Controller with Integrated Machine Learning Tweaks Fusion Plasmas in Real Time

The conditions for high fusion performance in fusion devices can result in damaging energy bursts called edge-localized modes (ELMs). ELMs can be stabilized through small adjustments to the magnetic confinement field, but this approach is usually limited to manual, preprogrammed responses. In this research, scientists integrated machine learning with adaptive control to achieve real-time adjustment capable of responding to the dynamic conditions of a fusion plasma in the DIII-D National Fusion Facility and Korea Institute of Fusion Energy KSTAR tokamaks.

Understanding the Outsized Effect of Hydrogen Isotopes

Creating a fusion plasma requires deep understanding of the behavior of various isotopes of hydrogen. But plasma scientists have long been puzzled by a mysterious contradiction– the disconnect between theoretical predictions and experimental observations of how fusion energy confinement varies with the mass of hydrogen isotopes used to fuel the plasma. A new analysis has helped unravel this mystery.