Machine learning application in complicated burning plasmas for future magnetic fusion exploration

Abstract

On October 24, 2021, the Chinese government unveiled a guiding document about state-level work to achieve carbon peaking and carbon neutrality goals under a new development philosophy, laying out key targets and measures for upcoming decades. The documment, titled “Working Guidance For Carbon Dioxide Peaking and Carbon Neutrality In Fully And Faithful Implementation Of The New Development Philosophy.” stated that China's carbon dioxide emissions per unit GDP would have dropped by 18% from the level in 2020 by 2025 and by more than 68% from the level in 2005 by 2030. Fusion energy, as an environment-friendly and intrinsically safe energy source with an abundant fuel supply, is considered an effective low-carbon ultimate solution. Fusion research has been conducted for 60 years, and has been demonstrated that magnetic confinement controlled fusion is very effective magnetic confinement fusion research has progressed significantly over the years. Generally, it is believed in the magnetic confinement fusion plasma community that the next pivotal step in magnetic confinement fusion plasma research is to create burning plasma, where the dominant method of heating plasmas is alpha particle self-sustaining heating. The research aims to understand the underlying physics of the confinement, heating, and instability of burning plasma and to explore the technologies associated with the power-producing fusion reactor. In future burning plasmas, multiple mode-number instabilities are very important but complicated physics issues because of the employment of strong power auxiliary heating schemes. Compared with the time-consuming large-scale computer simulation, the machine learning can bring an advantage for the investigations concerning multiple mode-number instabilities, especially in predictions compared to the time-consuming large-scale computer simulation.