DeepMind’s AI can control superheated plasma inside a fusion reactor 

In nuclear fusion, the atoms of a hydrogen atom are forced together to form heavier atoms, such as helium. This produces a lot of energy compared to a small amount of fuel, which makes it a very efficient source of energy. It is cleaner and safer than fossil fuels or conventional nuclear power, created by fission – separating the nuclei. It is also the process that powers the stars.

However, nuclear fusion on Earth is difficult to control. The problem is that the atomic nuclei repel each other. This can only be done by breaking them together inside the reactor at extremely high temperatures, which often reach millions of degrees – which is warmer than the center of the sun. At this temperature, matter is neither solid, liquid nor gas. It enters the fourth stage, known as plasma: particle rolling, superheated broth.

The function is to hold the plasma inside the reactor together long enough to extract energy from it. Inside the stars, plasma is held together by gravity. On Earth, researchers use a variety of tricks, including lasers and magnets. In a magnet-based reactor known as a talkmak, the plasma gets trapped inside the electromagnetic cage, forcing it to hold its shape and preventing it from touching the reactor walls, which cools the plasma and damages the reactor.

Controlling the plasma requires constant monitoring and manipulation of the magnetic field. The team trained its reinforcement-learning algorithm to do this within the simulation. Once he learned how to control and change the shape of the plasma inside the virtual reactor, the researchers gave him control of the magnet in the experimental reactor, the Variable Configuration TalkMock (TCV) in Lausanne. They found that AI was able to control the actual reactor without any additional fine-tuning. Altogether, the AI ​​controlled the plasma for only two seconds પરંતુ but this is long enough for the TCV reactor to run before it gets too hot.

Quick reactions

Ten thousand times per second, the trained neural network takes 90 different measurements describing the shape and position of the plasma and adjusts the voltage in 19 magnets in response. This response loop is much faster than the previous reinforcement-learning algorithm encountered. To speed things up, AI was split into two neural networks. A huge network, called Critic, learned through trial and error how to control a reactor within a simulation. The critic’s ability was then encoded into a small, fast network, called an actor, which runs on reactors.

“It’s a very powerful method,” says Jonathan Citrin of the Dutch Institute for Fundamental Energy Research, who was not involved. “It’s an important first step in a very exciting direction.”

Researchers believe that using AI to control plasma will make it easier to experiment with different conditions inside the reactor, help them understand the process and potentially accelerate the development of commercial nuclear fusion. AI also learned how to control plasma by adjusting magnets in a way humans had not previously tried, suggesting that there may be new reactor configurations to explore.

“We can take risks with this kind of control system that we would not dare to take otherwise,” said Ambrogio Fasoli, director of the Swiss Plasma Center and chairman of the Eurofusion Consortium. Human operators are often unwilling to push plasma beyond certain limits. “There are incidents that we should avoid altogether because it damages the device,” he says. “If we are sure we have a control system that takes us closer to the limit but not beyond that, we can explore more possibilities. We can speed up research.”

Similar Posts

Leave a Reply

Your email address will not be published.