English 
量子科学技術研究開発機構と日本電信電話株式会社は、大型核融合装置のプラズマ閉じ込め磁場に適用するAI予測手法を確立しました。
QSTとNTTは2020年に連携協定を締結し、世界に先駆けた革新的な環境エネルギー技術の創出を目指して共同研究を進めてきました。
This time, they applied a method called Mixture of Experts (MoE), which weights and integrates the optimal AI model according to the changing situation, to establish a technology to predict plasma with high accuracy, and evaluated the actual plasma confinement magnetic field of the world’s largest tokamak type superconducting plasma experimental device JT-60SA. As a result, for the first time in the world, they succeeded in reproducing the position and shape of plasma that depend on the magnetic field structure with the accuracy required for actual plasma control. Conventional analytical reconstruction methods based on the laws of physics could in principle control the position and shape of the plasma boundary (periphery) which changes sequentially. However, this method has made it possible to control multiple control quantities in real time, including the distribution of current and pressure inside the plasma, which is important for avoiding plasma instability, something that was not possible with conventional methods.
こちらもお読みください： JTOWERとメトロウェザー、世界最大のドップラーライダー精度検証センターを開設
この成果は、今後JT-60SAで行われる加熱実験において、高温プラズマのリアルタイム制御に挑戦する上で有用であり、ETERや原型炉のように、より大きなプラズマを少数の計測器で制御する核融合炉のプラズマ予測制御につながる画期的な成果です。この結果を受けて QST そして NTT は、2020年に締結した協力協定をさらに延長することに合意し、核融合エネルギーの早期実用化に向けて引き続き協力していきます。
Tokamak reactors, which are the most advanced in the world toward the prototype reactor of fusion energy, an innovative form of environmental energy, require a continuous flow of electric current in the plasma because the confinement magnetic field is formed by the current flowing in the plasma itself. However, instabilities may occur due to the current and pressure. To utilize a stable prototype reactor, it is important to predict these instabilities before they occur and control them appropriately, and the challenge was to establish a method to reconstruct the plasma confinement magnetic field required for control in real time with high accuracy from measurement signals. As one solution, this joint research has developed a method to evaluate the plasma confinement magnetic field with high accuracy using AI technology that is good at optimization problems, with the aim of utilizing it for the control of the world’s largest tokamak-type superconducting plasma experimental device JT-60SA.
ソース NTT
日本語