Search NASA⌕ Search

SEARCH · Search NASA

Results for “Operator 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 1,603 records · Page 89
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

Incentivizing Cooperative Merging Control: Insights from Multi-Agent Deep Reinforcement Learning

Cooperative driving automation enables connected and automated vehicles (CAVs) to devise cooperative merging control, introducing great potentials to alleviate traffic congestion, reduce energy consumption, and enhance safety for highway on-ramp operations. Although numerous CAV cooperative merging algorithms have been developed to improve energy and traffic performance, the agreement-seeking among CAV users and their local benefits have been understudied. This can lead to rejections of cooperative merging plans and jeopardizing CAV performance, as a cooperation may entail certain CAVs to sacrifice their local benefits to achieve a system optimum. To address this issue, the study first leverages multi-agent deep reinforcement learning (MADRL) factoring both local reward and regional reward to demonstrate the discrepancies between CAV users’ local benefits and system optimum. Next, the existence of a correlated equilibrium is proved to characterize the convergence of MADRL training. This further facilitates the incorporation of incentives (computed based on reward discrepancies) to compensate for CAV users’ local benefits and facilitate system-optimal agreements in cooperative merging operations.

Zhou, Anye [ORNL] (ORCID:0000000301455579)↗
DOE OSTI2020

Analyzing Deep Learning Model Inferences for Image Classification using OpenVINO

It may be desirable to execute deep learning model inferences on an integrated GPU at the edge. While such GPUs are much less powerful than discrete GPUs, it is able to deliver higher floating-point operations per second than a CPU located on the same die. For edge devices, the benefit of moving to lower precision with minimal loss of accuracy to obtain higher performance is also attractive. Hence, we chose 14 deep learning models for image classification to evaluate their inference performance with the OpenVINO toolkit. Then, we analyzed the implementation of the fastest inference model of all the models. The experimental results are promising. Compared to the performance of full-precision (FP32) models, the speedup of the 8-bit (INT8) quantization ranges from 1.02 to 1.56 on an Intel (R) Xeon (R) 4-core CPU, and the speedup of the FP16 models ranges from 1.1 to 2 on an Intel (R) Iris (TM) Pro GPU. For the FP32 models, the GPU is on average 1.5X faster than the CPU.

CPU↗
DOE OSTI2024

A physics-based ensemble machine-learning approach to identifying a relationship between lightning indices and binary lightning hazard

To convert lightning indices generated by numerical weather prediction experiments into binary lightning hazard, a machine-learning tool was developed. This tool, consisting of parallel multilayer perceptron classifiers, was trained on an ensemble of planetary boundary layer schemes and microphysics parameterizations that generated four different lightning indices over 1 week. In a subsequent week, the multi-physics ensemble was applied and the machine-learning tool was used to evaluate the accuracy. Unintuitively, the machine-learning tool performed better on the testing dataset than the training dataset. Much of the error may be attributed to mischaracterizing the convection. The combination of the machine learning model and simulations could not differentiate between cloud-to-cloud lightning and cloud-to-ground lightning, despite being trained on cloud-to-ground lightning. It was found that the simulation most representative of the local operational model was the most accurate simulation tested.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2026

Self-supervised and multi-fidelity learning for extended predictive soil spectroscopy

Infrared spectroscopy is a cost-effective, non-destructive, and environmentally benign technology that is increasingly recognized as an important solution for meeting the global demand for soil data. While both near-infrared (NIR) and mid-infrared (MIR) diffuse reflectance spectroscopy enable rapid estimation of soil properties, they present a significant trade-off: NIR offers superior scalability and lower operational costs, whereas MIR provides higher analytical fidelity by capturing fundamental molecular vibrations. In this study, we propose a self-supervised, multi-fidelity learning framework designed to bridge this gap. Our approach leverages large-scale MIR spectral libraries to learn a compact, transferable latent representation, into which NIR spectra are subsequently aligned for downstream prediction. The workflow consists of pretraining a latent model on a large MIR library, adapting the representation using a smaller paired NIR–MIR dataset, and evaluating generalization on an independent external test set. Across a range of chemical and physical soil properties, we found that MIR-derived embeddings improved prediction accuracy relative to baseline models that used raw MIR inputs. Predictions derived from the spectrum conversion (NIR to MIR) task did not match the performance of the original MIR spectra but were similar or superior to predictive performance of NIR-only models, suggesting the unified spectral latent space can effectively leverage the larger and more diverse MIR dataset for prediction of soil properties not well represented in current NIR libraries.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2025

Lessons Learned from AskGDR: Usage and Impact Analysis of the Geothermal Data Repository's AI Research Assistant: Preprint

In October of 2024, the Department of Energy's (DOE) Geothermal Data Repository (GDR) team officially launched AskGDR, an AI research assistant resulting from the integration of a Large Language Model (LLM) with the metadata and supporting documents associated with GDR datasets. AskGDR allows GDR users to ask deeper questions about the origin of datasets, the methods used to collect them, and the findings they help support. Using Retrieval Augmented Generation (RAG), AskGDR can be used to summarize findings spread across dozens of papers and technical reports or to extract relevant information describing a single data field. However, generative AI is experimental. The National Renewable Energy Laboratory (NREL) has been collecting metrics on AskGDR and documenting lessons learned during its deployment. This paper will outline the efficacy and impact of AskGDR through analysis of its use, operating costs, number and types of questions asked, and the quality of answers provided.

15 GEOTHERMAL ENERGY↗
DOE OSTI2023

Battery state-of-health diagnostics during fast cycling using physics-informed deep-learning

Rapid, in-situ Li-ion battery state-of-health (SOH) quantification is challenging. Li-ion battery aging can vary significantly with chemistry, operating conditions, cycling demands, electrode design, and operation history. As a cell ages, optimal and safe operating conditions need to be adapted to account for battery degradation by tracking critical aging modes such as loss-of-lithium-inventory (LLI), loss-of-active-material (LAM) in either electrode, and/or impedance rise. This manuscript describes a framework for identifying battery aging modes in-operando using fast-rate voltage charge/discharge responses. The framework uses a physically based Li-ion battery model to produce synthetic high-rate responses at aged states. The aging model is calibrated against experimental data from cells with different electrode loadings and cycled under a variety of fast-charging conditions (1 h, 15 min, 10 min, and 7 min charging). The synthetically generated high-rate responses at aged states are then used to train a deep-learning model to identify real cell state-of-health from fast charge/discharge battery voltage responses. The synthetically trained deep-learning model performance is validated by comparing to standard incremental capacity analysis and half-cell measurements. Finally, the framework demonstrates the benefits of using high-rate physics-based models to generate synthetic data for training deep-learning models.

25 ENERGY STORAGE↗
DOE OSTI2022

Correlating Time-Resolved Pressure Measurements With Rim Sealing Effectiveness for Real-Time Turbine Health Monitoring

Purge flow is bled from the upstream compressor and supplied to the under-platform region to prevent hot main gas path ingress that damages vulnerable under-platform hardware components. A majority of turbine rim seal research has sought to identify methods of improving sealing technologies and understanding the physical mechanisms that drive ingress. While these studies directly support the design and analysis of advanced rim seal geometries and purge flow systems, the studies are limited in their applicability to real-time monitoring required for condition-based operation and maintenance. As operational hours increase for in-service engines, this lack of rim seal performance feedback results in progressive degradation of sealing effectiveness, thereby leading to reduced hardware life. To address this need for rim seal performance monitoring, this study utilizes measurements from a one-stage turbine research facility operating with true-scale engine hardware at engine-relevant conditions. Time-resolved pressure measurements collected from the rim seal region are regressed with sealing effectiveness through the use of common machine learning techniques to provide real-time feedback of sealing effectiveness. Two modeling approaches are presented that use a single sensor to predict sealing effectiveness accurately over a range of two turbine operating conditions. Here, the results show that an initial purely data-driven model can be further improved using domain knowledge of relevant turbine operations, which yields sealing effectiveness predictions within 3% of measured values.

42 ENGINEERING↗
DOE OSTI2021

Toward active disruption avoidance via real-time estimation of the safe operating region and disruption proximity in tokamaks

This paper describes a real-time capable algorithm for identifying the safe operating region around a tokamak operating point. The region is defined by a convex set of linear constraints, from which the distance of a point from a disruptive boundary can be calculated. The disruptivity of points is calculated from an empirical machine learning predictor that generates the likelihood of disruption. While the likelihood generated by such empirical models can be compared to a threshold to trigger a disruption mitigation system, the safe operating region calculation enables active optimization of the operating point to maintain a safe margin from disruptive boundaries. The proposed algorithm is tested using a random forest disruption predictor fit on data from DIII-D. The safe operating region identification algorithm is applied to historical data from DIII-D showing the evolution of disruptive boundaries and the potential impact of optimization of the operating point. Real-time relevant execution times are made possible by parallelizing many of the calculation steps and implementing the algorithm on a graphics processing unit. Lastly, a real-time capable algorithm for optimizing the target operating point within the identified constraints is also proposed and simulated.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2026

Machine Learning Based Metamodel for Faster Life Cycle Assessment of Large Portfolio of Buildings

Managing a large portfolio of buildings involves decisions on reuse, retrofit, renovation, rehabilitation, and new construction, influenced by trade-offs between performance metrics such as cost, time, and operational flexibility over the building's life cycle. Traditional life cycle assessment tools for evaluating these metrics can be labor- and compute-intensive, requiring extensive data and modeling for each building. Metamodels (or surrogate models) using machine learning have been explored as faster alternatives, but training these models has been hindered by the limited availability of comprehensive data on key life cycle metrics. Recent advancements in machine learning, particularly deep learning techniques like zero-shot and few-shot learning, allow models to learn from sparse or limited data. We propose a machine learning-based metamodel that leverages these techniques for rapid estimation of key building life cycle metrics. This presentation will cover the model architecture, data collection, training, and validation processes, along with an ongoing case study applied to a large portfolio of buildings. We will discuss the model's performance in terms of accuracy, compute time, limitations, and its potential for expanding to additional life cycle metrics. This data-driven approach offers a promising direction for the rapid evaluation of large building portfolios.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2026

Rolling Root Mean Square Based Multimodal Anomaly Detection for Real Time Monitoring of Smart Grid

Reliable real-time monitoring is valuable for maintaining the operational integrity of modern electrical smart grids. Deployment of heterogeneous sensing technologies in substations has enabled high-resolution, multichannel waveform monitoring, but also introduces challenges for anomaly detection due to noise, baseline drift, and modality-dependent signal characteristics. In this work, we present a computationally efficient unsupervised method for multimodal event detection based on Rolling Root Mean Square based Event Detection (RRMSED). The method is developed using in-house, field deployed sensors collecting data at a utility substation. The sensing system comprises voltage and current sensors, triaxial accelerometers, and magnetometers, collectively capturing electrical, vibrational, and magnetic waveform measurements at high temporal resolution. RRMSED operates by extracting rolling RMS energy features and their first-order temporal differences from consecutive waveform segments for each channel and then applying channel-specific statistical thresholds learned from historical data. A persistence-based exceedance logic is employed to robustly identify transient events while suppressing impulsive noise, and to provide precise temporal localization with high resolution. The framework is designed for continuous server-side operation and can be deployed in real time without requiring complex models. Experiments on simulated waveform data with known ground truth demonstrate low false positive (FP) and false negative (FN) rates. Application to real substation data shows RRMSED to identify events that are not captured by conventional monitoring indicators including fast transient detection algorithm currently deployed in the system. These results indicate that rolling RMS based features provide an effective and practical basis for real-time multimodal event detection in smart-grid substations.

Mukherjee, Subrata [ORNL] (ORCID:0000000309930338)↗
DOE OSTI2020

Adaptive Power System Emergency Control using Deep Reinforcement Learning

Power system emergency control is generally regarded as the final safety net for grid security and resiliency. Existing emergency control schemes are usually designed off-line based on either the conceived “worst” case scenarios or a few typical operation scenarios. These schemes are facing significant adaptiveness and robustness issues as increasing uncertainties and variations occur in modern electrical grids. To address these challenges, for the first time, this paper proposes a novel adaptive emergency control scheme using deep reinforcement learning (DRL), by leveraging the high-dimensional feature extraction and non-linear generalization capabilities DRL has for complex systems with high-dimensional variations. Furthermore, an open-source platform named DeepGrid has been designed for the first time to assist the DRL development and benchmarking processes in power system emergency control. Details of the platform, DRL, and emergency control schemes that use dynamic braking or under-voltage load shedding are presented. Extensive case studies performed in both two-area four-machine system and IEEE 39-Bus system have demonstrated the excellent performance and robustness of the proposed schemes.

Deep reinforcement learning, emergency control, lo↗
DOE OSTI2021

Deep learning inversion of gravity data for detection of CO 2 plumes in overlying aquifers

In this work, we developed an effective U-Net based deep learning (DL) model for inversion of surface gravity data on a rectangular grid to predict 2-D high-resolution subsurface CO 2 distribution along a vertical cross-section due to CO 2 leakage through a wellbore within a deep CO 2 storage reservoir. We used synthetic data to model two types of CO 2 leakage scenarios: one CO 2 plume in a shallow aquifer (single plume case), and two plumes present at different depths (double plume case). The 3-D synthetic plume samples were created by sampling among predetermined CO 2 plume depths, saturations, and volumes. The corresponding surface gravity data on a rectangular grid were generated by a 3-D forward model. The U-Net model detected 72% of single-plume samples, and one or both plumes in 75% of double-plume samples. Most of the undetected single plumes have small gravity field strengths below the typical noise level of 5 μGal. This model generated reproducible, reliable predictions with acceptable errors and demonstrated improved spatial resolution over the conventional least-squares inversion. In contrast to the conventional least-squares inversion, which often overestimates the size of its target and underestimates its density, this U-Net model accurately delineated the boundary of a target. Furthermore, this DL inversion detected deep, small, or low saturation CO 2 plumes that are often more difficult to resolve with conventional gravity inversion methods. We note the limitations of this feasibility study, including the use of synthetic data with regular CO 2 plume shapes, and the prediction of a 2-D plume cross-section rather than the full 3-D plume, as well, we recognize the lower detection fraction for double-plume scenarios. Nevertheless, this study demonstrates that DL gravity inversion is a promising and potentially superior method to conventional least-squares inversion. Our U-Net based deep learning inversion approach may be adapted for inversion of other types of geophysical data. DL inversion can facilitate near real-time monitoring of geologic carbon sequestration to provide site operators with prompt information about subsurface CO 2 distribution for risk management and mitigation.

58 GEOSCIENCES↗
DOE OSTI2020

Automatic and rapid calibration of urban building energy models by learning from energy performance database

Urban building energy modeling (UBEM) is attracting increasing attention in the energy modeling filed. Unlike modeling a single building using detailed building systems information, UBEM generally uses limited high-level building stock data to infer default assumptions about building characteristics and operations. Additionally, this practice inherently brings uncertainty to UBEM. This study introduced a novel method of automatic and rapid calibration of UBEM based on the annual electricity and natural gas energy use data by learning the correlations between crucial model input parameters and the building energy use from the reference building models. A case study was presented to calibrate 72 large office buildings built before 1978 in San Francisco. Seventeen model parameters were selected and Monte Carlo sampling was used to create 1000 samples that reasonably represent the parameter space. Then 1000 simulations were performed for the reference building model to create an energy performance database. The results showed that by learning from the energy performance database, it took less than four simulation runs on average to calibrate a building model. After the calibration, the distributions of each parameter were obtained to replace their single predefined default values. For example, the default lighting power density of 21.39 W/m 2 was calibrated to be 7.50 W/m 2 on average. The case study successfully demonstrated the effectiveness of the novel calibration method for UBEM in the mild climate. The method will be further tested in future for other climate zones and other building types.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2020

High precision control and deep learning-based corn stand counting algorithms for agricultural robot

This paper presents high precision control and deep learning-based corn stand counting algorithms for a low-cost, ultra-compact 3D printed and autonomous field robot for agricultural operations. Currently, plant traits, such as emergence rate, biomass, vigor, and stand counting, are measured manually. This is highly labor-intensive and prone to errors. The robot, termed TerraSentia, is designed to automate the measurement of plant traits for efficient phenotyping as an alternative to manual measurements. In this paper, we formulate a Nonlinear Moving Horizon Estimator that identifies key terrain parameters using onboard robot sensors and a learning-based Nonlinear Model Predictive Control that ensures high precision path tracking in the presence of unknown wheel-terrain interaction. Moreover, we develop a machine vision algorithm designed to enable an ultra-compact ground robot to count corn stands by driving through the fields autonomously. The algorithm leverages a deep network to detect corn plants in images, and a visual tracking model to re-identify detected objects at different time steps. We collected data from 53 corn plots in various fields for corn plants around 14 days after emergence (stage V3 - V4). The robot predictions have agreed well with the ground truth with C robot =1.02×C human -0.86 and a correlation coefficient R=0.96. The mean relative error given by the algorithm is -3.78%, and the standard deviation is 6.76%. These results indicate a first and significant step towards autonomous robot-based real-time phenotyping using low-cost, ultra-compact ground robots for corn and potentially other crops.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Machine Learning-Assisted Stability Boundary Determination of Multiport Autonomous Reconfigurable Solar Power Plants

The multiport autonomous reconfigurable solar (MARS) power plant is a promising solution to integrate renewable resources and energy storage systems into the alternating current (ac) power grid and an high-voltage direct current (HVdc) link. In the MARS system, various input power sources are connected to the individual submodules (SMs) through direct current (dc)–dc converters. However, the presence of external power sources can result in unbalanced capacitor voltages of SMs, thereby violating stability constraints under multiple/diverse operating conditions. This article aims to address the gap by accurately determining the stability boundary of the MARS system. As such, a novel machine learning (ML)-assisted energy balancing control (EBC) criterion is proposed. Further, in conjunction with a refined EBC, this approach ensures balanced capacitor voltages across various types of SMs, significantly enhancing the overall system efficiency. The proposed EBC criterion effectively controls EBC activation and deactivation, achieving remarkable accuracy. Both power systems computer aided design (PSCAD)/electromagnetic transients including direct current (EMTDC) simulations and control hardware-in-the-loop (cHIL) tests are conducted to validate the feasibility and efficiency of the proposed method. By combining the EBC and ML-assisted EBC criterion, efficient energy management is achieved for systems featuring multiple input power sources, such as MARS. This approach enables the system to fully exploit its potential across an expanded operational range while upholding high-efficiency standards.

14 SOLAR ENERGY↗
DOE OSTI2020

Learning to Solve Large-Scale Security-Constrained Unit Commitment Problems

Security-constrained unit commitment (SCUC) is a fundamental problem in power systems and electricity markets. In practical settings, SCUC is repeatedly solved via mixed-integer linear programming (MIP), sometimes multiple times per day, with only minor changes in input data. In this work, we propose a number of machine learning techniques to effectively extract information from previously solved instances in order to significantly improve the computational performance of MIP solvers when solving similar instances in the future. Based on statistical data, we predict redundant constraints in the formulation, good initial feasible solutions, and affine subspaces where the optimal solution is likely to lie, leading to a significant reduction in problem size. Computational results on a diverse set of realistic and large-scale instances show that using the proposed techniques, SCUC can be solved on average 4.3 times faster with optimality guarantees and 10.2 times faster without optimality guarantees, with no observed reduction in solution quality. Out-of-distribution experiments provide evidence that the method is somewhat robust against data-set shift. Summary of Contribution. The paper describes a novel computational method, based on a combination of mixed-integer linear programming (MILP) and machine learning (ML), to solve a challenging and fundamental optimization problem in the energy sector. The method advances the state-of-the-art, not only for this particular problem, but also, more generally, in solving discrete optimization problems via ML. We expect that the techniques presented can be readily used by practitioners in the energy sector and adapted, by researchers in other fields, to other challenging operations research problems that are solved routinely.

Machine Learning↗
DOE OSTI2019

Deep learning and radiomics: the utility of Google TensorFlow™ Inception in classifying clear cell renal cell carcinoma and oncocytoma on multiphasic CT

Purpose: Currently, all solid enhancing renal masses without microscopic fat are considered malignant until proven otherwise and there is substantial overlap in the imaging findings of benign and malignant renal masses, particularly between clear cell RCC (ccRCC) and benign oncocytoma (ONC). Radiomics has attracted increased attention for its utility in pre-operative work-up on routine clinical images. Radiomics based approaches have converted medical images into mineable data and identified prognostic imaging signatures that machine learning algorithms can use to construct predictive models by learning the decision boundaries of the underlying data distribution. The TensorFlow™ framework from Google is a state-of-the-art open-source software library that can be used for training deep learning neural networks for performing machine learning tasks. The purpose of this study was to investigate the diagnostic value and feasibility of a deep learning-based renal lesion classifier using open-source Google TensorFlow™ Inception in differentiating ccRCC from ONC on routine four-phase MDCT in patients with pathologically confirmed renal masses. Methods: With institutional review board approval for this 1996 Health Insurance Portability and Accountability Act compliant retrospective study and a waiver of informed consent, we queried our institution’s pathology, clinical, and radiology databases for histologically proven cases of ccRCC and ONC obtained between January 2000 and January 2016 scanned with a an intravenous contrast-enhanced four-phase renal mass protocol (unenhanced (UN), corticomedullary (CM), nephrographic (NP), and excretory (EX) phases). To extract features to be used for the machine learning model, the entire renal mass was contoured in the axial plane in each of the four phases, resulting in a 3D volume of interest (VOI) representative of the entire renal mass. We investigated thirteen different approaches to convert the acquired VOI data into a set of images that adequately represented each tumor which was used to train the final layer of the neural network model. Training was performed over 4000 iterations. In each iteration, 90% of the data were designated as training data and the remaining 10% served as validation data and a leave-one-out cross-validation scheme was implemented. Accuracy, sensitivity, specificity, positive (PPV) and negative predictive (NPV) values, and CIs were calculated for the classification of the thirteen processing modes. Results: We analyzed 179 consecutive patients with 179 lesions (128 ccRCC and 51 ONC). The ccRCC cohort had a mean size of 3.8 cm (range 0.8–14.6 cm) and the ONC cohort had a mean lesion size of 3.9 cm (range 1.0–13.1 cm). The highest specificity and PPV (52.9% and 80.3%, respectively) were achieved in the EX phase when we analyzed the single mid-slice of the tumor in the axial, coronal and sagittal plane, and when we increased the number of mid-slices of the tumor to three, with an accuracy of 75.4%, which also increased the sensitivity to 88.3% and the PPV to 79.6%. Using the entire tumor volume also showed that classification performance was best in the EX phase with an accuracy of 74.4%, a sensitivity of 85.8% and a PPV of 80.1%. When the entire tumor volume, plus mid-slices from all phases and all planes presented as tiled images, were submitted to the final layer of the neural network we achieved a PPV of 82.5%. Conclusions: The best classification result was obtained in the EX phase among the thirteen classification methods tested. Our proof of concept study is the first step towards understanding the utility of machine learning in the differentiation of ccRCC from ONC on routine CT images. We hope this could lead to future investigation into the development of a multivariate machine learning model which may augment our ability to accurately predict renal lesion histology on imaging.

62 RADIOLOGY AND NUCLEAR MEDICINE↗