AI is set to aid the development of nuclear fusion reactors in a big way by predicting when major disruptions could halt reactions and damage the reactor.
While some claim that a stable nuclear fusion reactor capable of producing near-limitless, clean energy could be just over a decade away, researchers are still very much in the experimental stage of development. The biggest obstacle to achieving commercial energy production is controlling the intense, highly unstable plasma within the reactor with attempts so far only lasting a matter of minutes.
However, a team at the US Department of Energy’s Princeton Plasma Physics Laboratory (PPPL) has found a way to use the power of deep learning artificial intelligence (AI) to predict any disruptions that halt fusion reactions and damage doughnut-shaped tokamak reactors.
Image: © Korn V./Stock.adobe.com
“This research opens a promising new chapter in the effort to bring unlimited energy to Earth,” said Steve Cowley, director of PPPL, about the study published to Nature. “AI is exploding across the sciences and now it’s beginning to contribute to the worldwide quest for fusion power.”
Crucial to this new deep learning algorithm – called the Fusion Recurrent Neural Network (FRNN) – has been its access to 2TB of data provided by two major fusion facilities: the DIII-D National Fusion Facility in California and the Joint European Torus (JET) in the UK.
These facilities the largest in the US and the world, respectively, and the PPPL trained the AI system using these vast databases to reliably predict disruptions on other tokamaks.
‘Fusion science is very exciting’
Speaking of the importance of deep learning to the project, the team said it can achieve what other forms of AI can’t within nuclear fusion development.
For example, while non-deep learning software might consider the temperature of a plasma at a single point in time, the FRNN considers profiles of the temperature developing in time and space.
So far, FRNN is able to predict true disruptions within the 30 millisecond warning time frame required by the International Thermonuclear Experimental Reactor (ITER), an international nuclear fusion megaproject underway in France. However, it is closing in on the additional requirement of 95pc correct predictions with fewer than 3pc false alarms.
Bill Tang, co-author of the research and a principal investigator at PPPL, said: “AI is the most intriguing area of scientific growth right now, and to marry it to fusion science is very exciting.
“We’ve accelerated the ability to predict with high accuracy the most dangerous challenge to clean fusion energy.”
The next step in the research will be to move from prediction to the control of disruptions, but this will be quite the challenge.
“We will combine deep learning with basic, first-principle physics on high-performance computers to zero in on realistic control mechanisms in burning plasmas,” Tang said. “By control, one means knowing which ‘knobs to turn’ on a tokamak to change conditions to prevent disruptions. That’s in our sights and it’s where we are heading.”