Search NASA⌕ Search

SEARCH · Search NASA

Results for “generative flow model”

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 37 records · Page 2
All sourcesNASA NTRSDOE OSTI
DOE OSTI2019

Memory embedded non-intrusive reduced order modeling of non-ergodic flows

Generating a digital twin of any complex system requires modeling and computational approaches that are efficient, accurate, and modular. Traditional reduced order modeling techniques are targeted at only the first two, but the novel nonintrusive approach we present in this study is an attempt at taking all three into account effectively compared to their traditional counterparts. Based on dimensionality reduction using proper orthogonal decomposition (POD), we introduce a long short-term memory neural network architecture together with a principal interval decomposition (PID) framework as an enabler to account for localized modal deformation. As an effective partitioning tool for breaking the Kolmogorov barrier, our PID framework, therefore, can be considered a key element in the accurate reduced order modeling of convective flows. Our applications for convection-dominated systems governed by Burgers, Navier-Stokes, and Boussinesq equations demonstrate that the proposed approach yields significantly more accurate predictions than the POD-Galerkin method and could be a key enabler toward near real-time predictions of unsteady flows.

42 ENGINEERING↗
DOE OSTI2024

Diffusion-Model-Assisted Supervised Learning of Generative Models for Density Estimation

Here, we present a supervised learning framework of training generative models for density estimation. Generative models, including generative adversarial networks (GANs), normalizing flows, and variational auto-encoders (VAEs), are usually considered as unsupervised learning models, because labeled data are usually unavailable for training. Despite the success of the generative models, there are several issues with the unsupervised training, e.g., requirement of reversible architectures, vanishing gradients, and training instability. To enable supervised learning in generative models, we utilize the score-based diffusion model to generate labeled data. Unlike existing diffusion models that train neural networks to learn the score function, we develop a training-free score estimation method. This approach uses mini-batch-based Monte Carlo estimators to directly approximate the score function at any spatial-temporal location in solving an ordinary differential equation (ODE), corresponding to the reverse-time stochastic differential equation (SDE). This approach can offer both high accuracy and substantial time savings in neural network training. Once the labeled data are generated, we can train a simple, fully connected neural network to learn the generative model in the supervised manner. Compared with existing normalizing flow models, our method does not require the use of reversible neural networks and avoids the computation of the Jacobian matrix. Compared with existing diffusion models, our method does not need to solve the reverse-time SDE to generate new samples. As a result, the sampling efficiency is significantly improved. We demonstrate the performance of our method by applying it to a set of 2D datasets as well as real data from the University of California Irvine (UCI) repository.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

MG-RAVENS: Working Group Update [Slides]

Problem Statement: Users in industry, who are planning the deployment and operation of microgrids, face a multi-domain problem that requires multiple engineering tools to solve; Existing tools lack interoperability, which requires tedious recreation and conversion of equipment models, and creates opportunities for errors in translation or through inconsistent assumptions; Microgrid modelers waste time reinventing the wheel because it is challenging to reuse existing distribution, load, generation, power flow, optimization models for new use cases.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2022

Dynamic Material Flow Analysis of Silicon Photovoltaic Modules to Support a Circular Economy Transition

Solar photovoltaics (PV) are the fastest growing renewable energy technologies for clean, cheap, and sustainable electricity generation. To prepare for rapid scale-up, the PV industry needs to project material requirements to build out all aspects of the supply chain appropriately and plan to handle large volumes of module waste. Impacts of deploying different material circularity strategies to reduce waste and conserve primary resources need to be quantified to inform sustainable material management. Here, we introduce the photovoltaic dynamic material flow analysis (PV DMFA) model based on PV electricity generation. The model quantifies material flows and stocks in the cradle-to-cradle life cycles of utility-scale c-Si PV systems in the United States through 2100. We present case studies for solar flat glass and aluminum frame materials under various scenarios to project the impacts of PV performance, reliability, and processing parameters, material circularity strategies, and module design shifts. In the absence of circularity measures, ~100 million MT of flat glass and ~12 million MT of aluminum would be needed for PV installations by 2100 to meet projected growth in domestic utility PV demand to nearly 1000 TWh in 2100. With optimistic but feasible improvements in efficiency, reliability, and circularity, material intensity and waste could be reduced by nearly 50%. Efficient module collection, minimally intrusive recycling, and careful scrap handling and cleaning could improve material circularity in the PV value chain. This model serves as a sustainability data support tool that may aid in the circular economy transition for PV systems.

circular economy↗
DOE OSTI2020

TEAL

TEAL is a financial performance calculator plugin for the RAVEN code, framework, resolving around the computation of Net Present Value and associated financial metrics. TEAL can make use of inflation rates, taxation, escalation factors, capital expenditure economy of scale scaling factors. The unique feature of TEAL is the capability to be linked with RAVEN external models and build corresponding cash flows using the variables computed by those external models. In addition to be able to use the capability to generate cash flows derived from complex physical models generated by RAVEN, another distinctive feature of TEAL is the capability to provide financial risk/probabilistic metrics that can empower RAVEN to perform optimization/analysis driven by financial risk augmentations. Optimization, robust optimization, parametric studies, large parallel simulations, sensitivity analysis, data mining, etc. are just some of the capabilities that can be leveraged.

Alfonsi, Andrea↗
DOE OSTI2024

Holistic Small-Signal Stability Analysis for Large-Scale Inverter-Intensive Power Systems with Coupled and Full-Order Dynamics from Control Systems and Power Networks

The increasing penetration of inverter-based resources (IBRs) into the existing power systems introduces tremendous benefits for enhanced sustainability but also poses inevitable challenges in terms of insufficient inertia, potential instability, and complex network dynamics, among others. However, the additional coupling introduced by the interactions among gridfollowing (GFL) and grid-forming (GFM) IBRs and the other components (i.e., synchronous generators [SGs], loads, and network, etc.) has not been clearly explored. A holistic, scalable, and quantitative stability analysis framework with the control systems and power networks is still missing. Here, in this paper, to fill in the technical gaps, a holistic small-signal model of the entire system with both rotating generation units and IBRs is established. An extended power flow model with operation dynamics from both generator control schemes and power networks is proposed to provide the varying steady-state operating points for small-signal modeling. The proposed method is compared with MATLAB solvers, and the results show that the proposed approach has a minimum calculation time, which can be less than 12 seconds for a large-scale power system with up to 2,000 buses. Furthermore, a quantitative method is developed to identify the impacts of IBRs on system performance with emphases on the potential stability issues with GFL IBRs, additional benefits of employing GFM IBRs, the feasibility of replacing SGs with GFM IBRs, and the impact of penetration level of different kinds of generation units. Finally, a field island power system is used to verify the proposed approach, and hardware-in-the-loop (HIL) tests are provided to further demonstrate the effectiveness of the proposed analysis.

14 SOLAR ENERGY↗
DOE OSTI2022

Translation and rotation equivariant normalizing flow (TRENF) for optimal cosmological analysis

ABSTRACT Our Universe is homogeneous and isotropic, and its perturbations obey translation and rotation symmetry. In this work, we develop translation and rotation equivariant normalizing flow (TRENF), a generative normalizing flow (NF) model which explicitly incorporates these symmetries, defining the data likelihood via a sequence of Fourier space-based convolutions and pixel-wise non-linear transforms. TRENF gives direct access to the high dimensional data likelihood p(x|y) as a function of the labels y, such as cosmological parameters. In contrast to traditional analyses based on summary statistics, the NF approach has no loss of information since it preserves the full dimensionality of the data. On Gaussian random fields, the TRENF likelihood agrees well with the analytical expression and saturates the Fisher information content in the labels y. On non-linear cosmological overdensity fields from N-body simulations, TRENF leads to significant improvements in constraining power over the standard power spectrum summary statistic. TRENF is also a generative model of the data, and we show that TRENF samples agree well with the N-body simulations it trained on, and that the inverse mapping of the data agrees well with a Gaussian white noise both visually and on various summary statistics: when this is perfectly achieved the resulting p(x|y) likelihood analysis becomes optimal. Finally, we develop a generalization of this model that can handle effects that break the symmetry of the data, such as the survey mask, which enables likelihood analysis on data without periodic boundaries.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2025

SAM Two-Phase Flow Model Development and Applications for Operational Transients in Advanced Reactors

As advanced nuclear technologies continue to develop, the need for the flexible operation and generation of these advanced reactors becomes necessary to maximize economic potential. As large-scale experiments are not always feasible, modeling and simulations of advanced reactors play a crucial role in design optimization and analysis. The SAM (System Analysis Module) code developed at Argonne National Laboratory is a state-of-the-art system-level thermal-hydraulic code aimed at simulating advanced reactor systems. Recent code developments have implemented two-phase flow modeling using the homogeneous equilibrium model, and a new steam generator component has been developed to utilize the two-phase flow implementation. In addition to verification tests, a load-following simulation was performed to model a realistic load-following transient in a proposed integrated system consisting of a conceptual advanced reactor known as the Advanced Burner Test Reactor (ABTR) and thermal energy storage (TES) tanks. The integrated system model uses two large TES tanks designed for sodium and a model helical coil steam generator to simulate the operational load-following transient. The flow rates of the feedwater and secondary loops are regulated to meet a prescribed steam generator load consistent with the electricity demand over a 24-h period. In conclusion, the results found the ABTR system was able to maintain stable reactor conditions and primary- and secondary-side characteristics over the course of the load-following transient.

Advanced Burner Test Reactor (ABTR)↗
DOE OSTI2024

Anomaly detection with flow-based fast calorimeter simulators

Recently, several normalizing flow-based deep generative models have been proposed to accelerate the simulation of calorimeter showers. Using caloflow as an example, we show that these models can simultaneously perform unsupervised anomaly detection with no additional training cost. As a demonstration, we consider electromagnetic showers initiated by one (background) or multiple (signal) photons. The caloflow model is designed to generate single-photon showers, but it also provides access to the shower likelihood. We use this likelihood as an anomaly score and study the showers tagged as being unlikely. As expected, the tagger struggles when the signal photons are nearly collinear but is otherwise effective. This approach is complementary to a supervised classifier trained on only specific signal models using the same low-level calorimeter inputs. While the supervised classifier is also highly effective at unseen signal models, the unsupervised method is more sensitive in certain regions, and, thus, we expect that the ultimate performance will require a combination of these approaches.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2023

A phased workflow to define permit‐ready locations for large volume CO 2 injection and storage

Abstract To‐date, only two UIC Class VI permits have been issued by the US Environmental Protection Agency. We illustrate a four‐phase workflow to first identify regional storage resources and then down‐select sites to yield permit‐ready locations that can accept and store large volumes of CO 2 . Specific permit requirements should guide objectives and define deliverables of respective workflow phases. In the first phase we used available regional data and screened structure and injection zones to locate resources that match CO 2 volumes planned to be captured. Available data were also used to assess presence and depth of usable groundwater, the key resource being protected via permitting. We then used advanced, closed‐form, analytical solutions (EASiTool) to estimate CO 2 injectivity into each hydrologically connected injection compartment. In the second phase we acquired and conditioned additional wireline logs and leased available seismic datasets. We interpreted the depositional systems from wireline well‐log character and mapped sandbody geometry to interpolate injection and confining‐zone distribution. Using available data, we mapped faults and locations of freshwater and overpressure (or other capacity‐limiting geologic parameters) in more detail. In the third phase, we used the augmented geologic data to develop a static model for the selected area, extracted the areas of highest interest, and generated and ran dynamic (flow) models. In a fourth phase, we reduced major uncertainties identified in earlier phases. Our case study indicates that to complete preparation of a permit application requires (1) improved lithologic characterization information (thicknesses and horizontal and vertical connectivity) and (2) better definition of poorly defined local faults. © 2023 The Authors. Greenhouse Gases: Science and Technology published by Society of Chemical Industry and John Wiley & Sons Ltd.

58 GEOSCIENCES↗
DOE OSTI2025

CaloChallenge 2022: a community challenge for fast calorimeter simulation

Here, we present the results of the ‘Fast Calorimeter Simulation Challenge 2022’—the CaloChallenge. We study state-of-the-art generative models on four calorimeter shower datasets of increasing dimensionality, ranging from a few hundred voxels to a few tens of thousand voxels. The 31 individual submissions span a wide range of current popular generative architectures, including variational autoencoders (VAEs), generative adversarial networks (GANs), normalizing flows, diffusion models, and models based on conditional flow matching. We compare all submissions in terms of quality of generated calorimeter showers, as well as shower generation time and model size. To assess the quality we use a broad range of different metrics including differences in one-dimensional histograms of observables, KPD/FPD scores, AUCs of binary classifiers, and the log-posterior of a multiclass classifier. The results of the CaloChallenge provide the most complete and comprehensive survey of cutting-edge approaches to calorimeter fast simulation to date. In addition, our work provides a uniquely detailed perspective on the important problem of how to evaluate generative models. As such, the results presented here should be applicable for other domains that use generative AI and require fast and faithful generation of samples in a large phase space.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2022

The channel-source hypothesis: Empirical evidence for in-channel sourcing of dissolved organic carbon to explain hysteresis in a headwater mountain stream

Catchment hydrologists have long puzzled over the question: How can catchments rapidly generate storm flows and pulses of solutes in response to storm events? Conceptual models viewing catchments as composed of discrete source areas generating flow at unique time scales and with unique chemical characteristics have been used to explain the observed changes in flow and water chemistry. Surprisingly, those conceptual models usually do not treat the stream channel as one of the potential source areas. We propose the channel-source hypothesis in which the stream itself should be considered as a potential source with the same rigour as other contributing areas. We pose this in the spirit of the scientific use of the word: a hypothesis is not a proven idea but rather a provisional supposition serving as the basis for further study. We suggest that the channel should be considered as a potential source for dissolved organic carbon (DOC). Channels store substantial amounts of organic matter, and stream ecologists have long studied stream carbon cycling. From those studies, we know that leaching and decomposition can generate DOC from particulate organic carbon (POC). Further, POC is stored in channel ‘dead-zones’—regions of low flow velocity - that can be activated as flow velocity increases, thus releasing accumulated DOC during storms. All catchments are different; there is no reason to assume that channel sources are always important, in every catchment, in every storm. Thus, the channel-source hypothesis does not replace existing conceptual models. Instead, it adds another potential mechanism that may explain DOC dynamics observed in streams. The channel-source hypothesis has substantial implications for catchment studies examining sources of DOC in stream water or using DOC as a tracer to determine the locations of, and proportional contributions of, different source areas for streamflow generation.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2020

Learning how structures form in drift-wave turbulence

Drift-wave turbulence produces anomalous transport via cross-correlations between fluctuations. This transport has profound implications for confinement, structure formation, and virtually all aspects of the non-linear turbulent dynamics. Herein, we use a data-driven method based on deep learning in order to study turbulent transport in the 2D Hasegawa–Wakatani system and infer a reduced mean-field model from numerical solution. In addition to the usual turbulent diffusion, we find an effect which couples the particle flux to the local gradient of vorticity, which tends to modulate the density profile. The direct coupling to the shear is relatively weak. In addition, the deep learning method finds a model for spontaneous zonal flow generation by negative viscosity, stabilized by non-linear and hyperviscous terms. We compare these results to analytic calculations using quasilinear theory and wave kinetics, finding qualitative agreement, though the calculations miss certain higher-order effects. A simplified, 1-D model for the evolution of the profile, flow, and intensity based on the deep learning results is solved numerically and compared to previous models for staircasing based on bistability. We see that the physics uncovered by the deep learning method provided simple explanations for the formation of zonal structures in the density, flow, and turbulence fields. We highlight the important role of symmetry in the deep learning method and speculate on the portability of the method to other applications.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2020

Energy Prediction of the Chicago Metropolitan Area Using Distributed Transportation MBSE Framework

Energy consumption impact of vehicle powertrain is traditionally assessed using standard and regulatory drive cycles. In this paper, energy consumption is determined by studying all the trips occurring in the Chicago Metropolitan Area in a 24 hour period. The data is generated by combining Polaris, an agent based traffic flow model with SVTrip, a stochastic model that generates 1 Hz vehicle speed profiles and Autonomie, a detailed vehicle level modeling tool used to predict fuel consumption. Different scenarios are constructed to define trips for different timeframes and assumptions. Vehicle fleet distribution, vehicle technologies and automation levels are applied to represent an appropriate vehicle fleet for each scenario. Analysis of energy consumption is then discussed with a particular emphasis on the impact of powertrain technology and vehicle types.

Autonomie↗
DOE OSTI2025

A comparison of probabilistic generative frameworks for molecular simulations

Generative artificial intelligence is now a widely used tool in molecular science. Despite the popularity of probabilistic generative models, numerical experiments benchmarking their performance on molecular data are lacking. Here, in this work, we introduce and explain several classes of generative models, broadly sorted into two categories: flow-based models and diffusion models. We select three representative models: neural spline flows, conditional flow matching, and denoising diffusion probabilistic models, and examine their accuracy, computational cost, and generation speed across datasets with tunable dimensionality, complexity, and modal asymmetry. Our findings are varied, with no one framework being the best for all purposes. In a nutshell, (i) neural spline flows do best at capturing mode asymmetry present in low-dimensional data, (ii) conditional flow matching outperforms other models for high-dimensional data with low complexity, and (iii) denoising diffusion probabilistic models appear the best for low-dimensional data with high complexity. Our datasets include a Gaussian mixture model and the dihedral torsion angle distribution of the Aib9 peptide, generated via a molecular dynamics simulation. We hope our taxonomy of probabilistic generative frameworks and numerical results may guide model selection for a wide range of molecular tasks.

Artificial intelligence↗