SEARCH · Search NASA

Results for “Integrated mean squared error”

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 37 records · Page 2

All sources NASA NTRS DOE OSTI

DOE OSTI2025

Stable Machine‐Learning Parameterization of Subgrid Processes in a Comprehensive Atmospheric Model Learned From Embedded Convection‐Permitting Simulations

Modern climate projections often suffer from inadequate spatial and temporal resolution due to computational limitations, resulting in inaccurate representations of sub-grid processes. A promising technique to address this is the multiscale modeling framework (MMF), which embeds a kilometer-resolution cloud-resolving model (CRM) within each atmospheric column of a host climate model to replace traditional convection and cloud parameterizations. Machine learning offers a unique opportunity to make MMF more accessible by emulating the embedded CRM and reducing its substantial computational cost. Although many studies have demonstrated proof-of-concept success of achieving stable hybrid simulations, it remains a challenge to achieve near operational-level success with real geography and comprehensive variable emulation that includes, for example, explicit cloud condensate coupling. In this study, we present a stable hybrid model capable of integrating for at least 5 years with near operational-level complexity, including coarse-grid geography, seasonality, explicit cloud condensate and wind predictions, and land coupling. Our model demonstrates skillful online performance, achieving a 5-year zonal mean tropospheric temperature bias within 2 K, water vapor bias within 1 g/kg, and a precipitation root mean square error of 0.96 mm/day. Key factors contributing to our online performance include an expressive U-Net architecture and physical thermodynamic constraints for microphysics. With microphysical constraints mitigating unrealistic cloud formation, our work is the first to demonstrate realistic multi-year cloud condensate climatology under the MMF framework. Despite these advances, online diagnostics reveal persistent biases in certain regions, highlighting the need for innovative strategies to further optimize online performance.

Hu, Zeyuan [NVIDIA Corporation, Santa Clara, CA (U↗

DOE OSTI2024

Prediction of the Cu oxidation state from EELS and XAS spectra using supervised machine learning

Abstract Electron energy loss spectroscopy (EELS) and X-ray absorption spectroscopy (XAS) provide detailed information about bonding, distributions and locations of atoms, and their coordination numbers and oxidation states. However, analysis of XAS/EELS data often relies on matching an unknown experimental sample to a series of simulated or experimental standard samples. This limits analysis throughput and the ability to extract quantitative information from a sample. In this work, we have trained a random forest model capable of predicting the oxidation state of copper based on its L-edge spectrum. Our model attains an R 2 score of 0.85 and a root mean square error of 0.24 on simulated data. It has also successfully predicted experimental L-edge EELS spectra taken in this work and XAS spectra extracted from the literature. We further demonstrate the utility of this model by predicting simulated and experimental spectra of mixed valence samples generated by this work. This model can be integrated into a real-time EELS/XAS analysis pipeline on mixtures of copper-containing materials of unknown composition and oxidation state. By expanding the training data, this methodology can be extended to data-driven spectral analysis of a broad range of materials.

36 MATERIALS SCIENCE↗

DOE OSTI2024

Sampling Size Optimization for Bioburden Density Estimation in Planetary Protection

Planetary protection (PP) is a discipline that focuses on minimizing the biological contamination of spacecraft to ensure compliance with international policy. Precise estimation of bioburden - the total number of microbes in or on spacecraft hardware – and the bioburden density are of utmost importance for PP. Such estimation is the way concordance with requirements is demonstrated, and it is critical for quantifying the potential risk of inadvertently contaminating other planetary bodies. Although a suite of molecular techniques have been used to thoroughly characterize and profile the microbiome of various cleanroom environments and spacecraft, the gold standard remains the physical enumeration of microbes via culturing of samples directly taken from spacecraft and associated surfaces. However, due to technical, budgetary, and programmatic constraints, only a manageable portion (around 10%) of the entire spacecraft surface is directly sampled with cotton swabs or wipes. To generate the bioburden current best estimate (CBE) for components not directly verifiable, the accepted approach is to apply a NASA-defined bioburden estimate based on the components’ manufacturing or assembly environment. This approach utilizes a prespecified bioburden density estimation that applies a maximum value across the total surface area of the specified component. For hardware components that underwent similar assembly processes, an implied bioburden is adopted for all components, based on a direct verification of a representative component within the same lot. Once all components have a CBE, the bioburden estimates are generated. In previous publication [ 1], we have shown that statistical risks quantifying the accuracy of the estimates for sampled, prespecified, and implied components can be derived and ranked. For mean squared error (MSE) function, the risks are available analytically and hence a cost function can be obtained to optimize the risks with respect to the sampling area and sampling cost. Since the sampling area and sampling cost are two complimentary variables, their sum will have a well-defined minimum. This paper presents the multivariate optimization of the integrated risk of an empirical Bayes estimator to determine the optimal sampling schedule for a given number of components. It is assumed that given a number of components, N, the bioburden density for each component can either be sampled, implied, or prespecified. The multivariate optimization searches through different options to sample, imply or prespecify the bioburden density for a component, and account for the component’s surface area and cost of sampling. The idea of the optimization is based on the observation that the statistical risk of using an estimator is a monotonically decreasing function of the sampled area. The larger the sampled area, the lower the risk of using the estimator as the estimator becomes more and more accurate as the sampling area increases. On the other hand, the cost of sampling is monotonically increasing as the sampled surface grows. This makes the risk and total cost of sampling complimentary variables which can be counterbalanced to achieve an optimal overall value with respect to the sampled surface. In this paper, the integrated risk has been used to quantify the accuracy of the estimator. This risk has been selected because it depends on neither the true value of the parameter nor on the collected data. The cost of each sample was also available to obtain the total cost of sampling of N components. The paper will present the results based on computer-simulated data as well as the data collected during the InSight mission. The computer-simulated data have N components with randomly generated total areas and each component assigned to one of the three categories according to the method of estimating of bioburden density: sampled, implied, or prespecified. The cost of sampling is also available. The cost of sampling is estimated based on a cost model provided by the planetary protection group at JPL. For this paper, the overall cost was assumed to be a linear function of exposure. The optimization process finds the allocation of the components to the three categories that minimizes the tradeoff between integrated risk and total cost. For the InSight data, a set of components is selected representing all three categories, and optimization is performed to determine if the performed allocation was optimal or if a better allocation could have been obtained. To the best of our knowledge, this work is the first attempt not only perform an accurate estimation of bioburden density but also do it in an optimal way.

97 - MATHEMATICS AND COMPUTING↗

DOE OSTI2025

A machine learning method of modern urban building energy modeling: A case study of Chicago

Urban-scale building energy modeling is vital for urban planning. However, it can be challenging to assimilate reliable non-geometry building data for urban-scale modeling without extensive investment. Here, this study introduces a novel approach to developing modern urban-scale building energy stock data using geographic information systems and machine learning algorithms without necessarily requiring pre-supplied non-geometric metadata. The proposed framework integrates building footprint and height data to estimate gross floor areas, and matches each building to a pool of candidate records from ComStock or ResStock—filtered to the same county and ranked by geometric similarity—demonstrate a proof-of-concept case study in Chicago for predicting energy use intensity (EUI) using scalable datasets. The model achieved a mean bias error (MBE) of 0.08 kWh/m² and root mean square error (RMSE) of 14.84 kWh/m² under full metadata input for EUI prediction. With only location inputs, the model captured 69.2 % of EUI within predicted ranges. These results demonstrate the model’s potential to support early-stage urban planning, identify candidates for energy-efficient retrofits. By removing the dependency on detailed pre-surveys or extensive building metadata, the approach overcomes a key barrier in traditional urban-scale building energy modeling, illustrating a pathway toward broader and more cost-effective application, though further multi-city validation and improved treatment of pre-1925 buildings are needed.

Energy Use Intensity↗

DOE OSTI2026

Applying Particle Swarm Optimization and Extended Kalman Filtering to Model Kaplan Generation Dynamics for Hydropower Systems

Variable renewable generation is increasing the need for hydropower plants to provide fast and flexible grid support, which places new demands on plant-level dynamic models used for monitoring, control, and operational decision-making. This need is especially important for hydroelectric systems, where turbine and generator dynamics are strongly coupled, nonlinear, and time-varying, making accurate real-time representation difficult. To address this problem, this paper develops a digital twin (DT) framework for a synchronous generator–Kaplan turbine system using an explicit separation of slow turbine dynamics and fast generator dynamics. The turbine subsystem is represented by a six-coefficient model, whose parameters are identified offline using particle swarm optimization, while the generator subsystem is updated online through an extended Kalman filter for real-time state and parameter estimation. These models are integrated within a closed-loop simulation that includes a proportional–integral–derivative–double-derivative governor and excitation system, allowing the DT to track plant behavior under realistic operating conditions. Unlike prior studies that treat turbine and generator modeling separately or rely mainly on simulated inputs, the proposed framework is validated using real operational data from a hydropower plant. Results show that the DT reproduces terminal voltage, active power, and reactive power with a normalized root mean square error of approximately 5%. This hybrid offline–online formulation constitutes the main contribution of the work, providing an adaptive and practically deployable DT for hydropower systems with direct relevance to control improvement, performance monitoring, and grid-support applications under high renewable penetration.

13 HYDRO ENERGY↗

DOE OSTI2025

Characterizing leaf-scale fluorescence with spectral invariants

Sun-induced chlorophyll fluorescence (SIF) is increasingly recognized as a non-destructive probe for tracking terrestrial photosynthesis. Emerging developments in spectral invariants theory provide an innovative and efficient approach for representing SIF radiative transfer processes at the canopy scale. However, modeling leaf-scale fluorescence based on the spectral invariants properties (SIP) remains underexplored. In this study, the spectral invariants theory is employed for the first time to model the leaf-scale total, backward and forward fluorescence (leaf-SIP SIF). The leaf-SIP SIF model separates the leaf-scale radiative transfer process into two distinct components: the wavelength-dependent one associated with leaf biochemical properties, and the wavelength-independent component linked to leaf structural characteristics. The leaf structure-related effects are characterized by two spectrally invariant parameters: the photon recollision probability (p) and the scattering asymmetry parameter (q), which are parameterized using the directly measurable leaf dry matter. Evaluation against field measurements shows that the proposed leaf-SIP SIF model has a good performance, with coefficient of determination (R 2 ) of 0.89, 0.89, 0.90 and root mean squared errors (RMSE) of 1.28, 0.69, 0.74 Wm -2 µm -1 sr -1 , respectively for the total, backward, and forward fluorescence (660–800 nm). The leaf-SIP SIF model with a more concise formulation demonstrates comparable performance with the widely used Fluspect model. Further, the leaf-SIP SIF model provides a simple and efficient approach for simulating leaf-scale fluorescence, with the potential to be integrated into a unified SIP-based model framework for simulating the radiative transfer processes across the soil-leaf-canopy-atmosphere continuum.

59 BASIC BIOLOGICAL SCIENCES↗

DOE OSTI2025

Detection of Diversion in a Realistic Heat Pipe Microreactor Using Supervised Machine Learning

Microreactors (MRs) pose new challenges for international safeguards. Here, their small size and mass reproducibility make them ideal for deployment in greater numbers and in remote locations, making the job of safeguards inspectors more challenging. Machine learning (ML) is currently being applied to many fields to augment human performance and increase automation; in particular, ML could be used to provide insight for international inspectors to help detect the diversion of nuclear fuel from MR cores. Four ML model types (k-nearest neighbors, decision tree, random forest, and histogram-based gradient boosted ensemble) were trained on integrated flux and critical control drum angle data generated with Serpent 2 for a realistic heat pipe MR design, achieving nearly 100% binary classification accuracy of nominal and diversion core configurations by the end of 1 full power year for three of the four model types. Regression model variants were also trained, using the same input data, for predicting the number of fuel pins diverted. Root-mean-square errors below 5% of the total number of fuel pins were achieved by the 1 full power year mark for all models.

22 - GENERAL STUDIES OF NUCLEAR REACTORS↗

DOE OSTI2025

Predictive Modeling and Uncertainty Quantification in Condition Monitoring of Active Components: A Reactor Coolant Pump Use Case

This work develops data-driven models for onset of thermal barrier leakage in reactor coolant pumps. It incorporates uncertainty quantification to enhance the reliability and robustness of pre- dictions. Using synthetic data generated by the Generic Pressurized Water Reactor simulator, realistic degradation scenarios were simulated across lifecycle stages—beginning, middle, and end of life. Key variables, including differential pressure, flow rate, vibration, and temperatures, were analyzed using machine learning framework. The fully connected neural network models demonstrated exceptional performance, achieving R2 scores exceeding 0.99 and root mean square errors as low as around 8.23 × 10-2 gallon per minute (gpm) for the three stages of the lifecy- cle. UQ analysis further validated the model’s robustness, with narrow uncertainty bounds during steady-state operations and appropriately wider bounds during transitional phases, reflecting the physical behavior of the system. This work addresses important gaps in real-time condition moni- toring and regulatory compliance by integrating advanced condition monitoring technologies with UQ into IST programs. The ability to detect thermal barrier leakage early and quantify prediction reliability supports optimizing maintenance strategies while ensuring nuclear power plants’ safe and reliable operation.

99 - GENERAL AND MISCELLANEOUS↗

DOE OSTI2025

S AP F LOWER : an automated tool for sap flow data preprocessing, gap-filling, and analysis using deep learning

Sap flow, a critical process in plant water use and ecosystem water cycles, is often measured using thermal dissipation probes (TDP) due to their ease of installation and continuous data collection. However, sap flow data frequently include noise, outliers, and gaps, creating challenges for analysis and requiring substantial manual processing. We developed S AP F LOWER , a tool that automates data preprocessing, model training, gap-filling, sapwood area scaling and modeling, and water use analysis. It integrates autocleaning, machine learning and deep learning models (e.g. random forest, Gaussian process regression, long short-term memory (LSTM), bidirectional LSTM (BiLSTM)), and efficient workflows to process sap flow data. S AP F LOWER can remove over 90% of noisy data while preserving legitimate variations and achieve high accuracy in gap-filling based on user-determined parameters. Random forest, LSTM, and BiLSTM models reduced root mean square error to 10% or less for long-term gaps. Model training and prediction can be performed efficiently within seconds. S AP F LOWER significantly enhances the efficiency and accessibility of TDP data analysis by automating complex tasks, enabling researchers without programming expertise to employ advanced techniques. Future improvements will focus on species-specific corrections for TDP and support for additional measurement methods. S AP F LOWER is openly available on GitHub (https://github.com/JiaxinWang123/SapFlower) and Zenodo (doi: 10.5281/zenodo.13665919).

ecosystem water balance↗

DOE OSTI2026

A dendritic strontium river isoscape for fisheries applications in the Sacramento River basin, California, USA

Objective Understanding the origins and movements of fish is fundamental to effective conservation and fisheries management. Strontium isotope ratios ( 87 Sr/ 86 Sr) in otoliths provide a powerful tracer of natal origin and migratory pathways. However, existing 87 Sr/ 86 Sr isoscapes for the Sacramento River basin, an ecosystem that supports ecologically and economically important salmon populations, rely on discrete classification approaches that overlook unsampled habitats and do not incorporate spatial uncertainty. Our objective was to develop a continuous, network-explicit 87 Sr/ 86 Sr isoscape with quantified uncertainty to fill in data gaps and enable probabilistic assignments of fish origin and movement. Methods We used river water 87 Sr/ 86 Sr data from 106 sites (1997–2021) to develop spatial stream network models that use dendritic connectivity and watershed characteristics (lithology, bedrock age, and land cover) to predict river water 87 Sr/ 86 Sr throughout the basin. Models were fitted using maximum and restricted likelihood and were evaluated via Akaike’s information criterion and leave-one-out cross validation. We produced both historical (pre-dam) and present-day (below-dam) isoscapes, delineated uncertainty-informed isotopic ranges using k -means clustering, and applied a proof-of-concept Bayesian assignment to estimate natal origins and early rearing habitats for two endangered winter-run Chinook Salmon Oncorhynchus tshawytscha. Results Cross validation indicated strong performance of the 87 Sr/ 86 Sr model (leave-one-out cross validation: R 2 = 0.91; root mean square error = 0.0005). Uncertainty-informed clustering identified 19 isotopic “suites” (reaches with indistinguishable 87 Sr/ 86 Sr values) in present-day anadromous habitats and 25 suites in the historical network. Example natal and early rearing assignments included predictions that challenged expectations for juvenile salmon migration based on predicted river 87 Sr/ 86 Sr compositions. Conclusions This study developed a continuous, network-explicit 87 Sr/ 86 Sr isoscape that integrates existing river data to predict 87 Sr/ 86 Sr in unsampled reaches and the likely achievable range and resolution of otolith-based origin and life history inference. The resulting river isoscape provides a valuable tool to predict salmon movements and identify habitats supporting their survival and growth that otherwise might remain undetected. Coupling these predictions with complementary approaches that ground-truth juvenile presence (e.g., targeted fish surveys) represents an important step toward science-informed restoration and management of critical habitats throughout the Sacramento River basin.

Environmental sciences↗

DOE OSTI2025

Primary and Low-Strain Creep Models for 9Cr Tempered Martensitic Steels Including the Effects of Irradiation Softening and High-Helium Re-Hardening

Primary and low-strain creep represents a very important integrity challenge to large, complex structures, like fusion reactors. Here, we develop a predictive empirical primary creep model for 9Cr tempered martensitic steels (TMS), relating the applied stress (σ) to strain (ε), time (t) and temperature (T). The most accurate model is based on the applied σ normalized by the steel’s T-dependent ultimate tensile stress (σo), σ/σo(T). The model, fit to 17 heats of 9Cr TMS, yielded a σ root mean square error (RMSE) of ≈±11 MPa. Notably, the model also provides robust predictions for all the other TMS, when calibrated only by the fusion candidate Eurofer97 database. The model was extended to explore two possible effects of neutron irradiation, which produces both displacements per atom (dpa) and helium (He in atomic parts per million, appm) damage. These effects, which have not been previously considered, include: (a) softening, as a function of dpa, at T > ≈400–450 °C, in low-He fission environments (<1 He/dpa); and (b) subsequent re-hardening in high-He (≥10 He/dpa) fusion first-wall environments. The irradiation effect models predict (a) accelerated primary creep due to irradiation softening; and (b) fully arrested creep due to high-He re-hardening.

Alam, Md Ershadul (ORCID:0000000345968026)↗

DOE OSTI2025

Establishing Models for Digital Twin of Hydropower Systems Using Probability Density Function Shaping

This paper introduces a digital twin modeling method for hydropower systems with Kaplan turbines using probability density function (PDF) shaping. We first use multilayer perceptron (MLP) model to build the discretized openloop Kaplan unit, where the MLP is trained by historical data. Then we use a proportional integral double derivative (PIDD) controller and a lead-lag exciter to test the obtained digital twin model in a closed-loop fashion. Simulation results show that the proposed digital twin modeling method can accurately capture the dynamics of the Kaplan hydropower unit. Finally, we show that the obtained digital twin can help to optimize the PIDD parameters. Compared with the original PIDD controller, the optimized one can achieve an over 90% improvement on the mean square tracking error.

Yin, Zhun [New York University]↗

DOE OSTI2025

Co-Firing Switchgrass and Waste Coal in A Power Plant: A Techno-Economic and Life Cycle Evaluation for The Ohio River Valley (SWITCH) (Final Technical Report for Ohio State/FE0032204)

Abandoned coal mine lands (AMLs) represent one of the most persistent environmental challenges in the United States. Prior to the enactment of the Surface Mining Control and Reclamation Act (SMCRA) in 1977, coal mining operations were not legally required to reclaim disturbed lands, leaving behind approximately 500,000 AML sites nationwide. These sites pose severe environmental and health risks, including acid mine drainage, soil and water contamination, and spontaneous combustion of waste coal piles. Millions of Americans live within one mile of these AMLs, underscoring the urgency of remediation. Traditional reclamation practices, such as planting cool-season grasses, often fail to fully restore ecological function or leverage the economic potential of these lands. This project addressed these challenges by developing integrated strategies for resource recovery, land reclamation, and sustainable energy production. This project evaluated an integrated strategy to convert this liability into an opportunity by recovering waste coal and co-firing it with switchgrass (Panicum virgatum L.) cultivated on reclaimed or marginal AML areas in existing coal-fired power plants. Switchgrass not only provides a renewable feedstock but also aids in land reclamation and carbon sequestration. 1) Remote Sensing and Machine Learning for Waste Coal Identification Using Sentinel-2 satellite imagery and supervised classification, we applied four machine learning models to detect historical waste coal piles. Random Forest achieved the highest accuracy (precision: 86%, recall: 77%). Time-series analysis revealed gradual vegetation recovery since 1986, indicating natural reclamation processes in historical sites, while active mining areas showed ongoing disturbance. This workflow enables scalable monitoring and prioritization of reclamation efforts. 2) UAS-Based Stockpile Volume Estimation To quantify recoverable waste coal, we evaluated Unmanned Aerial Systems (UAS) equipped with Light Detection and Ranging (LiDAR) and multispectral sensors. Structure-from-Motion (SfM) photogrammetry combined with interpolated Digital Terrain Models (DTMs) achieved strong agreement with LiDAR reference volumes (Root Mean Square Error (RMSE) ≈147 m 3 , Mean Absolute Percentage Error (MAPE) ≈2%). Sensitivity analysis confirmed that spatial resolution significantly influences accuracy, emphasizing the need for high-resolution data for precise volume estimation. This approach offers a scalable, cost-effective, and accurate alternative to conventional ground-based surveys. 3) Switchgrass Cultivation for Bioenergy and Water Quality Improvement We assessed the hydrological and water quality impacts of converting AMLs to switchgrass production areas using the Soil and Water Assessment Tool (SWAT). Results showed that converting 10% of the watershed area into the switchgrass production zone reduced streamflow by 3.1%, total suspended solids by 18.1%, total nitrogen by 7.6%, and total phosphorus by 6.2%, while achieving biomass yields of 8.6–9.2 metric tons per hectare. These findings highlight switchgrass as a dual-benefit strategy for land reclamation and bioenergy feedstock production. 4) Integrated Co-Firing and CCS for Carbon-Negative Power Generation We modeled co-firing scenarios using the Power Plant Flexible Model (PPFM) to evaluate plant efficiency, greenhouse gas (GHG) emissions, and levelized cost of electricity (LCOE). Without carbon capture and storage (CCS), increasing switchgrass co-firing ratios reduced LCOE from $\$$150/MWh at 0% biomass to $\$$110/MWh at full substitution. Under CCS, costs remained higher (~$\$$250/MWh at 0% biomass) but decreased to $\$$200/MWh at 100% biomass, while enabling net-zero or carbon-negative electricity due to switchgrass sequestration benefits. Although CCS introduced efficiency penalties, pairing it with biomass co-firing offset these impacts and maximized climate benefits. Overall, optimizing co-firing ratios between 60-100%, supported by reliable logistics and storage strategies, emerged as a practical pathway to balance affordability, sustainability, and net-zero or negative GHG emissions while promoting productive reuse of AMLs.

01 COAL, LIGNITE, AND PEAT↗

DOE OSTI2025

Dual‐Transformer Deep Learning Framework for Seasonal Forecasting of Great Lakes Water Levels

Abstract The Great Lakes of North America form one of the largest freshwater systems on Earth, and their lake‐wide average water levels (lake levels) can fluctuate by more than 0.5 m on a seasonal scale. These fluctuations pose substantial challenges for coastal resilience, flood risk management, and navigation planning. Accurate seasonal forecasting of lake levels using traditional mechanistic models is challenging due to the complex physical mechanisms and coupled hydroclimatic processes involved. Recently, deep learning has gained prominence in geoscience applications for its ability to recognize intricate patterns within multiphysical data sets. Here, we introduce a novel Dual‐Transformer deep learning framework, tested on the Great Lakes. This architecture integrates two modified Transformer models: the Prophet, which predicts underlying trends, and the Critic, which refines the Prophet's predictions. The final lake level prediction is derived by weighting the outputs of both models through a multi‐layer perceptron, jointly trained with the Prophet and Critic to enhance overall accuracy. Our results demonstrate that the innovative learning framework achieves the highest prediction accuracy compared to established deep learning models when using identical input features. It attains a root mean square error of 4–7 cm in predicting lake levels up to 6 months in advance across the lakes. Additionally, the Dual‐Transformer model runs six orders of magnitude faster than conventional mechanistic models, producing results in less than one second on a typical personal computer. These findings suggest that our deep learning framework has strong potential to advance lake level prediction and carries important implications for water management and disaster mitigation, thereby enhancing the quality of life in coastal regions.

Chen, Yi [Great Lakes Research Center Michigan Tec↗

DOE OSTI2025

Machine learning and process-based modeling of spatiotemporal changes in active layer thickness across Alaska

Permafrost degradation poses a growing threat to infrastructure stability and ecosystem resilience in the rapidly warming Arctic. We investigated the spatiotemporal dynamics of active layer thickness (ALT) across Alaska by integrating field observations, environmental datasets, a physically based Stefan model, and machine learning (ML) techniques. Using weather projections from the Coupled Model Intercomparison Project Phase 6 under two Shared Socioeconomic Pathways (SSP 2-4.5 and SSP 5-8.5), we assessed ALT sensitivity to projected future weather conditions. The random forest (RF) model outperformed the Stefan approach in predicting ALT on the training dataset (R² = 0.84 vs. 0.53) but demonstrated lower generalizability on the test dataset (R² = 0.24 vs. 0.54). The root mean square error (RMSE) for the RF model for training and testing ranged from 14 to 22 cm, compared to 17 and 18 cm for the Stefan model. Variable importance analysis revealed that mean annual temperature and slope angle were the strongest predictors of ALT, accounting for 19% and 18% of the variance, respectively, followed by sediment transport index (14%) and stream power index (11%). Comparative analysis of baseline ALT predictions showed the Stefan model tended to project a thicker active layer (mean ± SD: 65 ± 16 cm), compared to the RF model (mean ± SD: 59 ± 8.8) cm). Both models indicated a latitudinal gradient in ALT, with shallower depths at higher latitudes. Projected ALT increases by 2100 were estimated at 3.3 ± 2.2 cm under SSP 2-4.5 and 5.9 ± 4.0 cm under SSP 5-8.5 for the ML model, whereas the Stefan model projected substantially larger increases of 13 ± 2.6 cm (SSP 2-4.5) and 28 ± 4.4 cm (SSP5-8.5). Spatial analysis showed the greatest ALT increases in northern Alaska, with relatively smaller changes in southern regions. These findings highlight the complex, multifactorial nature of ALT dynamics and the value of hybrid modeling approaches. As rising temperatures accelerate permafrost thaw, changes in ALT can disrupt ecosystems, damage infrastructures, and enhance the release of stored soil carbon, highlighting the urgent need for improved predictive capabilities to inform adaptation strategies in the Arctic.

Climate sciences↗

DOE OSTI2025

Active Learning of Microgrid Frequency Dynamics Using Neural Ordinary Differential Equations

Accurate frequency modelling of inverter‐based resource (IBR)‐dominated power systems is crucial for ensuring stable, reliable and resilient operations, particularly given their inherent low‐inertia characteristics and fast dynamics that traditional swing equation‐based models inadequately capture. This paper explores neural ordinary differential equations (Neural ODEs) as a computationally efficient, data‐driven framework for modelling power system frequency dynamics, specifically within microgrids integrating high penetrations of distributed energy resources (DERs). The developed neural ODEs framework incorporates a neural network architecture designed to capture input dynamics. By actively perturbing the system with a known signal, the Python‐based neural ODEs framework was trained using measured system states and inputs, without the need for detailed system information. The framework, tested on a model of the Cordova, AK, microgrid, achieved a goodness of fit ranging from 60% to 99% across different state variables and maintained a mean square error in the 10 -6 p.u. range under square and step excitation signals. The proposed approach demonstrated robustness to measurement noise and initial condition variations while maintaining low computational complexity suitable for real‐time power system control applications. Furthermore, transfer learning enabled the neural ODEs model to adapt to the following changes in system topology or generator dispatch, highlighting its effectiveness for dynamic microgrids with frequently evolving configurations and diverse DERs.

Aryal, Tara [South Dakota State Univ., Brookings, ↗

DOE OSTI2025

Navigating the Noise: Bringing Clarity to ML Parameterization Design With O $\boldsymbol{\mathcal{O}}$(100) Ensembles

Abstract Machine‐learning (ML) parameterizations of subgrid processes (here of turbulence, convection, and radiation) may one day replace conventional parameterizations by emulating high‐resolution physics without the cost of explicit simulation. However, uncertainty about the relationship between offline and online performance (i.e., when integrated with a large‐scale general circulation model) hinders their development. Much of this uncertainty stems from limited sampling of the noisy, emergent effects of upstream ML design decisions on downstream online hybrid simulation. Our work rectifies the sampling issue via the construction of a semi‐automated, end‐to‐end pipeline for size ensembles of hybrid simulations, revealing important nuances in how systematic reductions in offline error manifest in changes to online error and online stability. For example, removing dropout and switching from a Mean Squared Error to a Mean Absolute Error loss both reduce offline error, but they have opposite effects on online error and online stability. Other design decisions, like incorporating memory, converting moisture input from specific humidity to relative humidity, using batch normalization, and training on multiple climates do not come with any such compromises. Finally, we show that ensemble sizes of may be necessary to reliably detect causally relevant differences online. By enabling rapid online experimentation at scale, we can empirically settle debates regarding subgrid ML parameterization design that would have otherwise remained unresolved in the noise.

Lin, Jerry [Department of Earth System Sciences Un↗

DOE OSTI2026

Digital Twin-Enabled Adaptive Control for Hydroelectric Systems: Turbine Governor and Voltage Regulation

This paper presents a comprehensive digital twin (DT) framework for hydroelectric systems that enables adaptive control of turbine governors and excitation systems without requiring detailed manufacturer specifications. The proposed framework integrates neural network-based system identification with stabilizing adaptive control laws for the installed turbine controller and middle-branch adaptive tuning for the installed voltage regulator. Using real operational data from Unit C-8 at Rocky Reach Dam (1,349 MW capacity), highfidelity neural network models are developed to capture turbine and generator dynamics without requiring detailed manufacturer specifications. The DT enables safe controller synthesis and validation in simulation before deployment. For turbine control, the proposed method achieves a 79.9% mean square error (MSE) reduction compared with that of an optimal controller. For voltage regulation, the adaptive excitation controller achieves approximately 42.6% MSE reduction while preserving installed protection logic. The results demonstrate that DT technology provides a practical pathway for modernizing hydropower control systems with minimal operational disruption.

Gui, Yonghao [ORNL] (ORCID:0000000250435534)↗