•

Avoiding fusion plasma tearing instability with deep reinforcement learning

Avoiding fusion plasma tearing instability with deep reinforcement learning - Nature

https://www.nature.com/articles/s41586-024-07024-9

Artificial intelligence control is used to avoid the emergence of disruptive tearing instabilities in the magnetically confined fusion plasma in the DIII-D tokamak reactor.

Avoiding fusion plasma tearing instability with deep reinforcement learning - Nature

1 comments

jarfil

•

As some sources have titled it: "AI fixes fusion!" /s

On a more serious note, I find it interesting that a 25ms control loop is enough, and that:

the tearing prediction model could forecast the instability 300 ms before the disruption

Good model.