SEARCH · Engineering Papers
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.
NREL is tackling the most pressing data center challenges - energy consumption, resilience, resource adequacy, and energy source systems integration - through cutting-edge research and collaborative development.
Stand-alone data centers, nuclear-powered data centers, and novel nuclear facilities such as fusion plant and small modular reactors (SMRs) are rapidly expanding, water-intensive sectors that present a significant challenge to the water-for-energy nexus. Their large-scale consumption of water for cooling systems not only reduces downstream water quantity but also degrades quality, demanding an integrated approach by stakeholders to understand and address these impacts through infrastructure planning and development.
Artificial intelligence (AI) training data centers with periodic load profiles can induce sustained grid oscillations across a wide frequency range, making accurate monitoring essential for reliable power system operation. This report evaluates the adequacy of existing measurement systems for monitoring such oscillations, focusing on phasor measurement units (PMUs) and point-on-wave (POW) measurement systems. The analysis shows that while PMUs are highly effective for monitoring low-frequency electromechanical oscillations, they have inherent limitations in accurately representing higher-frequency oscillations due to constraints imposed by reporting rates and the bandwidth of phasor estimation filters. Even when configured with higher reporting rates, the filtering inherent in the phasor estimation process can significantly attenuate oscillation magnitudes, potentially leading to underestimation of oscillatory behavior. This has important implications for compliance and performance monitoring of large loads. To address the limitations associated with PMU-based monitoring, the report examines the use of high-resolution POW measurements, which can capture oscillations across a broader frequency range. However, continuous POW monitoring introduces practical challenges related to large data volumes, communication bandwidth, and real-time data processing. For this reason, the report also discusses emerging approaches that use POW measurements as a complementary capability alongside PMUs to improve observability of oscillations from large data center loads.
This report examines how battery energy storage systems (BESS) can help utilities accommodate large load growth while protecting affordability for existing ratepayers. Rapid growth in electricity demand from artificial intelligence (AI) data centers is straining the U.S. grid. Furthermore, many new data centers are entering rural markets, which could offer economic benefits but may also pose implementation challenges for smaller utilities. At the same time, retail electricity prices are increasing faster than inflation, elevating customer affordability as a key challenge. While data centers have not been the primary driver of increases in residential prices to date, they have pushed wholesale energy and capacity prices higher in several markets. Fundamental utility cost-allocation principles show that data center growth can be rate-positive for existing customers only if new peak demand grows faster than the costs a utility must incur to serve it. Several factors, including a utility’s degree of wholesale market exposure, forecast uncertainty and stranded-asset risk, and tariff design can determine the outcome of load growth on retail rates. Energy storage can make several affordability contributions in the face of this landscape of uncertainty and market volatility, including deferral of higher-cost grid investments through improved utilization of existing assets and flexibility of new large loads, insulation from volatile wholesale prices through peak shaving, and reliability support to address grid risks stemming from the behavior of AI data center loads. Different potential BESS deployment pathways—utility-scale front-of-the-meter systems, aggregated small-scale storage installations, and data center-sited behind-the-meter storage—are compared against each other and against conventional capacity alternatives. This framework is intended as a conceptual resource to utilities, particularly smaller public utilities with rural service territories, who may be considering the role that energy storage can play in insulating existing ratepayers from data center cost shifts.
This report serves as a user manual for the accompanying EMT model library developed by the National Laboratory of the Rockies (NLR) for various equipment in large data centers. The EMT model library enables detailed modeling of large data center loads for conducting grid stability studies. The EMT model library for data centers include detailed models of a 5.5 kW power supply unit (PSU), a 2.5 uninterruptible power supply (UPS), a 260 MW gas turbine-generator, a 500 kW motor load, and a 33 kW IT rack. These components represent all major equipment in data centers that need to be modeled for performing grid stability studies for data centers.
By recent estimates, data center energy demands are projected to consume between 6.7% and 12% of U.S. annual electricity generation by the year 2028, driven primarily by expanded demands from cloud services, big data analytics, and Artificial Intelligence (AI) (Shehabi et al., 2024). As much as 40% of data center total energy consumption are loads associated with the site infrastructure cooling systems, and these are often highly water consumptive (Aljbour et al., 2024). For energy system planners, this presents significant challenges to meeting and managing the anticipated loads, and especially the peak loads of projected data center deployments. Geothermal technologies offer two unique solutions to these challenges: 1) by serving loads through the deployment of new conventional and/or next-generation geothermal power technologies such as EGS and 2) through an often-overlooked opportunity to reduce data center peak cooling loads. The latter is the focus of this paper which explores Cold Underground Thermal Energy Storage ("Cold UTES") as an emerging industrial-scale geothermal cooling solution. This cooling solution is energy efficient, non-water-consumptive, and utilizes long duration energy storage (LDES) on both diurnal and seasonal time scales. Cold UTES has the potential to also function as a virtual power plant (VPP). The US Department of Energy's Geothermal Technologies Office is supporting R&D to understand the grid and system-wide value, costs, and impacts of deploying this emergent cooling solution at scale.
The Strategic Defense Initiative Organization has created data centers for midcourse, plumes, and backgrounds phenomenologies. The Backgrounds Data Center (BDC) has been designated as the prime archive for data collected by SDIO programs. The BDC maintains a Summary Catalog that contains 'metadata,' that is, information about data, such as when the data were obtained, what the spectral range of the data is, and what region of the Earth or sky was observed. Queries to this catalog result in a listing of all data sets (from all experiments in the Summary Catalog) that satisfy the specified criteria. Thus, the user can identify different experiments that made similar observations and order them from the BDC for analysis. On-site users can use the Science Analysis Facility (SAFE for this purpose. For some programs, the BDC maintains a Program Catalog, which can classify data in as many ways as desired (rather than just by position, time, and spectral range as in the Summary Catalog). For example, data sets could be tagged with such diverse parameters as solar illumination angle, signal level, or the value of a particular spectral ratio, as long as these quantities can be read from the digital record or calculated from it by the ingest program. All unclassified catalogs and unclassified data will be remotely accessible.
NASA National Space Science Data Center operation
This white paper evaluates the potential of a new energy system configuration: Strategic colocation of data centers - on airport property or adjacent to airports - in order to realize significant operational, economic, and environmental benefits to airports, data centers, and the communities they serve. While increasing demand from both airports and data centers requires both parties to make significant infrastructure investments, colocation can mitigate some of the energy, land use, permitting, and connectivity challenges of siting and powering these industries.
The global importance of data center water use is increasing with the rapid growth of digitalization and artificial intelligence. This study analyzes the factors influencing workload-level water use, measured in liters consumed per workload, to guide water-saving strategies in data centers. Our findings reveal workload-level water use variations exceeding 10,000-fold, driven by over 1000-fold differences in water consumption per kilowatt hour of server electricity consumed and approximately 10-fold differences in server workload efficiency. Key determinants are ranked as server efficiency, electrical grid water consumption factors, server utilization, cooling system type, infrastructure efficiency, climate zone, inactive server percentage, and server refresh cycle. Notably, there is no single recipe for minimizing water use; instead, optimal outcomes depend on tailored combinations of these factors. This analysis addresses critical knowledge gaps by identifying the determinants of data center water use and exploring their achievable minima under diverse site-specific constraints.
Nuclear energy is increasingly being considered for such targeted energy applications as data centers in light of their high capacity factors and low carbon emissions. This paper focuses on assessing the tradeoffs between economies of scale versus mass production to identify promising reactor sizes to meet data center demands. A framework is then built using the best cost estimates from the literature to identify ideal reactor power sizes for the needs of the given data center. Results should not be taken to be deterministic but highlight the variability of ideal reactor power output against the required demand. While certain advocates claim that with the gigawatts of clean, firm energy needed, large plants are ideal, others advocate for SMRs that can be deployed in large quantities and reap the benefits from learning effects. The findings of this study showcase that identifying the optimal size for a reactor is likely more nuanced and dependent on the application and its requirements. Overall, the study does show potential economic promise for coupling nuclear reactors to data centers and industrial heat applications under certain key conditions and assumptions.
Coprocessors, especially GPUs, will be a vital ingredient of data production workflows at the HL-LHC. At CMS, the GPU-as-a-service approach for production workflows is implemented by the SONIC project (Services for Optimized Network Inference on Coprocessors). SONIC provides a mechanism for outsourcing computationally demanding algorithms, such as neural network inference, to remote servers, where requests from multiple clients are intelligently distributed across multiple GPUs by a load-balancing service. This talk highlights the recent progress in deploying SONIC at selected U.S. CMS Tier-2 data centers. Using realistic CMS Run3 data processing workflows, such as those containing transformer-based algorithms, we demonstrate how SONIC is integrated into the production-like environment to enable accelerated inference offloading. We will present developments from both the client and server sides, including production job and data center configurations for NVIDIA and AMD GPUs. We will also present performance scaling benchmarks and discuss the challenges of operating SONIC in CMS production, such as server discovery, GPU saturation, fallback server logic, etc.
This report summarizes the outcomes of the Data Center Cohort under the Department of Energy’s Technical Assistance for Digital Assurance (TADA) initiative, aimed at enhancing grid resilience through cybersecurity, supply chain risk management (SCRM), and Cyber-Informed Engineering (CIE). The cohort engaged 17 organizations across utilities, data center operators, vendors, and technology providers in three sessions combining presentations, discussions, and exercises. Key topics included AI-driven load behavior, cybersecurity vulnerabilities in UPS/BESS and cooling systems, governance gaps at utility–data center boundaries, and supply chain integrity. Five cross-cutting themes emerged: interconnection architecture vulnerabilities, fragmented governance, AI-driven stability risks, lack of regulatory frameworks, and long-term supply chain concerns. Actionable recommendations were developed, including implementing DMZ segmentation, formalizing vendor access agreements, designing AI workload limits, and advancing standards through NERC and state-level programs. These strategies aim to strengthen resilience, clarify responsibilities, and ensure secure integration of data centers into the grid.
Liquid cooling is critical for thermal management in high-density data centers with the rising AI workloads. However, machine learning-based controllers are essential to unlock greater energy efficiency and reliability, promoting sustainability. We present LC-Opt, a Sustainable Liquid Cooling (LC) benchmark environment, for reinforcement learning (RL) control strategies in energy-efficient liquid cooling of high-performance computing (HPC) systems. Built on the baseline of a high-fidelity digital twin of Oak Ridge National Lab's Frontier Supercomputer cooling system, LC-Opt provides detailed Modelica-based end-to-end models spanning site-level cooling towers to data center cabinets and server blade groups. RL agents optimize critical thermal controls like liquid supply temperature, flow rate, and granular valve actuation at the IT cabinet level, as well as cooling tower (CT) setpoints through a Gymnasium interface, with dynamic changes in workloads. This environment creates a multi-objective real-time optimization challenge balancing local thermal regulation and global energy efficiency, and also supports additional components like a heat recovery unit (HRU). We benchmark centralized and decentralized multi-agent RL approaches, demonstrate policy distillation into decision and regression trees for interpretable control, and explore LLM-based methods that explain control actions in natural language through an agentic mesh architecture designed to foster user trust and simplify system management. LC-Opt democratizes access to detailed, customizable liquid cooling models, enabling the ML community, operators, and vendors to develop sustainable data center liquid cooling control solutions.
A general overview is given of the National Space Science Data Center (NSSDC) Standard Information Retrieval System. A description, in general terms, the information system that contains the data files and the software system that processes and manipulates the files maintained at the Data Center. Emphasis is placed on providing users with an overview of the capabilities and uses of the NSSDC Standard Information Retrieval System (SIRS). Examples given are taken from the files at the Data Center. Detailed information about NSSDC data files is documented in a set of File Users Guides, with one user's guide prepared for each file processed by SIRS. Detailed information about SIRS is presented in the SIRS Users Guide.
The rapid expansion of artificial intelligence (AI) has triggered an unprecedented surge in electricity demand, with US data center energy use projected to double or triple 2023 levels by 2028. This exponential growth places strain on grid infrastructure, which can hinder timely construction of desired computing capacity. To bridge this supply-demand gap, utilities and AI developers are increasingly turning to demand flexibility, a strategy that incentivizes shifting or reducing power use during peak periods of grid stress. Data centers are uniquely equipped for flexible operations due to their digital workloads, built-in redundancy, and onsite energy assets. This article outlines four primary mechanisms to enable data center flexibility: computational load flexibility (shifting tasks temporally or geographically), flexible use of core facility infrastructure adjustments, energy storage utilization, and onsite electricity generation. To encourage adoption, utilities are deploying new tariff designs, including voluntary interruptible service riders, mandated flexibility requirements, and streamlined interconnection processes for flexible loads. For the highly capitalized and rapidly growing AI industry, the primary motivators for embracing these strategies are expediting facility interconnection, satisfying emerging regulatory mandates, and mitigating community resistance. While demand flexibility cannot substitute the long-term need for new bulk power generation, it serves as an essential, immediate solution for enabling near-term deployment. By transforming data centers from grid stressors into stabilizing assets, flexible operations can ensure reliable grid integration, ease market pressures, and support a resilient power system.
This is a poster for the INL Data centers workshop. It shows how power lines were identified as good candidates for reconductoring with advanced conductors to better connect data centers sited on federal lands.
Data centers (DCs) serve as critical infrastructure for powering the growth and evolution of AI. Next-generation AI DCs present unique challenges in thermal management driven by unprecedented computational demands. This paper provides a comprehensive summary of key stakeholder perspectives on technology gaps, infrastructure requirements, test bed needs, emerging opportunities, and preliminary solutions related to thermal management for AI DCs. It establishes six strategic pillars of thermal management for next generation AI DC: reliability, deployability, efficiency, resilience, measurability, and valorization. The discussion spans a range of critical topics, including advanced cooling technologies, thermal strategies for emerging modular and edge DCs, system-level optimization and control frameworks, infrastructure planning and grid integration designs, benchmarking approaches, and pathways for waste heat recovery and reuse. The proposed research, development, and demonstration efforts are aimed at accelerating the deployment of AI DCs while ensuring energy efficiency, reliability, safety, and regulatory compliance.