Google’s AI has learned to control the plasma inside a nuclear fusion reactor, a very hot gas riddled with ions and electrons trapped by a magnetic field inside a doughnut-shaped bowl called a tokamak. The result that opens up new scenarios for exploiting this future source of clean energy is published In nature from the Federal Institute of Technology in Lausanne (Epfl) in collaboration with the British company DeepMind from Google.
One of the biggest difficulties in carrying out nuclear fusion is the need to confine the plasma to the tokamak, preventing it from contacting its walls and degrading it. To prevent this inconvenience, researchers at the Swiss Plasma Center at EPFL (one of the few centers in the world with a tokamak in operation) are using to experiment with control system configurations on a simulator, developed in more than 20 years of research and constantly updating.
“However – explains Federico Felici, researcher at SPC and co-author of the study – very long computations are still required to determine the correct value for each control system variable. This is where our research project began. In partnership with DeepMind.” In collaboration with Google experts, researchers have developed an algorithm that can control magnetic coils to produce and maintain a variety of plasma configurations. The system has also been tested live on SPC tokamak to assess its performance in the real world.
The algorithm controlled magnetic fields to create not only traditional (elongated) plasma shapes, but also those that resembled triangles and snowflakes. It was also able to simultaneously keep separate plasmas in suspension. According to the study authors, this new approach could help improve the design and management of future fusion reactors, maximizing their performance.
Reproduction is reserved © Copyright ANSA
“Explorer. Devoted travel specialist. Web expert. Organizer. Social media geek. Coffee enthusiast. Extreme troublemaker. Food trailblazer. Total bacon buff.”