South KoreaPublic sector2024 – ongoingKorea Institute of Fusion Energy
Machine-learning adaptive control of KSTAR plasma edge instabilities
Researchers used machine-learning adaptive control to suppress plasma edge bursts on 2 tokamaks, including KSTAR
Problem
High-performance fusion plasmas could produce damaging bursts at their edges. Conventional magnetic-field adjustments relied on manual tuning or preprogrammed responses.
Approach
Researchers combined machine learning with adaptive control on KSTAR and the US DIII-D tokamak. The controller adjusted magnetic perturbations in response to changing plasma conditions. The US Department of Energy reported suppression of edge bursts while maintaining high fusion performance on both devices.
Results
No published figures
Takeaway
The controller worked during experiments on different tokamaks. The result did not establish operation in a commercial fusion power plant.
Categories
- Industry
- Public sector
- Business function
- IT and R&D
- Innovation type
- Process innovation
- AI technology
- Predictive analytics and MLOptimization
- Adoption
- Partner co-development
- Goal
- Quality and risk
- Evidence
- Independent study
Sources
Source summaries
DOE described experiments with machine-learning adaptive control on KSTAR and DIII-D. Researchers achieved nearly edge-burst-free plasmas while maintaining high fusion performance.
Full details are in the original sources.