Search NASA⌕ Search

SEARCH · Search NASA

Results for “validation dataset”

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 109 records · Page 6
All sourcesNASA NTRSDOE OSTI
DOE OSTI2026

Direct Measurement of ICRF-Enhanced Plasma Potentials on WEST Using Reciprocating Emissive Probes

An extensive documentation of ICRF-enhanced plasma potentials has been conducted over two experimental campaigns on the WEST tokamak using reciprocating emissive probes magnetically connected to two ICRF antennas. The collected data spans a wide range of antenna electrical settings (coupled power, toroidal phasing, left–right power balance) and plasma parameters (density at the antenna limiter above and below the lower hybrid resonance, plasma current, minority fraction). By scanning the edge safety factor across multiple probe plunges, the magnetic connection between the probe and the antenna varied, enabling the construction of a 2D map of the plasma and floating potentials around an active ICRF antenna. This dataset will be used to validate RF simulation tools equipped with the sheath boundary condition and used to predict RF rectified potentials and ICRF-induced impurity sputtering in future machines. This paper presents the diagnostic and some initial measurements, while the rest will be reported elsewhere.

Diab, Raymond [Massachusetts Inst. of Technology (↗
DOE OSTI2024

Estimating UV-B, UV-Erithemic, and UV-A Irradiances From Global Horizontal Irradiance and MERRA-2 Ozone Column Information

The ground ultraviolet (UV) solar radiation is relevant due to its impacts on plastics degradation (mainly UVA) and on human health (UVB and erithemic UV (UVE)). UV ground measurements are not as ubiquitous as the relatively common global horizontal irradiance (GHI) measurements. Three simple models that estimate the UVA, UVB, and UVE components of solar irradiance from GHI and ozone column information are locally adjusted and validated. Five one-minute datasets from three sites in southeastern South America and two in the United States are used for simultaneous solar irradiance and UV data. All sites correspond to temperate mid-latitude regions. Simultaneous atmospheric total ozone column information is obtained from the reanalysis modern-era retrospective analysis for research and applications (MERRA-2) database for each site. Aside from locally adjusted models, average models with a single set of coefficients are also evaluated. For instance, the best average model is able to estimate UVE with a typical uncertainty below 12% and mean biases between +-3%, relative to the average of the measurements. Similar results are reported for the UVB and UVA components. These results, which can be useful in regions with similar climate and geography, provide a simple way to estimate UV irradiance under all-sky conditions with known uncertainty. This is an alternative to global satellite-based UV estimates, which can have high uncertainties at specific locations. Because MERRA-2 information has a global coverage, when coupled with good satellite-based estimates for GHI, UV irradiances can be estimated by this method over a large territory.

environmental UV radiation↗
DOE OSTI2026

Effects of 9.5 Years of Whole-Soil Warming on the Fatty Acid and n-Alkanes Composition in Bulk Soil and Density Fractions at Blodgett Experimental Forest, California, USA

Original data of molecular data (fatty acids and n-alkanes) including concentrations and calculated molecular proxies in a whole-soil warming experiment at the Blodgett Forest Research Station after 9.5 years of warming. The study site has a Mediterranean climate with annual average temperature of 12.5 ℃ and annual average precipitation of 1774 mm. The study site is characterized by a mesic Ultic Alfisol formed from granitic parent material, corresponding to a Dystric Cambisol under the World Reference Base for Soil Resources (WRB) classification system. Experimental warming is applied throughout the soil profile to a depth of 1 m using vertically embedded heating cables that raise soil temperature by 4 °C relative to ambient conditions. Soil samples were collected on 1 May 2023, after the experiment had been operating continuously for about 9.5 years since its initiation in January 2014.The data has been processed from raw data and cross-validated by other peers. The dataset includes: - Bulk_Fattyacid_9.5-year_Soil_Warming_Blodgett, California, USA: fatty acid concentrations and proxies including Carbon Preference Index (CPI) and Average Chain Length (ACL) of bulk soil organic carbon; - Fractions_Fattyacid_9.5-year_Soil_Warming_Blodgett, California, USA: fatty acid concentrations and proxies including CPI and ACL of free particulate organic matter (fPOM) and mineral-associated organic matter (MAOM); - Bulk_Alkanes_9.5-year_Soil_Warming_Blodgett, California, USA: n-alkanes concentrations and proxies including CPI and ACL of bulk soil organic carbon; - Fractions_Alkanes_9.5-year_Soil_Warming_Blodgett, California, USA: n-alkanes concentrations and proxies including CPI and ACL of fPOM and MAOM; - n-Alkanes_All_Monomer_Concentration_9.5-year_Soil_Warming_Blodgett, California, USA: concentration of all the n-alkane monomers identified and integrated for bulk soil, fPOM and MAOM; - Fattyacid_All_Monomer_Concentration_9.5-year_Soil_Warming_Blodgett, California, USA: concentration of all the fatty acid monomers including diacids identified and integrated for bulk soil, fPOM, and MAOM. All data are provided in CSV format and can be viewed using Microsoft Excel. We specifically look at fatty acids (FA) and n-alkanes in bulk soil, fPOM and MAOM and calculated molecular proxies such as CPI and ACL to understand the source of oragnic carbon (with ACL) and degree of decomposition (CPI) of each soil fraction. Due to lack of long-chain fatty acids (carbon number ⩾ 20), microorganism-derived organic carbon is characterized by shorter ACL in comparison to plant-derived organic carbon. Fresh SOC is characterized by even-over-odd dominance for fatty acids and odd-over-even dominance for n-alkanes. Therefore, CPI indicates whether soil organic carbon (SOC) represents fresh input (CPI > 10) or is strongly decomposed (close to 1). The research questions should be then, after 9.5-year warming: 1. whether the relative contribution between microorganism-derived and plant-derived SOC in each soil fraction? 2. whether fPOM became more decomposed whereas MAOM remained relatively persistent in each soil fraction across the soil depth?

Carbon↗
DOE OSTI2026

Hydrogen Dispersion Modeling for Development of Smart Distributed Monitoring

Studying hydrogen dispersion is crucial for ensuring the safe and effective deployment of hydrogen as an energy carrier. This study presents a comprehensive CFD modeling framework for simulating hydrogen dispersion at a real-world hydrogen production, storage, and utilization facility. Utilizing the Hydrogen Research Facility under the Advanced Research on Integrated Energy Systems (ARIES) at the National Renewable Energy Laboratory's (NREL) Flatirons campus, controlled hydrogen releases at 27 kg-H2/hr were simulated. The model incorporated site-specific atmospheric conditions, including hourly wind speeds and temperatures recorded between 8 AM and 8 PM from October to December 2023. To reduce computational demands, a statistical reduction technique was applied to condense the dataset to 100 representative scenarios, validated by statistical tests for wind speeds and power law coefficients. Simulations were conducted using the Reynolds-Averaged Navier-Stokes equations. Results demonstrated that wind speed substantially influences hydrogen dispersion, with low wind conditions forming concentrated clouds and higher wind speeds stretching the plume. Additionally, clustering analysis informed optimal sensor placement at various elevations with up to 10 sensor locations on each elevation. This framework offers a robust approach for understanding hydrogen behavior in ambient conditions and informing detection strategies.

08 HYDROGEN↗
DOE OSTI2025

Protein-ligand binding affinity prediction using multi-instance learning with docking structures

Recent advances in 3D structure-based deep learning approaches demonstrate improved accuracy in predicting protein-ligand binding affinity in drug discovery. These methods complement physics-based computational modeling such as molecular docking for virtual high-throughput screening. Despite recent advances and improved predictive performance, most methods in this category primarily rely on utilizing co-crystal complex structures and experimentally measured binding affinities as both input and output data for model training. Nevertheless, co-crystal complex structures are not readily available and the inaccurate predicted structures from molecular docking can degrade the accuracy of the machine learning methods. We introduce a novel structure-based inference method utilizing multiple molecular docking poses for each complex entity. Our proposed method employs multi-instance learning with an attention network to predict binding affinity from a collection of docking poses. We validate our method using multiple datasets, including PDBbind and compounds targeting the main protease of SARS-CoV-2. The results demonstrate that our method leveraging docking poses is competitive with other state-of-the-art inference models that depend on co-crystal structures. This method offers binding affinity prediction without requiring co-crystal structures, thereby increasing its applicability to protein targets lacking such data.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Geothermal well testing pressure prediction by using a hybrid transformer model system: FORGE well use case

Geothermal has huge potential to become an indispensable component in achieving the goal of sustainable energy economy, given its capability to provide consistent baseload power to the electric grid. Injection tests are crucial in geothermal energy system as they naturally help to evaluate reservoir properties, understand fluid flow and even enhance reservoir performance. In this research, we developed a hybrid model system that integrates machine learning (ML) regression, a physics-based mathematical model, and transformer deep learning. Trained and validated using FORGE injection test dataset, this system can forecast the pressure variations both upward and downward over time. The pressure prediction achieved prediction accuracy within 3-6% variance of true pressure values. The system can significantly save time and reduce costs by testing only a few cycles and then using model predictions for further analysis, instead of conducting additional real injection cycle tests. The developed model system also holds promise for designing injection test processes and maintaining well production in geothermal energy. Presented at the IMAGE ‘25 Conference led by Shell.

FORGE↗
DOE OSTI2025

Alaska's Rural Building Stock: a Validation Study Using ResStock and Field Data

The availability of accurate national data on demographics, building stock, and energy use is vital for modeling residential buildings and evaluating decarbonization strategies. However, rural and Indigenous populations, including those in rural Alaska, are typically underrepresented in these datasets. These communities face unique challenges due to their remote locations, severe weather conditions, and limited access to resources, resulting in high energy burden. This report examines how rural Alaskan communities are underrepresented in the ResStock housing model and highlights the need for improved data to address their unique housing and energy challenges. Thus, this report examines the representation of rural Alaskan communities within the national housing stock model, ResStock. A validation study was conducted, considering ResStock, Field Data and Aerial and 3D-view data collection (A3DDC) datasets. The validation process started by using the down selecting approach on the ResStock building stock dataset. For the purpose of this study, only the rural Alaska Boroughs and Census areas located in ASHRAE IECC Climate Zone 8 were considered to ensure a more accurate and fair comparison with the field data, which was collected in rural areas located in climate zone 8, specifically within the Nome Census area. While ResStock may accurately represent several characteristics of the building stock for rural Alaska, some differences between modeled, field data, and aerial and 3D-view data collection datasets were identified. The following building characteristics have a high impact on modeled energy consumption and demonstrated large differences: Revisit heating setpoints and consider a substantially higher setpoint distribution, it could potentially address "missing loads" if this is the case. Develop and include Toyo heating in future modeling for ResStock and EnergyPlus. Remove natural gas as a water heater fuel type outside of North Slope County. Foundation type updated to have more crawlspaces rather than basements. Infiltration rates need reexamination for a larger distribution toward higher infiltration rates. Include more vinyl and less brick in exterior wall type and revise wall color for greater proportion of light rather than dark color. Roof material revised from majority shingles to majority metal. Update number of occupants to higher number of occupant count. Building orientation represents a higher proportion of south facing buildings rather than relatively equal. The findings suggest that updating ResStock's probability logic could better represent rural Alaskan buildings. ResStock can be utilized to identify the best upgrades or energy efficiency and energy efficiency improvements, helping community leaders in making more informed decisions.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2026

Scatter Removal using Black Body Grids and Determining the Minimum Resolvable Hydrogen Concentration at MARS

This dataset was used to develop and validate a scatter-correction pipeline for neutron radiographs acquired at the MARS beamline, based on the methodology of Carminati et al. (2019). Scatter removal was performed using 36 gadolinium black bodies to characterize and remove the spatially varying neutron scatter field that cannot be eliminated by conventional open-beam normalization. The dataset comprises eight samples - four single-crystal nickel and four polycrystalline austenitic 316L stainless steel. Within each material system, samples were pre-charged with hydrogen gas at four charging pressures (1.5, 5, 12.5, and 20 kpsi). Following scatter correction, the pipeline was applied to all radiographs to recover quantitative attenuation maps. After subtracting the known attenuation contribution of the metal matrix, spatial maps of the hydrogen attenuation coefficient were obtained for each sample. This data was then used to construct a calibration curve relating hydrogen attenuation to hydrogen concentration and to determine the minimum resolvable hydrogen concentration at the MARS beamline. The resulting dataset provides a pressure-resolved calibration standard for quantitative hydrogen mapping using neutron radiography.

36 MATERIALS SCIENCE↗
DOE OSTI2026

Meta-analysis of North American Arctic and boreal aboveground biomass datasets: assessing accuracy, dynamics, and similarities

The North American arctic and boreal regions (ABRs) are rapidly warming and experiencing intensifying disturbances. Accurately quantifying aboveground biomass (AGB) is critical for understanding the impacts of these changes on the carbon cycle and for designing climate change mitigation strategies. Several AGB maps have been developed for the North American ABRs, including recent contributions from National Aeronautics and Space Administration’s Arctic-Boreal Vulnerability Experiment (ABoVE) campaign. However, these maps differ widely in training data, methodology, and resulting AGB density estimates. Presently, a comprehensive comparative evaluation is lacking, making it difficult for users to select datasets suited to their research or management needs. Here, in this study, we conducted a comparative analysis of nine AGB density datasets across North American ABRs, specifically for Alaska and Canada. We (1) summarized AGB by ecoregion and Canadian provinces, (2) evaluated their accuracy against field-based measurements, (3) analyzed spatial and temporal similarities among datasets, and (4) assessed their ability to capture disturbance (fire and harvest) impacts on AGB. We found substantial variation in regional and local AGB estimates across datasets, with overall accuracy ranging from R 2 = 0.25–0.62 and Bias% from −47.8% to 69.9% when validated against field plots. Despite these differences, most datasets have comparatively consistent spatial patterns in AGB (r > 0.8 for most cases). In contrast, agreement on the temporal patterns of AGB change is generally low. We found datasets with spatial resolutions ⩽300 m are capable of capturing disturbance impacts on AGB dynamics, though sensitivity varies across products. Our findings and dataset summary provide guidance for selecting appropriate AGB datasets for different applications within our study area. Our analysis also highlights the need to decrease map bias and increase capability to detect temporal change to decrease uncertainty of AGB datasets potentially by using training data which is representative of major plant functional types within the mapped area.

ABoVE↗
DOE OSTI2025

Datasets of Faults in Variable Air Volume Terminal Units in a Multi-Zone Commercial Building

Faults in HVAC systems can decrease system efficiency and equipment lifespan, leading to 5%–30% of energy consumption being wasted in commercial buildings. We identified two common faults in HVAC variable air volume systems: a stuck damper fault in the variable air volume terminal unit and a discharge airflow sensor fault. We conducted three sets of damper stuck tests and two sets of airflow sensor tests, each including a fault-free scenario and scenarios with varying levels of faults, over one day. The faults were implemented in Oak Ridge National Laboratory’s two-story Flexible Research Platform building to generate a high-quality, well-controlled dataset covering fault-induced and fault-free scenarios. The test building, fault test scenarios, and data validation are described here. The open-source dataset includes 1 min intervals of weather and building data on the presence and absence of building faults. This dataset can be used to analyze the effects of HVAC system faults on system operation and indoor building conditions, and to develop or evaluate a fault detection and diagnosis algorithm.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2025

Galaxy cluster matter profiles - I. Self-similarity, mass calibration, and observable-mass relation validation employing cluster mass posteriors

We present a study of the weak lensing inferred matter profiles ΔΣ(R) of 698 South Pole Telescope (SPT) thermal Sunyaev-Zel’dovich effect (tSZE) selected and MCMF optically confirmed galaxy clusters in the redshift range 0.25 < z < 0.94 that have associated weak gravitational lensing shear profiles from the Dark Energy Survey (DES). Rescaling these profiles to account for the mass dependent size and the redshift dependent density produces average rescaled matter profiles ΔΣ(R/R200c)/(ρcritR200c) with a lower dispersion than the unscaled ΔΣ(R) versions, indicating a significant degree of self-similarity. Galaxy clusters from hydrodynamical simulations also exhibit matter profiles that suggest a high degree of self-similarity, with RMS variation among the average rescaled matter profiles with redshift and mass falling by a factor of approximately six and 23, respectively, compared to the unscaled average matter profiles. We employed this regularity in a new Bayesian method for weak lensing mass calibration that employs the so-called cluster mass posterior P(M200|ζ̂, λ̂, z), which describes the individual cluster masses given their tSZE (ζ̂) and optical (λ̂, z) observables. This method enables simultaneous constraints on richness λ-mass and tSZE detection significance ζ-mass relations using average rescaled cluster matter profiles. We validated the method using realistic mock datasets and present observable-mass relation constraints for the SPT×DES sample, where we constrained the amplitude, mass trend, redshift trend, and intrinsic scatter. Our observable-mass relation results are in agreement with the mass calibration derived from the recent cosmological analysis of the SPT×DES data based on a cluster-by-cluster lensing calibration. Our new mass calibration technique offers a higher efficiency when compared to the single cluster calibration technique. We present new validation tests of the observable-mass relation that indicate the underlying power-law form and scatter are adequate to describe the real cluster sample but that also suggest a redshift variation in the intrinsic scatter of the λ-mass relation may offer a better description. In addition, the average rescaled matter profiles offer high signal-to-noise ratio (S/N) constraints on the shape of real cluster matter profiles, which are in good agreement with available hydrodynamical ΛCDM simulations. This high S/N profile contains information about baryon feedback, the collisional nature of dark matter, and potential deviations from general relativity.Key words: gravitational lensing: weak / galaxies: clusters: general / large-scale structure of Universe

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2026

A generative machine learning model for designing metal hydrides applied to hydrogen storage

Developing new metal hydrides is a critical step toward efficient hydrogen storage in carbon-neutral energy systems. However, existing materials databases, such as the Materials Project, contain a limited number of well-characterized hydrides, which constrains the discovery of optimal candidates. This work presents a framework that integrates causal discovery with a lightweight generative machine learning model to generate novel metal hydride candidates that may not exist in current databases. Using a dataset of 450 samples (270 training, 90 validation, and 90 testing), the model generates 1000 candidates. After ranking and filtering, six previously unreported chemical formulas and crystal structures are identified, four of which are validated by density functional theory simulations and show strong potential for future experimental investigation. Overall, the proposed framework provides a scalable and time-efficient approach for expanding hydrogen storage datasets and accelerating materials discovery.

generative model↗
DOE OSTI2025

Solar Radiation Measurements (CRADA Final Report)

Long-term solar radiation measurements are important for understanding solar resource availability and variability for systems design and deployment of cost-effective solar resource technologies. Additionally high-quality traceable solar measurements are required to validate satellite-based solar resource datasets that provide high-resolution long-term data covering the US. The purpose and intent of this agreement is to collect long-term solar radiation and meteorological measurements from the state-of-the-art facility at the University of Arizona to meet the above needs.

14 SOLAR ENERGY↗
DOE OSTI2024

SCEC/USGS Community Stress-Drop Validation Study: How Spectral Fitting Approaches Influence Measured Source Parameters

Spectral source parameters used to estimate an earthquake’s stress drop (⁠Δσ⁠) can vary significantly across measurement approaches. The Statewide California Earthquake Center/U.S. Geological Survey Community Stress‐Drop Validation Study was initiated to compare source parameter estimates, focusing initially on a dataset from the 2019 Ridgecrest earthquake sequence. As part of that validation effort, here we focus on one potential source of uncertainty: whether spectral fitting approaches alone, applied to a common set of spectra from the 2019 Ridgecrest sequence result in different source parameter estimates. By using a common set of benchmark spectra analyzed across a consistent frequency band of 1–40 Hz, we eliminate many sources of variability. A subgroup of validation study participants volunteered to estimate the low‐frequency displacement (⁠Ω 0 ⁠) and corner frequency (⁠ƒ c ⁠) by fitting a smooth function to benchmark displacement spectra. Participants used linear‐ or log‐sampled spectra, assumed a Brune or Boatwright spectral model, and applied different misfit criteria. We compare 17 approaches used to estimate ⁠Ω 0 ⁠, ƒ c ⁠, and Δσ for 54 earthquake spectra. Our results reveal that 35% of events have Δσ estimates within a factor of two, whereas others exhibit variations exceeding an order of magnitude. The variability in and can largely be attributed to whether a spectrum is consistent with the smooth function of an idealized simple crack model. The trade‐off between Ω 0 and ƒ c may be more pronounced when using linearly sampled spectra, as higher frequency spectral bumps control the fits. As expected, methods that assumed a Boatwright model tended to have lower Ω 0 and somewhat higher ƒ c compared to those assuming a Brune model, although resulting Δσ estimates are similar. Finally, when compared to the overall validation study results, the fitting approach alone may account for between 5% and 90% (25% on average) of the total variability in spectral Δσ⁠.

58 GEOSCIENCES↗
DOE OSTI2026

A physically based mechanical model for Mullins effect in thermoplastic polyurethanes

Despite decades of research, connecting the chemical and physical structure of thermoplastic polyurethanes to their mechanical properties remains highly challenging. Of particular note are their large-deformation and rate-dependent behaviors, which vary greatly with molecular chemistry, including the type and relative content of soft and hard segments. In this work, we develop a physically motivated mechanical theory for predicting the behavior of thermoplastic polyurethanes. The theory incorporates a representation of microstructural evolution during mechanical deformation, which captures the signatures of stress softening over cyclic loading (commonly referred to as the Mullins effect). There are only eight physically motivated fitting parameters, including a direct dependence on the hard segment fraction. The model predicts that increasing the hard segment fraction leads to higher stiffness and greater energy dissipation, in quantitative agreement with published experimental data. Furthermore, we provide a comprehensive analysis of the model and validate its predictions across several independent datasets focused on mechanical characterization. Direct comparisons to experimental data demonstrate its predictive capability on the effect of loading rate, cyclic deformations, and applied tension or compression. Altogether, this work establishes a predictive framework that connects polymer chemistry and microstructure to emergent mechanical behaviors.

36 MATERIALS SCIENCE↗
DOE OSTI2024

Toroidal injection angle dependence of EC assisted plasma initiation at DIII-D

An experimental scan of the electron cyclotron waves (EC) toroidal injection angle in plasma breakdown is performed at the DIII-D tokamak. The second harmonic, extraordinary mode EC is used for the study. The dependence of n e and T e on the EC injection angle cannot be conclusively drawn from this study due to the large error bars in the n e and T e measurements. On the other hand, high T e data points are observed in some discharges which can be explained by nonlinear heating. The D α emission measurement shows a clear relation between the breakdown time and the injection angle. An experimental investigation of the cause of the dependence of breakdown delay on the EC injection angle suggests that when the injection angle is oblique, the EC heating after the reflection at the inboard wall may become ineffective and cause the breakdown delay even when the EC heating before and directly upon injection remains effective. A preliminary run of the heat and transport balance code DYON indicates that the obtained dataset is suitable for a quantitative validation of EC absorption models.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2024

Linking Threat Agents to Targeted Organizations: A Pipeline for Enhanced Cybersecurity Risk Metrics

In this study, we present a methodology leveraging Large Language Models (LLMs) to transform Cybersecurity Threat Intelligence (CTI) narratives into actionable insights for individual organizations. Our approach automates the extraction of machine-readable adversary SKRAM (Skills, Knowledge, Resources, Authorities, and Motivation) attributes from open-source reports, extending LLM utility beyond typical interactions. This innovation enables precise, automated assessments of cybersecurity risks posed by various adversaries. Using a chain-of-thought and multi-shot prompting strategy, our methodology advances the automation of cybersecurity feature extraction for new machine-learning models that predict the risk of adversary targeting. This approach is refined using a substantial dataset of over 150 analyst-validated threat reports and synthetic organizational data from 900 companies. Here, by bootstrapping the training data with a rule-based heuristic over synthetic data, we have developed a high-accuracy machine-learning model that allows entities to dynamically prioritize threats and defensive actions.

Cyber Threat Intelligence↗
DOE OSTI2025

Model Residuals as Shields: A Two-Level Formulation to Defend Smart Grids From Poisoning Attacks

The advancement of smart grids presents both vast opportunities and heightened cybersecurity risks. Data-driven defense mechanisms, though designed as a shield against these threats, can fall prey to poisoning attacks. We delve into regression settings, underscoring the imperative to fortify defenses against a spectrum of poison ratios, notably those above 0.5—an issue scarcely addressed in prior studies. Recognizing the susceptibilities of smart grids and their manipulable sensors, we exploit the very intent of poisoning attacks, compromising model accuracy, as our defense mechanism. Our proposed two-level optimization framework discerns between poisoned and authentic data based on model residuals, outperforming or matching existing methods in 72% to 77% of precision and 75% to 80% of recalls across various poisoning attacks, poison ratios, and datasets. Once the authentic data are identified, the trained model is adaptable for a variety of applications. Comprehensive evaluations on different smart grid datasets, pitted against myriad poisoning schemes, validate our methodology’s edge over existing methods. Here, we also shed light on the implications of model misspecification originating from temporal auto-correlation, a common feature in Internet of Things and smart grid data.

Adversarial machine learning (ML)↗