SEARCH · Engineering Papers

Results for “distributed algorithms”

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.

SourceResearch connectionFrom yearThrough year

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

At least 1,423 records · Page 79

All sources NASA NTRS DOE OSTI

DOE OSTI2022

Efficient reconstruction and validation of heterogeneous microstructures for energy applications

The digital reconstruction of microstructures is necessary for simulations in fields ranging from geology to electrochemistry, but the state-of-the-art digital reconstruction techniques often compromise between resolution and field of view. It is challenging to retain detailed microstructure information in large-scale reconstructions. Here, this study investigates different aspects of the Yeong-Torquato algorithm based on correlation functions to make it more efficient. We achieve this goal by reducing the computational complexity of the chord-length distribution function and the two-point correlation function, applying the random sphere-packing method as the initial condition, and restricting potential voxel swaps to interfaces. In addition, a novel superposition parallel scheme is introduced to aid in searching for potential voxel swaps. The algorithm proposed is validated by comparing the pore-size distributions of reconstructed 3D custom battery electrodes from a sample dataset obtained from transmission X-ray microscopy. From a sample image with 200 x 200 pixels, the code can reconstruct a 300 x 300 x 300 structure in under 22 h and reconstruct a 400 x 400 x 400 structure in 43 h with eight cores.

42 ENGINEERING↗

DOE OSTI2020

Augmenting machine learning of energy landscapes with local structural information

We present a machine learning approach for accurately predicting formation energies of binary compounds in the context of crystal structure predictions. The success of any machine learning model depends significantly on the choice of representation used to encode the relevant physical information into machine-learnable data. We test different representation schemes based on partial radial and angular distribution functions (RDF+ADF) on Al–Ni and Cd–Te structures generated using our genetic algorithm for structure prediction. We observe a remarkable improvement in predictive accuracy upon transitioning from global to atom-centered representations, resulting in a threefold decrease in prediction errors. We show that a support vector regression model using a combination of atomic radial and angular distribution functions performs best at the formation energy prediction task, providing small root mean squared errors of 3.9 meV/atom and 10.9 meV/atom for Al–Ni and Cd–Te, respectively. We test the performance of our models against common traditional descriptors and find that RDF- and ADF-based representations significantly outperform many of those in the prediction of formation energies. The high accuracy of predictions makes our machine learning models great candidates for the exploration of energy landscapes.

Honrao, Shreyas J. (ORCID:0000000166155257)↗

DOE OSTI2024

TrustDER: Trusted, Private and Scalable Coordination of Distributed Energy Resources

In this project, the Stanford and SLAC Teams have developed a Trusted, Private and Scalable platform for coordinating Coordination of Distributed Energy Resources (TrustDER). This is a layered system that ensures private, trusted and scalable coordination and monitoring of DERs. It accommodates a variety of resources, such as solar generation, gensets and loads, with a particular focus on battery systems-based resources, as they are a transformational technology experiencing fast growth in adoption by large critical facilities. The platform can be used as standalone or added to existing aggregation systems to enable trust, privacy and resilience. TrustDER consists of layers that address each of the shortcomings of the existing state of the art. Each layer in the platform can operate independently but provides information to the layers above it to enable a novel form of overall coordination architecture. The project consists of several tasks, with each task dedicated to the design of each layer. Task 2 Resource Virtualization defined a software abstraction layer for distributed energy resources (DERs). The goal of this abstraction was to simplify the implementation of algorithms utilizing cooperation of DERs resources in a variety of use cases. Task 3 is on Secure ID for Asset Authentication. Identity Management Systems (IDMS) are a foundational infrastructure for interactions between entities (organizations, users, devices, and services). Secure ID is blockchain-based a distributed identity management system allowing (1) identity provisioning, (2) authentication, (3) authorization, and (4) identity data sharing for IoT-enabled assets on the electricity grid. In this project, the SLAC team focused on designing and testing Keymaker, a protocol for authenticating device identity managed by Secure ID. Task 5 Private and Safe Integration is focused on the design and evaluation of a DER cooperation scheme which allows for the aggregation of DERs without impacting network reliability. The approach is designed based on realistic assumptions regarding data availability, communication infrastructure limitations, and privacy. Task 6 Scalable Distributed Privacy for Information explored how virtualized batteries could be managed privately. Specifically, it examined the case in which a principal provides a partitioned battery to multiple clients. Task 7 Use Cases was to ensure that this technology was applied in relevant situations and scenarios. Primarily, this means that virtualization needed to be employed in a manner that either improved flexibility, bolstered security or privacy, or decreased costs.

25 ENERGY STORAGE↗

DOE OSTI2015

SU-E-P-16: A Feasibility Study of Using Eclipse AAA for SRS Treatement

Purpose: To commission Varian Eclipse AAA for SRS treatment and compare the accuracy with Brainlab iPlan system for clinical cases measured with radiochromic film. Methods: A 6MV AAA clinical model for a Varian TrueBeam STx is used as baseline. The focal spot and field size of the baseline model(BASE) are (1.75,0.75) and 40×40cm{sup 2} respectively. Maximum field sizes, output factors(S{sub t}), FWHM focal spot and secondary source sizes are systematically adjusted to obtain an optimized model(OPT) by comparing the calculated PDD’s, profiles, and output factors with measurements taken with a stereotactic diode(SD) and, cc01 and cc04 ion chambers in Blue Phantom. In-phantom dose distributions of clinical SRS fields are calculated using the OPT and the clinical Brainlab iPlan pencil-beam. Within the 90% isodose-line(ROI), the average dose difference between the calculations and radiochromic film measurements are assessed. Results: The maximum field, focal spot and secondary source sizes for the OPT are 15×15cm{sup 2}, (0,0), and 32.3mm respectively. The OPT St input at 1.0 and 2.0cm fields are increased by 4.5% and 1.5% from BASE. The calculated output of the BASE and OPT underestimate by 16.1%–3.2% respectively at 0.5×0.5cm{sup 2} field and 3.1%−0.02% respectively at 1.0×1.0cm{sup 2} field. The depth doses at 10cm are within 3.5% and 0.4% of measurements for 0.5×0.5 and 1.0×1.0cm{sup 2}. The ROI dose of OPT and iPlan are within 1.6% and 0.6% of film measurements for 3.0cm clinical fields. For 1.0cm fields, the ROI dose of OPT underestimate 0.0–2.0% and iPlan overestimates 1.7–2.9% relative to measurements. Conclusion: The small field dose calculation of Eclipse AAA algorithm can be significantly improved by carefully adjusting the input parameters. The larger deviation of the OPT for 0.5×0.5cm{sup 2} field from measurements can be attributed to the lowest 1.0cm field size input limit of AAA. The OPT compares reasonably well with the iPlan pencil-beam and measurements.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2023

An Online Joint Optimization–Estimation Architecture for Distribution Networks

Here in this article, we propose an optimal joint optimization-estimation architecture for distribution networks, which jointly solves the optimal power flow (OPF) problem and static state estimation (SE) problem through an online gradient-based feedback algorithm. The main objective is to enable a fast and timely interaction between the OPF decisions and state estimators with limited sensor measurements. First, convergence and optimality of the proposed algorithm are analytically established. Then, the proposed gradient-based algorithm is modified by introducing statistical information of the inherent estimation and linearization errors for an improved and robust performance of the online OPF decisions. Overall, the proposed method eliminates the traditional separation of operation and monitoring, where optimization and estimation usually operate at distinct layers and different time scales. Hence, it enables a computationally affordable, efficient, and robust online operational framework for distribution networks under time-varying settings.

24 POWER TRANSMISSION AND DISTRIBUTION↗

DOE OSTI2017

REMOVING BIASES IN RESOLVED STELLAR MASS MAPS OF GALAXY DISKS THROUGH SUCCESSIVE BAYESIAN MARGINALIZATION

Stellar masses of galaxies are frequently obtained by fitting stellar population synthesis models to galaxy photometry or spectra. The state of the art method resolves spatial structures within a galaxy to assess the total stellar mass content. In comparison to unresolved studies, resolved methods yield, on average, higher fractions of stellar mass for galaxies. In this work we improve the current method in order to mitigate a bias related to the resolved spatial distribution derived for the mass. The bias consists in an apparent filamentary mass distribution and a spatial coincidence between mass structures and dust lanes near spiral arms. The improved method is based on iterative Bayesian marginalization, through a new algorithm we have named Bayesian Successive Priors (BSP). We have applied BSP to M51 and to a pilot sample of 90 spiral galaxies from the Ohio State University Bright Spiral Galaxy Survey. By quantitatively comparing both methods, we find that the average fraction of stellar mass missed by unresolved studies is only half what previously thought. In contrast with the previous method, the output BSP mass maps bear a better resemblance to near-infrared images.

79 ASTRONOMY AND ASTROPHYSICS↗

DOE OSTI2020

Projected Future Changes in Tropical Cyclones Using the CMIP6 HighResMIP Multimodel Ensemble

Future changes in tropical cyclone properties are an important component of climate change impacts and risk for many tropical and midlatitude countries. In this study we assess the performance of a multimodel ensemble of climate models, at resolutions ranging from 250 to 25 km. We use a common experimental design including both atmosphere-only and coupled simulations run over the period 1950–2050, with two tracking algorithms applied uniformly across the models. There are overall improvements in tropical cyclone frequency, spatial distribution, and intensity in models at 25 km resolution, with several of them able to represent very intense storms. Projected tropical cyclone activity by 2050 generally declines in the South Indian Ocean, while changes in other ocean basins are more uncertain and sensitive to both tracking algorithm and imposed forcings. Coupled models with smaller biases suggest a slight increase in average TC 10 m wind speeds by 2050.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2013

Development of XFCT imaging strategy for monitoring the spatial distribution of platinum-based chemodrugs: Instrumentation and phantom validation

Purpose: Developing an imaging method to directly monitor the spatial distribution of platinum-based (Pt) drugs at the tumor region is of critical importance for early assessment of treatment efficacy and personalized treatment. In this study, the authors investigated the feasibility of imaging platinum (Pt)-based drug distribution using x-ray fluorescence (XRF, a.k.a. characteristic x ray) CT (XFCT). Methods: A 5-mm-diameter pencil beam produced by a polychromatic x-ray source equipped with a tungsten anode was used to stimulate emission of XRF photons from Pt drug embedded within a water phantom. The phantom was translated and rotated relative to the stationary pencil beam in a first-generation CT geometry. The x-ray energy spectrum was collected for 18 s at each position using a cadmium telluride detector. The spectra were then used for the K-shell XRF peak isolation and sinogram generation for Pt. The distribution and concentration of Pt were reconstructed with an iterative maximum likelihood expectation maximization algorithm. The capability of XFCT to multiplexed imaging of Pt, gadolinium (Gd), and iodine (I) within a water phantom was also investigated. Results: Measured XRF spectrum showed a sharp peak characteristic of Pt with a narrow full-width at half-maximum (FWHM) (FWHM{sub K{alpha}1}= 1.138 keV, FWHM{sub K{alpha}2}= 1.052 keV). The distribution of Pt drug in the water phantom was clearly identifiable on the reconstructed XRF images. Our results showed a linear relationship between the XRF intensity of Pt and its concentrations (R{sup 2}= 0.995), suggesting that XFCT is capable of quantitative imaging. A transmission CT image was also obtained to show the potential of the approach for providing attenuation correction and morphological information. Finally, the distribution of Pt, Gd, and I in the water phantom was clearly identifiable in the reconstructed images from XFCT multiplexed imaging. Conclusions: XFCT is a promising modality for monitoring the spatial distribution of Pt drugs. The technique may be useful in tailoring tumor treatment regimen in the future.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

DOE OSTI2021

Distributional Deep Reinforcement Learning-Based Emergency Frequency Control

Emergency frequency control is one of the most critical approaches to maintain power system stability after major disturbances. With the increasing number of grid-connected renewable energy sources, existing model-based methods of frequency control are facing up with challenges of computational speed and scalability for large-scale systems. In this paper, the emergency frequency control problem is formulated as a Markov Decision Process (MDP) and solved through a novel Distributional Deep Reinforcement Learning (DDRL) method, namely the distributional soft actor critic (DSAC) method. Compared with other RL methods that only estimate the mean value, the proposed DSAC model estimates the distribution of value function over returns. This advancement can lead to more insights and knowledge for the agent, with the benefit of a much faster and more stable learning process, and the improved frequency control performance. Here, the simulation results on IEEE 39-bus and IEEE 118-bus systems demonstrate the effectiveness and robustness of proposed models, as well as the advantage compared to other state-of-the-art DRL algorithms.

24 POWER TRANSMISSION AND DISTRIBUTION↗

DOE OSTI2024

Model Formulations: Integrating Distributed Energy Resources (DER) using Advanced Unit Commitment Models and DER Aggregation Methodologies

A distribution energy resource aggregator (DERA) constitutes a group of distribution energy resources with small generation capacities which meet the threshold to participate in the electricity wholesale market. This document provides the proposed DERA model formulation that will be implemented in the SCUC simulation’s architecture for the SCUC-DER project. Different economical assessment methodologies have been developed to incorporated bids for individual distributed resources, which include solar cost dispatch and cost model, BESS opportunity cost offer algorithm, and price sensitive demand response model. Detailed methods are proposed to aggregate individual cost offers to a DERA cost curve to bid in SCUC market while three methods are proposed to simulate DER actual dispatch. Based on the DERA models in this document, the SCUC-DER project will be able to assess the impacts of DERA on the distribution system’s operation and reliability.

24 POWER TRANSMISSION AND DISTRIBUTION↗

DOE OSTI2012

Multiparticle Simulation of Intrabeam Scattering for SuperB

In this communication we present the structure of a multiparticle tracking code to investigate intrabeam scattering effects in low emittance colliders. Simulation results obtained with particular reference to the SuperB parameters are compared with those of conventional IBS theories.and with those of a novel semi-analythical model able to predict IBS effect in terms of emittance growths. Intrabeam scattering (IBS) is associated with multiple small angle scattering events leading to emittance growth. In most electron storage rings, the growth rates arising from IBS are usually much longer than damping times due to synchrotron radiation, and its effect is not observed. However, IBS growth rates increase with bunch charge density, and for machines such as SuperB, that operate with high bunch charges and very low emittances, the IBS growth rates can be large enough to observe significant emittance increase. Several formalisms have been developed for calculating IBS growth rates in storage rings, notably those by Piwinski, Bjorken and Mtingwa, and their high energy approximations. Calculations show that IBS should be manageable in both SuperB rings. However these analytical models, based on Gaussian bunch distributions, cannot investigate some interesting aspects of IBS such as its impact during the damping process and its effect on the beam distribution. We developed a multiparticle tracking code, based on the Zenkevich-Bolshakov algorithm, to investigate these effects. In this communication we present the structure of the code and some simulation results obtained with particular reference to the SuperB parameters. Simulation results are compared with those of conventional IBS theories.

43 PARTICLE ACCELERATORS↗

DOE OSTI2023

BigNeuron: a resource to benchmark and predict performance of algorithms for automated tracing of neurons in light microscopy datasets

BigNeuron is an open community bench-testing platform with the goal of setting open standards for accurate and fast automatic neuron tracing. We gathered a diverse set of image volumes across several species that is representative of the data obtained in many neuroscience laboratories interested in neuron tracing. Here, we report generated gold standard manual annotations for a subset of the available imaging datasets and quantified tracing quality for 35 automatic tracing algorithms. The goal of generating such a hand-curated diverse dataset is to advance the development of tracing algorithms and enable generalizable benchmarking. Together with image quality features, we pooled the data in an interactive web application that enables users and developers to perform principal component analysis, t-distributed stochastic neighbor embedding, correlation and clustering, visualization of imaging and tracing data, and benchmarking of automatic tracing algorithms in user-defined data subsets. The image quality metrics explain most of the variance in the data, followed by neuromorphological features related to neuron size. Furthermore, we observed that diverse algorithms can provide complementary information to obtain accurate results and developed a method to iteratively combine methods and generate consensus reconstructions. The consensus trees obtained provide estimates of the neuron structure ground truth that typically outperform single algorithms in noisy datasets. However, specific algorithms may outperform the consensus tree strategy in specific imaging conditions. Finally, to aid users in predicting the most accurate automatic tracing results without manual annotations for comparison, we used support vector machine regression to predict reconstruction quality given an image volume and a set of automatic tracings.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2013

SCADA Protocol Anomaly Detection Utilizing Compression (SPADUC) 2013

There is a significant need to protect the nation’s energy infrastructures from malicious actors using cyber methods. Supervisory, Control, and Data Acquisition (SCADA) systems may be vulnerable due to the insufficient security implemented during the design and deployment of these control systems. This is particularly true in older legacy SCADA systems that are still commonly in use. The purpose of INL’s research on the SCADA Protocol Anomaly Detection Utilizing Compression (SPADUC) project was to determine if and how data compression techniques could be used to identify and protect SCADA systems from cyber attacks. Initially, the concept was centered on how to train a compression algorithm to recognize normal control system traffic versus hostile network traffic. Because large portions of the TCP/IP message traffic (called packets) are repetitive, the concept of using compression techniques to differentiate “non-normal” traffic was proposed. In this manner, malicious SCADA traffic could be identified at the packet level prior to completing its payload. Previous research has shown that SCADA network traffic has traits desirable for compression analysis. This work investigated three different approaches to identify malicious SCADA network traffic using compression techniques. The preliminary analyses and results presented herein are clearly able to differentiate normal from malicious network traffic at the packet level at a very high confidence level for the conditions tested. Additionally, the master dictionary approach used in this research appears to initially provide a meaningful way to categorize and compare packets within a communication channel.

24 POWER TRANSMISSION AND DISTRIBUTION↗

DOE OSTI2012

Characterization of tissue magnetic susceptibility-induced distortions for MRIgRT

Purpose: MR image geometric integrity is one of the building blocks of MRI-guided radiotherapy. In particular, tissue magnetic susceptibility-induced effects are patient-dependent and their behavior is difficult to assess and predict. In this study, the authors investigated in detail the characteristics of susceptibility ({chi}) distortions in the context of MRIgRT, including the case of two common MR-linac system configurations. Methods: The magnetic field distortions were numerically simulated for several imaging parameters and anatomical sites, i.e., brain, lung, pelvis (with air pockets), and prostate. The simulation process consisted of (a) segmentation of patient CT data into susceptibility relevant anatomical volumes (i.e., soft-tissue, bone and air/lung), (b) conversion of CT data into susceptibility masks by assigning bulk {chi} values to the structures defined at (a), (c) numerical computations of the local magnetic fields by using a finite difference algorithm, and (d) generation of the geometric distortion maps from the magnetic field distributions. For each patient anatomy, the distortions were quantified at the interfaces of anatomical structures with significantly different {chi} values. The analysis was performed for two specific orientations of the external main magnetic field (B{sub 0}) characteristic to the MR-linac systems, specifically along the z-axis for a bore MR scanner and in the (x,y)-plane for a biplanner magnet. The magnetic field local perturbations were reported in ppm. The metrics used to quantify the geometric distortions were the maximum, mean, and range of distortions. The numerical simulation algorithm was validated using phantom data measurements. Results: Susceptibility-induced distortions were determined for both quadratic and patient specific geometries. The numerical simulations showed a good agreement with the experimental data. The measurements were acquired at 1.5 and 3 T and with an encoding gradient varying between 3 and 20 mT/m by using an annular phantom mimicking the water-air and water-oil {chi} interfaces. For quadratic geometries, the magnitude of field distortion increased rapidly with the size of the inhomogeneity up to about 10 mm and then tended to plateau. This trend became more evident for materials with a larger {Delta}{chi} relative to water. The simulations showed only a slight increase in the maximum distortion values when the B{sub 0} orientation was varied with regard to the shape of the {chi} inhomogeneity. In the case of patient anatomy, the largest distortion values arose at the air-soft-tissue interface. Considering the two MR-linac system configurations and comparing the field distortion values corresponding to all organ structures, the distortions tended to be larger for the biplanar magnet. The authors provide a reference table with ppm values which can be used to easily evaluate the geometric distortions for patient data as a function of B{sub 0} and the strength of the encoding gradient. Conclusions: The susceptibility distortions were quantified as a function of multiple parameters such as the {chi} inhomogeneity size and shape, the magnitude of B{sub 0} and the readout gradient, and the orientation of B{sub 0} with respect to the sample geometry. The analysis was performed for several anatomical sites and corresponding to two B{sub 0} orientations as featured by MR-linac systems.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2016

WE-AB-207B-05: Correlation of Normal Lung Density Changes with Dose After Stereotactic Body Radiotherapy (SBRT) for Early Stage Lung Cancer

Purpose: To investigate correlation of normal lung CT density changes with dose accuracy and outcome after SBRT for patients with early stage lung cancer. Methods: Dose distributions for patients originally planned and treated using a 1-D pencil beam-based (PB-1D) dose algorithm were retrospectively recomputed using algorithms: 3-D pencil beam (PB-3D), and model-based Methods: AAA, Acuros XB (AXB), and Monte Carlo (MC). Prescription dose was 12 Gy × 4 fractions. Planning CT images were rigidly registered to the followup CT datasets at 6–9 months after treatment. Corresponding dose distributions were mapped from the planning to followup CT images. Following the method of Palma et al .(1–2), Hounsfield Unit (HU) changes in lung density in individual, 5 Gy, dose bins from 5–45 Gy were assessed in the peri-tumor region, defined as a uniform, 3 cm expansion around the ITV(1). Results: There is a 10–15% displacement of the high dose region (40–45 Gy) with the model-based algorithms, relative to the PB method, due to the electron scattering of dose away from the tumor into normal lung tissue (Fig.1). Consequently, the high-dose lung region falls within the 40–45 Gy dose range, causing an increase in HU change in this region, as predicted by model-based algorithms (Fig.2). The patient with the highest HU change (∼110) had mild radiation pneumonitis, and the patient with HU change of ∼80–90 had shortness of breath. No evidence of pneumonitis was observed for the 3 patients with smaller CT density changes (<50 HU). Changes in CT densities, and dose-response correlation, as computed with model-based algorithms, are in excellent agreement with the findings of Palma et al. (1–2). Conclusion: Dose computed with PB (1D or 3D) algorithms was poorly correlated with clinically relevant CT density changes, as opposed to model-based algorithms. A larger cohort of patients is needed to confirm these results. This work was supported in part by a grant from Varian Medical Systems, Palo Alto, CA.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2015

SU-E-J-58: Comparison of Conformal Tracking Methods Using Initial, Adaptive and Preceding Image Frames for Image Registration

Purpose: Accounting for tumor motion during radiation therapy is important to ensure that the tumor receives the prescribed dose. Increasing the field size to account for this motion exposes the surrounding healthy tissues to unnecessary radiation. In contrast to using motion-encompassing techniques to treat moving tumors, conformal radiation therapy (RT) uses a smaller field to track the tumor and adapts the beam aperture according to the motion detected. This work investigates and compares the performance of three markerless, EPID based, optical flow methods to track tumor motion with conformal RT. Methods: Three techniques were used to track the motions of a 3D printed lung tumor programmed to move according to the tumor of seven lung cancer patients. These techniques utilized a multi-resolution optical flow algorithm as the core computation for image registration. The first method (DIR) registers the incoming images with an initial reference frame, while the second method (RFSF) uses an adaptive reference frame and the third method (CU) uses preceding image frames for registration. The patient traces and errors were evaluated for the seven patients. Results: The average position errors for all patient traces were 0.12 ± 0.33 mm, −0.05 ± 0.04 mm and −0.28 ± 0.44 mm for CU, DIR and RFSF method respectively. The position errors distributed within 1 standard deviation are 0.74 mm, 0.37 mm and 0.96 mm respectively. The CU and RFSF algorithms are sensitive to the characteristics of the patient trace and produce a wider distribution of errors amongst patients. Although the mean error for the DIR method is negatively biased (−0.05 mm) for all patients, it has the narrowest distribution of position error, which can be corrected using an offset calibration. Conclusion: Three techniques of image registration and position update were studied. Using direct comparison with an initial frame yields the best performance. The authors would like to thank Dr.YeLin Suh for making the Cyberknife dataset available to us. Scholarship funding from the Natural Sciences and Engineering Research Council of Canada (NSERC) and CancerCare Manitoba Foundation is acknowledged.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2022

Post-Disaster Microgrid Formation for Enhanced Distribution System Resilience

This paper proposes a deep reinforcement learning (DRL) based approach for post-disaster critical load restoration in active distribution systems to form microgrids through network reconfiguration to minimize critical load curtailments. Distribution networks are represented as graph networks, and optimal network configurations with microgrids are obtained by searching for the optimal spanning forest. The constraints to the research question being explored are the radial topology and power balance. Unlike existing analytical and population-based approaches, which necessitate the repetition of entire analyses and computation for each outage scenario to find the optimal spanning forest, the proposed approach, once properly trained, can quickly determine the optimal, or near-optimal, spanning forest even when outage scenarios change. When multiple lines fail in the system, the proposed approach forms microgrids with distributed energy resources in active distribution systems to reduce critical load curtailment. The proposed DRL-based model learns the action-value function using the REINFORCE algorithm, which is a model-free reinforcement learning technique based on stochastic policy gradients. A case study was conducted on a 33-node distribution test system, demonstrating the effectiveness of the proposed approach for post-disaster critical load restoration.

active distribution systems↗

DOE OSTI2024

Deployment of inference as a service at the US CMS Tier-2 data centers

Coprocessors, especially GPUs, will be a vital ingredient of data production workflows at the HL-LHC. At CMS, the GPU-as-a-service approach for production workflows is implemented by the SONIC project (Services for Optimized Network Inference on Coprocessors). SONIC provides a mechanism for outsourcing computationally demanding algorithms, such as neural network inference, to remote servers, where requests from multiple clients are intelligently distributed across multiple GPUs by a load-balancing service. This talk highlights the recent progress in deploying SONIC at selected U.S. CMS Tier-2 data centers. Using realistic CMS Run3 data processing workflows, such as those containing transformer-based algorithms, we demonstrate how SONIC is integrated into the production-like environment to enable accelerated inference offloading. We will present developments from both the client and server sides, including production job and data center configurations for NVIDIA and AMD GPUs. We will also present performance scaling benchmarks and discuss the challenges of operating SONIC in CMS production, such as server discovery, GPU saturation, fallback server logic, etc.

Holzman, Burt↗