School of Science,China University of Geosciences(Beijing);College of Physics and Optoelectronic Engineering,Shenzhen University;Shandong Computer Science Center(National Supercomputer Center in Jinan);Shanghai WoOrange Information Technology Co.,Ltd.;Institute of Nuclear and New Energy Technology,Tsinghua University;
[Significance] The rapid development of big data and artificial intelligence(AI) technologies has greatly advanced research onmagnetic confinement fusion(MCF), which is a promising approach to achieving sustainable fusion energy. Because of the complexity ofplasma dynamics, including nonlinear and high-dimensional processes, conventional methods for simulation, monitoring, and control havefaced significant limitations in accuracy and computational efficiency. By integrating AI and big data, researchers can now address manyof these challenges, leading to significant improvements in the precision and speed of simulations as well as real-time analysis ofexperimental data. The application of these technologies has become essential for advancing fusion research, particularly in facilitating thecontrol and optimization of plasma confinement, a key factor in achieving sustained nuclear fusion reactions. [Progress] The application ofAI in MCF has led to numerous advances across different areas. In plasma simulation, AI-based methods, such as machine learning-drivensurrogate models, have significantly reduced computational costs while maintaining or even enhancing accuracy. These models enable therapid prediction of plasma behavior in response to varying conditions, which is crucial for optimizing experimental parameters andensuring the stability of plasma confinement. The capability of AI to manage large-scale, high-dimensional data has proven particularlybeneficial for multiscale simulations that involve complex interactions between physical processes. In experimental monitoring and control,AI, combined with big data analytics, has enabled real-time processing of sensor data from fusion devices. Through predictive modelingand adaptive control mechanisms, AI algorithms can detect potential anomalies and make autonomous adjustments to operationalparameters, thereby improving the reliability and safety of fusion experiments. The dynamic nature of plasma requires precise andimmediate responses to fluctuations, and the capability of AI to analyze past experimental data and predict future behavior has enabledmore effective management of plasma instabilities, thereby enhancing the overall system robustness and contributing to the optimization ofplasma performance. AI has also been instrumental in the design and optimization of fusion devices. By employing AI to model magneticfield configurations and predict material performance under extreme conditions, researchers have been able to improve the durability andefficiency of critical reactor components. These advancements include optimizing the design of superconducting magnets andplasma-facing materials, both of which are essential for the long-term operation of fusion reactors. AI-driven optimization has resulted inimproved magnetic confinement configurations, ensuring better plasma stability and enhanced confinement performance, which arenecessary for achieving continuous fusion reactions. Furthermore, AI facilitates interdisciplinary collaboration by integrating data fromdiverse fields, such as plasma physics, materials science, and computational modeling. The use of AI in cross-disciplinary research fostersinnovation and accelerates progress in addressing key challenges in fusion research. Moreover, AI has contributed to the development ofintelligent educational platforms and virtual experimentation environments, enabling researchers and students to gain hands-on experiencethrough simulations and virtual experiments. These platforms are crucial for advancing knowledge and skills in plasma physics and fusiontechnology and help cultivate the next generation of fusion researchers. [Conclusions and Prospects] The future of MCF research will beincreasingly shaped by the integration of AI and big data technologies. The capability of AI to enhance simulation accuracy, optimizeexperimental design, and improve real-time control systems will play a central role in overcoming existing technical barriers in fusionresearch. Furthermore, AI-driven materials research will contribute to the discovery and design of new materials capable of withstandingthe harsh conditions inside fusion reactors, thus ensuring longer operational lifespans and increased reactor efficiency. As AI technologiescontinue to evolve, they are expected to play a more significant role in all levels of fusion research, from experimental planning toreal-time plasma control and material optimization. These advancements will not only accelerate progress toward realizing practical fusionenergy but also contribute to the development of novel technologies that support the broader scientific community. In addition, AI-powerededucational platforms will continue to provide researchers with advanced tools for learning and experimentation, helping them bridge thegap between theory and practical application. The continued development of AI and big data in this field holds great promise for thesuccessful realization of MCF as a viable energy source for the future.
| 496 | 0 | 3 |
| Downloads | Citas | Reads |
[1] WANG Z X, LIU T, WEI L. Nonlinear evolution and control of neo-classical tearing mode in reversed magnetic shear Tokamak plasmas[J]. Reviews of Modern Plasma Physics, 2022, 6(1):14.
[2] CAI H S, LI D. Recent progress in the interaction between energetic particles and tearing modes[J]. National Science Review, 2022, 9(11):019.
[3] ZHANG X X, CAI H S, WANG Z X. Influence of deeply trapped energetic ions on tearing modes[J]. Physics of Plasmas,2019, 26(6):062505.
[4] ZHANG X X, GAO B F, CAI H S, et al. Excitation of 2/1fishbone-like modes by trapped energetic ions[J]. Plasma Physics and Controlled Fusion, 2020, 62(8):085010.
[5] QIU Z Y, CHEN L, ZONCA F, et al. Nonlinear excitation of a geodesic acoustic mode by toroidal Alfvén eigenmodes and the impact on plasma performance[J]. Nuclear Fusion, 2019, 59(6):066031.
[6] YANG Z. Real-time disruption prediction in the plasma control system of HL-2A based on deep learning[J]. Fusion Engineering and Design, 2022, 182:113223.
[7] MENEGHINI O. Neural-network accelerated coupled core-pedestal simulations with self-consistent transport of impurities and compatible with ITER IMAS[J]. Nuclear Fusion,2021, 61(2):026006.
[8] CHURCHILL R M, BOYER M D, COWLEY S C. Accelerating fusion energy with AI[M]//ALOK C, GEOFFREY F, TONY H.Artificial intelligence for science:A deep learning revolution.Singapore:World Scientific Publishing Company, 2023:271–284.
[9] SHOUSHA R. Machine learning-based real-time kinetic profile reconstruction in DIII-D[J]. Nuclear Fusion, 2023, 64(2):026006.
[10] ASAHI Y, LATU G, INA T, et al. Optimization of fusion kernels on accelerators with indirect or strided memory access patterns[J]. IEEE Transactions on Parallel and Distributed Systems, 2017, 28(7):1974–1988.
[11] CHEW J, GIBBON P, BROMMEL D, et al. First principles simulation of early stage plasma initiation process in ITERscale tokamak[J]. Nuclear Fusion, 2023, 64(1):016003.
[12] DONG G, WEI X, BAO J, et al. Deep learning based surrogate models for first-principles global simulations of fusion plasmas[J]. Nuclear Fusion, 2021, 61(12):126061.
[13] KIM S K. Highest fusion performance without harmful edge energy bursts in Tokamak[J]. Nature Communications, 2024,15(1):3990.
[14] SPINOSA A G, BUSCARINO A, FORTUNA L, et al. Datadriven order reduction in Hammerstein-Wiener models of plasma dynamics[J]. Engineering Applications of Artificial Intelligence, 2021, 100:104180.
[15] FROIO A, BONIFETTO R, CARLI S. Design and optimization of artificial neural networks for the modelling of superconducting magnets operation in Tokamak fusion reactors[J]. Journal of Computational Physics, 2016, 321:476–491.
[16] LAO L. Application of machine learning and artificial intelligence to extend EFIT equilibrium reconstruction[J]. Plasma Physics and Controlled Fusion, 2022, 64(7):074001.
[17] KRUGER S. Overview and progress of the EFIT-AI project[C]//American Physical Society. 64th Annual Meeting of the APS Division of Plasma Physics, 2022:11.
[18] TOMASSETTI G, MARZI DE G, FIAMOZZI ZIGNANI C. A methodological approach for the optimal design of the toroidal field coils of a Tokamak device using artificial intelligence[J].Superconductor Science&Technology, 2022, 35(1):014002.
[19] GUO B H, CHEN D L, SHEN B, et al. Disruption prediction on EAST Tokamak using a deep learning algorithm[J]. Plasma Physics and Controlled Fusion, 2021, 63:115007.
[20] HUMPHREYS D KUPRESANIN A, BOYER M D, et al.Advancing fusion with machine learning research needs workshop report[J]. Journal of Fusion Energy, 2020, 39(4):123–155.
[21] CHURCHILL R M, TOBIAS B, ZHU Y, et al. Deep convolutional neural networks for multi-scale time-series classification and application to Tokamak disruption prediction using raw, high temporal resolution diagnostic data[J]. Physics of Plasmas, 2020, 27(6):062510.
[22] HUMBIRD K D, PETERSON J L, SALMONSON J, et al.Cognitive simulation models for inertial confinement fusion:Combining simulation and experimental data[J]. Physics of Plasmas, 2021, 28(4):042709.
[23] PUSZTAI I EKMARK I, BERGSTRM H, et al. Bayesian optimization of massive material injection for disruption mitigation in Tokamaks[J]. Journal of Plasma Physics, 2023,89(2):905890204.
[24] J?RVINEN A E, FULOP T, HIRVIJOKI T, et al. Bayesian approach for validation of runaway electron simulations[J].Journal of Plasma Physics, 2022, 88(6):905880612.
[25] CHEN W, ZHU X L, WANG F, et al. Resonant interaction of tearing modes with energetic-ions resulting in fishbone activities on HL-2A[J]. Nuclear Fusion, 2019, 59(9):096037.
[26] CAI H S. Frequency chirping of neoclassical tearing modes by energetic ions[J]. Nuclear Fusion, 2021, 61(12):126012.
[27] PAJARES A, THOME K E, SCHUSTER E, et al. Nonlinear,real-time optimization for actuator management in tokamaks[J].Fusion Engineering and Design, 2024, 198:114085.
[28] VENTURINI C. Machine learning approaches for plasma state monitoring in Tokamaks[Z/OL].[2024-09-10]. https://thesis.unipd.it/handle/20.500.12608/36025.
[29] MOHAPATRA D, SUBUDHI B, DANIEL R. Real-time sensor fault detection in Tokamak using different machine learning algorithms[J]. Fusion Engineering and Design, 2020, 151:111401.
[30] WIESEN S, DASBACH, KIT A, et al. Data-driven models in fusion exhaust:AI methods and perspectives[J]. Nuclear Fusion,2024, 64(8):086046.
[31] TANG W, FEIBUSH E, DONG G, et al. AI-machine learningenabled Tokamak digital twin[Z/OL].(2024-09-04)[2024-09-10].https://doi.org/10.48550/arXiv.2409.03112.
[32] STARON C, LE BORGNE H, MITTEAU R, et al. Detection of thermal events by semi-supervised learning for Tokamak first wall safety[J]. IEEE Transactions on Instrumentation and Measurement, 2024:5012409.
[33] ABBATE J. AI-based prediction and control of Tokamaks:Combining simulations and experimental data[Z/OL].[2024-09-10].http://arks.princeton.edu/ark:/88435/dsp0147429d506.
[34] ZHENG W, XUE F M, SHEN C S, et al. Overview of machine learning applications in fusion plasma experiments on J-TEXT Tokamak[J]. Plasma Science and Technology, 2022, 24(12):124003.
[35] SEO J, KIM S, JALALVAND A, et al. Avoiding fusion plasma tearing instability with deep reinforcement learning[J]. Nature,2024, 626(8000):746–751.
[36] DEGRAVE J, FELICI F, BUCHLI J, et al. Magnetic control of Tokamak plasmas through deep reinforcement learning[J].Nature, 2022, 602:414–419.
[37] MORGAN D, PILANIA G, COUET A, et al. Machine learning in nuclear materials research[J]. Current Opinion in Solid State and Materials Science, 2022, 26(2):100975.
[38] ZHU X L, LI H, ZHAO Y F, et al. Machine learning application in complicated burning plasmas for future magnetic fusion exploration[J]. iEnergy, 2022, 1(3):272–276.
[39] KWON J M, LEE K S, WOO M,et al. Development of a virtual Tokamak platform[J]. Fusion Engineering and Design, 2022,184:113281.
Basic Information:
DOI：10.16791/j.cnki.sjg.2025.04.001
China Classification Code:TL631;TP311.13;TP18
Citation Information:
[1]张小西,段思哲,高宝峰等.大数据与AI技术在磁约束聚变领域的应用与展望[J].实验技术与管理,2025,42(04):1-13.DOI:10.16791/j.cnki.sjg.2025.04.001.
Fund Information:
国家自然科学基金青年科学基金(12405261,12405263); 三束材料改性教育部重点实验室开放基金(KF2502); 齐鲁工业大学(山东省科学院)人才科研项目(2023RCKY140)