Search NASA⌕ Search

SEARCH · Search NASA

Results for “Heterogeneous inputs”

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 19 records
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

D2NO: Efficient handling of heterogeneous input function spaces with distributed deep neural operators

Neural operators have been applied in various scientific fields, such as solving parametric partial differential equations, dynamical systems with control, and inverse problems. However, challenges arise when dealing with input functions that exhibit heterogeneous properties, requiring multiple sensors to handle functions with minimal regularity. To address this issue, discretization-invariant neural operators have been used, allowing the sampling of diverse input functions with different sensor locations. However, existing frameworks still require an equal number of sensors for all functions. We propose a novel distributed approach to further relax the discretization requirements and solve the heterogeneous dataset challenges. Our method involves partitioning the input function space and processing individual input functions using independent and separate neural networks. A centralized neural network is used to handle shared information across all output functions. This distributed methodology reduces the number of gradient descent back-propagation steps, improving efficiency while maintaining accuracy. Here, we demonstrate that the corresponding neural network is a universal approximator of continuous nonlinear operators and present three numerical examples to validate its performance.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Data and Code for Understanding Generative AI Content with Embedding Models

This repository contains code for the experiments in the paper "Understanding Generative AI Content with Embedding Models". Constructing high-quality features is critical to any quantitative data analysis. While feature engineering was historically addressed by carefully hand-crafting data representations based on domain expertise, deep neural networks (DNNs) now offer a radically different approach. DNNs implicitly engineer features by transforming their input data into hidden feature vectors called embeddings. For embedding vectors produced by foundation models -- which are trained to be useful across many contexts -- we demonstrate that simple and well-studied dimensionality-reduction techniques such as Principal Component Analysis uncover inherent heterogeneity in input data concordant with human-understandable explanations. Of the many applications for this framework, we find empirical evidence that there is intrinsic separability between real samples and those generated by artificial intelligence (AI).

Vargas, Max [Pacific Northwest National Laboratory↗
DOE OSTI2024

Integrating multi-modal remote sensing, deep learning, and attention mechanisms for yield prediction in plant breeding experiments

In both plant breeding and crop management, interpretability plays a crucial role in instilling trust in AI-driven approaches and enabling the provision of actionable insights. The primary objective of this research is to explore and evaluate the potential contributions of deep learning network architectures that employ stacked LSTM for end-of-season maize grain yield prediction. A secondary aim is to expand the capabilities of these networks by adapting them to better accommodate and leverage the multi-modality properties of remote sensing data. In this study, a multi-modal deep learning architecture that assimilates inputs from heterogeneous data streams, including high-resolution hyperspectral imagery, LiDAR point clouds, and environmental data, is proposed to forecast maize crop yields. The architecture includes attention mechanisms that assign varying levels of importance to different modalities and temporal features that, reflect the dynamics of plant growth and environmental interactions. The interpretability of the attention weights is investigated in multi-modal networks that seek to both improve predictions and attribute crop yield outcomes to genetic and environmental variables. This approach also contributes to increased interpretability of the model's predictions. The temporal attention weight distributions highlighted relevant factors and critical growth stages that contribute to the predictions. The results of this study affirm that the attention weights are consistent with recognized biological growth stages, thereby substantiating the network's capability to learn biologically interpretable features. Accuracies of the model's predictions of yield ranged from 0.82-0.93 R 2 ref in this genetics-focused study, further highlighting the potential of attention-based models. Further, this research facilitates understanding of how multi-modality remote sensing aligns with the physiological stages of maize. The proposed architecture shows promise in improving predictions and offering interpretable insights into the factors affecting maize crop yields, while demonstrating the impact of data collection by different modalities through the growing season. By identifying relevant factors and critical growth stages, the model's attention weights provide valuable information that can be used in both plant breeding and crop management. The consistency of attention weights with biological growth stages reinforces the potential of deep learning networks in agricultural applications, particularly in leveraging remote sensing data for yield prediction. To the best of our knowledge, this is the first study that investigates the use of hyperspectral and LiDAR UAV time series data for explaining/interpreting plant growth stages within deep learning networks and forecasting plot-level maize grain yield using late fusion modalities with attention mechanisms.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2024

Improved Fuel Cycle Capability of Griffin for Fast Reactor Applications

Griffin is a MOOSE based reactor multiphysics analysis application jointly developed by Idaho National Laboratory and Argonne National Laboratory under the Department of Energy Office of Nuclear Energy Nuclear Energy Advanced Modeling and Simulation Program. This fiscal year, the fuel cycle capability has been significantly extended by improving the assembly shuffling option to allow flexible fuel reloading in the multi-cycle depletion calculation and incorporating decay between cycles. An equilibrium core calculation capability was also implemented to find an equilibrium core. Additionally, an enrichment search capability was added to determine the enrichment condition that allows a core to reach an equilibrium cycle with the end-of-cycle k-effective meeting a user-specified target value. The updated fuel management capability has been extensively tested using the three-dimensional ABTR problem with different batch schemes, exhibiting reasonable solutions in terms of manual shuffling, equilibrium cycle, and enrichment search calculations. The cross section generation workflow capability for fast reactors was further verified to produce microscopic cross sections as well as Griffin core inputs for an ring-heterogeneous configuration. Rigorous verification tests using the ABTR problem demonstrated that the RH core calculations, with cross sections and Griffin inputs generated from the cross section workflow of Griffin, produced accurate solution for fast reactor problems. In addition, an option to convert delay neutron parameter data generated from MC 2 -3 in the DLAYXS format into XML format was added to support transient calculations using MC 2 -3-generated data.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2025

Diel partitioning in microbial phosphorus acquisition in the Sargasso Sea

The daily cycle of photosynthetic primary production at the base of marine food webs is often limited by the availability of scarce nutrients. Microbial competition for these scarce resources can be alleviated insofar as the intensity of nutrient uptake and assimilation activities are distributed heterogeneously across organisms over periodic input cycles. Recent analysis of community transcriptional dynamics in the nitrogen-limited subtropical North Pacific gyre revealed evidence of temporal partitioning of nitrogen uptake and assimilation between eukaryotic phytoplankton, cyanobacteria, and heterotrophic bacteria over day-night cycles. Here, we present results from a Lagrangian metatranscriptomic time series survey in the Sargasso Sea and demonstrate temporally partitioned phosphorus uptake in this phosphorus-limited environment. In the Sargasso, heterotrophic bacteria, eukaryotic phytoplankton, and cyanobacteria express genes for phosphorus assimilation during the morning, day, and dusk, respectively. In conclusion, these results support the generality of temporal niche partitioning as an emergent mechanism that can structure uptake of limiting nutrients and facilitate coexistence of diverse microbes in open ocean ecosystems.

biological oceanography↗
DOE OSTI2024

Sensitivity of mesoscale modeling to urban morphological feature inputs and implications for characterizing urban sustainability

We examine the differences in meteorological output from the Weather Research and Forecasting (WRF) model run at 270 m horizontal resolution using 10 m, 100 m and 1 km resolution 3D neighborhood morphological inputs and with no morphological inputs. We find that the spatial variability in temperature, humidity, and other meteorological variables across the city can vary with the resolution and the coverage of the 3D urban morphological input, and that larger differences occur between simulations run without 3D morphological input and those run with some type of 3D morphology. We also find that the inclusion of input-building-defined roughness length calculations would improve simulation results further. We show that these inputs produce different patterns of heat wave spatial heterogeneity across the city of Washington, DC. These findings suggest that understanding neighborhood level urban sustainability under extreme heat waves, especially for vulnerable neighborhoods, requires attention to the representation of surface terrain in numerical weather models.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2025

Future Intensity‐Duration‐Frequency Curves of Extreme Precipitation in the Midwest United States From Convection‐Permitting Modeling

Abstract During the last four decades, global warming has statistically significant intensified extreme precipitation events in the Midwestern United States (defined here as the region covering Illinois, Indiana, Ohio, and Kentucky), leading to increased risks to human life, property, and infrastructure. To enable climate change adaptation and resilience across various economic and social sectors in this region, updated information about future climate changes, specifically at finer spatial scales, is essential. Leveraging a new 150‐year dynamical downscaling data set at convection‐permitting resolution, this study introduces a framework to construct the projected future intensity‐duration‐frequency (IDF) curves of heavy precipitation, which are prominent tools for infrastructure design and water resources management. This framework generates IDF curves at both sub‐daily and multi‐day duration utilizing hourly in situ observations as well as quantile‐based statistical techniques in bias‐correction and return levels selection. The assumption of non‐stationarity in the distribution parameter fitting process is also implemented in this workflow. Compared to historical IDF curves for 1980–2022, future projected IDF curves for 2058–2100 under Representative Concentration Pathway (RCP) 4.5 and RCP 8.5 scenarios indicate an average intensity increase of approximately 15% and 25%, respectively, across 74 stations, considering both annual and seasonal timescales. Future projections suggest that extreme precipitation events may become more severe across six investigated return periods, with longer return periods showing a greater increase. The frequency of future extreme precipitation events in the Midwest region is also projected to double. Furthermore, current results reveal spatial heterogeneity of future trends across stations owing to the high‐resolution input data set. Plain Language Summary This study investigates the evolving nature of extreme precipitation events in the Midwestern United States under a changing climate. By leveraging a high‐resolution dynamical downscaling data set, we construct projected intensity‐duration‐frequency (IDF) curves for future extreme rainfall events. These curves serve as vital tools for infrastructure planning and water resource management. Our analysis reveals a significant increase in both the intensity and frequency of extreme precipitation events in the region. Future projected IDF curves for the late century indicate an average intensity increase of approximately 15%–25% compared to historical values. Moreover, the frequency of such events is expected to double. Spatial heterogeneity in future trends is observed across different stations within the Midwest, highlighting the importance of high‐resolution modeling in capturing localized climate variability. These findings underscore the urgent need for climate adaptation strategies to mitigate the increasing risks associated with extreme precipitation events in the region. Key Points This study introduces a workflow to construct future intensity‐duration‐frequency (IDF) curves over the Midwest United States using a new convection‐permitting modeling data set The current IDF construction workflow reproduces well the historical observed IDF 30 curves in summer months with median relative errors of 2.4% among 74 stations and 6 investigated durations The projected IDF curves show diverse future trends of extreme precipitation across stations, with intensity increases of approximately 15% and 25% under RCP4.5 and RCP8.5 climate scenarios, respectively, and a doubling of frequency on average

Nguyen, Trung↗
DOE OSTI2024

In Situ TEM for Structural and Chemical Evolutions of Bimetallic Pt–Ni Nanoparticles at Elevated Temperatures: Implications for Heterogeneous Catalysis

Platinum-based bimetallic nanoparticles (NPs) are of great interest for their applications in catalysis. The catalytic properties of these NPs are significantly dependent on their morphology, structure, and composition, whose response to thermal input remains challenging to be fully understood. This study investigates the thermally induced structural and chemical evolutions of single-crystalline Pt-Ni NPs using in situ transmission electron microscopy. The observed morphological evolution includes the facet development from a truncated octahedron to a spherical-like isotropic shape, followed by the formation of pancake-like ellipsoid shape at high temperatures due to surface atom migration and interfacial wetting enabled by the particle-substrate interaction. Comparative investigations by in situ scanning transmission electron microscopy with energy dispersive X-ray spectroscopy mapping elucidate that the solid-solution compositional configuration can be retained over a large temperature range, while core-shell NPs undergo irreversible solid-solution transitions through chemical homogenization at elevated temperatures. Finally, these findings elucidate the effect of thermal input on structural evolution and compositional redistribution of Pt-Ni bimetallic NPs, offering valuable insights into the design of heterogeneous catalysts.

25 ENERGY STORAGE↗
DOE OSTI2024

3-D Geological Modeling for Numerical Flow Simulation Studies of Gas Hydrate Reservoirs at the Kuparuk State 7-11-12 Pad in the Prudhoe Bay Unit on the Alaska North Slope

Accurate reservoir evaluation requires reliable three-dimensional (3-D) geological models. Here, this study conducted 3-D geological modeling for numerical flow simulation of the B1 sand gas hydrate reservoir at the Kuparuk State 7-11-12 pad, Prudhoe Bay Unit, Alaska North Slope. The model integrates well logs, core, and seismic data to address spatial heterogeneity in geological structures and reservoir properties. Two modeling types were performed: structural framework modeling and petrophysical property modeling. For structural framework modeling, seismic data and well log markers were used to reproduce subsurface structures characterized by a normal fault system. A volume-based modeling algorithm and stair-stepping grid were applied. The resulting 3-D model comprised 2,640,000 grid cells across 264 layers, including seven fault grids. For petrophysical property modeling, total porosity was initially modeled using sequential Gaussian simulation with collocated cokriging. To reproduce the upward coarsening of the B1 sand, upscaled log-derived total porosity and a three-dimensional (3-D) trend depicting total porosity variation were used as primary and secondary data, respectively. Gas hydrate saturation distribution was modeled similarly, with secondary data from estimated porosity distribution and seismic-derived acoustic impedance map enhancing accuracy. Results indicate higher gas hydrate saturation in the upper part of the B1 sand and areas with higher acoustic impedance. Intrinsic permeability was modeled from the total porosity and clay-bound water volume, and effective permeability was derived from the gas hydrate saturation and intrinsic permeability distributions based on the “Tokyo model”. Effective permeability distributions were influenced by the total porosity, gas hydrate saturation, and intrinsic permeability. Within the same layer, higher gas hydrate saturation leads to decreased effective permeability. In total, 100 sets of multiple scenarios were prepared, providing input data for dynamic flow simulations to evaluate the effects of lateral heterogeneity in reservoir properties and the hydraulic characteristics of faults on production behavior for preassessment before the long-term production test.

58 GEOSCIENCES↗
DOE OSTI2025

Longitudinal Patterns in Nitrate Revealed Through Drone‐Based Measurements in an Agriculturally Influenced Midwestern U.S. River

Nitrate concentrations in streams and rivers in the Midwestern United States are often elevated, reflecting the predominance of agriculture in the surrounding landscape. Recent advances in technology, including surface water drones and more precise sensors, provide opportunities to investigate nitrate dynamics with high spatial and temporal resolution. We deployed an aquatic drone, the AquaBOT, in a sixth-order, agriculturally influenced river to examine longitudinal patterns in water quality. Our goal was to measure the spatial and temporal heterogeneity in nitrate and nitrate removal processes and determine the influence of tributary inputs on main stem chemistry. We navigated the drone along a 12-km reach of the Des Moines River (Iowa, USA) nine times between June 2021 and August 2022. Across the deployments, mean nitrate concentration was positively related to discharge and was nearly two orders of magnitude higher in spring than summer. We observed contrasting patterns in main stem nitrate, which decreased downstream during some runs (e.g., 3.1–2.7 mg N L −1 in June 2021), demonstrating net nitrate uptake along the reach, and remained constant on other dates. Similarly, tributaries to the Des Moines had a varied influence on riverine nitrate. Tributaries either increased or decreased main stem nitrate concentrations depending on the tributary and the date. Nitrate removal rates were spatially and temporally variable but showed some consistency at the subreach (2 km) scale, with two subreaches often showing elevated rates of nitrate removal across dates. In conclusion, our study reveals nuanced heterogeneity in nitrate dynamics of the Des Moines River despite the homogeneity of agricultural land cover in the watershed.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2026

Data for KETCHUP: Parameterizing of Large-Scale Kinetic Models Using Multiple Datasets with Different Reference States

Repository for Kinetic Estimation Tool Capturing Heterogeneous Datasets Using Pyomo (KETCHUP), a flexible parameter estimation tool that leverages a primal-dual interior-point algorithm to solve a nonlinear programming (NLP) problem that identifies a set of parameters capable of recapitulating the steady-state fluxes and concentrations in wild-type and perturbed metabolic networks. KETCHUP can use K-FIT [2] input files. Example K-FIT input files are located in the K-FIT repository at https://github.com/maranasgroup/K-FIT.

Metabolomics↗
DOE OSTI2026

Application of artificial intelligence methods in the international roughness index prediction of rigid and composite pavements: a systematic review

The International Roughness Index (IRI) is a widely adopted metric for quantifying pavement roughness, directly influencing vehicle safety, ride comfort, and overall roadway performance. In recent years, the use of Machine Learning (ML) models for IRI prediction has gained momentum, with the goal of improving the allocation of maintenance and rehabilitation resources by enabling accurate assessments of pavement conditions. Most prior reviews, however, have concentrated on flexible pavements, leaving a notable gap regarding rigid and composite pavements. To address this gap, the present study conducts a systematic review of Artificial Intelligence (AI) methods applied to IRI prediction for rigid and composite pavements. Literature published between 2004 and 2025 is synthesized to highlight prevailing trends, methodological contributions, and directions for future research. Particular attention is given to the types of models employed, the datasets used for training and validation, and the role of input variables and data-processing strategies. Across the included studies, ensemble learning methods (especially gradient boosting variants such as XGBoost), artificial neural networks, and hybrid architectures frequently achieved high predictive skill, with several models reporting test-set coefficients of determination approaching 0.9–0.96, indicating strong potential for capturing the influence of traffic, pavement structure, and climatic factors. Since these results are obtained from heterogeneous datasets and evaluation protocols, they are interpreted qualitatively rather than as strict cross-study rankings. Analysis of input variables revealed that pavement age and initial IRI were included in 91% (21 of 23) and 78% (18 of 23) of studies, respectively. Climatic variables such as the freezing index appeared in 57% (13 of 23), while traffic-related factors were considered in 65% (15 of 23). The findings underscore the importance of standardized, high-quality datasets, such as those from the Long-Term Pavement Performance (LTPP) program, along with data consistency, model interpretability, computational efficiency, and replicability in enhancing IRI prediction. Future research should focus on incorporating input variable selection techniques to identify the most influential predictors, thereby improving accuracy and robustness. Integrating these approaches with advanced non-linear data-driven models, coupled with robust hyperparameter optimization, holds considerable promise for strengthening the reliability of IRI prediction and supporting resilient pavement management strategies.

42 ENGINEERING↗
DOE OSTI2024

On learning what to learn: Heterogeneous observations of dynamics and establishing possibly causal relations among them

Abstract Before we attempt to (approximately) learn a function between two sets of observables of a physical process, we must first decide what the inputs and outputs of the desired function are going to be. Here we demonstrate two distinct, data-driven ways of first deciding “the right quantities” to relate through such a function, and then proceeding to learn it. This is accomplished by first processing simultaneous heterogeneous data streams (ensembles of time series) from observations of a physical system: records of multiple observation processes of the system. We determine (i) what subsets of observables are common between the observation processes (and therefore observable from each other, relatable through a function); and (ii) what information is unrelated to these common observables, therefore particular to each observation process, and not contributing to the desired function. Any data-driven technique can subsequently be used to learn the input–output relation—from k-nearest neighbors and Geometric Harmonics to Gaussian Processes and Neural Networks. Two particular “twists” of the approach are discussed. The first has to do with the identifiability of particular quantities of interest from the measurements. We now construct mappings from a single set of observations from one process to entire level sets of measurements of the second process, consistent with this single set. The second attempts to relate our framework to a form of causality: if one of the observation processes measures “now,” while the second observation process measures “in the future,” the function to be learned among what is common across observation processes constitutes a dynamical model for the system evolution.

Sroczynski, David W.↗
DOE OSTI2024

Anticipating Technical Expertise and Capability Evolution in Research Communities Using Dynamic Graph Transformers

The ability to anticipate global technical expertise and capability evolution trends is essential for national and global security, especially in safety-critical domains such as nuclear nonproliferation (NN) and rapidly emerging fields like artificial intelligence (AI). Here, in this work, we extend traditional statistical relational learning approaches (e.g., link prediction in collaboration networks) and formulate a problem of anticipating technical expertise and capability evolution using dynamic heterogeneous graph representations. We develop novel capabilities to forecast collaboration patterns, authorship behavior, and technical capability evolution at different granularities (e.g., scientist and institution levels) in two distinct research fields. We implement a dynamic graph transformer (DGT) neural architecture, which pushes the state-of-the-art graph neural network models by: 1) forecasting heterogeneous (rather than homogeneous) nodes and edges; and 2) relying on both discrete- and continuous-time inputs. We demonstrate that our DGT models predict collaboration, partnership, and expertise patterns with 0.26, 0.73, and 0.53 mean reciprocal rank values for AI and 0.48, 0.93, and 0.22 for NN domains. DGT model performance exceeds the best-performing static graph baseline models by 30%–80% across AI and NN domains. Our findings demonstrate that DGT models boost inductive task performance when previously unseen nodes appear in the test data for the domains with emerging collaboration patterns (e.g., AI). Specifically, models accurately predict which established scientists will collaborate with early career scientists and vice versa in the AI domain.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2026

AI-Based Analytics and Energy Modeling Framework for Characterizing Urban Energy Systems

Developing location-specific district energy models is essential for understanding energy patterns and supporting efficient management and planning decisions. However, accurately characterizing these models remains challenging due to gaps in building characteristics and labor-intensive traditional modeling workflows. To address these challenges, we develop an AI-based framework that integrates top-down and bottom-up building energy data to automate urban energy model characterization. The framework trains multimodal deep learning models using heterogeneous ResStockTM datasets to infer missing building characteristics from varying levels of known information and generate simulation-ready inputs for district-scale energy modeling. It also employs a conditioning-based injection approach to generate ”what-if” scenarios, enabling users to explore retrofit, efficiency, and technology-upgrade pathways. Integrated within URBANoptTM, a bottom-up district energy modeling platform for simulating co-located buildings, the framework infers detailed building-level inputs required for bottom-up simulations. Both localized and generalized AI models are developed to learn relationships across categorical, numerical, and time-series data, enabling reconstruction of missing attributes and generation of targeted upgrade scenarios. We demonstrate this methodology on a residential neighborhood in Baltimore, MD, assessing internal consistency against ResStock reference data and URBANopt simulation, and comparing selected attributes against real-world building characteristics. Results show strong overall predictive accuracy in data completion and scenario generation, with localized and generalized models offering complementary trade-offs between precision and scalability. Overall, our automated framework streamlines energy modeling and provides a reliable framework for urban building energy characterization.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2025

Verifying infectious disease scenario planning for geographically diverse populations

In the face of the COVID-19 pandemic, the literature saw a spike in publications for epidemic models, and a renewed interest in capturing contact networks and geographic movement of populations. There remains a general lack of consensus in the modeling community around best practices for spatiotemporal epi-modeling, specifically as it pertains to the infection rate formulation and the underlying contact or mixing model. We mathematically verify several common modeling assumptions in the literature, to prove when certain choices can provide consistent results across different geographic resolutions, population densities and patterns, and mixing assumptions. The most common infection rate formulation, a computationally low cost per capita infection rate assumption, fails the consistency tests for heterogeneous populations and gravity-weighting assumptions. Future modeling efforts in spatiotemporal disease modeling should be wary of this limitation, particularly when working with more heterogeneous or sparse populations. Our results provide guidance for testing that a model preserves desirable properties even when model inputs mask potential problems due to symmetry or homogeneity. We also provide a recipe for performing this type of verification, strengthening decision support tools.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2025

Developing machine learning for heterogeneous catalysis with experimental and computational data

Machine learning techniques have emerged as a useful tool for identifying complex patterns and correlations in large datasets, such as associating catalyst performance to its physicochemical properties. In the heterogeneous catalysis communities, machine learning models have mostly been developed using high-throughput quantum chemistry calculations, with only a few case studies resulting in experimentally validated catalyst improvements. This limited success may be due to the use of simplified catalyst structures in computational studies and the lack of comprehensive experimental datasets. In this Review, we bring together studies integrating high-throughput approaches and machine learning for the advancement of solid heterogeneous catalysis, leveraging both experimental and computational data. We systematically analyze trends in the field, based on the descriptors used as model input and output; the materials, devices, or reactions investigated; the dataset size; and the overall achievements. Furthermore, for models reporting unitless R 2 values, we compare the performances based on these mentioned trends.

Computational chemistry↗
DOE OSTI2025

CO2 Plume Imaging with Accelerated Deep Learning-based Data Assimilation Considering Multiple Realizations: Application to the Illinois Basin-Decatur Carbon Sequestration Project

We propose a fast and efficient deep learning workflow for near real-time data assimilation, forecasting and visualization of CO2 plume evolution in saline aquifer and demonstrate its application at a field site. Unlike the previous work, this study incorporates the impact of spatial heterogeneity using multiple realizations. In the proposed workflow, a neural network model utilizes available monitoring data such as downhole pressure measurements as input and predicts the propagating pressure ‘front’ using the diffusive time of flight (DTOF) map which is considered as representative reservoir image of the flow field. The DTOF is the arrival time of pressure front propagation, which can be computed by the Fast Marching Method rapidly without flow simulations. Reservoir model calibration can be implemented by selecting the training data samples that describe the predicted DTOF map based on observed data. The power and efficacy of our workflow is demonstrated by application to the Illinois Basin-Decatur Project.

CO2 plume imaging↗