Learning plasma dynamics and robust rampdown trajectories with predict-first experiments at TCV | Nature Communications

The authors report on the implementation of a data-efficient machine learning approach to predict plasma dynamics. This enables offline design of robust trajectories to terminate the plasma without disruptive instabilities. Experimental results at the TCV tokamak show statistically significant improvements in key figures of merit and the ability to a priori predict the dynamics of key plasma properties.

Download PDF Subjects Magnetically confined plasmas Nuclear fusion and fission Abstract The rampdown phase of a tokamak pulse is difficult to simulate and often exacerbates multiple plasma instabilities. To reduce the risk of disrupting operations, we leverage advances in Scientific Machine Learning (SciML) to combine physics with data-driven models, developing a neural state-space model (NSSM) that predicts plasma dynamics during Tokamak à Configuration Variable (TCV) rampdowns. The NSSM efficiently learns dynamics from a modest dataset of 311 pulses with only five pulses in a reactor-relevan

related reading

• Accelerating fusion science through learned plasma control — Google DeepMinddeepmind.com
• Trending Papers - Hugging Facepaperswithcode.com
• Science Needs AI Data Stocktakesaipolicyperspectives.com
• Is Frontier Asynchronous RL Solved? — Luke J. Huangluk-huang.github.io
• Competing with sampling — Alignment Research Centeralignment.org
• Researchbactra.org
• Plasma Focus Radiative Model: Review of the Lee Model Code | Journal of Fusion Energy | Springer Nature Linklink.springer.com
• ModLaNets: Learning Generalisable Dynamics via Modularity and Physical Inductive Biasproceedings.mlr.press
• Neural Data Transformerarxiv.org
• [1805.12114] Deep Reinforcement Learning in a Handful of Trials using Probabilistic Dynamics Modelsar5iv.labs.arxiv.org
• Training agents to plan in latent space — a technical overview | by Lukas Bierling | Mediummedium.com
• Trainloop AItrainloop.ai