Search NASA⌕ Search

SEARCH · Search NASA

Results for “adaptive learning”

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 379 records · Page 21

A Generation-Storage Coordination Dispatch Strategy for Power System Based on Causal Reinforcement Learning

In the backdrop of global energy transformation, power systems integrating high proportions of renewable energy sources are facing unprecedented challenges in operational stability and dispatch efficiency. To address these challenges, this study introduces a generation-storage coordination real-time dispatch strategy based on Causal Power System Dynamic Reinforcement Learning (CPSDRL). Diverging from traditional reinforcement learning approaches, CPSDRL innovatively incorporates causal inference within the state prediction model - the crux of model-based reinforcement learning - thereby establishing the Power Causal Dynamic Model (PCDM). Assisted by the prior knowledge of power systems, the model significantly enhances prediction accuracy and reliability through a two-stage training process. Utilizing PCDM, this study further applies a direct policy search algorithm to optimize the real-time dispatch strategy. Experimental results indicate that the proposed method improves the stability of generation-storage coordination real-time dispatch and exhibits competitive advantages in sample efficiency and computational speed, compared to traditional model-based and model-free reinforcement learning algorithms. This method is expected to enhance the practicality and adaptability of causal reinforcement learning techniques in power system scheduling and control.

causal reinforcement learning↗

Adaptive stretching of representations across brain regions and deep learning model layers

Prefrontal cortex (PFC) is known to modulate the visual system to favor goal-relevant information by accentuating task-relevant stimulus dimensions. Does the brain broadly re-configures itself to optimize performance by stretching visual representations along task-relevant dimensions? We considered a task that required monkeys to selectively attend on a trial-by-trial basis to one of two dimensions (color or motion direction) to make a decision. Although effects were most prominent in frontal areas, representations stretched along task-relevant dimensions in all sites considered: V4, MT, lateral PFC, frontal eye fields (FEF), lateral intraparietal cortex (LIP), and inferotemporal cortex (IT). Spike timing was crucial to this code. A deep learning model was trained on the same visual input and rewards as the monkeys. Despite lacking an explicit selective attention or other control mechanism, by minimizing error during learning, the model’s representations stretched along task-relevant dimensions, indicating that stretching is an adaptive strategy.

59 BASIC BIOLOGICAL SCIENCES↗

Machine learning surrogate for charged particle beam dynamics with space charge based on a recurrent neural network with aleatoric uncertainty

In this work, we develop a machine learning (ML) model with aleatoric uncertainty for the low energy beam transport (LEBT) region of the LANSCE linear accelerator in which we model the transport of a space-charge-dominated 750 keV proton beam through a lattice of 22 quadrupole magnets. Our ML model is developed based on data generated by a Kapchinsky–Vladimirsky (KV) envelope model of beam transport. We show that a recurrent neural network can be used as a dynamical surrogate model for fast prediction of the LEBT beam envelope. Furthermore, we endow the model with the prediction of aleatoric uncertainty and compare three different approaches. We demonstrate that the ML-based uncertainty quantification models are well calibrated and produce good estimates of the regions where the model is less certain about its predictions. This ML framework is a necessary step in the development of a real-time virtual diagnostic tool with uncertainty quantification that can be integrated into more complex downstream tasks (e.g., adaptive control or learning flexible control policies via reinforcement learning) for improved efficiency in beam operations. In future work, we plan to expand on this preliminary study by considering more realistic envelope models that include longitudinal momentum spread and dispersive effects in bending magnets, as well as particle tracking codes with 3D space charge (such as and ). Published by the American Physical Society 2024

43 PARTICLE ACCELERATORS↗

Indirect learning control for nonlinear dynamical systems

In a previous paper, learning control algorithms were developed based on adaptive control ideas for linear time variant systems. The learning control methods were shown to have certain advantages over their adaptive control counterparts, such as the ability to produce zero tracking error in time varying systems, and the ability to eliminate repetitive disturbances. In recent years, certain adaptive control algorithms have been developed for multi-body dynamic systems such as robots, with global guaranteed convergence to zero tracking error for the nonlinear system euations. In this paper we study the relationship between such adaptive control methods designed for this specific class of nonlinear systems, and the learning control problem for such systems, seeking to converge to zero tracking error in following a specific command repeatedly, starting from the same initial conditions each time. The extension of these methods from the adaptive control problem to the learning control problem is seen to be trivial. The advantages and disadvantages of using learning control based on such adaptive control concepts for nonlinear systems, and the use of other currently available learning control algorithms are discussed.

Ryu, Yeong Soon↗

Geologic CO 2 sequestration monitoring design: A machine learning and uncertainty quantification based approach

Monitoring is a crucial aspect of geologic carbon dioxide (CO 2 ) sequestration risk management. Effective monitoring is critical to ensure CO 2 is safely and permanently stored throughout the life-cycle of a geologic CO 2 sequestration project. Effective monitoring involves deciding: (i) where is the optimal location to place the monitoring well(s), and (ii) what type of data (pressure, temperature, CO 2 saturation, etc.) should be measured taking into consideration the uncertainties at geologic sequestration sites. We have developed a filtering-based data assimilation procedure to design effective monitoring approaches. To reduce the computational cost of the filtering-based data assimilation process, a machine-learning algorithm: Multivariate Adaptive Regression Splines is used to derive computationally effcient reduced order models from results of full-physics numerical simulations of CO 2 injection in saline aquifer and subsequent multi-phase fluid flow. We use example scenarios of CO 2 leakage through legacy wellbore and demonstrate a monitoring strategy can be selected with the aim of reducing uncertainty in metrics related to CO 2 leakage. We demonstrate the proposed framework with two synthetic examples: a simple validation case and a more complicated case including multiple monitoring wells. The examples demonstrate that the proposed approach can be effective in developing monitoring approaches that take into consideration uncertainties.

58 GEOSCIENCES↗

Quantifying Groundwater Response and Uncertainty in Beaver‐Influenced Mountainous Floodplains Using Machine Learning‐Based Model Calibration

Abstract Beavers ( Castor canadensis ) alter river corridor hydrology by creating ponds and inundating floodplains, and thereby improving surface water storage. However, the impact of inundation on groundwater, particularly in mountainous alluvial floodplains with permeable gravel/cobble layers overlain by a soil layer, remains uncertain. Numerical modeling across various floodplain structures considers topographic and sediment complexity and multidirectional flow, linking inundation to groundwater response. This study develops a model‐data integration workflow to address uncertainty in groundwater response to beaver‐induced inundations in a mountainous alluvial floodplain in the Upper Colorado River Basin. Uncertain factors include seasonal hydrologic dynamics, hydraulic conductivities, floodplain structures, and meteorological forcings. We employed an ensemble of groundwater models, based on geophysical and hydrologic data, with machine learning‐based calibration using a neural density estimator. This allowed us to quantify the vertical flux from the soil layer to the permeable gravel bed, the down‐valley underflow within the gravel bed, and their ratios. Results show a significant increase in the vertical flux relative to down‐valley underflow, from 2 during dry pond periods to 20 during wet periods, serving as an analogy for conditions without and with beaver ponds. The study highlights the influence of floodplain structure on groundwater storage, water balance, and water quality impacted by beaver ponds. A thick gravel bed layer, with a large down‐valley underflow, minimizes the effect of beaver‐induced inundation on water quality. We emphasize the need for field‐scale measurements of floodplain structure and improved characterization of evapotranspiration changes to reduce uncertainty in groundwater response. Plain Language Summary Beavers change the flow of water in river corridors by creating ponds, expanding wetlands, and flooding floodplains. This increases surface water area, promotes plant growth, and enhances biodiversity. However, the impact of this flooding on groundwater flow is not well understood, especially in mountainous areas with gravel layers where water moves easily beneath soil. In this study, we used numerical modeling to investigate how beaver ponds influence groundwater in a mountainous floodplain of the Upper Colorado River Basin. We adapted a machine learning method to validate our numerical models using multiple field data sets. Our findings show that beaver ponds significantly increase vertical water flow from the soil to the gravel during wet periods, compared to when the ponds are fully drained. The study also highlights the importance of floodplain structure in controlling both water flow in gravel layers along the river direction and vertical flow from the soil to the gravel with the presence of beavers. To reduce uncertainty in groundwater response, we emphasize the need for more field‐scale measurements of floodplain structure, hydraulic properties, and evapotranspiration changes. Key Points Floodplain structures and hydraulic conductivities are important for groundwater response with beaver ponds in mountainous floodplains Large down‐valley underflow in permeability‐stratified floodplains reduces beaver‐induced impacts on groundwater storage and water quality Machine learning‐based model calibration methods are effective for estimating posterior distributions of groundwater model parameters

Wang, Lijing↗

Fuzzy Adaptive Control for Intelligent Autonomous Space Exploration Problems

The principal objective of the research reported here is the re-design, analysis and optimization of our newly developed neural network fuzzy adaptive controller model for complex processes capable of learning fuzzy control rules using process data and improving its control through on-line adaption. The learned improvement is according to a performance objective function that provides evaluative feedback; this performance objective is broadly defined to meet long-range goals over time. Although fuzzy control had proven effective for complex, nonlinear, imprecisely-defined processes for which standard models and controls are either inefficient, impractical or cannot be derived, the state of the art prior to our work showed that procedures for deriving fuzzy control, however, were mostly ad hoc heuristics. The learning ability of neural networks was exploited to systematically derive fuzzy control and permit on-line adaption and in the process optimize control. The operation of neural networks integrates very naturally with fuzzy logic. The neural networks which were designed and tested using simulation software and simulated data, followed by realistic industrial data were reconfigured for application on several platforms as well as for the employment of improved algorithms. The statistical procedures of the learning process were investigated and evaluated with standard statistical procedures (such as ANOVA, graphical analysis of residuals, etc.). The computational advantage of dynamic programming-like methods of optimal control was used to permit on-line fuzzy adaptive control. Tests for the consistency, completeness and interaction of the control rules were applied. Comparisons to other methods and controllers were made so as to identify the major advantages of the resulting controller model. Several specific modifications and extensions were made to the original controller. Additional modifications and explorations have been proposed for further study. Some of these are in progress in our laboratory while others await additional support. All of these enhancements will improve the attractiveness of the controller as an effective tool for the on line control of an array of complex process environments.

Esogbue, Augustine O.↗

Regional-scale soil carbon predictions can be enhanced by transferring global-scale soil–environment relationships

Accurate modelling and mapping soil organic carbon are crucial for supporting soil health restoration and climate change mitigation at both regional and global scales. However, regional soil predictions often suffer from data scarcity and high prediction uncertainty. Utilizing a pre-trained global-to-regional soil carbon predictive model can be a potential solution to address this challenge. Despite its promise, how to construct and apply the global-scale model to enhance regional-scale soil carbon mapping remains largely unexplored. Here, we propose the Global Soil Carbon Pre-trained Model (GSoilCPM), a deep-learning-based domain adaptative model, to enhance regional-scale soil carbon predictions. Based on large amount of environmental covariate data and 106,167 soil samples across the globe, we verify our hypothesis of the effectiveness of this 'global-to-regional' modelling strategy. The pre-trained model can be then transferred and fine-tuned to bridge the regional- and global-scale soil–environment relationships. We applied and validated this modelling strategy in four regional-scale study areas, three in the Northern Hemisphere and one in the Southern Hemisphere, each with distinct environmental background. Compared to traditional modelling approaches as a baseline, four case studies all demonstrated significant improvement in prediction accuracy across diverse environments and varying data availabilities. The average percentage improvement across all regions is 10.93% (absolute values decreased by 1.20 g kg−1 averagely) in MAE and 29.04% (absolute values increased by 0.10 averagely) in CCC. The applicability and future horizons of using GSoilCPM were further discussed. We further reveal that regions with fewer soil samples or lower baseline accuracy benefit more from the pre-trained global model. Our findings highlight the advantages of leveraging the generalized knowledge from global models to enhance specifically localized soil modelling, positioning a potential paradigm shift in digital soil mapping, and far-reaching implications for soil monitoring and land management.

Deep learning↗

An overview of 3D field optimization for control of transport and edge instabilities on KSTAR

An international team from several laboratories and universities has made key advances over the last few years in the control of plasma transport and edge instabilities with applied 3D fields in the KSTAR tokamak to optimize long pulse operation scenarios. This overview begins with the optimization of both core and edge resonant magnetic perturbations (RMP) to improve fast ion confinement to avoid excessive limiter heat loads due to fast ion losses and successful modeling of the experimental results. Integrated and advanced plasma control techniques with machine learning (ML) and adaptive control were then used to optimize the 3D field spectrum in real-time to control edge localized modes (ELMs) while avoiding core locked modes that could disrupt the plasma. Accelerating the offline model of 3D fields with a surrogate ML model can optimize ELM suppression in the edge while limiting the impact of the applied RMP fields deeper in the plasma core in real-time. In addition, the impact of the 3D fields on the divertor heat load has been modeled and compared with experimental measurements. An analysis of a multi-machine database including KSTAR has been performed to better understand the metrics for the observed RMP thresholds for ELM suppression and the resulting plasma performance. Predictive modeling of the operational space for ELM suppression and density pumpout due to RMP has shown the importance of magnetic islands in the plasma edge and their impact on plasma turbulence. This research has culminated in the development of successful long pulse operational scenarios on KSTAR while attempting to overcome challenges of the new tungsten divertor.

3D fields↗

Sparse Hardware Embedding of Spiking Neuron Systems for Community Detection

We study the applicability of spiking neural networks and neuromorphic hardware for solving general optimization problems without the use of adaptive training or learning algorithms. We leverage the dynamics of Hopfield networks and spin-glass systems to construct a fully connected spiking neural system to generate synchronous spike responses indicative of the underlying community structure in an undirected, unweighted graph. Mapping this fully connected system to current generation neuromorphic hardware is done by embedding sparse tree graphs to generate only the leading-order spiking dynamics. Here, we demonstrate that for a chosen set of benchmark graphs, the spike responses generated on a current generation neuromorphic processor can improve the stability of graph partitions and non-overlapping communities can be identified even with the loss of higher-order spiking behavior if the graphs are sufficiently dense. For sparse graphs, the loss of higher-order spiking behavior improves the stability of certain graph partitions but does not retrieve the known community memberships.

97 MATHEMATICS AND COMPUTING↗

Sequential Learning with Active Partial Labeling for Building Metadata

Modern buildings are instrumented with thousands of sensing and control points. The ability to automatically extract the physical context of each point, e.g., the type, location, and relationship with other points, is the key to enabling building analytics at scale. However, this process is costly as it usually requires domain expertise with a deep understanding of the building system and its point naming scheme. In this study, we aim to reduce the human effort required for mapping sensors to their context, i.e., metadata mapping. We formulate the problem as a sequential labeling process and use the conditional random field to exploit the regular and dependent structures observed in the metadata. We develop a suite of active learning strategies to adaptively select the most informative subsequences in point names for human labeling, which significantly reduces the inputs from domain experts.We evaluated our approach on three different buildings and observed encouraging performance in metadata mapping from the proposed solution.

99 GENERAL AND MISCELLANEOUS↗

STEMDL

The code "STEMDL" provides highly-optimized multi-gpu implementations of DL models of applicability in electron microscopy via the use of mixed-precision arithmetic, native graph construction and evaluation in Tensorflow, and adaptive scaling of learning parameters to avoid numerical instabilities due to reduced precision. STEMDL also uses low-level methods from the Horovod library for gradient communication and reduction and adds an additional optimization by allowing the user to scheduling the gradient reduction operations based on their position within the layers of a neural net. STEMDL provides optimized I/O routines for data staging and data reading/writing in both Tensorflow and Pytorch using multi-processing and asynchronous transfer between host and device. .

Laanait, Nouamane↗

HydraGNN_GFM_FineTuning4Materials v1.0

This repository enables fine-tuning of the HydraGNN Predictive GFM 2026 — an open-source ensemble of pre-trained graph foundation models for atomistic materials modeling, developed at Oak Ridge National Laboratory. The GFM 2026 is freely available and downloadable via Globus from the OLCF Data Constellation (DOI: 10.13139/OLCF/2562660). Starting from these pre-trained weights, this repository provides a complete transfer learning pipeline for adapting the GFM ensemble to domain-specific molecular and materials property prediction tasks. It includes: 1) Utilities for ensemble fine-tuning with task-specific output heads 2) Example pipelines for eight widely-used materials and molecular datasets 3) Tools for model adaptation and head configuration 4) Data preprocessing utilities for each supported dataset 5) Benchmarking and evaluation scripts

Ungerboeck, Linda↗

The Role of Internal Variability in Springtime Arctic Amplification from 1980 to 2022

Arctic amplification (AA) refers to the enhanced warming of the Arctic relative to the global average due to rising greenhouse gases, measured as the ratio of Arctic-mean to global-mean surface air temperature (SAT) trends. From 1980 to 2022, annual-mean AA reached 4.2 (Arctic defined as north of 70°N). Climate models simulate AA but fail to reproduce its magnitude. Sweeney et al. attributed much of this model–observation discrepancy to internal variability. AA shows seasonality and so does the discrepancy. Spring (March–May) shows the largest gap: Observed AA is 4.2, while the multimodel mean is 2.7. This raises several questions: 1) What role does internal variability play in observed spring AA? 2) How does simulated spring AA compare to observations when internal variability is removed? 3) If internal variability is significant, what mechanisms drive it? To address these, we adapted the machine learning algorithm from Sweeney et al., training on simulated multidecadal spring SAT and sea level pressure (SLP) trend maps. Our results show that internal variability enhanced spring Arctic warming by 37% and reduced global warming by 10%. Removing internal variability reconciles the spring AA discrepancy. The estimated internal contribution to Arctic spring warming is supported by an independent dynamical adjustment approach. We identify an atmospheric circulation pattern in observations associated with this internal warming. Observed internal Siberian SAT and SLP trends follow the simulated SAT–SLP relationship but lie at the distribution’s extreme, suggesting models generally underestimate internal variability unless the observed configuration reflects a rare real-world realization.

Arctic↗

Employing artificial intelligence to steer exascale workflows with colmena

Computational workflows are a common class of application on supercomputers, yet the loosely coupled and heterogeneous nature of workflows often fails to take full advantage of their capabilities. We created Colmena to leverage the massive parallelism of a supercomputer by using Artificial Intelligence (AI) to learn from and adapt a workflow as it executes. Colmena allows scientists to define how their application should respond to events (e.g., task completion) as a series of cooperative agents. In this paper, we describe the design of Colmena, the challenges we overcame while deploying applications on exascale systems, and the science workflows we have enhanced through interweaving AI. The scaling challenges we discuss include developing steering strategies that maximize node utilization, introducing data fabrics that reduce communication overhead of data-intensive tasks, and implementing workflow tasks that cache costly operations between invocations. These innovations coupled with a variety of application patterns accessible through our agent-based steering model have enabled science advances in chemistry, biophysics, and materials science using different types of AI. In conclusion, our vision is that Colmena will spur creative solutions that harness AI across many domains of scientific computing.

Workflows↗

Improving North American Wildfire Prediction by Integrating a Machine-Learning Fire Model in a Land Surface Model

Wildfires have shown increasing trends in both frequency and severity across the Contiguous United States (CONUS). However, process-based fire models have difficulties in accurately simulating the burned area over the CONUS due to a simplification of the physical process and cannot capture the interplay among fire, ignition, climate, and human activities. The deficiency of burned area simulation deteriorates the description of fire impact on energy balance, water budget, and carbon fluxes in the Earth System Models (ESMs). Alternatively, machine learning (ML) based fire models, which capture statistical relationships between the burned area and environmental factors, have shown promising burned area predictions and corresponding fire impact simulation. We develop a hybrid framework (ML4Fire-XGB) that integrates a pretrained eXtreme Gradient Boosting (XGBoost) wildfire model with the Energy Exascale Earth System Model (E3SM) land model (ELM) version 2.1. A Fortran-C-Python deep learning bridge is adapted to support online communication between ELM and the ML fire model. Specifically, the burned area predicted by the ML-based wildfire model is directly passed to ELM to adjust the carbon pool and vegetation dynamics after disturbance, which are then used as predictors in the ML-based fire model in the next time step. Evaluated against the historical burned area from Global Fire Emissions Database 5 from 2001-2020, the ML4Fire-XGB model outperforms process-based fire models in terms of spatial distribution and seasonal variations. Sensitivity analysis confirms that the ML4Fire-XGB well captures the responses of the burned area to rising temperatures. The ML4Fire-XGB model has proved to be a new tool for studying vegetation-fire interactions, and more importantly, enables seamless exploration of climate-fire feedback, working as an active component in E3SM.

54 ENVIRONMENTAL SCIENCES↗

Challenges and Opportunities for Electric Utility Modeling and Asset Valuation Frameworks: Case Study on Valuing New Pumped Storage Hydropower

Asset valuation by electric utilities is becoming increasingly difficult in the rapidly changing electric sector. Rapid deployment of variable generation and inverter-based storage systems along with uncertain demand growth, climate, policies, and other factors create a challenging environment for understanding the value proposition of a new potential asset. This report describes an effort between the Tennessee Valley Authority (TVA) and three U.S. Department of Energy laboratories to perform a detailed review of utility modeling and analysis practices for asset valuation and identify challenges and opportunities for advancing its methods into the future. It focuses on a case study of new potential pumped storage hydropower (PSH) because of growing interest in new PSH capacity to provide energy balancing, firm capacity, and a range of ancillary services. Staff from the DOE labs conducted systematic interviews about current practices in capacity expansion modeling, production-cost modeling, hydrological modeling, and transmission stability modeling while also discussing how scenario analysis is conducted and how models and data are integrated. The effort resulted in a set of model, integration, and scenario recommendations that could be valuable to TVA, other utilities, system operators, and other stakeholders conducting integrated grid analysis. Individual model recommendations suggest exploring computational tradeoffs with detail and resolution across spatiotemporal structure, supply- and demand-side details, transmission overlays, market interactions, and ancillary services. Automated processes to pass data between models and conduct larger scenario suites could also enhance valuation practices by enabling a more consistent study of asset value across a broader range of uncertain future grid conditions where PSH could be particularly valuable. TVA and other industry stakeholders can learn from and adapt applied research-grade methods developed by DOE laboratories and other research institutions to improve decision making and accelerate progress towards a reliable, economic, sustainable energy system.

13 HYDRO ENERGY↗

Simulated wildfire burned area over the CONUS during 2001-2020

Wildfires have shown increasing trends in both frequency and severity across the Contiguous United States (CONUS). However, process-based fire models have difficulties in accurately simulating the burned area over the CONUS due to a simplification of the physical process and cannot capture the interplay among fire, ignition, climate, and human activities. The deficiency of burned area simulation deteriorates the description of fire impact on energy balance, water budget, and carbon fluxes in the Earth System Models (ESMs). Alternatively, machine learning (ML) based fire models, which capture statistical relationships between the burned area and environmental factors, have shown promising burned area predictions and corresponding fire impact simulation. We develop a hybrid framework (ML4Fire-XGB) that integrates a pretrained eXtreme Gradient Boosting (XGBoost) wildfire model with the Energy Exascale Earth System Model (E3SM) land model (ELM). A Fortran-C-Python deep learning bridge is adapted to support online communication between ELM and the ML fire model. Specifically, the burned area predicted by the ML-based wildfire model is directly passed to ELM to adjust the carbon pool and vegetation dynamics after disturbance, which are then used as predictors in the ML-based fire model in the next time step. Evaluated against the historical burned area from Global Fire Emissions Database 5 from 2001-2020, the ML4Fire-XGB model outperforms process-based fire models in terms of spatial distribution and seasonal variations. Sensitivity analysis confirms that the ML4Fire-XGB well captures the responses of the burned area to rising temperatures. The ML4Fire-XGB model has proved to be a new tool for studying vegetation-fire interactions, and more importantly, enables seamless exploration of climate-fire feedback, working as an active component in E3SM.

Liu, Ye↗