AI helps contain Fusion Plasma

 

"Power of the sun in the palm of my hand"

Many may remember this line from Toby Maguire's Spiderman 2 uttered by Dr. Otto Octavius. Well, the technology of the power of the sun is not far off in real-life. Usable fusion power has been dreamed of for decades and we get closer to it every year.

Practical fusion reactors of today apply heat to atoms to generate fusion plasma. This plasma, when heated to the requisite temperatures (hundreds of millions of degrees), begin to cause atoms to fuse and release large amounts of energy. One day, we hope, the amount of energy used to run the reactor will be surmounted by the energy released, thus providing a clean, unlimited, eco-friendly power source, one to match the sun itself.

In order to contain the fusion plasma (after all, earthly materials are not going to be capable of withstanding the heat), magnetic containment is necessary. To help control the delicate process of confinement of ultra-hot plasma, AI techniques are now being used.

Google's London-based Deep Mind and EPFL in Switzerland are experimenting with a single controller to maintain effective control over a plasma's vertical and radial positions as well as its current. The AI technique being used is termed "Reinforcement Learning". Reinforcement Learning is, broadly, the process of training an agent (or controller) to either reach an objective or maintain an objective by feeding appropriate reward and punishment for actions taken by the agent in a simulated environment. The trained controller can then be used in the actual real-life environment.

In the case of our fusion plasma controller, the algorithmic “agent” receives measured and target values of the plasma parameters – a long list, including numerous spatial dimensions, electrical currents and magnetic fluxes within the fusion chamber – as its inputs. After processing these values it issues outputs corresponding to the voltage levels on each of the magnets. Once the magnets are adjusted and the new feedback from sensors is received the cycle starts again – being repeated some 10,000 times a second.

The researchers say that their new AI-based scheme could improve reactor performance, with its open-ended nature perhaps allowing power output to be maximised. More broadly, they say, the technology might lead to new reactor designs by allowing the joint optimisation of several device parameters – including plasma shape, wall design and heat load.

To read the research article on this advancement, please go here.