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 685 records · Page 38
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Conceptual Design of Integrated Energy Systems with Market Interaction Surrogate Models

Most integrated energy system (IES) optimization frameworks employ the price-taker approximation, which ignores important interactions with the market and can result in overestimated economic values. In this work, we propose a machine learning surrogate-assisted optimization framework to quantify IES/market interactions and thus go beyond price-taker. We use time series clustering to generate representative IES operation profiles for the optimization problem and use machine learning surrogate models to predict the IES/market interaction. We quantify the accuracy of the time series clustering and surrogate models in a case study to optimally retrofit a nuclear power plant with a polymer electrolyte membrane electrolyzer to co-produce electricity and hydrogen.

Chen, Xinhe↗
DOE OSTI2025

Deep-learning-based domain adaptation for cavity fault prediction at Jefferson Laboratory

Superconducting radio-frequency (SRF) cavities are the core components of the Continuous Electron Beam Accelerator Facility (CEBAF) at Jefferson Lab, providing high-power electron beams for nuclear physics experiments. The facility comprises 418 SRF cavities, and any fault in these cavities can lead to interruptions in the electron beam supply. Cavity faults are the leading cause of beam trips in CEBAF. Predicting and mitigating those faults before onset can help maintain normal operation. Existing models face challenges in distinguishing between normal and fault signals when changes occur in the underlying time-series data, from changes in control software, operational parameters, or the environment. This work proposes a deep learning domain adaptation model that leverages transfer learning to address fault prediction challenges by improving accuracy. The model is trained and fine-tuned using a dataset collected for faulty and normal operation using a data acquisition system in CEBAF. Our deep learning-based domain adaptation model achieves a prediction accuracy of 89.61% of the fault and normal signals. The developed model effectively predicts normal running signals compared to the baseline approach without domain adaptation. This capacity is essential for the fault prediction task in the CEBAF because of heavily imbalanced data containing vast amounts of normal signals. The model performs well for predicting faults several hundred milliseconds before the fault onset compared to other models where no adaptation is applied. Incorporating deep learning-based domain adaptation techniques will significantly improve the fault prediction performance.

Rahman, Md Monibor [Old Dominion Univ., Norfolk, V↗
DOE OSTI2024

SUBTASK 1.6 – BASIN ELECTRIC CARBON STORAGE RESEARCH PROJECT: NOVEL MONITORING TECHNIQUES

The Energy & Environmental Research Center (EERC) conducted baseline activities associated with an applied research project at Basin Electric Power Cooperative’s (Basin’s) carbon capture and storage (CCS) site in Beulah, North Dakota, to establish novel carbon storage-monitoring techniques as commercial methods under Cooperative Agreement No. DE-FE0024233, Subtask 1.6. The following report summarizes the baseline activities performed and briefly describes the subsequent (operational monitoring) activities that have been proposed to the U.S. Department of Energy (DOE) as part of the overall project to develop and demonstrate novel monitoring techniques at North America’s largest permitted CCS operation. Dakota Gasification Company (DGC), a wholly owned subsidiary of Basin, owns and operates the Great Plains Synfuels Plant (GPSP) approximately 5 miles northwest of the town of Beulah, North Dakota (Figure 1). In 2023, DGC received approval from the North Dakota Industrial Commission (NDIC) to develop a storage facility on-site for injecting a stream of carbon dioxide (CO2) captured from GPSP. DGC will transport the captured CO2 stream with approximately 6.8 miles of transmission lines that extend north of GPSP and inject >1 million tonnes (MMt) of CO2 annually (>1 MMt/yr) over a 12-year period with up to six underground injection control (UIC) Class VI-compliant injection wells completed in the Broom Creek Formation, a predominantly sandstone reservoir and saline aquifer underlying GPSP. The Broom Creek Formation lies approximately 5900 feet (ft) below ground surface (bgs) at GPSP. The commercial scale (i.e., >1 MMt/yr) of DGC’s permitted carbon storage project is ideal for developing and testing the novel monitoring techniques included within Subtask 1.6. The goals of this project are to demonstrate 1) the cost-effectiveness of novel monitoring technologies included as part of this research, 2) technology capability for tracking the CO2 plume and/or associated pressure response in the subsurface and monitoring out-of-zone migration, and 3) compliance with UIC Class VI program requirements. The research activities proposed for the overall project include 1) design of an automated, integrated, modular (AIM) monitoring station; 2) time-lapse electromagnetic (EM) field surveys; 3) drone-based surveillance studies; 4) time-lapse monitoring with seismic methods; 5) advanced wellbore-monitoring methods; 6) deployment of an AIM monitoring network; 7) EM monitoring of CO2 with real-time data processing; 8) continued seasonal drone-based surveillance studies; 9) seismic monitoring with passive and active surveys; and 10) wellbore monitoring with nuclear magnetic resonance (NMR) for near-surface characterization. Completion of Activities 1.0–5.0 (baseline activities) are described in this report. Upon authorization of funding by DOE, the EERC will initiate Activities 6.0– 10.0 (operational monitoring activities). Current state-of-the-art (SOA) carbon storage-monitoring techniques require countless labor hours dedicated to the acquisition of data. Once data are gathered, these SOA techniques often rely on commercial facilities to process raw data from the field. However, it is anticipated that next-generation monitoring techniques, such as those being demonstrated, will lower acquisition footprints, be less operationally intensive, and improve data acquisition efficiencies. These new techniques are more conducive to the application of machine learning, artificial intelligence, and automation, thus providing a pathway for integration into active control systems, informing site operability, and improving the integration of data for future CCS projects across the United States. Additionally, reclaimed and active mining lands are present within the project site, creating a unique opportunity to demonstrate the effectiveness of remote sensing and surface-based geophysics monitoring techniques at similar project sites that may include disturbed, unconsolidated, or actively excavated near-surface environments. The efforts included in the overall project will produce necessary designs, learnings, and data acquired during the baseline and operational monitoring periods that are necessary for time-lapse demonstration and validation of the described monitoring techniques. In addition, it is anticipated that the monitoring technologies included in this study will be compliant with UIC Class VI requirements to enable the potential for implementation at other CCS sites across the United States.

42 ENGINEERING↗
DOE OSTI2022

An Efficient Distributed Reinforcement Learning for Enhanced Multi-Microgrid Management

Economic dispatch in multi-microgrid (MMG) systems requires coordinating distributed energy resources (DERs) of different microgrids, which leads to a significant increase in the number of states for energy management. In these cases, traditional reinforcement learning (RL) approaches become computationally expensive or output a solution that causes extra-operating costs for the system. This paper proposes an RL approach that employs local learning agents to interact with microgrid environments in a distributed manner and aggregates the outcomes to train the global agent to learn the policy for the MMG system. This distributed exploration and aggregation process provides an effective solution and guides the global agent to learn the dispatch policy efficiently. Case studies are performed on a system with three microgrids with different types of DERs. Results obtained using the proposed RL and comparisons with conventional methods substantiate the effectiveness of the proposed approach in terms of operation costs, computation time, and peak-to-average ratio.

Das, Avijit↗
DOE OSTI2019

Investigation and Recovery Following the Drum Event at INL's ARP-V Facility - 19517

Contractors working under the Department of Energy's (DoE's) Idaho Field Office have been exhuming and processing buried legacy waste from early weapons development and productions operations since 1984. Since operations began, nearly 400,000 drums have passed through this process. On April 11, 2018, repackaged drums were staged in an airlock (radiological buffer area) awaiting transfer to another facility for characterization. Late in the evening after operations had ended, four of the drums over pressurized causing an energetic displacement of their lids. A significant fraction of drum contents was ejected and dispersed throughout the airlock. The event activated the fire detection system in the facility initiating an emergency response from the site fire department, radiological protection, and operations. Initial intervention by fire department personnel was unsuccessful at quenching the reaction, so the responders egressed to a safe distance and monitored the situation. After determining the event had terminated and it was safe to enter, the facility was assessed revealing significant contamination in the airlock from the ejected waste. The area surrounding the facility was surveyed and it was determined that no contamination was released outside the facility. The Fluor Idaho, LLC (Fluor Idaho) Program Manager commissioned an investigation and recovery effort, which focused on three aspects: (1) technical investigation, (2) event and causal analysis, and (3) facility recovery. A team of highly qualified individuals from around the DOE Complex was assembled to support the technical investigation. The team included representatives from several Fluor Idaho, Fluor Corporate, Battelle Energy Alliance, LLC (INL), LANL, SRNL, South West Research Institute, DOE Carlsbad Field Office, and Defense Nuclear Facility Safety Board, and DOE-Headquarters. This team addressed the reaction mechanism of the event by guiding the sampling activities and coordinating analysis and testing by the analytical laboratories. Independent and experienced investigators were brought in to conduct a rigorous event and cause analysis. The Causal Analysis Team pursued direct, root, and contributing causes. Interviews were conducted, and documents were reviewed to develop an event and causal analysis. The Accelerated Retrieval Project V Operations team mobilized to decontaminate and recover the facility. Significant progress has been made in this effort, area and airborne contamination levels have been significantly reduced allowing limited activities to resume. In addition, this group supported sample collection efforts as directed by the Technical Investigation Team. This paper reviews key findings and milestones from the technical investigation, the event and causal analysis, and facility recovery operations. From these elements of the investigation, lessons learned are identified that would benefit other organizations involved with legacy waste management and disposition. Waste management is common to all sites in the DOE Complex and the events and recovery efforts described in this paper may provide beneficial insights to these operations. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗
DOE OSTI2019

International MNSR Training Center

Plans for the conversion of the Chinese origin miniature neutron source reactors (MNSRs) to low-enriched uranium (LEU), and the return of the high- enriched uranium (HEU) core to China, were initiated in February 2011 when an international MNSR Working Group (WG) was established to coordinate these activities. Based on a Ghana Atomic Energy Commission (GAEC) decision, the MNSR core at the GHARR-1 facility was selected for pilot shipment. Working on the preparation of the HEU MNSR materials’ removal from the GHARR-1 facility as a training tool, a mock-up MNSR vessel was built for operator training. Experiencing the usefulness of the mock-up vessel in operator training in the Ghana case has since been developed further into a full-scale training facility, named the International MNSR Training Center (IMTC), which is now available for training operators from MNSR operating countries. The main objective of this Book is to document the IMTC as a full-scale MNSR training facility in operator training for HEU MNSR core discharge and packaging. The goal is to describe the IMTC in its full complexity — that is: (1) to present the IMTC’s buildings with its internal and outer areas; (2) to provide a full listing of the installed equipment, as well as a functional description, available at the IMTC site; (3) to introduce the training materials, tutorials, and training program; and (4) to present the training sessions and Dry-run demonstration of the recently completed NIRR-1 operator training, with their experiences and lessons-learned included. This Book serves as a kind of user manual for the IMTC, which contains all of the information to launch training courses at the site at any time. The target audience for this Book is the MNSR community, especially MNSR operators and regulators who are facing HEU-LEU core conversion and HEU material transportation, who require operator training to carry out their jobs safely and professionally.

99 - GENERAL AND MISCELLANEOUS↗
DOE OSTI2016

Lessons Learned about Liquid Metal Reactors from FFTF Experience

The Fast Flux Test Facility (FFTF) is the most recent liquid-metal reactor (LMR) to operate in the United States, from 1982 to 1992. FFTF is located on the DOE Hanford Site near Richland, Washington. The 400-MWt sodium-cooled, low-pressure, high-temperature, fast-neutron flux, nuclear fission test reactor was designed specifically to irradiate Liquid Metal Fast Breeder Reactor (LMFBR) fuel and components in prototypical temperature and flux conditions. FFTF played a key role in LMFBR development and testing activities. The reactor provided extensive capability for in-core irradiation testing, including eight core positions that could be used with independent instrumentation for the test specimens. In addition to irradiation testing capabilities, FFTF provided long-term testing and evaluation of plant components and systems for LMFBRs. The FFTF was highly successful and demonstrated outstanding performance during its nearly 10 years of operation. The technology employed in designing and constructing this reactor, as well as information obtained from tests conducted during its operation, can significantly influence the development of new advanced reactor designs in the areas of plant system and component design, component fabrication, fuel design and performance, prototype testing, site construction, and reactor operations. The FFTF complex included the reactor, as well as equipment and structures for heat removal, containment, core component handling and examination, instrumentation and control, and for supplying utilities and other essential services. The FFTF Plant was designed using a “system” concept. All drawings, specifications and other engineering documentation were organized by these systems. Efforts have been made to preserve important lessons learned during the nearly 10 years of reactor operation. A brief summary of Lessons Learned in the following areas will be discussed: Acceptance and Startup Testing of FFTF FFTF Cycle Reports

Wootan, David W.↗
DOE OSTI2021

Probing effective field theory operators in the associated production of top quarks with a Z boson in multilepton final states at $\sqrt{s}$ = 13 TeV

A search for new top quark interactions is performed within the framework of an effective field theory using the associated production of either one or two top quarks with a Z boson in multilepton final states. The data sample corresponds to an integrated luminosity of 138 fb -1 of proton-proton collisions at √s = 13 TeV collected by the CMS experiment at the LHC. Five dimension-six operators modifying the electroweak interactions of the top quark are considered. Novel machine-learning techniques are used to enhance the sensitivity to effects arising from these operators. Distributions used for the signal extraction are parameterized in terms of Wilson coefficients describing the interaction strengths of the operators. All five Wilson coefficients are simultaneously fit to data and 95% confidence level intervals are computed. All results are consistent with the SM expectations.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2024

Physics-informed machine learning of the correlation functions in bulk fluids

The Ornstein–Zernike (OZ) equation is the fundamental equation for pair correlation function computations in the modern integral equation theory for liquids. In this work, machine learning models, notably physics-informed neural networks and physics-informed neural operator networks, are explored to solve the OZ equation. The physics-informed machine learning models demonstrate great accuracy and high efficiency in solving the forward and inverse OZ problems of various bulk fluids. The results highlight the significant potential of physics-informed machine learning for applications in thermodynamic state theory.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Postdisaster Routing of Movable Energy Resources for Enhanced Distribution System Resilience: A Deep Reinforcement Learning-Based Approach

The deployment of movable energy resources (MERs) can be an effective strategy to restore critical loads to enhance power system resilience when no other energy sources are available after the occurrence of an extreme event. Since the optimal locations of MERs following an extreme event are dependent on system operating states (e.g., the loads at each node, on/off status of system branches, and so on), existing analytical and population-based approaches must repeat the entire analysis and calculation when the system operating states change. On the contrary, if deep reinforcement learning (DRL)-based algorithms are sufficiently trained with a wide range of scenarios, they can quickly find optimal or near-optimal locations irrespective of changes in system states. A deep Q-learning-based approach is proposed for optimal MER deployment to enhance power system resilience. MERs can be also utilized to complement other types of resources, if available. The proposed approach operates in two stages after the occurrence of extreme events. In the first stage, the distribution network is represented as a graph, and the network is then reconfigured using tie switches by using Kruskal’s spanning forest search algorithm (KSFSA). To maximize critical load recovery, the optimal or near-optimal locations of MERs are chosen in the second stage. Further, case studies on a 33-node distribution system and a modified IEEE 123-node system demonstrate the effectiveness of the proposed approach for postdisaster routing of MERs.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2024

Frequency-domain computing using nonlinear acoustic-wave device on lithium niobate

Abstract Multiply-accumulation are crucial computing operations in signal processing, numerical simulations, and machine learning. In recent years, optical analog approaches have demonstrated higher computing performance and better power efficiency than their digital counterparts. However, analog computing chips usually need large areas and complex structures for parallel computing, as a single device element only executes one computing operation at a single time. Here, we demonstrate frequency-domain computing using the nonlinear acoustic-wave devices on lithium niobate, featuring a normalized external second-harmonic generation conversion efficiency of ~ 5.7 × 10-4 W-1. The second-order sum-frequency nonlinear process of lithium niobate enables multiplication of inputs encoded in the frequency domain. Compared to the analog schemes, our device features a notably simpler design, and nanofabrication requires only one lift-off. Using a single acoustic-wave device within an area of 0.03 mm2, we can simultaneously conduct over 130,000 multiply-accumulation operations. Our acoustic-wave device shows applications in real and complex vector convolutions and image processing. This demonstration sets the stage for experimental realizations into frequency-domain integrated nonlinear acoustic computing systems, potentially shaping future developments in acoustic neural networks and quantum computing.

chai, mingzhao (ORCID:0009000466226341)↗
DOE OSTI2025

Accelerating Hamiltonian Monte Carlo for Bayesian inference in neural networks and neural operators

Hamiltonian Monte Carlo (HMC) is a powerful and accurate method to sample from the posterior distribution in Bayesian inference. However, HMC techniques are computationally demanding for Bayesian neural networks due to the high dimensionality of the network’s parameter space and the non-convexity of their posterior distributions. Therefore, various approximation techniques, such as variational inference (VI) or stochastic gradient MCMC, are often employed to infer the posterior distribution of the network parameters. Such approximations introduce inaccuracies in the inferred distributions, resulting in unreliable uncertainty estimates. In this work, we propose a hybrid approach that combines inexpensive VI and accurate HMC methods to efficiently and accurately quantify uncertainties in neural networks and neural operators. The proposed approach leverages an initial VI training on the full network. We examine the influence of individual parameters on the prediction uncertainty, which shows that a large proportion of the parameters do not contribute substantially to uncertainty in the network predictions. This information is then used to significantly reduce the dimension of the parameter space, and HMC is performed only for the subset of network parameters that strongly influence prediction uncertainties. This yields a framework for accelerating the full batch HMC for posterior inference in neural networks. We demonstrate the efficiency and accuracy of the proposed framework on deep neural networks and operator networks, showing that inference can be performed for large networks with tens to hundreds of thousands of parameters. Finally, we show that this method can effectively learn surrogates for complex physical systems by modeling the operator that maps from upstream conditions to wall-pressure data on a cone in hypersonic flow.

Bayesian inference↗
DOE OSTI2022

DSO+T: Transactive Energy Coordination Framework (DSO+T Study: Volume 3)

This report describes a transactive energy coordination scheme designed to integrate into existing day-ahead and real-time wholesale energy markets. This scheme was evaluated in the Distribution System Operator with Transactive (DSO+T) study to assess the engineering and economic performance of the transactive energy coordination of a large-scale deployment of distributed energy resources (DER). Transactive agents were developed for a range of DERs (heating, ventilation, and air conditioning units, water heaters, batteries, and electric vehicles) that optimize flexibility over a 48-hour horizon and adjust their strategy in response to changes in real-time prices. A transactive energy coordination scheme, executed by a DSO retail market operator, aggregates these DER bids from participating customers and clears them against a DSO supply curve using a double auction market mechanism. The process of constructing the price-quantity DSO supply curve includes distribution-level transportation constraints (for example, substation congestion limits) and forecast locational marginal price of the DSO’s connected transmission node. The resulting day-ahead and real-time quantities are then bid into a competitive wholesale market operated by an independent system operator. This report also details additional capabilities for proper marketplace simulation such as wholesale price, weather, and load forecasting. The report concludes with a discussion of lessons learned and key design features required to ensure successful operation.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗
DOE OSTI2022

Demonstration and Evaluation of the Human-Technology Integration Guidance for Plant Modernization

The significance of nuclear power in its role producing carbon-free electricity to the U.S. cannot be overstated. However, with changes in the energy market coupled with changes in incentives given to certain resources like solar and wind, the operating and maintenance costs for these sources have seen a significant reduction, which has consequently negatively impacted the economic viability of the existing U.S. nuclear power plant fleet. Digital instrumentation and control (I&C) and control room modernization is a major critical work domain to reduce operating and maintenance costs. Existing nuclear power plants are commonly configured with mostly legacy analog I&C as well as isolated pockets of digital I&C (a plant process computer, digital recorders, etc.). One challenge with this analog I&C is that replacement parts are becoming prohibitively more expensive and difficult to obtain. Moreover, a significant challenge with the existing analog I&C is that the way in which plants are currently operated and maintained is no longer competitive with other electricity generating sources, like natural gas, where advanced digital I&C technologies are commonplace. This gap between the waynuclear power plants are operated compared to other electricity generating sources significantly challenges the economic viability of the nuclear industry. Indeed, digital I&C systems can fundamentally change the way the plant is operated (i.e., the concept of operation). The introduction of digital I&C technologies offers a wealth of benefits to the nuclear industry. However, it is important to emphasize that, to realize these benefits, a careful understanding of how to integrate technology in a collaborative way that leverages the capabilities of people and technologies is necessary. Human-technology integration applies human factors engineering methods and tools to ensure the safe and reliable use of these technologies while ensuring that the inherent features of the technologies that provide economic value are not missed. The scope of this work documents the demonstration of the recently developed human and technology integration methodology, as applied to developing a new vision and concept of operations for a major U.S. nuclear power plant fleet. This report focuses on the methodological aspects of developing a vision and concept of operations. This report shares the tools, activities, and lessons learned during a modernization currently underway for industry as a whole to consider when planning any significant digital modification and developing a new vision and concept of operations.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2014

Archiving Data from New Survey Technologies: Lessons Learned on Enabling Research with High-Precision Data While Preserving Participant Privacy: Preprint

During the past 15 years, increasing numbers of organizations and planning agencies have begun collecting high-resolution Global Positioning System (GPS) travel data. Despite the significant effort and expense to collect it, privacy concerns often lead to underutilization of the data. To address this dilemma of providing data access while preserving privacy, the National Renewable Energy Laboratory, with support from the U.S. Department of Transportation and U.S. Department of Energy, established the Transportation Secure Data Center (TSDC). Lessons drawn from best-practice examples from other data centers have helped shape the structure and operating procedures for the TSDC, which functions under the philosophy of first and foremost preserving privacy, but doing so in a way that balances security with accessibility and usability of the data for legitimate research. This paper provides details about the TSDC approach toward achieving these goals, which has included creating a secure enclave with no external access for backing up and processing raw data, a publicly accessible website for downloading cleansed data, and a secure portal environment through which approved users can work with detailed spatial data using a variety of tools and reference information. This paper also describes lessons learned from operating the TSDC with respect to improvements in GPS data handling, processing, and user support, along with plans for continual enhancements to better support the needs of both data providers and users and to thus advance the research value derived from such valuable data.

33 ADVANCED PROPULSION SYSTEMS↗
DOE OSTI2016

Lessons Learned Concerning the Onsite Disposal Facility at the Fernald Preserve, Harrison, Ohio - 16176

An onsite disposal facility (OSDF) was constructed at the Fernald Preserve, Ohio, for the disposal of contaminated soil and debris generated during a Comprehensive Environmental Response, Compensation, and Liability Act site cleanup. Completed in October of 2006, the 30 ha facility contains 2.26 million m{sup 3} of waste materials, has a maximum height of approximately 19.8 m, and consists of eight individual cells. The dimensions of cells 1-7 are approximately 213 m x 122 m each, or 26,000 m{sup 2} (2.60 ha). The dimensions of cell 8 (the last cell constructed) are larger than those of the other cells (approximately 3.8 ha). Each cell was constructed with a leachate collection system (LCS) that collected rainwater and storm-water runoff during waste placement and prevented it from entering the underlying environment. Other engineered features include a multilayer composite liner system, a leak detection system (LDS) positioned beneath the primary liner, and a multilayer composite cover that was placed over each cell following the completion of waste-placement activities. The LCS and LDS layers are designed to convey (by gravity) any leachate/fluid that enters the system through pipes to the west side of the facility and into tanks located in valve houses. From there, leachate can be sampled and conveyed to treatment. Groundwater is monitored beneath each cell via monitoring wells. In November 2006, the US DOE Office of Legacy Management (LM) assumed responsibility for the long-term care and maintenance of the OSDF. LM has nine years of monitoring and maintenance experience with the Fernald OSDF. Monitoring data is reported each year in the Fernald Annual Site Environmental Report. The data collected and inspections conducted indicate that the facility is operating as designed. The past nine years has provided LM with the opportunity to evaluate how well many of the design and as-built features of the facility have functioned over time as they pertain to environmental monitoring, valve-house operation and maintenance, and leachate transmission. Lessons learned concerning the long-term care and maintenance of the OSDF at the Fernald Preserve are presented. The main lessons learned include the following: How bivariate plots can be used to address environmental monitoring challenges created by installing an engineered disposal facility in a location where contamination is above the background level but below cleanup standards. How valve-house design decisions can help to alleviate long-term monitoring challenges. How scale buildup can lead to leachate transmission issues. Some additional minor lessons learned are also presented. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗
DOE OSTI2019

A Supervised Machine Learning Approach to Control Energy Storage Devices

This paper introduces a supervised machine learning approach to predict and schedule the real-time operation mode of the next operation interval for residential PV/Battery systems controlled by mode-based controllers. The performance of the mode-based economic model-predictive control (EMPC) approach is used as the benchmark. The residential load and PV data used in the paper are 1-minute data downloaded from the the Pecan Street Project website. The optimal operation mode for each control interval is first derived from the historical data used as the training set. Then, four machine learning algorithms (i.e. neural network, support vector machine, logistic regression, and random forest algorithms) are applied. We compared the performance of the four algorithms when using different number of features and length of the training sets extracted from different months of the year. Simulation results show that using the machine learning approach can effectively improve the performance of the mode-based control system and reduce the computation effort of local controllers because the training can be completed on a cloud-based Machine Learning engine. Furthermore, the work presented in this paper paves the way for using a shared-learning platform to design controllers of residential PV/storage systems. This may significantly reduce the cost for implementing such systems.

25 ENERGY STORAGE↗
DOE OSTI2025

From disorganized data to emergent dynamic models: Questionnaires to partial differential equations

Starting with sets of disorganized observations of spatially varying and temporally evolving systems, obtained at different (also disorganized) sets of parameters, we demonstrate the data-driven derivation of parameter dependent, evolutionary partial differential equation (PDE) models capable of generating the data. This tensor type of data is reminiscent of shuffled (multidimensional) puzzle tiles. The independent variables for the evolution equations (their “space” and “time”) as well as their effective parameters are all emergent , i.e. determined in a data-driven way from our disorganized observations of behavior in them. We use a diffusion map based questionnaire approach to build a smooth parametrization of our emergent space/time/parameter space for the data. This approach iteratively processes the data by successively observing them on the “space,” the “time” and the “parameter” axes of a tensor. Once the data become organized, we use machine learning (here, neural networks) to approximate the operators governing the evolution equations in this emergent space. Our illustrative examples are based (i) on a simple advection–diffusion model; (ii) on a previously developed vertex-plus-signaling model of Drosophila embryonic development; and (iii) on two complex dynamic network models (one neuronal and one coupled oscillator model) for which no obvious smooth embedding geometry is known a priori. This allows us to discuss features of the process like symmetry breaking, translational invariance, and autonomousness of the emergent PDE model, as well as its interpretability.

generative models↗