Nuclear fusion is a promising energy source for carbon neutrality, with recent experiments demonstrating success in producing more energy than injected. Tokamaks are key for fusion research, achieving milestones like sustaining plasma at high temperatures and breaking fusion energy records. However, challenges like plasma disruption remain, addressed with AI for prediction and mitigation. An AI controller using deep reinforcement learning optimizes plasma pressure while avoiding tearing instability in tokamaks by controlling actuators. Experimental results show successful tearing avoidance with adaptive and active AI control, emphasizing the importance of optimal threshold values for stable plasma operation.
https://www.nature.com/articles/s41586-024-07024-9