2024年能源人工智能报告.pdf

April 2024ADVANCED RESEARCH DIRECTIONS ONAI FOR ENERGYReport on Winter 2023 WorkshopsClaus Daniel Argonne National LaboratoryJess C.Gehin Idaho National LaboratoryKirsten Laurin-Kovitz Argonne National LaboratoryBryan Morreale National Energy Technology LaboratoryRick Stevens Argonne National LaboratoryWilliam Tumas National Renewable Energy LaboratoryANL-23/69 AI FOR ENERGY i Advanced Research Directions on AI for Energy Report on the U.S.Department of Energy(DOE)Winter 2023 Workshop Series on Artificial Intelligence(AI)for Energy Program Committee Claus Daniel Associate Laboratory Director,Argonne National Laboratory Kirsten Laurin-Kovitz Associate Laboratory Director,Argonne National Laboratory Rick Stevens Associate Laboratory Director,Argonne National Laboratory U.S.Department of Energy Ceren Susut-Bennett Program Manager,U.S.Department of Energy Keith Benes Senior Fellow,U.S.Department of Energy Kenneth Ham Technology Manager,U.S.Department of Energy Tassos Golnas Technology Manager,U.S.Department of Energy Mike C.Robinson Senior Technology Advisor,U.S.Department of Energy Key Contributors Argonne National Laboratory Mihai Anitescu,Alec Poczatek,Andrew Siegel,Sibendu Som,Richard Vilim Brookhaven National Laboratories Meng Yue Idaho National Laboratory Ahmad Al Rashdan,Christopher Ritter Lawrence Berkeley National Laboratory Mary Ann Piette,Tianzhen Hong Lawrence Livermore National Laboratory John Grosh,Brian Van Essen Los Alamos National Laboratory Hari Viswanathan National Renewable Energy Laboratory Ray Grout,Benjamin Kroposki National Energy Technology Laboratory Kelly Rose Oak Ridge National Laboratory Prasanna Balaprakash,Prashant Jain,Teja Kuruganti Pacific Northwest National Laboratory Court Corley,Robert Rallo Sandia National Laboratories Matt Reno Editorial Frank Alexander Director AI Research and Strategic Development,Argonne National Laboratory Emily M.Dietrich Strategic Program Communications Lead,Argonne National Laboratory Special Thanks To the Argonne National Laboratory Communications and Public Affairs Divisions Writing Center of Excellence,including key support from Andrea Manning and Lorenza Salinas AI FOR ENERGY ii CONTENTS Executive Summary.1 Introduction.3 Key Findings for Establishing the Cross-cutting Aspects of AI Supremacy Needed to Ensure Success in Energy Mission Areas.5 High-Consequence.5 Urgency.6 Complexity.7 01.Nuclear Energy.8 1.1 Grand Challenges.8 1.2 Advances in the Next Decade.11 1.3 Accelerating Development.13 1.4 Expected Outcomes.15 1.5 References.15 02.Power Grid.18 2.1 Grand Challenges.18 2.2 Advances in the Next Decade.19 2.3 Accelerating Development.21 2.4 Expected Outcomes.24 2.5 References.25 03.Carbon Management.26 3.1 Grand Challenges.26 3.2 Advances in the Next Decade.29 3.3 Accelerating Development.31 3.4 Expected Outcomes.32 3.5 References.32 04.Energy Storage.34 4.1 Grand Challenges.34 4.2 Advances in the Next Decade.36 4.3 Accelerating Development.39 4.4 Expected Outcomes.42 4.5 References.42 05.Energy Materials.44 5.1 Grand Challenges.44 5.2 Advances in the Next Decade.46 5.3 Accelerating Development.47 5.4 Expected Outcomes.49 5.5 References.49 Appendixes.51 AA.Agenda.52 AB.Workshop Participants.54 AC.Acronyms and Abbreviations.56 AD.References by Chapter.58 EXECUTIVE SUMMARY AI FOR ENERGY 1 EXECUTIVE SUMMARY Artificial intelligence(AI)provides a transformational opportunity to rapidly deploy new clean energy,secure critical grid energy assets from threat actors,and reduce capital and operational costs of next-generation energy technologies and the connected systems that embody the demand side of the transformation.The United States will need to invest trillions of dollars in energy infrastructure to reach the nations clean,resilient goals by 2050.At the Department of Energy(DOE)national laboratories,AI has incredible potential across nuclear,renewable,and carbon management domains due to the ability to represent unprecedented system model sizes,provide intense computational resources,and capture knowledge from a workforce of the nations top scientists.In aggregate,AI could reduce the cost to design,license,deploy,operate,and maintain energy infrastructure by hundreds of billions of dollars if the following applied energy challenges are realized.AI provides a breakthrough opportunity to accelerate the design,deployment,and licensing of new energy capacity.Commercial powerplant design and licensing are a multi-year effort that can account for up to 50%of time to market for new energy deployments.DOE estimates the onboarding of 1.6 TW of new solar capacity and 200 GW of new nuclear capacity,while enabling hydrogen,geothermal,critical minerals,and other clean energy resources by 2050,with a cost that could approach trillions of dollars in national investment to meet growing global clean energy demand.Additionally,DOE estimates the need to reduce costs to less than$100/net metric ton of CO2 equivalent for both carbon capture and storage to address carbon pollution.AI has the potential to reduce schedules by approximately 20%across new clean energy designs,with potential savings in the hundreds of billions of dollars by 2050.Additionally,AI can augment and extend the energy development workforce that will be in high demand.The energy grids generation capabilities and demand-side needs are experiencing rapid changes in requirements for secure,reliable,and resilient planning and operations controls.The increasing volumes of communications,controls,data,and information are growing the digital landscape,increasing flexibility and improving the reliability and agility of the grid by increasing visibility to operators and consumers.Integrating energy systems together across grid operations could save billions of dollars annually by automatically optimizing generation and demand-side needs.Autonomous operation technologies can provide monitoring,control,and maintenance automation across various clean energy technologies.Distributed,consumer-sited technologies are changing the power load with electric vehicles(EVs),distributed storage,smart buildings,and appliances adding new intelligence to loads while also requiring the integration of consumer-sited controllability.Furthermore,new advanced nuclear technologies,such as microreactors,will likely need to operate autonomously to realize economies of scale.Delivering AI capabilities across the operations and maintenance lifecycle can transform safety,efficiency,and innovation within national energy production and distribution infrastructure.The siting of new energy capacity is a complex challenge balancing energy generation options,community needs,environmental factors,and resiliency considerations.AI could aid community energy planning based on a comprehensive dataset and a trained community energy foundation model that captures characteristics of and interactions between physical infrastructure,human behavior,and climate/weather impacts.AI tools can achieve national clean energy goals by democratizing community-level clean energy resources and facilitating the identification of energy transition pathways that reflect local objectives,demographics,and legacy infrastructure.Natural disasters and human-caused events are occurring more frequently and with more intensity,delivering significant impacts to the nation.Adverse weather events are increasingly disrupting supply chains,damaging property and assets,and making certain areas less habitable.The U.S.experienced a record 28 unique weather/climate disasters that cost at least$1 billion in 2023.Climate change,urbanization,population growth,aging infrastructure,and deferred maintenance increase risks to communities and human survival.An AI-based,all-hazards global response system that has ingested global and EXEMPLAR GRAND CHALLENGES FROM THE CHAPTERS OF THE AI FOR ENERGY REPORT 01 Nuclear Energy:Accelerating the Licensing and Regulatory Process 02 Power Grid:Building Cyber-and All-Hazards Resilient and Secure Energy Systems 03 Carbon Management:Realizing A Virtual Subsurface Earth Model 04 Energy Storage:Equitable and Accessible Deployment 05 Energy Materials:Advancing Beyond Material Properties and Performance to Achieve Lifecycle-Aware Materials Design EXECUTIVE SUMMARY AI FOR ENERGY 2 stakeholder datasets,facilitating international preparation,response,and recovery,can enhance preparedness and resilience solutions and inform faster recovery.Science-based models enhanced with AI multi-modeling approaches can improve predictions of subsurface properties and systems to improve resource discovery for domestic critical materials,geothermal reservoirs,uranium,and water opportunities.This capability could create a national subsurface AI and data testbed to enable responsible commercial,regulatory,and science-based discovery and development.AI can improve the forecasting and prediction of subsurface properties and systems,informing and transforming our ability to reduce risks and responsibly interact with the subsurface.Energy material innovation is key to realizing national clean energy goals.Increasing automation in materials laboratories,such as autonomous laboratories,can transform the design and discovery of new materials.AI can also accelerate materials qualification through automation of materials testing,leading to new energy technologies such as advanced nuclear reactors and new battery certifications.In addition to these cross-cutting opportunities,there are unique use cases in nuclear,renewable,and carbon management energy systems.For example,while emissions,prediction,measurement,and mitigation are uniquely important to carbon management,the underlying computational infrastructure could be shared across grand challenges.Unattended operation of nuclear reactors has unique life-safety considerations;however,many plant-level digital twins of piping,valve,heat exchanger,and cooling towers could be shared across applied energy domains.A DOE consortium model from all energy domains,integrated with expertise from subject-matter experts from the laboratories,could help ensure and drive efficiency across research challenges.To accomplish these grand challenges,key developments are needed.The laboratories must establish a leadership computing ecosystem to train and host data and foundation models at ever-increasing scales.Fine-tuned models need to be developed for each domain that are coupled,where possible,with ground-truth,first-principles physics.Although the laboratories have hundreds of petabytes worth of data,only small amounts of these data are cataloged,warehoused,and ready for AI model ingestion.Curation of one-of-a-kind,ground-truth data coupled with energy industry data will be essential to building models at these scales.Most important,partnerships across laboratories,government,industry,and academia are essential to realizing the transformational benefits of AI for energy.This AI for Energy report further details grand challenges that provide significant opportunities for energy applications across nuclear energy,the power grid,carbon management,energy storage,and energy materials over the next decade.The main conclusions and opportunities from this study are available in the Key Findings section of this report.INTRODUCTION AI FOR ENERGY 3 INTRODUCTION An important aspect of the U.S.Department of Energys(DOE)mission is to ensure the nations energy independence and security both in the short and long term.Key to meeting this challenge are continued advancements in artificial intelligence(AI),especially in the context of energy.As an initial step toward addressing these challenges,a group of about 100 experts on AI/machine learning(ML)and applied energy convened at Argonne National Laboratory in December 2023 over the course of two days to map out future needs related to utilizing AI.The goal of the meeting was to detail pressing technical challenges and propose AI-assisted solutions.Five domain areas were identified(detailed below),along with potential paths forward.DOE is ideally positioned to address challenges associated with energy independence and security due to its unique set of assets.These assets include a highly skilled workforce with relevant domain expertise(nuclear engineering,chemistry,materials science,networked systems,etc.),and an array of world-leading experimental facilities for making advances in materials,chemistry,etc.These include synchrotron light sources,nanocenters,high-performance computing resources,and autonomous laboratories.By integrating these resources with other AI capabilities outlined in the previous AI for Science,Energy,and Security(AI4SES)report,the DOE can leverage AI to stay at the forefront of the rapidly evolving landscape.The applied energy focus described in this report centers on five areas vital to the energy future of the U.S.,as well as underscores the critical role that AI can play in shaping our worldhighlighting the urgency and importance of being leaders in AI to ensure impactful solutions to global energy needs.These areas include Nuclear Power,Power Grid,Carbon Management,Energy Storage,and Energy Materials.It will be essential to integrate these together and with other efforts in AI for science and technology.Complexity,the large-scale effort involved,real-time decision making required,robustness of systems,and safety implications all pose extra challenges.The grand challenges described in this report span multiple disciplines and have not been solved by conventional methods.The power of AI for solving such problems lies in its capacity to simultaneously handle multiple system characteristics while incorporating both data and specific domain(e.g.,physics,chemistry,etc.)models and to do so on a scale and at a complexity otherwise not possible.Nuclear energy plays a pivotal role in the clean energy landscape of the U.S.,representing about half of its clean electricity generation.To achieve its full potential,the nuclear industry must adopt and,where required,advance the latest AI tools and technologies.AIs transformative potential is particularly relevant in methodologies which could drastically improve the economics of nuclear system design and operation.These challenges span multiple scientific and engineering disciplines and require AIs unique ability to process vast amounts of data and integrate physics models on a scale previously unattainable.This integration must be carried out in a seamless manner.AI can facilitate this coordination,potentially reducing costs significantly compared to traditional nuclear energy development and deployment approaches.Recent Generation III reactor commissionings have experienced notable delays and cost overruns,often due to premature construction starts.AI,developed under science and technology initiatives,can mitigate such issues by enhancing design completion and process efficiency.The intricate interdependencies within the nuclear energy sector pose challenges well