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 91 records · Page 5
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Deep Nonparametric Estimation of Operators between Infinite Dimensional Spaces

Learning operators between infinitely dimensional spaces is an important learning task arising in machine learning, imaging science, mathematical modeling and simulations, etc. This paper studies the nonparametric estimation of Lipschitz operators using deep neural networks. Non-asymptotic upper bounds are derived for the generalization error of the empirical risk minimizer over a properly chosen network class. Under the assumption that the target operator exhibits a low dimensional structure, our error bounds decay as the training sample size increases, with an attractive fast rate depending on the intrinsic dimension in our estimation. Our assumptions cover most scenarios in real applications and our results give rise to fast rates by exploiting low dimensional structures of data in operator estimation. We also investigate the influence of network structures (e.g., network width, depth, and sparsity) on the generalization error of the neural network estimator and propose a general suggestion on the choice of network structures to maximize the learning efficiency quantitatively.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2026

Deep nonparametric estimation of operators between infinite dimensional spaces

Learning operators between infinitely dimensional spaces is an important learning task arising in machine learning, imaging science, mathematical modeling and simulations, etc. This paper studies the nonparametric estimation of Lipschitz operators using deep neural networks. Non-asymptotic upper bounds are derived for the generalization error of the empirical risk minimizer over a properly chosen network class. Under the assumption that the target operator exhibits a low dimensional structure, our error bounds decay as the training sample size increases, with an attractive fast rate depending on the intrinsic dimension in our estimation. Our assumptions cover most scenarios in real applications and our results give rise to fast rates by exploiting low dimensional structures of data in operator estimation. We also investigate the influence of network structures (e.g., network width, depth, and sparsity) on the generalization error of the neural network estimator and propose a general suggestion on the choice of network structures to maximize the learning efficiency quantitatively.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS2010

Operational Lessons Learned from the Ares I-X Flight Test

The Ares I-X flight test, launched in 2009, is the first test of the Ares I crew launch vehicle. This development flight test evaluated the flight dynamics, roll control, and separation events, but also provided early insights into logistical, stacking, launch, and recovery operations for Ares I. Operational lessons will be especially important for NASA as the agency makes the transition from the Space Shuttle to the Constellation Program, which is designed to be less labor-intensive. The mission team itself comprised only 700 individuals over the life of the project compared to the thousands involved in Shuttle and Apollo missions; while missions to and beyond low-Earth orbit obviously will require additional personnel, this lean approach will serve as a model for future Constellation missions. To prepare for Ares I-X, vehicle stacking and launch infrastructure had to be modified at Kennedy Space Center's Vehicle Assembly Building (VAB) as well as Launch Complex (LC) 39B. In the VAB, several platforms and other structures designed for the Shuttle s configuration had to be removed to accommodate the in-line, much taller Ares I-X. Vehicle preparation activities resulted in delays, but also in lessons learned for ground operations personnel, including hardware deliveries, cable routing, transferred work and custodial paperwork. Ares I-X also proved to be a resource challenge, as individuals and ground service equipment (GSE) supporting the mission also were required for Shuttle or Atlas V operations at LC 40/41 at Cape Canaveral Air Force Station. At LC 39B, several Shuttle-specific access arms were removed and others were added to accommodate the in-line Ares vehicle. Ground command, control, and communication (GC3) hardware was incorporated into the Mobile Launcher Platform (MLP). The lightning protection system at LC 39B was replaced by a trio of 600-foot-tall towers connected by a catenary wire to account for the much greater height of the vehicle. Like Shuttle, Ares I-X will be stacked on a MLP and rolled out to the pad on a Saturn-era crawler-transporter. While Ares I-X was only held in place by the four hold-down posts on its aft skirt during rollout, a new vehicle stabilization system (VSS) attached to the vertical service structure kept the vehicle from undue swaying prior to launch at the pad, LC 39B. Following the launch, the flight test vehicle first stage was recovered with the aid of new parachutes resized to accommodate the five-segment-long first stage, which had a much greater length and mass than the Shuttle s reusable solid rocket boosters. After splashdown, recovery divers exercised extra care when handling the first stage to ensure that the flight data recorders in the fifth segment simulator were not damaged by exposure to sea water. The data recovered from the Ares I-X flight test will be very valuable in verifying the predicted environments and models used to design the vehicle. Lessons learned from Ares I-X will be shared with the Ares Projects through written and verbal reports and through integration of mission team members into the Project workforce.

Davis, Stephan R.↗
NASA NTRS2012

CloudSat Anomaly Recovery and Operational Lessons Learned

In April 2011, NASA's pioneering cloud profiling radar satellite, CloudSat, experienced a battery anomaly that placed it into emergency mode and rendered it operations incapable. All initial attempts to recover the spacecraft failed as the resultant power limitations could not support even the lowest power mode. Originally part of a six-satellite constellation known as the "A-Train", CloudSat was unable to stay within its assigned control box, posing a threat to other A-Train satellites. CloudSat needed to exit the constellation, but with the tenuous power profile, conducting maneuvers was very risky. The team was able to execute a complex sequence of operations which recovered control, conducted an orbit lower maneuver, and returned the satellite to safe mode, within one 65 minute sunlit period. During the course of the anomaly recovery, the team developed several bold, innovative operational strategies. Details of the investigation into the root-cause and the multiple approaches to revive CloudSat are examined. Satellite communication and commanding during the anomaly are presented. A radical new system of "Daylight Only Operations" (DO-OP) was developed, which cycles the payload and subsystem components off in tune with earth eclipse entry and exit in order to maintain positive power and thermal profiles. The scientific methodology and operational results behind the graduated testing and ramp-up to DO-OP are analyzed. In November 2011, the CloudSat team successfully restored the vehicle to consistent operational collection of cloud radar data during sunlit portions of the orbit. Lessons learned throughout the six-month return-to-operations recovery effort are discussed and offered for application to other R&D satellites, in the context of on-orbit anomaly resolution efforts.

cloud profiling radar satellite↗
NASA NTRS2006

Leadership Challenges in ISS Operations: Lessons Learned from Junior and Senior Mission Control Personnel

The International Space Station (ISS) is operated by a multi-national, multi-organizational team that is dispersed across multiple locations, time zones, and work schedules. At NASA, both junior and senior mission control personnel have had to find ways to address the leadership challenges inherent in such work, but neither have had systematic training in how to do so. The goals of this study were to examine the major leadership challenges faced by ISS mission control personnel and to highlight the approaches that they have found most effective to surmount them. We pay particular attention to the approaches successfully employed by the senior personnel and to the training needs identified by the junior personnel. We also evaluate the extent to which responses are consistent across the junior and senior samples. Further, we compare the issues identified by our interview survey to those identified by a standardized questionnaire survey of mission control personnel and a contrasting group of space station crewmembers. We studied a sample of 14 senior ISS flight controllers and a contrasting sample of 12 more junior ISS controllers. Data were collected using a semi-structured qualitative interview and content analyzed using an iterative process with multiple coders and consensus meetings to resolve discrepancies. To further explore the meaning of the interview findings, we also conducted new analyses of data from a previous questionnaire study of 13 American astronauts, 17 Russian cosmonauts, and 150 U.S. and 36 Russian mission control personnel supporting the ISS or Mir space stations. The interview data showed that the survey respondents had substantial consensus on several leadership challenges and on key strategies for dealing with them, and they offered a wide range of specific tactics for implementing these strategies. Interview data from the junior respondents will be presented for the first time at the meeting. The questionnaire data showed that the US mission control sample reported a level of support from their management that compared favorably to national norms. American mission control personnel and Russian crewmembers reported higher supervisor support than American crewmembers and Russian mission control personnel. We will present the specific issues underlying these findings and compare and contrast the results from the two datasets. Although specific to space station personnel, our results are consistent with recent management, cultural, and aerospace research. We aim to use our results to improve training for current and future mission control personnel.

Clement, James L.↗
NASA NTRS2010

The Evolution of Extravehicular Activity Operations to Lunar Exploration Based on Operational Lessons Learned During 2009 NASA Desert RATS Field Testing

This paper will present options to evolutionary changes in several philosophical areas of extravehicular activity (EVA) operations. These areas will include single person verses team EVAs; various loss of communications scenarios (with Mission Control, between suited crew, suited crew to rover crew, and rover crew A to rover crew B); EVA termination and abort time requirements; incapacitated crew ingress time requirements; autonomous crew operations during loss of signal periods including crew decisions on EVA execution (including decision for single verses team EVA). Additionally, suggestions as to the evolution of the make-up of the EVA flight control team from the current standard will be presented. With respect to the flight control team, the major areas of EVA flight control, EVA Systems and EVA Tasks, will be reviewed, and suggested evolutions of each will be presented. Currently both areas receive real-time information, and provide immediate feedback during EVAs as well as spacesuit (extravehicular mobility unit - EMU) maintenance and servicing periods. With respect to the tasks being performed, either EMU servicing and maintenance, or the specific EVA tasks, daily revising of plans will need to be able to be smoothly implemented to account for unforeseen situations and findings. Many of the presented ideas are a result of lessons learned by the NASA Johnson Space Center Mission Operations Directorate operations team support during the 2009 NASA Desert Research and Technology Studies (Desert RATS). It is important that the philosophy of both EVA crew operations and flight control be examined now, so that, where required, adjustments can be made to a next generation EMU and EVA equipment that will complement the anticipated needs of both the EVA flight control team and the crews.

Bell, Ernest R., Jr.↗
DOE OSTI2020

First Ever Polymer Flood Field Pilot to Enhance the Recovery of Heavy Oils on Alaska North Slope – Producer Responses and Operational Lessons Learned

The US Department of Energy and Hilcorp Alaska, LLC are jointly cosponsoring a field pilot project entitled “First Ever Field Pilot on Alaska’s North Slope to Validate the Use of Polymer Floods for Heavy Oil Enhanced Oil Recovery (EOR).” The objective of this pilot is to perform a field experiment to validate the use of an advanced polymer flooding technology to unlock the vast heavy oil resources on Alaska North Slope (ANS). Two pairs of horizontal injection and production wells have been deployed in an isolated fault block of the Schrader Bluff heavy oil reservoir at the Milne Point Field to conduct a polymer flood pilot which was designed to acquire field performance data to optimize polymer flood design in the Schrader Bluff heavy oil reservoirs on ANS. Hydrolyzed polyacrylamide (HPAM) polymer injection started on August 28, 2018 using a custom made polymer mixing and pumping unit. Polymer solution properties such as concentration, viscosity, and filter ratio, have been closely monitored and adjusted as necessary. Operational procedures have been improved to ensure adequate injectivity and polymer propagation through the formation. Oil and water production is tested frequently to monitor water cut performance, and produced water is sampled and analyzed for polymer content. Tracers were pumped pre and post polymer injection startup to compare polymer flood breakthrough timing and sweep efficiency. Polymer injectivity of the horizontal wells drilled in the Schrader Bluff NB sand is found to be sufficient to replace the reservoir production voidage although there were some declines as high viscosity polymer swept the near wellbore region. Production data show significant reduction in water cut and the oil production rate has stabilized. No polymer production has been confirmed from the two horizontal producers after 22 months of polymer injection into the two supporting horizontal injectors. Tracer data show very low sweep efficiencies with waterflood alone, whereas significant improvement has been achieved with polymer flood. Many operational lessons have been learned and improvements have been made regarding polymer mixing, pumping, and onsite quality control, particularly in a challenging Arctic environment. Adequate polymer injection rate can be achieved with diligent monitoring and quality assurance of the properties of the injected polymer solution.

Ning, Samson↗
NASA NTRS2006

Constellation Operations: Lessons Learned For Future Exploration

The Earth science community has long advocated placing numerous instruments in space to study the Earth and its environment. Space agencies from many countries have responded to this call with a wide range of orbiting satellites. Scientists also envisioned placing some satellites in constellations, to enable diverse remote sensing instruments to observe the same part of the Earth (or its atmosphere) at about the same time, thereby increasing the opportunities for coincident science observations. The Earth Science Afternoon Constellation is answering this call, but there have been unique challenges on the way to its deployment. Currently, the Afternoon Constellation is to comprise six satellites. Three are currently on orbit: NASA's Earth Observing System (EOS)-Aqua (2002) and EOS-Aura (2004), and CNES's Polarization & Anisotropy of Reflectances for Atmospheric Sciences coupled with Observations from a Lidar (PARASOL) (2004). Two more satellites, the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) and Cloudsat, are to be jointly launched in late 2005, followed by the Orbiting Carbon Observatory (OCO) in 2008. The Afternoon Constellation is unlike most satellite constellations in that: 1) It is not a homogenous mix of identical satellites; rather it comprises several satellites with complementary observational capabilities; 2) The satellites are not spaced around the Earth to provide instantaneous, global coverage (as for a communications satellite constellation); rather they orbit in close proximity so observations occur at about the same time over approximately the same region; and 3) Lastly, the satellites are not managed and controlled by one organization; rather the list of organizations is diverse: CNES in France, NASA Centers at Goddard, Langley, and the Jet Propulsion Laboratory, and the US Air Force facility in New Mexico. The PARASOL launch and early orbit (L&EO) phase proved to be a learning experience for constellation members (including Constellation management). Prior to launch, all members signed an operations coordination document that spelled out basic requirements for keeping the constellation safe and resolving non-nominal events. Once PARASOL reached orbit and the mission teams gained experience using the newly-developed constellation monitoring tools, it became clear that some of the guidelines in the signed agreements had to be adjusted. This paper presents the L&EO lessons learned and how they were used to prepare for the next phase - the period following the CloudSat/CALIPSO launch.

Kelly, Angelita C.↗
DOE OSTI2023

On the influence of over-parameterization in manifold based surrogates and deep neural operators

Constructing accurate and generalizable approximators (surrogate models) for complex physico-chemical processes exhibiting highly non-smooth dynamics is challenging. The main question is what type of surrogate models we should construct and should these models be under-parameterized or over-parameterized. In this work, we propose new developments and perform comparisons for two promising approaches: manifold-based polynomial chaos expansion (m-PCE) and the deep neural operator (DeepONet), and we examine the effect of over-parameterization on generalization. While m-PCE enables the construction of a mapping by first identifying low-dimensional embeddings of the input functions, parameters, and quantities of interest (QoIs), a neural operator learns the nonlinear mapping via the use of deep neural networks. Here, we demonstrate the performance of these methods in terms of generalization accuracy by solving the 2D time-dependent Brusselator reaction-diffusion system with uncertainty sources, modeling an autocatalytic chemical reaction between two species. We first propose an extension of the m-PCE by constructing a mapping between latent spaces formed by two separate embeddings of the input functions and the output QoIs. To further enhance the accuracy of the DeepONet, we introduce weight self-adaptivity in the loss function. We demonstrate that the performance of m-PCE and DeepONet is comparable for cases of relatively smooth input-output mappings. However, when highly non-smooth dynamics is considered, DeepONet shows higher approximation accuracy. We also find that for m-PCE, modest over-parameterization leads to better generalization, both within and outside of distribution, whereas aggressive over-parameterization leads to over-fitting. In contrast, an even highly over-parameterized DeepONet leads to better generalization for both smooth and non-smooth dynamics. Furthermore, we compare the performance of the above models with another recently proposed operator learning model, the Fourier Neural Operator, and show that its over-parameterization also leads to better generalization. Taken together, our studies show that m-PCE can provide very good accuracy at very low training cost, whereas a highly over-parameterized DeepONet can provide better accuracy and robustness to noise but at higher training cost. In both methods, the inference cost is negligible.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS2000

Lessons Learned on Operating and Preparing Operations for a Technology Mission from the Perspective of the Earth Observing-1 Mission

The New Millennium Program's first Earth-observing mission (EO-1) is a technology validation mission. It is managed by the NASA Goddard Space Flight Center in Greenbelt, Maryland and is scheduled for launch in the summer of 2000. The purpose of this mission is to flight-validate revolutionary technologies that will contribute to the reduction of cost and increase of capabilities for future land imaging missions. In the EO-1 mission, there are five instrument, five spacecraft, and three supporting technologies to flight-validate during a year of operations. EO-1 operations and the accompanying ground system were intended to be simple in order to maintain low operational costs. For purposes of formulating operations, it was initially modeled as a small science mission. However, it quickly evolved into a more complex mission due to the difficulties in effectively integrating all of the validation plans of the individual technologies. As a consequence, more operational support was required to confidently complete the on-orbit validation of the new technologies. This paper will outline the issues and lessons learned applicable to future technology validation missions. Examples of some of these include the following: (1) operational complexity encountered in integrating all of the validation plans into a coherent operational plan, (2) initial desire to run single shift operations subsequently growing to 6 "around-the-clock" operations, (3) managing changes in the technologies that ultimately affected operations, (4) necessity for better team communications within the project to offset the effects of change on the Ground System Developers, Operations Engineers, Integration and Test Engineers, S/C Subsystem Engineers, and Scientists, and (5) the need for a more experienced Flight Operations Team to achieve the necessary operational flexibility. The discussion will conclude by providing several cost comparisons for developing operations from previous missions to EO-1 and discuss some details that might be done differently for future technology validation missions.

Mandl, Dan↗
NASA NTRSDate not supplied

OSIRIS-REx Earth Return & Entry: Navigation Operations & Lessons Learned

The Origins, Spectral Interpretation, Resource Identification, and Security–Regolith Explorer (OSIRIS-REx) spacecraft successfully returned to Earth on September24, 2024, safely delivering its Sample Return Capsule (SRC) to the Utah Test and Training Range (UTTR). This paper describes the navigation operations that occurred between the departure from Bennu and the return of the SRC. An overview is given of the Flight Dynamics System (FDS) that includes tracking, orbit determination (OD), maneuver planning, and interfaces with entry, descent, and landing. Operational details on the SRC release criteria and conjunction assessment considerations are also provided. Lessons learned are presented that may help future sample return or interplanetary entry missions.

Kenneth M. Getzandanner↗
DOE OSTI2025

Network Slicing for Federated Learning in Operational Technology Environment

Industrial Control Systems (ICS) and Supervisory Control and Data Acquisition (SCADA) environments are essential to modern infrastructure, facing challenges in ensuring low-latency, high-throughput communication while mitigating cyber threats. This paper presents a framework integrating Federated Learning (FL) and network slicing with Quality of Service (QoS) to enable real-time monitoring without disrupting OT operations. Leveraging digital twin technology and Network Function Virtualization (NFV), the architecture supports predictive analytics and Industry 4.0 requirements. FL facilitates decentralized model training, preserving data privacy and scalability, though it introduces potential throughput constraints. Network slicing addresses this by creating dedicated virtualized segments optimized for performance and security. Advanced fault tolerance at the container and instance levels enhances system reliability. The proposed architecture ensures high throughput, low latency, and secure orchestration for real-time anomaly detection in OT networks. Performance evaluations validate its efficiency in throughput, deployment, and learning accuracy, providing a robust foundation for future ICS automation and data-driven decision-making.

Delgado, Brian G. Rodiles [University of Texas at ↗
NASA NTRSDate not supplied

Progress in Red vs. Green Propellant Operations: ASCENT Load Operations Lessons Learned

The Green Propellant Infusion Mission (GPIM) [1] is a NASA Space Technology Mission Directorate (STMD) effort to establish via demonstration ASCENT propellant (Advanced Space Craft Energetic Non-Toxic propellant, formerly referred to as AF-M315E) ground operations through on –orbit performance. ASCENT was developed by Air Force Research Laboratory (AFRL) as an increased energy density, reduced toxicity and reduced hazard monopropellant formulation relative to hydrazine. The GPIM team was led by Ball Aerospace and Technologies Corporation (BATC) as Program Management and Principal Investigator with AFRL, Aerojet-Rocketdyne (AR), the United States Space Force Space and Missile Systems Center (USSF/SMC), NASA Glenn Research Center (GRC), NASA Goddard Space Flight Center (GSFC), and NASA Kennedy Space Center (KSC) as Co-Investigators. The AFRL co-investigator role in the GPIM program was to develop the Propellant Ground Support Equipment (PGSE), associated Standard Operating Procedures (SOP) and Check Lists (CL), propellant load, propellant specification standard conventions and analysis methods, decontamination and waste disposal practices. This paper will discuss lessons learned in documentation, crew composition and readiness, equipment readiness, propellant readiness and potential improvements in the PGSE design.

Anthony P Zuttarelli↗
NASA NTRS2001

Using Model-Based Reasoning for Autonomous Instrument Operation - Lessons Learned From IMAGE/LENA

Model-based reasoning has been applied as an autonomous control strategy on the Low Energy Neutral Atom (LENA) instrument currently flying on board the Imager for Magnetosphere-to-Aurora Global Exploration (IMAGE) spacecraft. Explicit models of instrument subsystem responses have been constructed and are used to dynamically adapt the instrument to the spacecraft's environment. These functions are cast as part of a Virtual Principal Investigator (VPI) that autonomously monitors and controls the instrument. In the VPI's current implementation, LENA's command uplink volume has been decreased significantly from its previous volume; typically, no uplinks are required for operations. This work demonstrates that a model-based approach can be used to enhance science instrument effectiveness. The components of LENA are common in space science instrumentation, and lessons learned by modeling this system may be applied to other instruments. Future work involves the extension of these methods to cover more aspects of LENA operation and the generalization to other space science instrumentation.

Johnson, Michael A.↗
NASA NTRS2010

Operational Lessons Learned from NASA Analog Missions

National Aeronautics and Space Administration s (NASA) efforts in human space flight are currently focused on the Space Shuttle and International Space Station (ISS) programs, with efforts beginning on the future exploration opportunities. Both the Space Shuttle and ISS programs are important to the development of a capability for human exploration beyond Low Earth Orbit (LEO). The ISS provides extensive research capabilities to determine how the human body reacts to long duration stays in space. Also, the ISS and Shuttle can serve as a limited testbed for equipment or entire systems that may be used on missions to the Moon, Mars, or to a near-Earth asteroid. It has been nearly 35 years since the Apollo astronauts visited the Moon. Future space explorers will have to re-learn how to work and live on planetary surfaces, and how to do that for extended periods of time. Exploration crews will perform a wide assortment of scientific tasks, including material sampling and emplacement of automated instruments. Surface mission operations include the activities of the crew living and working, mission support from the Earth, and the operation of robotic and other remotely commanded equipment on the surface and in planetary orbit. Other surface activities will include the following: exploring areas surrounding a habitat; using rovers to collect rock and soil samples; setting up experiments on the surface to monitor the radiation environment and any seismic or thermal activity; and conducting scientific analyses and experiments inside a habitat laboratory. Of course, the astronauts will also have to spend some of their surface time "doing chores" and maintaining their habitat and other systems. In preparation for future planetary exploration, NASA must design the answers to many operational questions. What will the astronauts do on the surface? How will they accomplish this? What tools will they require for their tasks? How will robots and astronauts work together? What vehicle and system capabilities are required to support the activities? How will the crew and the Earth-based mission control team interact? During the initial phases of manned planetary exploration, one challenge in particular is virtually the same as during the Apollo program: How can scientific return be maximized during a relatively short surface mission? Today, NASA is investigating solutions to these challenges by conducting analog missions. These Earth-based missions possess characteristics that are analogous to missions on the Moon or Mars. These missions are excellent for testing operational concepts, and the design, configuration, and functionality of spacesuits, robots, rovers, and habitats. Analog mission crews test specific techniques and procedures for surface field geology, biological sample collection, and planetary protection. The process of actually working an analog mission reveals a myriad of small details, which either contribute to or impede efficient operations, many of which would never have been thought about otherwise. It also helps to define the suite of tools, containers, and other small equipment that surface explorers will use. This paper focuses on how analog missions have addressed selected operational considerations for future planetary missions.

Arnold, Larissa S.↗
DOE OSTI2023

Multifidelity deep operator networks for data-driven and physics-informed problems

Operator learning for complex nonlinear systems is increasingly common in modeling multi-physics and multi-scale systems. However, training such high-dimensional operators requires a large amount of expensive, high-fidelity data, either from experiments or simulations. In this work, we present a composite Deep Operator Network (DeepONet) for learning using two datasets with different levels of fidelity to accurately learn complex operators when sufficient high-fidelity data is not available. Additionally, we demonstrate that the presence of low-fidelity data can improve the predictions of physics-informed learning with DeepONets. We demonstrate the new multi-fidelity training in diverse examples, including modeling of the ice-sheet dynamics of the Humboldt glacier, Greenland, using two different fidelity models and also using the same physical model at two different resolutions.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Data Curation for Machine Learning Applied to Geothermal Power Plant Operational Data for GOOML: Geothermal Operational Optimization with Machine Learning: Preprint

Geothermal Operational Optimization with Machine Learning (GOOML) is a transferable and extensible component-based geothermal asset modeling framework that considers complex steamfield relationships and identifies optimization prospects using a data-driven approach to physics-guided, data-centric machine learning. This framework has been used to develop digital twins that provide steamfield operators with operational environments to analyze and understand historical and forecasted power production, explore new steamfield configuration possibilities, and seek optimal asset management in real world applications. To create, test, and apply the GOOML framework, diverse time-series datasets spanning multiple years were sourced from various geothermal power plant components within several complex real-world geothermal operations. These operations are based in the United States and New Zealand and include a variety of technologies, end-uses and configurations, collectively covering nearly all relevant operating conditions for modern geothermal fields. Datasets were acquired from multiple sources to ensure that machine learning experiments generalized properly to various operating conditions. It was found that the data varied in quality, format, and completeness. To ensure consistency between the various datasets, a standardized data curation process was developed to reliably streamline data preparation. This paper will discuss best practices as learned from the GOOML data curation process which takes the following steps: 1) acquisition of large quantities of data from power plant operators, 2) digestion of data to gain an initial understanding of what is included, 3) data transformation, which includes converting the data into a standardized machine-readable format so that they can be visualized, quality checked, and cleaned, 4) quality assurance and quality control, involving identification of significant data gaps and apparent anomalies through mapping of data features to real world componentry via the GOOML historical model, followed by discussion with modelers and power plant operators to identify additional data needs and to resolve issues, 5) use in machine learning algorithms, and 6) repetition of steps one through five until all data needs are met and data are deemed suitable for producing trustworthy modeling results which may be disseminated, ideally along with the curated dataset. This iterative process is focused on improving the quality of the data rather than tuning machine learning model parameters and supports a shift towards data-centric AI as a means to improving real-world applicability of geothermal machine learning projects.

access↗