Results for “CLASSIFICATION” — Page 58

DOE OSTI2022

TULIP: An RNA-seq-based Primary Tumor Type Prediction Tool Using Convolutional Neural Networks

Background: With cancer as one of the leading causes of death worldwide, accurate primary tumor type prediction is critical in identifying genetic factors that can inhibit or slow tumor progression. There have been efforts to categorize primary tumor types with gene expression data using machine learning, and more recently with deep learning, in the last several years. Methods In this paper, we developed four 1-dimensional (1D) Convolutional Neural Network (CNN) models to classify RNA-seq count data as one of 17 highly represented primary tumor types or 32 primary tumor types regardless of imbalanced representation. Additionally, we adapted the models to take as input either all Ensembl genes (60,483) or protein coding genes only (19,758). Unlike previous work, we avoided selection bias by not filtering genes based on expression values. RNA-seq count data expressed as FPKM-UQ of 9,025 and 10,940 samples from The Cancer Genome Atlas (TCGA) were downloaded from the Genomic Data Commons (GDC) corresponding to 17 and 32 primary tumor types respectively for training and validating the models. Results: All 4 1D-CNN models had an overall accuracy of 94.7% to 97.6% on the test dataset. Further evaluation indicates that the models with protein coding genes only as features performed with better accuracy compared to the models with all Ensembl genes for both 17 and 32 primary tumor types. For all models, the accuracy by primary tumor type was above 80% for most primary tumor types. Conclusions: We packaged all 4 models as a Python-based deep learning classification tool called TULIP (TUmor CLassIfication Predictor) for performing quality control on primary tumor samples and characterizing cancer samples of unknown tumor type. Further optimization of the models is needed to improve the accuracy of certain primary tumor types.

Jones, Sara↗

DOE OSTI2024

Development and Evaluation of Ensemble Learning-based Environmental Methane Detection and Intensity Prediction Models

The environmental impacts of global warming driven by methane (CH 4 ) emissions have catalyzed significant research initiatives in developing novel technologies that enable proactive and rapid detection of CH 4 . Several data-driven machine learning (ML) models were tested to determine how well they identified fugitive CH 4 and its related intensity in the affected areas. Various meteorological characteristics, including wind speed, temperature, pressure, relative humidity, water vapor, and heat flux, were included in the simulation. We used the ensemble learning method to determine the best-performing weighted ensemble ML models built upon several weaker lower-layer ML models to (i) detect the presence of CH 4 as a classification problem and (ii) predict the intensity of CH 4 as a regression problem. The classification model performance for CH 4 detection was evaluated using accuracy, F1 score, Matthew’s Correlation Coefficient (MCC), and the area under the receiver operating characteristic curve (AUC ROC), with the top-performing model being 97.2%, 0.972, 0.945 and 0.995, respectively. The R 2 score was used to evaluate the regression model performance for CH 4 intensity prediction, with the R 2 score of the best-performing model being 0.858. The ML models developed in this study for fugitive CH 4 detection and intensity prediction can be used with fixed environmental sensors deployed on the ground or with sensors mounted on unmanned aerial vehicles (UAVs) for mobile detection.

Majumder, Reek↗

DOE OSTI2024

Development of message passing-based graph convolutional networks for classifying cancer pathology reports

Abstract Background Applying graph convolutional networks (GCN) to the classification of free-form natural language texts leveraged by graph-of-words features (TextGCN) was studied and confirmed to be an effective means of describing complex natural language texts. However, the text classification models based on the TextGCN possess weaknesses in terms of memory consumption and model dissemination and distribution. In this paper, we present a fast message passing network (FastMPN), implementing a GCN with message passing architecture that provides versatility and flexibility by allowing trainable node embedding and edge weights, helping the GCN model find the better solution. We applied the FastMPN model to the task of clinical information extraction from cancer pathology reports, extracting the following six properties: main site, subsite, laterality, histology, behavior, and grade. Results We evaluated the clinical task performance of the FastMPN models in terms of micro- and macro-averaged F1 scores. A comparison was performed with the multi-task convolutional neural network (MT-CNN) model. Results show that the FastMPN model is equivalent to or better than the MT-CNN. Conclusions Our implementation revealed that our FastMPN model, which is based on the PyTorch platform, can train a large corpus (667,290 training samples) with 202,373 unique words in less than 3 minutes per epoch using one NVIDIA V100 hardware accelerator. Our experiments demonstrated that using this implementation, the clinical task performance scores of information extraction related to tumors from cancer pathology reports were highly competitive.

59 BASIC BIOLOGICAL SCIENCES↗

DOE OSTI2021

CATMoS: Collaborative Acute Toxicity Modeling Suite

Background: Humans are exposed to tens of thousands of chemical substances that need to be assessed for their potential toxicity. Acute systemic toxicity testing serves as the basis for regulatory hazard classification, labeling, and risk management. However, it is cost- and time-prohibitive to evaluate all new and existing chemicals using traditional rodent acute toxicity tests. In silico models built using existing data facilitate rapid acute toxicity predictions without using animals. Objectives: The U.S. Interagency Coordinating Committee on the Validation of Alternative Methods Acute Toxicity Workgroup organized an international collaboration to develop in silico models for predicting acute oral toxicity based on five different endpoints: LD50 value, U.S. Environmental Protection Agency hazard categories, Globally Harmonized System for Classification and Labelling hazard categories, very toxic chemicals (LD50 =50 mg/kg), and non-toxic chemicals (LD50 >2000 mg/kg). Methods: An acute oral toxicity data inventory for 11,992 chemicals was compiled, split into training and evaluation sets, and made available to 35 participating international research groups that submitted a total of 139 predictive models. Predictions that fell within the applicability domains of the submitted models were evaluated using external validation sets. These were then combined into consensus models to leverage strengths of individual approaches. Results: The resulting consensus predictions, which leverage the collective strengths of each individual model, form the Collaborative Acute Toxicity Modeling Suite (CATMoS). CATMoS demonstrated high performance in terms of accuracy and robustness when compared to in vivo results. Discussion: CATMoS is being evaluated by regulatory agencies for its utility and applicability as a potential replacement for in vivo rat acute oral toxicity studies. CATMoS predictions for over 800,000 chemicals have been made available via the NTP’s Integrated Chemical Environment. The models are also implemented in a free, standalone open-source tool, OPERA, which allows predictions of new and untested chemicals to be made.

63 RADIATION, THERMAL, AND OTHER ENVIRON. POLLUTAN↗

DOE OSTI2024

Role of depth in optical diffractive neural networks

Free-space all-optical diffractive neural networks have emerged as promising systems for neuromorphic scene classification. Understanding the fundamental properties of these systems is important to establish their ultimate performance. Here we consider the case of diffraction by subwavelength apertures and study the behavior of the system as a function of the number of diffractive layers by employing a co-design modeling approach. We show that adding depth allows the system to achieve high classification accuracies with a reduced number of diffractive features compared to a single layer, but that it does not allow the system to surpass the performance of an optimized single layer. The improvement from depth is found to be limited to the first few layers. These properties originate from the constraints imposed by the physics of light, in particular the weakening electric field with distance from the aperture.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

DOE OSTI2026

An improved dataset for predicting mammal infecting viruses from genetic sequence information

There have been several attempts to develop machine learning (ML) models to identify human infecting viruses from their genomic sequences, with varying degrees of success. Direct comparison between models is problematic, because these models are typically trained and evaluated on different datasets with alternative data splitting schemes, features, and model performance metrics. In this paper we present a standardized dataset of mammal infecting and non-infecting viral pathogens, refined from the previous work of Mollentze et al. to include the latest literature evidence, roughly doubling the number of curated host-virus records available to the community, and new host target labels, primate and mammal. The new host labels were included for several reasons, including previous reports that classification performance is better at broader taxonomic ranks and the idea that there may be more data for primate infection that might serve as a suitable proxy for zoonotic potential and avoidance of false positives for human infection due to absence of evidence. On this dataset, we report the performance of eight machine learning models for predicting mammal-infecting viruses from their genomic sequences. We find that randomly assigning cases in our improved dataset to training/testing sets, when compared to the original assignments into training/testing in Mollentze et al., increases the overall average ROC AUC of prediction of human infection from 0.663 ± 0.070 to 0.784 ± 0.013, consistent with the reduction in phylogenetic distance between train and test sets (relative entropy change from 3.00 to 0.08). The broadest host category of mammal infection can be predicted most reliably at 0.850 ± 0.020. We share our improved dataset and code to enable standardized comparisons of machine learning methods to predict human host infections. Overall, we have presented preliminary evidence that classification of virus host infection is more tractable at higher taxonomic ranks, that unsurprisingly reducing the phylogenetic distance between training and test sets can improve predictive performance, that peptide kmer features appear to be harmful to out of sample model performance, and we are left with the question of whether models for virus host prediction can reasonably be expected to perform well in out of sample scenarios given the likelihood that viruses do not share a common ancestor. Consistent with this concern, when the data is resampled such that there is no overlap between viral families in training and test sets (relative entropy > 24), models perform no better than random chance at prediction of human infection regardless of whether kmers are included (ROC AUC 0.50 ± 0.08) or not (ROC AUC 0.50 ± 0.04).

59 BASIC BIOLOGICAL SCIENCES↗

DOE OSTI2024

Evaluation of WSR-88D Level III and MRMS Rainfall Estimates Against Rain Gauge Observations at the Savannah River Site

Rainfall data for the Savannah River Site (SRS) has been historically measured by rain gauges. These instruments serve as ground truth for most climatological and weather applications; however, gauge measurements are prone to errors or biases under certain weather conditions. Rainfall estimates from radar reflectivity values have been developed and improved over the years and serve as an alternative or supplement for gauge measurements. This study compares measurements from tipping bucket rain gauges with rainfall estimates from the NOAA NWS WSR-88D Level III hourly rainfall product and the Multi-Radar Multi-Sensor (MRMS) gauge-corrected precipitation estimate. Results show good agreement between radar derived amounts and ground measurements, with MRMS values showing correlation coefficients of 0.60-0.95 and RMSE values of less than 1.0 cm (0.4 in). The WSR-88D Level III estimates result in correlation coefficients of 0.46-0.86 and RMSE values of less than 1.3 cm (0.5 in). Few outliers are observed for each data pair and are evaluated against precipitation classification products (WSR-88D Hybrid Hydrometeor Classification product and MRMS Precipitation Flag product). The rain gauges used in this study are not part of the Hydrometeorological Automated Data System (HADS) network used to correct the MRMS estimates and therefore, results of this work provide an independent validation of the MRMS gauge correction scheme.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2025

Hydrologic Model Data for the East Fork Poplar Creek Watershed Simulated with the Advanced Terrestrial Simulator (ATS): Streamflow and Network Expansion–Contraction Dynamics

This dataset supports hydrologic modeling and stream network expansion–contraction analysis for the East Fork Poplar Creek (EFPC) Watershed in Tennessee. It includes a Jupyter notebook for model setup, model configuration files, simulation outputs, and derived products used to evaluate model performance and investigate stream dynamics under varying hydrologic conditions. The dataset was generated using the Watershed Workflow Python package and the Advanced Terrestrial Simulator (ATS), enabling integrated surface–subsurface hydrologic simulations using a stream-aligned mesh. Outputs include high-resolution time series of streamflow, active network length, water table depth, and related hydrologic variables. Also included are spatially explicit stream persistency indices and classifications of reaches as perennial or non-perennial. These data facilitate reproducibility and support further research on stream intermittency and variability in network extent.The model data archive is organized in following directories:1) model_setup_inputsContains the Watershed Workflow Jupyter notebooks (accessed through any open source code editor), selected input datasets, and resulting ATS input files, including XML files (access through any open source code editor), computational mesh (.exo files can be viewed using Paraview), and meteorological forcing files (.h5 files can be accessed through h5py python package and HDFView open source software). 2) model_outputsIncludes ATS simulation outputs relevant to this study. Time series of spatially integrated or averaged variables (e.g., streamflow, water table depth) are provided as CSV files. Select spatial fields (e.g., ponded depth and water table depth) are saved as pickled Python objects to reduce file size, and can be accessed through pickle package in Python. Key geometry objects from Watershed Workflow—such as the surface mesh and river tree—are also included to support analysis of streamflow persistency and expansion–contraction dynamics. These files can also be accessed through Watershed Workflow Python package.3) model_evaluationProvides observed streamflow time series and field survey-based flow regime classifications used to evaluate model performance. Jupyter notebooks for processing ATS outputs and comparing model predictions with observations to build confidence in the model prior to scientific analysis are also included.4) Q_L_relationshipsContains workflows for generating time series of discharge, active network length, and related hydrologic variables used in the stream network expansion–contraction analysis. Includes routines for delineating baseflow-dominated periods. For each catchment, notebooks and processed data (as pickled DataFrames accessed through Pandas Python package) are provided. 5) figure_scriptsProvides the Jupyter notebooks used to generate the figures presented in the paper.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2023

Understanding oxidation of Fe-Cr-Al alloys through explainable artificial intelligence

Abstract The oxidation resistance of FeCrAl based on alloying composition and oxidizing conditions is predicted using a combinatorial experimental and artificial intelligence approach. A neural network (NN) classification model was trained on the experimental FeCrAl dataset produced at GE Research. Furthermore, using the SHapley Additive exPlanations (SHAP) explainable artificial intelligence (XAI) tool, we explore how the NN can showcase further material insights that are unavailable directly from a black-box model. We report that high Al and Cr content forms protective oxide layer, while Mo in FeCrAl creates thick unprotective oxide scale that is vulnerable to spallation due to thermal expansion. Graphical abstract

Materials Science↗

DOE OSTI2023

Experimental Observations of the Topology of Convolutional Neural Network Activations

Topological data analysis (TDA) is a branch of computational mathematics, bridging algebraic topology and data science, that provides compact, noise-robust representations of complex structures. Deep neural networks (DNNs) learn millions of parameters associated with a series of transformations defined by the model architecture resulting in high-dimensional, difficult to interpret internal representations of input data. As DNNs become more ubiquitous across multiple sectors of our society, there is increasing recognition that mathematical methods are needed to aid analysts, researchers, and practitioners in understanding and interpreting how these models' internal representations relate to the final classification. In this paper we apply cutting edge techniques from TDA with the goal of gaining insight towards interpretability of convolutional neural networks used for image classification. We use two common TDA approaches to explore several methods for modeling hidden layer activations as high-dimensional point clouds, and provide experimental evidence that these point clouds capture valuable structural information about the model's process. First, we demonstrate that a distance metric based on persistent homology can be used to quantify meaningful differences between layers and discuss these distances in the broader context of existing representational similarity metrics for neural network interpretability. Second, we show that a mapper graph can provide semantic insight as to how these models organize hierarchical class knowledge at each layer. These observations demonstrate that TDA is a useful tool to help deep learning practitioners unlock the hidden structures of their models.

topological data analysis, deep learning↗

DOE OSTI2020

Semisupervised Learning for Seismic Monitoring Applications

The impressive performance that deep neural networks demonstrate on a range of seismic monitoring tasks depends largely on the availability of event catalogs that have been manually curated over many years or decades. However, the quality, duration, and availability of seismic event catalogs vary significantly across the range of monitoring operations, regions, and objectives. Semisupervised learning (SSL) enables learning from both labeled and unlabeled data and provides a framework to leverage the abundance of unreviewed seismic data for training deep neural networks on a variety of target tasks. We apply two SSL algorithms (mean-teacher and virtual adversarial training) as well as a novel hybrid technique (exponential average adversarial training) to seismic event classification to examine how unlabeled data with SSL can enhance model performance. In general, we find that SSL can perform as well as supervised learning with fewer labels. We also observe in some scenarios that almost half of the benefits of SSL are the result of the meaningful regularization enforced through SSL techniques and may not be attributable to unlabeled data directly. Lastly, the benefits from unlabeled data scale with the difficulty of the predictive task when we evaluate the use of unlabeled data to characterize sources in new geographic regions. Finally, in geographic areas where supervised model performance is low, SSL significantly increases the accuracy of source-type classification using unlabeled data.

58 GEOSCIENCES↗

DOE OSTI2025

Exploring Data Set Bias and Decision Support with Predictive Uncertainty Through Bayesian Approximations and Convolutional Neural Networks

Individual seismic catalogs can contain multiscale observations from fault level to global scales and associated waveforms from discrete events reflect crustal structure across many different scales and locations. Seismic network aperture, geographic location, and observation distance may not provide informative guidance or intuition on how different catalogs will behave across models trained under different conditions. We rely on uncertainty to provide guardrails for when to trust model decisions, but understanding when our uncertainty is trustworthy is an open challenge. Here, in this work, we explore Bayesian approximation methods for assigning predictive uncertainty in seismic event classification problems. We find that computationally expensive Bayesian approximations do not outperform simple ensemble methods. We also find that when exploiting multiple seismic event catalogs, joint training with data from all the catalogs combined with Bayesian approximations and supervised training for classification can obscure bias and result in less robust uncertainty while also not providing substantial performance benefits compared to training individual models for each catalog.

58 GEOSCIENCES↗

DOE OSTI2021

TX$^2$: Transformer eXplainability and eXploration

The Transformer eXplainability and eXploration (Martindale & Stewart, 2021), or TX 2 software package, is a library designed for artificial intelligence researchers to better understand the performance of transformer models (Vaswani et al., 2017) used for sequence classification. The tool is capable of integrating with a trained transformer model and a dataset split into training and testing populations to produce an ipywidget (Project Jupyter Contributors, 2021) dashboard with a number of visualizations to understand model performance with an emphasis on explainability and interpretability. The TX 2 package is primarily intended to integrate into a workflow centered around Jupyter Notebooks (Kluyver et al., 2016), and currently assumes the use of PyTorch (Paszke et al., 2019) and Hugging Face transformers library (Wolf et al., 2020). The dashboard includes visualization and data exploration features to aid researchers, including an interactive UMAP embedding graph (McInnes et al., 2018) to understand classification clusters, a word salience map that can be updated as researchers alter textual entries in near real time, a set of tools to understand word frequency and importance based on the clusters in the UMAP embedding graph, and a set of traditional confusion matrix analysis tools.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2020

HEP ML/Optimization Go Quantum – QuantISED Pilot

This poster includes sections on: Quantum annealers for classification, Quantum gate models for classification, Software and Workflows, and Quanutm optimization for LHC physics.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

DOE OSTI2020

Machine-Learning Enabled Evaluation of Probability of Piping Degradation In Secondary Systems of Nuclear Power Plants

The transition to condition-based, risk-informed automated maintenance will contribute to a significant reduction of operations and maintenance costs that account for the majority of nuclear power generation costs. Furthermore, of the operations and maintenance costs in U.S. plants, approximately 80% are labor costs. To address the issue of rising operating costs and economic viability, technologies used to perform online monitoring of piping and other secondary system structural components in commercial nuclear power plants (NPPs) are under evaluation. These online monitoring systems have the potential to identify when a more detailed inspection is needed using real time measurements, rather than at a pre-determined inspection interval thus reducing the maintenance cost. This paper describes distributed high-temperature stable fiber sensors fabricated in optical fibers through a roll-to-roll laser direct writing process using femtosecond lasers. Using phase-sensitive optical time domain reflectometry, distributed acoustic and vibration sensors can be developed and deployed to critical components and systems in NPPs to perform active measurements with spatial resolution down to 0.5-meter throughout the piping systems. Complex acoustic and vibration signatures harnessed by distributed sensors are registered and analyzed by artificial intelligence algorithms for degradation detection and flaw identification. Piping elbows with machined-in flaws were instrumented with fiber sensors. High-spatial-resolution data were used to develop and validate machine learning algorithms, including both linear and nonlinear regression, and classification. Additionally, classification and sensor analysis were also performed for data analysis. The paper concludes with recommendations and future work on applications of machine learning enabled high-resolution fiber sensors for piping degradation monitoring in current or future NPPs.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

DOE OSTI2020

Removal of High Specific Activity Fission Products from Uranyl Sulfate Waste Solutions

The Savannah River National Laboratory (SRNL) is currently providing support to SHINE Medical Technologies (SHINE) which plans to deploy a low energy, accelerator-based neutron source to fission low enriched U in a uranyl sulfate target solution for 99 Mo production. The 99 Mo is initially separated from the fission products and target solution by an extraction column. Subsequent washing of the column will generate waste solutions containing residual U and fission product activity. A small number of high specific activity fission products (e.g., 90 Sr, 137 Cs, and 144 Ce) in these streams will likely control the classification of the low level waste (LLW). If a sufficient amount of the high specific activity isotopes are separated from the SHINE waste streams and concentrated into a waste form, it would be possible to treat a majority of the wash solutions from the column operations as a lower class of LLW (Class A versus Class B or C or Class B versus Class C). The high specific activity fission product elements could then be dispositioned as a much smaller volume of waste rather than requiring the disposal of the entire waste stream at the higher waste classification. The Savannah River Site (SRS) has experience with using monosodium titanate (MST) and crystalline silicotitanate (CST) to remove Cs and Sr from high salt content waste solutions generated during the reprocessing of nuclear fuels and targets. Both of these materials have worked very well for their intended purposes at the SRS where the fission product elements are removed from highly alkaline waste. On the other hand, SHINE waste streams from the extraction column contain H 2 SO 4 which makes the solution acidic. Additionally, the SRS waste streams do not contain other fission product elements such as transition metals or lanthanides because they precipitate upon neutralization of the SRS waste and are not present in the supernate which is dispositioned as LLW following treatment. As such, there are inherent differences between SHINE and SRS waste treatment strategies. Savannah River National Laboratory was tasked with performing scoping studies to see if MST and CST would remove Sr, Cs, and Ce from an acidic mixed metal simulant solution. Batch contact experiments were performed using MST and two CST type materials. The MST material is a 15 wt % powder in 0.15 M NaOH slurry. The MST showed low adsorption for elements of interest from acidic solution. Furthermore, the powder size makes MST non-ideal for column operations. A CST IE-911 ion exchange material had high Cs adsorption, moderate Sr, and marginal Ce adsorption. Based on adsorption of all species, the ion exchange capacity was found to be 0.032 meq/mL. A bench-top column experiment to measure elemental breakthrough curves was performed using CST IE-911 where chromatographic separations of the mixed simulant were expected to occur. While most elements behaved as expected, the lanthanide series, containing Ce, broke though the column earlier than expected. The second CST material, CST R9120, displayed high adsorption for all elements in the acidic mixed simulant solutions in a batch contact study, and had a calculated loading capacity of 0.091 meq/mL. Future studies to develop a waste treatment flowsheet should focus on CST R9120 to treat SHINE waste solutions.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

DOE OSTI2020

Characterization and Analysis of the Energy-Reporting Accuracy of Connected Devices

Emerging energy-efficient building systems increasingly exhibit greater functionality, often requiring multiple operating modes (e.g. white-tunability for lighting products and data traffic for devices with networked, integrated sensors). This increased functionality makes energy consumption estimates more complex. Given that these functions consume energy, the energy performance of such building systems is dependent on what operating modes they use and how much time they spend in each mode. Devices and systems that can report their own energy consumption mitigate this energy-performance uncertainty. This study explores the energy-reporting accuracy of market-available connected electrical outlets. The study considers two residential-market products (five units each, one outlet per unit) and three commercial-market products (two units each, 18 to 24 outlets per unit) with the ability to report power drawn and/or energy consumed by devices connected to their receptacles. The products were purchased through typical market channels. Pacific Northwest National Laboratory (PNNL) conducted testing in December 2018 at its Connected Lighting Test Bed (CLTB), using a custom-developed test setup and method adapted from industry standards. The setup collected energy-consumption data reported by the outlet devices under test (DUTs) at one-minute intervals and compared that data with measurements taken by a reference meter over a range of test conditions. The residential products reported power draw but not interval or cumulative energy consumption. The commercial products reported both power draw and cumulative energy consumption. Relative reporting error (RRE) was calculated for all measurements, and analysis of the results revealed variations across devices and test conditions. The total number of measurements (50 for each residential product, 60 for each commercial product) offers an appreciable comparison of performance at the make/model level. The average RRE of the residential products derived from reported power draw was -0.02% and -1.20%. The average RRE for two of the three the commercial products derived from reported power draw was worse than those of the residential products (-2.40%, -2.72%, -0.36%). The internal integration of power over time, used to calculate cumulative energy consumption, typically occurs at current and voltage sampling rates much higher than once per minute. This suggests that the average commercial-product RRE derived from reported energy consumption should be very consistent and better than performance based on reported power draw. However, the RRE derived from reported energy consumption varied significantly across the three makes of commercial-market products and was uniformly less accurate than performance based on reported power draw. Subsequent analysis identified a number of root causes for this decrease in performance, most of which were related to reporting resolution. The goals of this study are to generate awareness of building systems capable of reporting their own energy consumption, further interest in the value of energy data for a variety of uses, draw attention to how the accuracy of reported metrics can be characterized, and quantify the performance variation found in marketavailable products. The results of this study and subsequent related work may be relevant to stakeholders in industry-specification and standards-development organizations. The methods this study employs could inform test and measurement procedures and performance classifications for connected outlets, lighting products, and other building systems capable of reporting their own energy consumption. The study concludes with stakeholder recommendations, including the following: • Energy-reporting device and system manufacturers developing products that report energy consumption should characterize the accuracy of reported metrics using a reference meter calibrated by an independent laboratory that was accredited by an ILAC MRA signatory (and whose scope of accreditation explicitly covers energy measurement), and should include this information on product data sheets. • Standards and specification development organizations should develop application-specific performance classifications that end users can understand and relate to their energy-data use needs (e.g., 2% accuracy class for utility streetlight energy billing needs, or 10% accuracy class for ESCO performance verification needs). • Current or potential owners, operators, and specifiers of energy-reporting building systems should rigorously analyze the dependency of current and planned energy-data use cases on accuracy, noting in particular the dependence (or lack thereof) on relative vs. absolute accuracy, and on trueness vs. precision (i.e., repeatability), and should communicate use-case needs to industry standards and specification organizations.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

DOE OSTI2020

Characterization and Analysis of the Energy-Reporting Accuracy of Connected Devices

Emerging energy-efficient building systems increasingly exhibit greater functionality, often requiring multiple operating modes (e.g. white-tunability for lighting products and data traffic for devices with networked, integrated sensors). This increased functionality makes energy consumption estimates more complex. Given that these functions consume energy, the energy performance of such building systems is dependent on what operating modes they use and how much time they spend in each mode. Devices and systems that can report their own energy consumption mitigate this energy-performance uncertainty. This study explores the energy-reporting accuracy of market-available connected electrical outlets. The study considers two residential-market products (five units each, one outlet per unit) and three commercial-market products (two units each, 18 to 24 outlets per unit) with the ability to report power drawn and/or energy consumed by devices connected to their receptacles. The products were purchased through typical market channels. Pacific Northwest National Laboratory (PNNL) conducted testing in December 2018 at its Connected Lighting Test Bed (CLTB), using a custom-developed test setup and method adapted from industry standards. The setup collected energy-consumption data reported by the outlet devices under test (DUTs) at one-minute intervals and compared that data with measurements taken by a reference meter over a range of test conditions. The residential products reported power draw but not interval or cumulative energy consumption. The commercial products reported both power draw and cumulative energy consumption. Relative reporting error (RRE) was calculated for all measurements, and analysis of the results revealed variations across devices and test conditions. The total number of measurements (50 for each residential product, 60 for each commercial product) offers an appreciable comparison of performance at the make/model level. The average RRE of the residential products derived from reported power draw was -0.02% and -1.20%. The average RRE for two of the three the commercial products derived from reported power draw was worse than those of the residential products (-2.40%, -2.72%, -0.36%). The internal integration of power over time, used to calculate cumulative energy consumption, typically occurs at current and voltage sampling rates much higher than once per minute. This suggests that the average commercial-product RRE derived from reported energy consumption should be very consistent and better than performance based on reported power draw. However, the RRE derived from reported energy consumption varied significantly across the three makes of commercial-market products and was uniformly less accurate than performance based on reported power draw. Subsequent analysis identified a number of root causes for this decrease in performance, most of which were related to reporting resolution. The goals of this study are to generate awareness of building systems capable of reporting their own energy consumption, further interest in the value of energy data for a variety of uses, draw attention to how the accuracy of reported metrics can be characterized, and quantify the performance variation found in marketavailable products. The results of this study and subsequent related work may be relevant to stakeholders in industry-specification and standards-development organizations. The methods this study employs could inform test and measurement procedures and performance classifications for connected outlets, lighting products, and other building systems capable of reporting their own energy consumption. The study concludes with stakeholder recommendations, including the following: • Energy-reporting device and system manufacturers developing products that report energy consumption should characterize the accuracy of reported metrics using a reference meter calibrated by an independent laboratory that was accredited by an ILAC MRA signatory (and whose scope of accreditation explicitly covers energy measurement), and should include this information on product data sheets. • Standards and specification development organizations should develop application-specific performance classifications that end users can understand and relate to their energy-data use needs (e.g., 2% accuracy class for utility streetlight energy billing needs, or 10% accuracy class for ESCO performance verification needs). • Current or potential owners, operators, and specifiers of energy-reporting building systems should rigorously analyze the dependency of current and planned energy-data use cases on accuracy, noting in particular the dependence (or lack thereof) on relative vs. absolute accuracy, and on trueness vs. precision (i.e., repeatability), and should communicate use-case needs to industry standards and specification organizations.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗