Nuclear fusion has become a very valuable ally to resolve the extremely complex stabilization of plasma

  • 6

nuclear fusion keep adding hits on its way to commercial exploitation. There is still a lot of work to be done to bring electricity generated in a fusion reactor to our homes, and researchers and engineers working in this area are the first to admit it, but it is unquestionable that during the last twenty years this technology has been developing at a frenetic pace.

Relevant innovations follow one another almost constantly, but this pace of development is not the result of chance; is the result of joint effort that some of the main research institutions on the planet are doing, as well as the countries with the greatest technical and scientific development.

The itinerary set by EUROfusion, which is the international consortium responsible for setting up ITER, foresees that if everything goes its way, commercial nuclear fusion will come in the 60’s. Reaching this all-important milestone, however, requires a number of challenges, one of the most complex being how to precisely control the plasma to stabilize it and sustain the reaction over time.

Deep learning, an essential tool in nuclear fusion science

ITER, the experimental fusion reactor that an international consortium is building in the French town of Cadarache, is a nuclear fusion reactor using magnetic confinement of the type tokamak. Its purpose is to demonstrate feasibility and profitability of this form of energy generation, and its strategy to obtain it consists of fusing the nuclei of deuterium and tritium that make up a gas at 150 million degrees Celsius.

Reaching that temperature is essential because in this way the ionized nuclei of deuterium and tritium acquire the necessary kinetic energy to overcome their natural electrical repulsion and merge. However, dealing with a plasma at such a high temperature is very complex.

An industry in the hands of TSMC and Asian factories: the map of world chip production

In magnetic confinement reactors, a sophisticated system of very high power magnets is responsible for confining this ionized fuel at a very high temperature to prevent it from coming into contact with the walls of the vacuum chamber. If it did, it would degrade them, and it would be impossible to sustain the fusion reaction any longer. The problem is that the plasma is subject to turbulence that originate naturally and are especially intense in the outermost layer of this gas.

It is essential to understand how plasma behaves at 150 million degrees Celsius in order to stabilize it

Understanding how the plasma behaves when nuclear fusion processes are initiated is essential on the way to a solution that allows us to control it precisely, and therefore stabilize it. A few weeks ago, a group of researchers from MIT developed a turbulence model capable of predicting its behavior, and, curiously, an essential piece of this mathematical model is deep learning. However, these are not the only scientists who have opted for this strategy.

And it is that another group of researchers has just published in Nature a very interesting article in which he describes in great detail a complex magnetic control system that seeks, precisely, to control the plasma with unprecedented precision, according to his tests. It is not necessary for us to delve into the more complex details of his proposal, but we are interested to know that his approach is to monitor in real time plasma fluctuations to act on it very quickly and give it the optimal shape, position and current.

modelarchitecturenuclear fusion

Source: Nature

Actually, this strategy is not new. The technicians involved in the development of nuclear fusion by magnetic confinement have been working on it for many years. What is new is the way these researchers propose to solve this challenge. In the image that we publish above these lines we can see in detail the architecture of the magnetic control system that they have designed to act in real time on the plasma at a very high temperature.

What makes this magnetic control system different is that it has been designed using reinforcement learning algorithms.

It is evident that the physical elements of a magnetic control system like this are very important; however, and here lies the originality of this solution, what makes it different from other control architectures is that it has been designed using reinforcement learning algorithms. This area of ​​machine learning is used in many other technical and scientific disciplines to solve optimization problems, but this is the first time it has been used to design the magnetic control system of a nuclear fusion reactor.

Time crystals: what they are, why they are so revolutionary and how quantum computers are already helping us achieve them

The researchers who have approached this approach assure in their article that their design significantly reduces the complexity of the magnetic control system. However, the most promising thing is that it has been tested, apparently with great success, in TCV (Tokamak with variable configuration), which is a reactor tokamak of nuclear fusion for research housed in the Federal Polytechnic School of Lausanne (Switzerland).

What makes this magnetic confinement reactor so attractive to researchers is that it allows the study of plasmas with different geometric configurations. This is important because the form of this gas profoundly conditions the performance of the fusion reactor. And, according to these researchers, its magnetic control system has managed to monitor the behavior of the plasma, act on its shape in real time and successfully stabilize it. It sounds really good.

His conclusions are very exciting: ‘The magnetic control of a reactor tokamak it is one of the most complex scenarios in which reinforcement learning has been used. This is a new and very promising way to address the design of magnetic control systems, and has the potential to accelerate nuclear fusion science and contribute to the development of future reactors tokamak‘. It seems reasonable to accept that this technology could prove valuable when high-power tests with deuterium and tritium start at ITER. Let’s cross our fingers for it to be so.

Cover image | ITER

More information | Nature

nuclear fusion keep adding hits on its way to commercial exploitation. There is still a lot of work to be…

nuclear fusion keep adding hits on its way to commercial exploitation. There is still a lot of work to be…

Leave a Reply

Your email address will not be published.