If we can master the physics inside the reactor, we will be able to achieve nuclear fusion and deliver a sustainable source of clean energy.
Scientists have been taking steps towards this goal for decades. A new approach shows how we can control the unstable and super-heated plasma in the reactor.
Scientists at the Swiss Plasma Center and DeepMind used a deep reinforcement learning system to study the nuances of plasma behavior and control inside a fusion tokamak.
It is not an easy balancing act, as the coils require a huge amount of subtle voltage adjustments, up to thousands of times per second, to successfully keep the plasma confined within magnetic fields.
A 3D model of a vacuum vessel. DeepMind/SPC/EPFL.
Nuclear fusion reactions involve keeping the plasma stable at hundreds of millions of degrees Celsius, hotter than the core of the Sun.
Researchers show in a new study that a single system can handle the task on its own.
The team explains in a DeepMind post that they used a learning architecture that combines deepRL and a simulation environment to create controllers that can both keep the plasma steady and be used to accurately sculpt it into different shapes.
The researchers trained their machine learning system in a tokamak simulation to find out how to navigate the complex magnetic confinement of plasma.
After its training window, the artificial intelligence moved to the next level, applying what it had learned in the simulation to the real world.
There is a visualization of controlled shapes. DeepMind/SPC/EPFL.
By controlling the variable configuration tokamak, the RL system sculpted plasma into a range of different shapes inside the reactor, including one that had never before been seen in the TCV.
In addition to conventional shapes, the artificial intelligence could create advanced configurations, such as negative triangularity and snowflake configurations.
If we can maintain nuclear fusion reactions, there are different kinds of potential for harvesting energy in the future. One of the configurations controlled by the system here could hold promise for future study by the International thermonuclear Experimental Reactor (ITER), the world's largest nuclear fusion reactor.
The magnetic mastery of these formations is one of the most challenging real-world systems to which reinforcement learning has been applied, and could establish a radical new direction in how real-world tokamaks are designed.
Some believe that what we are seeing here will change the future of fusion control systems.
The only way forward is through the use of an artificial intelligence, according to physicist Gianluca Sarri from Queen's University Belfast.
A small change in one of the variables can cause a big change in the final output. It is a very lengthy process if you try to do it manually.
Nature reported the findings.
All Rights Reserved © 2024
