Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Distributed Estimation”

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 1,099 records · Page 61
All sourcesNASA NTRSDOE OSTI
NASA NTRS2012

A Tracking Sun Photometer Without Moving Parts

This innovation is small, lightweight, and consumes very little electricity as it measures the solar energy attenuated by gases and aerosol particles in the atmosphere. A Sun photometer is commonly used on the Earth's surface, as well as on aircraft, to determine the solar energy attenuated by aerosol particles in the atmosphere and their distribution of sizes. This information is used to determine the spatial and temporal distribution of gases and aerosols in the atmosphere, as well as their distribution sizes. The design for this Sun photometer uses a combination of unique optics and a charge coupled device (CCD) array to eliminate moving parts and make the instrument more reliable. It could be selfcalibrating throughout the year. Data products would be down-welling flux, the direct-diffuse flux ratio, column abundance of gas phase constituents, aerosol optical depth at multiple-wavelengths, phase functions, cloud statistics, and an estimate of the representative size of atmospheric particles. These measurements can be used to obtain an estimate of aerosol size distribution, refractive index, and particle shape. Incident light is received at a light-reflecting (inner) surface, which is a truncated paraboloid. Light arriving from a hemispheric field of view (solid angle 2 steradians) enters the reflecting optic at an entrance aperture at, or adjacent to, the focus of the paraboloid, and is captured by the optic. Most of this light is reflected from an inner surface. The light proceeds substantially parallel to the paraboloid axis, and is detected by an array detector located near an exit aperture. Each of the entrance and exit apertures is formed by the intersection of the paraboloid with a plane substantially perpendicular to the paraboloid axis. Incident (non-reflected) light from a source of limited extent (the Sun) illuminates a limited area on the detector array. Both direct and diffuse illumination may be reflected, or not reflected, before being received on the detector array. As the Sun traverses a path in the sky over some time interval, the track of the Sun can be traced on the detector array. A suitably modified Sun photometer might be used to study the dynamics of an environment on another planet or satellite with an atmosphere.

Strawa, Anthony W.↗
DOE OSTI2020

Recent Development of Frequency Estimation Methods for Future Smart Grid

The frequency estimated by the Phasor Measurement Unit (PMU) is a critical index of power system status and supports many smart grid applications. The future smart grid features high penetration of renewables and more fast-moving power electronics inverters but raises challenges to the reliable frequency estimation. This article presents three methods to address these challenges. First, an enhanced zero-crossing algorithm was developed to track the fast-changing frequency in system dynamics. Second, we propose a technology that can tolerate the system transient and suppress the outliers. Third, an algorithm was developed to export high time-resolution frequency estimations with minimum computational effort. All of the proposed methods are realized in hardware and compared with classical frequency estimation methods. The testing results indicate that the proposed methods have excellent performance. They can be used in future PMUs and provide reliable and high time resolution data for smart grid applications.

24 POWER TRANSMISSION AND DISTRIBUTION↗
NASA NTRS1995

Determination of Eros Physical Parameters for Near Earth Asteroid Rendezvous Orbit Phase Navigation

Navigation of the orbit phase of the Near Earth steroid Rendezvous (NEAR) mission will re,quire determination of certain physical parameters describing the size, shape, gravity field, attitude and inertial properties of Eros. Prior to launch, little was known about Eros except for its orbit which could be determined with high precision from ground based telescope observations. Radar bounce and light curve data provided a rough estimate of Eros shape and a fairly good estimate of the pole, prime meridian and spin rate. However, the determination of the NEAR spacecraft orbit requires a high precision model of Eros's physical parameters and the ground based data provides only marginal a priori information. Eros is the principal source of perturbations of the spacecraft's trajectory and the principal source of data for determining the orbit. The initial orbit determination strategy is therefore concerned with developing a precise model of Eros. The original plan for Eros orbital operations was to execute a series of rendezvous burns beginning on December 20,1998 and insert into a close Eros orbit in January 1999. As a result of an unplanned termination of the rendezvous burn on December 20, 1998, the NEAR spacecraft continued on its high velocity approach trajectory and passed within 3900 km of Eros on December 23, 1998. The planned rendezvous burn was delayed until January 3, 1999 which resulted in the spacecraft being placed on a trajectory that slowly returns to Eros with a subsequent delay of close Eros orbital operations until February 2001. The flyby of Eros provided a brief glimpse and allowed for a crude estimate of the pole, prime meridian and mass of Eros. More importantly for navigation, orbit determination software was executed in the landmark tracking mode to determine the spacecraft orbit and a preliminary shape and landmark data base has been obtained. The flyby also provided an opportunity to test orbit determination operational procedures that will be used in February of 2001. The initial attitude and spin rate of Eros, as well as estimates of reference landmark locations, are obtained from images of the asteroid. These initial estimates are used as a priori values for a more precise refinement of these parameters by the orbit determination software which combines optical measurements with Doppler tracking data to obtain solutions for the required parameters. As the spacecraft is maneuvered; closer to the asteroid, estimates of spacecraft state, asteroid attitude, solar pressure, landmark locations and Eros physical parameters including mass, moments of inertia and gravity harmonics are determined with increasing precision. The determination of the elements of the inertia tensor of the asteroid is critical to spacecraft orbit determination and prediction of the asteroid attitude. The moments of inertia about the principal axes are also of scientific interest since they provide some insight into the internal mass distribution. Determination of the principal axes moments of inertia will depend on observing free precession in the asteroid's attitude dynamics. Gravity harmonics are in themselves of interest to science. When compared with the asteroid shape, some insight may be obtained into Eros' internal structure. The location of the center of mass derived from the first degree harmonic coefficients give a direct indication of overall mass distribution. The second degree harmonic coefficients relate to the radial distribution of mass. Higher degree harmonics may be compared with surface features to gain additional insight into mass distribution. In this paper, estimates of Eros physical parameters obtained from the December 23,1998 flyby will be presented. This new knowledge will be applied to simplification of Eros orbital operations in February of 2001. The resulting revision to the orbit determination strategy will also be discussed.

Miller, J. K.↗
DOE OSTI2025

Graph-based Simulation Framework for Power Resilience Estimation and Enhancement

The increasing frequency of extreme weather events poses significant risks to power distribution systems, leading to widespread outages and severe economic and social consequences. This paper presents a novel simulation framework for assessing and enhancing the resilience of power distribution networks under such conditions. Resilience is estimated through Monte Carlo simulations, which simulate extreme weather scenarios and evaluate the impact on infrastructure fragility. Due to the proprietary nature of power network topology, a distribution network is synthesized using publicly available data. To generate the weather scenarios, an extreme weather generation method is developed. To enhance resilience, renewable resources such as solar panels and energy storage systems (batteries in this study) are incorporated. A customized Genetic Algorithm is proposed to determine the optimal locations and capacities for solar panels and battery installations, maximizing resilience while balancing cost constraints. Experiment results demonstrate that on a large-scale synthetic distribution network with more than 300,000 nodes and 300,000 edges, the proposed framework can efficiently evaluate the resilience, and enhance the resilience through the installations of distributed energy resources (DERs), providing utilities with valuable insights for community-level power system resilience estimation and enhancement.

Wang, Xuesong [Wayne State Univ., Detroit, MI (Uni↗
NASA NTRS2013

Jovian Plasmas Torus Interaction with Europa. Plasma Wake Structure and Effect of Inductive Magnetic Field: 3D Hybrid Kinetic Simulation

The hybrid kinetic model supports comprehensive simulation of the interaction between different spatial and energetic elements of the Europa moon-magnetosphere system with respect to a variable upstream magnetic field and flux or density distributions of plasma and energetic ions, electrons, and neutral atoms. This capability is critical for improving the interpretation of the existing Europa flyby measurements from the Galileo Orbiter mission, and for planning flyby and orbital measurements (including the surface and atmospheric compositions) for future missions. The simulations are based on recent models of the atmosphere of Europa (Cassidy et al., 2007; Shematovich et al., 2005). In contrast to previous approaches with MHD simulations, the hybrid model allows us to fully take into account the finite gyroradius effect and electron pressure, and to correctly estimate the ion velocity distribution and the fluxes along the magnetic field (assuming an initial Maxwellian velocity distribution for upstream background ions). Photoionization, electron-impact ionization, charge exchange and collisions between the ions and neutrals are also included in our model. We consider the models with Oþ þ and Sþ þ background plasma, and various betas for background ions and electrons, and pickup electrons. The majority of O2 atmosphere is thermal with an extended non-thermal population (Cassidy et al., 2007). In this paper, we discuss two tasks: (1) the plasma wake structure dependence on the parameters of the upstream plasma and Europa's atmosphere (model I, cases (a) and (b) with a homogeneous Jovian magnetosphere field, an inductive magnetic dipole and high oceanic shell conductivity); and (2) estimation of the possible effect of an induced magnetic field arising from oceanic shell conductivity. This effect was estimated based on the difference between the observed and modeled magnetic fields (model II, case (c) with an inhomogeneous Jovian magnetosphere field, an inductive magnetic dipole and low oceanic shell conductivity).

Jovian magnetosphere↗
DOE OSTI2022

Preliminary Results for Using Uncertainty and Out-of-distribution Detection to Identify Unreliable Predictions.

As machine learning (ML) models are deployed into an ever-diversifying set of application spaces, ranging from self-driving cars to cybersecurity to climate modeling, the need to carefully evaluate model credibility becomes increasingly important. Uncertainty quantification (UQ) provides important information about the ability of a learned model to make sound predictions, often with respect to individual test cases. However, most UQ methods for ML are themselves data-driven and therefore susceptible to the same knowledge gaps as the models themselves. Specifically, UQ helps to identify points near decision boundaries where the models fit the data poorly, yet predictions can score as certain for points that are under-represented by the training data and thus out-of-distribution (OOD). One method for evaluating the quality of both ML models and their associated uncertainty estimates is out-of-distribution detection (OODD). We combine OODD with UQ to provide insights into the reliability of the individual predictions made by an ML model.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS1994

Stratospheric aerosol increase after eruption of Pinatubo observed with lidar and aureolemeter

An increase in the amount of stratospheric aerosol due to the Pinatubo eruption (June 12-15, 1991, 15.14 deg N, 120.35 deg E) was observed from the end of June, 1991 by a lidar in NIES (National Institute for Environmental Studies), Tsukuba (36.0 deg N, 140.1 deg E). After large fluctuations in summer of 1991, the amount of the aerosols increased in mid-September as a result of enhanced transportation from the subtropical region. In autumn and winter of 1991, dense aerosol layers were continuously observed. Aureolemeter (scanning spectral radiometer) measurements were also carried out with lidar measurements and columnar size distribution of stratospheric aerosols was estimated for some cases. Collaborative measurements with the lidar and aureolemeter provided some information on height distribution of the surface area of aerosols in late 1991.

Hayashida, Sachiko↗
NASA NTRS1983

Temperature-time distribution and thermal stresses on the RTG fins and shell during water cooling

Radioisotope thermoelectric generator (RTG) packages designed for space missions generally do not require active cooling. However, the heat they generate cannot remain inside of the launch vehicle bay and requires active removal. Therefore, before the Shuttle bay door is closed, the RTG coolant tubes attached to the heat rejection fins must be filled with water, which will circulate and remove most of the heat from the cargo bay. There is concern that charging a system at initial temperature around 200 C with water at 24 C can cause unacceptable thermal stresses in the RTG shell and fins. A computer model is developed to estimate the transient temperature distribution resulting from such charging. The thermal stresses resulting from the temperature gradients do not exceed the elastic deformation limit for the material. Since the simplified mathematical model for thermal stresses tends to overestimate stresses, it is concluded that the RTG can be cooled by introducing water at 24 C to the initially hot fin coolant tubes while the RTG is in the Shuttle cargo bay.

Turner, R. H.↗
DOE OSTI2020

Scaling carbon fluxes from eddy covariance sites to globe: synthesis and evaluation of the FLUXCOM approach

FLUXNET comprises globally distributed eddy-covariance-based estimates of carbon fluxes between the biosphere and the atmosphere. Since eddy covariance flux towers have a relatively small footprint and are distributed unevenly across the world, upscaling the observations is necessary to obtain global-scale estimates of biosphere–atmosphere exchange. Based on cross-consistency checks with atmospheric inversions, sun-induced fluorescence (SIF) and dynamic global vegetation models (DGVMs), here we provide a systematic assessment of the latest upscaling efforts for gross primary production (GPP) and net ecosystem exchange (NEE) of the FLUXCOM initiative, where different machine learning methods, forcing data sets and sets of predictor variables were employed. Spatial patterns of mean GPP are consistent across FLUXCOM and DGVM ensembles (R 2 >0.94 at 1° spatial resolution) while the majority of DGVMs show, for 70 % of the land surface, values outside the FLUXCOM range. Global mean GPP magnitudes for 2008–2010 from FLUXCOM members vary within 106 and 130 PgC yr -1 with the largest uncertainty in the tropics. Seasonal variations in independent SIF estimates agree better with FLUXCOM GPP (mean global pixel-wise R 2 ~0.75) than with GPP from DGVMs (mean global pixel-wise R 2 ~0.6). Seasonal variations in FLUXCOM NEE show good consistency with atmospheric inversion-based net land carbon fluxes, particularly for temperate and boreal regions (R 2 >0.92). Interannual variability of global NEE in FLUXCOM is underestimated compared to inversions and DGVMs. The FLUXCOM version which also uses meteorological inputs shows a strong co-variation in interannual patterns with inversions (R 2 =0.87 for 2001–2010). Mean regional NEE from FLUXCOM shows larger uptake than inversion and DGVM-based estimates, particularly in the tropics with discrepancies of up to several hundred grammes of carbon per square metre per year. These discrepancies can only partly be reconciled by carbon loss pathways that are implicit in inversions but not captured by the flux tower measurements such as carbon emissions from fires and water bodies. We hypothesize that a combination of systematic biases in the underlying eddy covariance data, in particular in tall tropical forests, and a lack of site history effects on NEE in FLUXCOM are likely responsible for the too strong tropical carbon sink estimated by FLUXCOM. Furthermore, as FLUXCOM does not account for CO 2 fertilization effects, carbon flux trends are not realistic. Overall, current FLUXCOM estimates of mean annual and seasonal cycles of GPP as well as seasonal NEE variations provide useful constraints of global carbon cycling, while interannual variability patterns from FLUXCOM are valuable but require cautious interpretation. Exploring the diversity of Earth observation data and of machine learning concepts along with improved quality and quantity of flux tower measurements will facilitate further improvements of the FLUXCOM approach overall.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2017

Temperature dependence of the photodissociation of CO 2 from high vibrational levels: 205-230 nm imaging studies of CO(X 1 Σ + ) and O( 3 P, 1 D) products

In this paper, the 205-230 nm photodissociation of vibrationally excited CO 2 at temperatures up to 1800 K was studied using Resonance Enhanced Multiphoton Ionization (REMPI) and time-sliced Velocity Map Imaging (VMI). CO 2 molecules seeded in He were heated in an SiC tube attached to a pulsed valve and supersonically expanded to create a molecular beam of rotationally cooled but vibrationally hot CO 2 . Photodissociation was observed from vibrationally excited CO 2 with internal energies up to about 20 000 cm -1 , and CO(X 1 Σ + ), O( 3 P), and O( 1 D) products were detected by REMPI. The large enhancement in the absorption cross section with increasing CO 2 vibrational excitation made this investigation feasible. The internal energies of heated CO 2 molecules that absorbed 230 nm radiation were estimated from the kinetic energy release (KER) distributions of CO(X 1 Σ + ) products in v" = 0. At 230 nm, CO 2 needs to have at least 4000 cm -1 of rovibrational energy to absorb the UV radiation and produce CO(X 1 Σ + ) + O( 3 P). CO 2 internal energies in excess of 16 000 cm -1 were confirmed by observing O( 1 D) products. It is likely that initial absorption from levels with high bending excitation accesses both the A 1 B 2 and B 1 A 2 states, explaining the nearly isotropic angular distributions of the products. CO(X 1 Σ + ) product internal energies were estimated from REMPI spectroscopy, and the KER distributions of the CO(X 1 Σ + ), O( 3 P), and O( 1 D) products were obtained by VMI. The CO product internal energy distributions change with increasing CO 2 temperature, suggesting that more than one dynamical pathway is involved when the internal energy of CO 2 (and the corresponding available energy) increases. The KER distributions of O( 1 D) and O( 3 P) show broad internal energy distributions in the CO(X 1 Σ + ) cofragment, extending up to the maximum allowed by energy but peaking at low KER values. Finally, although not all the observations can be explained at this time, with the aid of available theoretical studies of CO 2 VUV photodissociation and O + CO recombination, it is proposed that following UV absorption, the two lowest lying triplet states, a 3 B 2 and b 3 A 2 , and the ground electronic state are involved in the dynamical pathways that lead to product formation.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
NASA NTRS2010

Jovian Plasma Torus Interaction with Europa: 3D Hybrid Kinetic Simulation. First results

The hybrid kinetic model supports comprehensive simulation of the interaction between different spatial and energetic elements of the Europa-moon-magnetosphere system with respect to variable upstream magnetic field and flux or density distributions of plasma and energetic ions, electrons, and neutral atoms. This capability is critical for improving the interpretation of the existing Europa flyby measurements from the Galileo orbiter mission, and for planning flyby and orbital measurements, (including the surface and atmospheric compositions) for future missions. The simulations are based on recent models of the atmosphere of Europa (Cassidy etal.,2007;Shematovichetal.,2005). In contrast to previous approaches with MHD simulations, the hybrid model allows us to fully take into account the finite gyro radius effect and electron pressure, and to correctly estimate the ion velocity distribution and the fluxes along the magnetic field (assuming an initial Maxwellian velocity distribution for upstream background ions).Non-thermal distributions of upstream plasma will be addressed in future work. Photoionization,electron-impact ionization, charge exchange and collisions between the ions and neutrals are also included in our model. We consider two models for background plasma:(a) with O(++) ions; (b) with O(++) and S(++) ions. The majority of O2 atmosphere is thermal with an extended cold population (Cassidyetal.,2007). A few first simulations already include an induced magnetic dipole; however, several important effects of induced magnetic fields arising from oceanic shell conductivity will be addressed in later work.

Lipatov, A. S.↗
NASA NTRS2008

A New Bond Albedo for Performing Orbital Debris Brightness to Size Transformations

We have developed a technique for estimating the intrinsic size distribution of orbital debris objects via optical measurements alone. The process is predicated on the empirically observed power-law size distribution of debris (as indicated by radar RCS measurements) and the log-normal probability distribution of optical albedos as ascertained from phase (Lambertian) and range-corrected telescopic brightness measurements. Since the observed distribution of optical brightness is the product integral of the size distribution of the parent [debris] population with the albedo probability distribution, it is a straightforward matter to transform a given distribution of optical brightness back to a size distribution by the appropriate choice of a single albedo value. This is true because the integration of a powerlaw with a log-normal distribution (Fredholm Integral of the First Kind) yields a Gaussian-blurred power-law distribution with identical power-law exponent. Application of a single albedo to this distribution recovers a simple power-law [in size] which is linearly offset from the original distribution by a constant whose value depends on the choice of the albedo. Significantly, there exists a unique Bond albedo which, when applied to an observed brightness distribution, yields zero offset and therefore recovers the original size distribution. For physically realistic powerlaws of negative slope, the proper choice of albedo recovers the parent size distribution by compensating for the observational bias caused by the large number of small objects that appear anomalously large (bright) - and thereby skew the small population upward by rising above the detection threshold - and the lower number of large objects that appear anomalously small (dim). Based on this comprehensive analysis, a value of 0.13 should be applied to all orbital debris albedo-based brightness-to-size transformations regardless of data source. Its prima fascia genesis, derived and constructed from the current RCS to size conversion methodology (SiBAM Size-Based Estimation Model) and optical data reduction standards, assures consistency in application with the prior canonical value of 0.1. Herein we present the empirical and mathematical arguments for this approach and by example apply it to a comprehensive set of photometric data acquired via NASA's Liquid Mirror Telescopes during the 2000-2001 observing season.

Mulrooney, Mark K.↗
DOE OSTI2023

Privacy-Preserving Robust Consensus for Distributed Microgrid Control Applications

Consensus-based distributed control has been proposed for coordinating distributed energy resources (DERs) in microgrids (MGs). As one key component, distributed average observers are used to estimate the average of a group of reference signals (e.g., voltage, current, or power). State-of-the-art distributed average observers could lead to loss of privacy due to information exchange on the communication channels. The DERs' reference signals, which contain private information, could be inferred by an eavesdropper. In this article, a privacy-preserving distributed average observer is proposed that is based on robust consensus and uses the state decomposition method to preserve privacy. Compared to the existing methods, the proposed observer does not require the knowledge of the reference signal's derivative and gives accurate and smooth estimation, and is thus applicable for MG distributed control applications. A detailed analysis regarding the convergence and privacy properties of the proposed observer is presented. Here, the proposed observer is implemented on hardware controllers and validated in the context of distributed MG control applications through hardware-in-the-loop (HIL) tests.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2023

Effect of Na Concentration on Cs Distribution with Crystalline Silicotitanate in Tank Waste Simulants

Crystalline silicotitanate (CST) ion exchange media is currently utilized in the tank side cesium removal (TSCR) system on the Hanford site to remove Cs-137 from the tank waste supernate. As the main dose contributor to the liquid supernate in Hanford tank waste, it is necessary to remove the Cs-137 to expedite low activity waste processing and immobilization. Ongoing batch contact experiments with CST have been done to tease out the impact of group II metals, anions, potassium, and temperature on Cs removal to better understand bounding conditions for operations as well as aid in development of a wholesome isotherm model to predict Cs behavior in the tank waste. This paper discusses the impact of Na concentration on Cs exchange and provides insight into the challenges associated with activity coefficient estimations for modeling Cs distribution in tank waste matrices.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2017

Fast automated analysis of strong gravitational lenses with convolutional neural networks

Quantifying image distortions caused by strong gravitational lensing—the formation of multiple images of distant sources due to the deflection of their light by the gravity of intervening structures—and estimating the corresponding matter distribution of these structures (the ‘gravitational lens’) has primarily been performed using maximum likelihood modelling of observations. Our procedure is typically time- and resource-consuming, requiring sophisticated lensing codes, several data preparation steps, and finding the maximum likelihood model parameters in a computationally expensive process with downhill optimizers. Accurate analysis of a single gravitational lens can take up to a few weeks and requires expert knowledge of the physical processes and methods involved. Tens of thousands of new lenses are expected to be discovered with the upcoming generation of ground and space surveys. We report the use of deep convolutional neural networks to estimate lensing parameters in an extremely fast and automated way, circumventing the difficulties that are faced by maximum likelihood methods. We also show that the removal of lens light can be made fast and automated using independent component analysis of multi-filter imaging data. Our networks can recover the parameters of the ‘singular isothermal ellipsoid’ density profile, which is commonly used to model strong lensing systems, with an accuracy comparable to the uncertainties of sophisticated models but about ten million times faster: 100 systems in approximately one second on a single graphics processing unit. These networks can provide a way for non-experts to obtain estimates of lensing parameters for large samples of data.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2020

Physics-guided Deep Learning for Power System State Estimation

In the past decade, dramatic progress has been made in the field of machine learning. This paper explores the possibility of applying deep learning in power system state estimation. Traditionally, physics-based models are used including weighted least square (WLS) or weighted least absolute value (WLAV). These models typically consider a single snapshot of the system without capturing temporal correlations of system states. In this paper, a physics-guided deep learning (PGDL) method is proposed. Specifically, inspired by autoencoders, deep neural networks (DNNs) are used to learn the temporal correlations. The estimated system states from DNNs are then checked against physics laws by running through a set of power flow equations. Hence, the proposed PGDL is both data-driven and physics-guided. The accuracy and robustness of the proposed PGDL method are compared with traditional methods in standard IEEE cases. Simulations show promising results and the applicability is further discussed.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2023

Image-based solar estimates

An example device is configured to determine, based on a sky image of a portion of sky over a power distribution network and using a convolutional neural network (CNN)-based image regression model, an estimated global horizontal irradiance (GHI) value and manage or control the power distribution network using the estimated GHI value. The device may also be configured to determine, based on GHI values and aggregate load values for at least a portion of the power distribution network, using a Bayesian Structural Time Series model, an estimated photovoltaic power output value for the at least a portion of the power distribution network. The device may manage or control the power distribution network using the estimated photovoltaic power output value.

Bernstein, Andrey↗
DOE OSTI2023

Uncertainty quantification of the convolutional neural networks on permeability estimation from micro-CT scanned sandstone and carbonate rock images

Rock permeability is one of the most crucial properties affecting subsurface fluid flow behaviors. To accurately and robustly estimate the permeability, Digital Rock Physics, including micro-CT scanning technology and direct flow simulations on scanned images, has prevailed in recent years. Besides, machine learning techniques such as convolutional neural networks (CNNs) have been widely adopted and achieved success in permeability estimations directly from rock images. However, existing ML methods used for permeability estimation from rock images lack uncertainty quantification that causes unreliable predictions and overconfident estimations on out-of-distribution (OOD) samples. Here, in this work, we propose a PI3NN-CNN framework to address this problem. PI3NN-CNN consists of a CNN model for absolute permeability estimation and a PI3NN method to quantify the estimation uncertainty. It is able to quantify the uncertainty for in-distribution (InD) data with a desired confidence level, and identify OOD samples to avoid overconfident predictions. We demonstrate the method using micro-CT scanned images from two sandstone and two carbonate rocks. We found that PI3NN-CNN generates accurate predictions for InD samples, while producing high-quality prediction uncertainties regardless of the prediction accuracy. Meanwhile, PI3NN-CNN identifies OOD samples using its special network initialization scheme. The unique feature of PI3NN-CNN makes it applicable to more complex real-world image-based data for robust learning and predictions without overconfident estimations when the ground-truth information is unavailable.

58 GEOSCIENCES↗