Real-time data-driven disruption prediction and its mitigation of MA-plasma experiments in KSTAR with a lower carbon divertor

Abstract

Real-time disruption prediction and its mitigation of mega-ampere (MA) plasma experiments have been performed in the 2022 KSTAR carbon lower divertor condition. Disruption of around 1 MA plasma current has severely affected the in-vessel components, especially through their high magnetic energy. In this study, we developed and implemented a data-driven real-time disruption detection model based on a neural network in the plasma control system. Additionally, to deal with the specific plasma instability only observed in few 1.2 MA plasma shots, an empirical threshold of plasma vertical position error was also used as a disruption precursor. A deuterium–neon mixture gas was utilized to mitigate disruption. Our integrated disruption prediction and mitigation system was applied to two types of sessions in the 2022 KSTAR campaign: high-density and MA plasma development experiments. The disruption prediction and mitigation experiments were successfully conducted, and a plasma termination methodology triggered by the disruption alarm was also studied to minimize potential harmful effects to tokamak structures.

Keywords

• KSTAR
• disruption prediction
• disruption mitigation
• mega-ampere plasma current