Machine Learning Applications Enabling Fusion Energy: recent developments

Over the last few years, machine learning helped to develop advanced capabilities for fusion energy over a broad range of domains. This includes advanced algorithms to extract information from fusion diagnostics, enhanced algorithms for plasma state estimation and control, accelerated simulation tools to improve predictive capabilities, and expanded modeling capabilities for fusion materials design. This topical collection covers recent developments in machine learning applied research further enabling the path to fusion energy; in particular it covers a wide breadth of fusion subfields – from inertial confinement fusion, to magnetically confined plasma, including high temperature superconducting magnet optimization. This editorial summarizes the collection while also providing a critical outlook on how machine learning can be used in the future to accelerate the development of fusion energy as a reliable energy source.

近年来，机器学习已在诸多领域推动聚变能源相关先进技术能力的发展。其涵盖的研究内容包括：用于从聚变诊断（fusion diagnostics）中提取信息的先进算法、用于等离子体状态估计与控制的优化算法、用以提升预测性能的加速模拟工具，以及面向聚变材料设计的拓展建模能力。本专题合集收录了机器学习应用研究的最新进展，这些进展进一步为聚变能源的研发路径提供支撑；其覆盖范围涵盖聚变领域的众多子方向——从惯性约束聚变（inertial confinement fusion）到磁约束等离子体（magnetically confined plasma），其中还包括高温超导磁体（high temperature superconducting magnet）优化相关研究。本编者按在对本次专题合集进行梳理总结的同时，还就未来如何利用机器学习加速推动聚变能源成为可靠能源这一议题，给出了批判性展望。