Researchers from EPFL’s Swiss Plasma Center (SPC) and the British firm DeepMind have developed an artificial intelligence (AI) -based algorithm that can control plasma – a state of matter at temperatures of millions of degrees – in a nuclear fusion reactor. They applied it to a real-world plasma for the first time in the SPC’s tokamak research facility, TCV.
The team hopes that the discovery will allow maximizing the production of electricity in still hypothetical power plants of the future based on the principle of nuclear fusion.
The SPC is a donut-shaped chamber that uses a powerful magnetic field to confine plasma at extremely high temperatures – hundreds of millions of degrees Celsius, even hotter than the sun’s core – so that nuclear fusion can occur between hydrogen atoms. The energy released from fusion is being studied for use in generating electricity.
The SPC’s tokamak allows for a variety of plasma configurations, hence its name: variable-configuration tokamak (TCV). Researchers here are continuously investigating new approaches for confining and controlling plasmas so that it doesn’t crash into the vessel walls and deteriorate.
Researchers at the SPC first test their control systems configurations on a simulator before using them in the TCV tokamak. “Our simulator is based on more than 20 years of research and is updated continuously,” says Federico Felici, an SPC scientist and co-author of the study. “But even so, lengthy calculations are still needed to determine the right value for each variable in the control system. That’s where our joint research project with DeepMind comes in.”
DeepMind developed a new AI algorithm that can create and maintain specific plasma configurations and trained it on the SPC’s simulator by having it attempt many different control strategies. Based on the collected experience, the algorithm was able to calculate control strategies for producing requested plasma configurations. Then the algorithm was called on to work the other way, identifying the right settings to produce a specific plasma configuration.
After being trained, the research team tested their new AI-based system directly on the tokamak to see how it would perform under real-world conditions. The system was able to create and control a wide range of plasma shapes, including elongated, conventional shapes, as well as advanced configurations, such as negative triangularity and ‘snowflake’ configurations.
“Our team’s mission is to research a new generation of AI systems – closed-loop controllers – that can learn in complex dynamic environments completely from scratch. Controlling a fusion plasma in the real world offers fantastic, albeit extremely challenging and complex, opportunities,” says Felici. Martin Riedmiller, control team lead at DeepMind and co-author of the study.
A week ago, engineers at the UK Atomic Energy Authority (UKAEA) announced that they produced the record-breaking 59 megajoules of sustained fusion energy in the form of neutrons during a five-second phase of a plasma discharge. This is double the previous energy record set in 1997. Nuclear fusion aims to reproduce what happens in the heart of the sun and is considered by its supporters as the energy of tomorrow.