First Nuclear Plasma Control with Digital Twin (Patreon)

[This is a transcript with links to references.]

Nuclear fusion is a great idea, in principle. In principle, it could solve the energy worries of the world beautifully. The problem is that whenever we’ve tried, getting nuclear fusion to work takes up more energy than it creates. But a team from Japan and the United States just got us a bit closer to our dream of clean energy. They recently succeeded in controlling nuclear plasma in a stellarator by creating a virtual twin. What’s a stellarator, what is digital twin and what did they actually do? Let’s have a look.

It used to be that there was basically only one approach to nuclear fusion, that was a tokamak. A Tokamak is a donut shaped device in which the hot nuclear fuel is held and heated with magnetic fields until the nuclei start fusing. The existing tokamaks are huge, for example, the one at ITER will be contained in a 70 meters high building, that’s about 20 floors. But in the past decade we have seen a lot of nuclear fusion startups that pursue different ideas.

Yes, most of them use tokamaks, but some of them use what’s called a stellarator, which is what they used for this new work. A stellarator is still a doughnut shape, but it has a more complicated magnetic field that allows the device to be smaller and should also make it harder for the plasma to escape thereby increasing efficiency. The biggest existing one is Wendelstein 7-x in Greifswald, Germany. The problem is, these stellarators are much more difficult to build than tokamaks, and they are also more difficult to model, so they haven’t been studied all that much.

Some companies use entirely different approaches, notably there’s what’s called inertial confinement. Inertial confinement basically means you shoot at a fuel pellet in the hope that it blows up, and then use the energy. You can shoot at it either with lasers, that’s what they do at the National Ignition Facility in the United States, or you can do it with a kind of bullet in what’s basically a big gun. That’s what they do for example at the startup First Light Fusion where they’ve called their device “Big Friendly Gun”. Yes, I can see it smiling, very friendly. And there are a few other startup ideas such as the z-pinch.

I covered all of this in a long episode last year.

But let’s come back to the new paper. The major issue with both standard tokamaks and stellarators  is that it’s difficult to control the plasma in these devices. Its motion is chaotic, which makes it a pain to steer. And if the plasma gets out of control it can damage the reactor vessel.  Because of this, when it looks like the plasma is becoming unstable, the process of heating and fusing must be quickly aborted. This is the biggest problem for making this approach to nuclear fusion energy efficient.

But this is an area where computing power can make a big difference through a method that’s called chaos control.

For example, already in 2019, a group of researchers from Harvard and Princeton  trained an artificially intelligent system on data from the Joint European Torus, that’s currently the largest tokamak in the world, and another tokamak, that’s currently the biggest in the United States.

They taught it to recognize data-patterns that signal an impending plasma instability. This was a hindsight analysis on data that had previously been collected, but they correctly identified an imminent instability one second ahead in somewhat more than 80 percent of cases; 30 milliseconds ahead, they saw almost all instabilities coming.

You then want to do this in real time, and indeed this was done last year by researchers from deepmind. They actively controlled plasma a tokamak device called TCV located at the Swiss Plasma Center in Lausanne. It’s  afairly small reactor with a size of just about two meters in each direction. In it, the plasma is held by strong magnetic fields that can be manipulated with a number of controllers To make this work, they first trained their artificial intelligence on a tokamak simulator, that has also been developed by the group in Lausanne. So: Take the AI, train it to control another software to save time, then take the trained AI to control the real thing.

And their AI control of the plasma worked out beautifully. In this movie on the left you see the measurement of the actual plasma inside the tokamak. On the right you see the reconstructed shape of the plasma. The DeepMind people were able to coax the plasma into a large number of different shapes including a triangular one, and two separate droplets.

This then brings me to the new paper which did a similar thing to what Deepmind did last year but with a new method on a different device. This was done by a team of researchers from Japan and the United States at the Large Helical Device in Japan. This is a stellarator, the second largest in the world after Wendelstein-X.

They used a new method for controlling fusion plasma by using a digital twin, that is a virtual replica of the plasma created on a computer, though I think “Digital Twin” is just a buzzword for model. One then uses real-time observations to update the twin and try to predict what happens next to control the plasma in return. The process is called data assimilation and it has more physics built in than the AI based approached used by deepmind.

In particular they used what’s known as ensemble forecasting in which you assume some amount of uncertainty for the exact current state and make a prediction for a set of initial values. Then you calculate the most likely one. It’s the same method that’s also used for making a weather forecast. In this experiment they controlled the temperature in the centre of the plasma and you can see here that their control method nicely reached the target they aimed at.

However, neither this nor the previous experiment by DeepMind  actually controlled an ongoing fusion process, they just steered the plasma without fusion. Clearly, actively controlling the fusion process will be next, and I’ll let you know when that happens, so stay tuned. You don’t want to miss the world revolution, do you.