University of Florida Advances Fusion Plasma Prediction with AI on HiPerGator

University of Florida researchers are working to improve predictions of plasma behavior in tokamaks for fusion energy using artificial intelligence on the HiPerGator supercomputer. Two projects funded by the National Science Foundation and the U.S. Department of Energy National Nuclear Security Administration support the effort. Christopher McDevitt, a plasma physicist and professor in the university Nuclear Engineering program, focuses on understanding predicting and preventing off normal plasma behaviors. Fusion research seeks to recreate processes that power the sun by fusing atoms at temperatures in excess of 100 million degrees Celsius. This promises an abundant carbon free source of energy. The plasma is an ionized gas held in magnetic fields inside a tokamak reactor. Inside the reactor the plasma extreme heat and magnetic confinement cause fuel nuclei to collide and fuse releasing energy absorbed as heat in the vessel walls. This heat can produce steam and electricity through turbines and generators similar to conventional power plants. Plasma instability poses risks according to McDevitt. Sudden loss of control at those temperatures could damage reactor materials and structure. Unintentional generation of energetic electrons could turn the fusion device into a particle accelerator. Researchers have turned to machine learning to simulate conditions inside the reactor and predict plasma anomalies without risking damage. McDevitt group uses HiPerGator to develop machine learning surrogates of complex plasma events. The recently upgraded supercomputer allows simulations that used to take days to be completed in a few minutes. The work is attributed to Harrem Monkhorst of the University of Florida.