrophic growth are evaluated, from conditions in which the NADP+/NADPH levels approach thermodynamic equilibrium to conditions in which the NADP+/NADPH ratio is far above the typical physiological values. Modeling predictions together with experimental measurements indicate that the redox poise of the cell results in large-scale changes in the activity of biosynthetic pathways and, thus, changes in cell macromolecule levels (DNA, RNA, proteins, and fatty acids). Furthermore, modeling predictions indicate that during phototrophic growth, reverse electron flow from the quinone pool is a minor contributor to the production of reduced cofactors (e.g., NAD(P)H) compared to other oxidative processes (H 2 and carbon substrate oxidation). Instead, the quinone pool primarily operates to aid ATP production. The high level of ATP, in turn, drives reduction processes even when NADPH levels are relatively low compared to NADP+ by coupling ATP hydrolysis to the reductive processes. The model, in agreement with experimental measurements of macromolecule ratios of cells growing on different carbon substrates, indicates that the dynamics of nucleotide versus lipid and protein production is likely a significant mechanism of balancing oxidation and reduction in the cell.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2025

Evaluating GHG Mitigation Potential from ESPC Projects [Slides]

This report explores the impact of implemented ESPC projects on the projected greenhouse gas (GHG) emissions reductions in the U.S. buildings sector, and the associated projected annual cost savings and marginal abatement costs. It is important to investigate the role that energy retrofits can play in achieving GHG emission reduction targets given the use of ESPC by public agencies to reduce emissions from energy use in addition to the historical use of ESPC to to achieve cost/energy savings and ancillary benefits (e.g., addressing deferred maintenance, aging infrastructure, etc.). The analysis draws from LBNL’s eProject Builder (ePB) database, which contains approximately 3,000 energy retrofit projects implemented by energy service companies (ESCOs), mostly energy savings performance contract (ESPC) projects. This is the first report to model GHG emission reductions from ongoing ESPC projects in the U.S public sector based on ePB project data. It is intended for policymakers, federal, state and local government officials and other potential ESPC customers, the ESCO industry, researchers, and other energy policy professionals.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗
DOE OSTI2025

A machine-learning-driven data labeling pipeline for scientific analysis in MLExchange

This study introduces a novel labeling pipeline to accelerate the labeling process of scientific data sets by using artificial intelligence (AI)-guided tagging techniques. This pipeline includes a set of interconnected web-based graphical user interfaces (GUIs), where Data Clinic and MLCoach enable the preparation of machine learning (ML) models for data reduction and classification, respectively, while Label Maker is used for label assignment. Throughout this pipeline, data can be accessed through a direct connection to a file system or through Tiled for access through Hypertext Transfer Protocol (HTTP). Our experimental results present three use cases where this labeling pipeline has been instrumental for the study of large X-ray scattering data sets in the area of pattern recognition, the remote analysis of resonant soft X-ray scattering data and the fine-tuning process of foundation models. These use cases highlight the labeling capabilities of this pipeline, including the ability to label large data sets in a short period of time, to perform remote data analysis while minimizing data movement and to enhance the fine-tuning process of complex ML models with human involvement.

Chavez, Tanny (ORCID:0000000193172896)↗
DOE OSTI2024

Overview of ASDEX upgrade results in view of ITER and DEMO

Experiments on ASDEX Upgrade (AUG) in 2021 and 2022 have addressed a number of critical issues for ITER and EU DEMO. A major objective of the AUG programme is to shed light on the underlying physics of confinement, stability, and plasma exhaust in order to allow reliable extrapolation of results obtained on present day machines to these reactor-grade devices. Concerning pedestal physics, the mitigation of edge localised modes (ELMs) using resonant magnetic perturbations (RMPs) was found to be consistent with a reduction of the linear peeling-ballooning stability threshold due to the helical deformation of the plasma. Conversely, ELM suppression by RMPs is ascribed to an increased pedestal transport that keeps the plasma away from this boundary. Candidates for this increased transport are locally enhanced turbulence and a locked magnetic island in the pedestal. The enhanced D-alpha (EDA) and quasi-continuous exhaust (QCE) regimes have been established as promising ELM-free scenarios. Here, the pressure gradient at the foot of the H-mode pedestal is reduced by a quasi-coherent mode, consistent with violation of the high-n ballooning mode stability limit there. This is suggestive that the EDA and QCE regimes have a common underlying physics origin. In the area of transport physics, full radius models for both L- and H-modes have been developed. These models predict energy confinement in AUG better than the commonly used global scaling laws, representing a large step towards the goal of predictive capability. A new momentum transport analysis framework has been developed that provides access to the intrinsic torque in the plasma core. In the field of exhaust, the X-Point Radiator (XPR), a cold and dense plasma region on closed flux surfaces close to the X-point, was described by an analytical model that provides an understanding of its formation as well as its stability, i.e., the conditions under which it transitions into a deleterious MARFE with the potential to result in a disruptive termination. With the XPR close to the divertor target, a new detached divertor concept, the compact radiative divertor, was developed. Here, the exhaust power is radiated before reaching the target, allowing close proximity of the X-point to the target. No limitations by the shallow field line angle due to the large flux expansion were observed, and sufficient compression of neutral density was demonstrated. With respect to the pumping of non-recycling impurities, the divertor enrichment was found to mainly depend on the ionisation energy of the impurity under consideration. In the area of MHD physics, analysis of the hot plasma core motion in sawtooth crashes showed good agreement with nonlinear 2-fluid simulations. This indicates that the fast reconnection observed in these events is adequately described including the pressure gradient and the electron inertia in the parallel Ohm’s law. Concerning disruption physics, a shattered pellet injection system was installed in collaboration with the ITER International Organisation. Thanks to the ability to vary the shard size distribution independently of the injection velocity, as well as its impurity admixture, it was possible to tailor the current quench rate, which is an important requirement for future large devices such as ITER. Progress was also made modelling the force reduction of VDEs induced by massive gas injection on AUG. The H-mode density limit was characterised in terms of safe operational space with a newly developed active feedback control method that allowed the stability boundary to be probed several times within a single discharge without inducing a disruptive termination. Regarding integrated operation scenarios, the role of density peaking in the confinement of the ITER baseline scenario (high plasma current) was clarified. The usual energy confinement scaling ITER98(p,y) does not capture this effect, but the more recent H20 scaling does, highlighting again the importance of developing adequate physics based models. Advanced tokamak scenarios, aiming at large non-inductive current fraction due to non-standard profiles of the safety factor in combination with high normalised plasma pressure were studied with a focus on their access conditions. A method to guide the approach of the targeted safety factor profiles was developed, and the conditions for achieving good confinement were clarified. Based on this, two types of advanced scenarios (‘hybrid’ and ‘elevated’

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2026

The AWAKEN wind farm benchmark, Part 2: Modeling results

Accurately modeling wind farm performance in complex atmospheric flows remains a challenge. This paper presents the modeling results of the American WAKE experimeNt (AWAKEN) wind farm benchmark, a collaborative effort involving 16 research groups from academia and industry within the International Energy Agency Wind Technology Collaboration Programme Task 57. The study evaluates a diverse suite of simulation tools, ranging from fast-running engineering wake models to high-fidelity large-eddy simulations, against a diurnal case study observed during the AWAKEN campaign. The benchmark utilized a three-phase structure to progressively assess model performance as observational data availability increased. Initial blind predictions showed that higher-fidelity models did not uniformly outperform simpler simulation tools. A distinct spatial bias was observed where models struggled to resolve the interplay between a low-level jet, wakes, and terrain-induced flow acceleration. In subsequent phases, leveraging additional measurements for model improvement led to a reduction in mean absolute error across the model ensemble; however, this effect was most pronounced in engineering wake models, where targeted calibration reduced error by up to 40~\%. Overall, the study demonstrates that inflow characterization remains a primary prerequisite for accuracy, particularly for models relying on coarse forcing datasets. While the limited ability to resolve local terrain-flow interactions under single-day conditions represent a recognized constraint, the overall findings on wake modeling and real-world validation still provide valuable guidance for model application and for mitigating this limitation.

Bodini, Nicola↗
DOE OSTI2024

Modeling and Analysis of a Heat Pump Clothes Dryer with Thermal Energy Storage

Clothes drying is an energy-intensive process that causes significant electricity consumption and carbon emissions in the US. Approximately 83% of Households in the US own a tumble clothes dryer at home and 80% of dryers are electrical resistance dryers with low energy efficiency. Heat pump technology makes it possible for highly efficient and clean drying. Additionally, the ventless design of heat pump clothes dryers (HPCD) provides more installation flexibility. HPCD involves three primary mediums: wet clothes, a closed air loop, and a refrigerant circuit. The evaporator is used to dehumidify the wet air and the condenser is used to re-heat the dry air. One of the critical technological barriers to HPCD market penetration is its long drying time, primarily due to the relatively low discharging temperature and the slow response during the warm-up period. In this study, the thermal energy storage (TES) technology is adopted to address this challenge by providing pre-heating of air prior to the condenser to increase the operating temperature of the process air. The heat pump will charge the phase change material (PCM) in the TES device with heating energy during clothes washing and the PCM will discharge the stored heat to facilitate air heating during clothes drying. To analyze the optimal design and potential for energy saving and drying time reduction, a mathematical model of the HPCD system was developed. The HPCD is a highly dynamic system with a coupled heat and mass transfer and heat pumping cycle. This paper provides solutions to simulate the transient behavior of the system while maintaining low computational cost. The modeling result indicates reduced energy consumption and drying time by integrating TES with HPCD, as compared to electrical resistance dryers. The study can provide significant insights into improving building flexibility with TES and smart appliances.

Liu, Xiaoli↗
DOE OSTI2024

CO 2 storage site characterization using ensemble-based approaches with deep generative models

Estimating spatially distributed properties such as permeability from available sparse measurements is a great challenge in efficient subsurface CO 2 storage operations. In this paper, a deep generative model that can accurately capture complex subsurface structure is tested with an ensemble-based inversion method for accurate and accelerated characterization of CO 2 storage sites. We chose Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP) for its realistic reservoir property representation and Ensemble Smoother with Multiple Data Assimilation (ES-MDA) for its robust data fitting and uncertainty quantification capability. WGAN-GP are trained to generate high-dimensional permeability fields from a low-dimensional latent space and ES-MDA then updates the latent variables by assimilating available measurements. Several subsurface site characterization examples including Gaussian, channelized, and fractured reservoirs are used to evaluate the accuracy and computational efficiency of the proposed method and the main features of the unknown permeability fields are characterized accurately with reliable uncertainty quantification. Furthermore, the estimation performance is compared with a widely-used variational, i.e., optimization-based, inversion approach, and the proposed approach outperforms the variational inversion method in several benchmark cases. We explain such superior performance by visualizing the objective function in the latent space: because of nonlinear and aggressive dimension reduction via generative modeling, the objective function surface becomes extremely complex while the ensemble approximation can smooth out the multi-modal surface during the minimization. This suggests that the ensemble-based approach works well over the variational approach when combined with deep generative models at the cost of forward model runs unless convergence-ensuring modifications are implemented in the variational inversion.

42 ENGINEERING↗