Researchtransformerstokamaksurrogate modelsreal time control
Transformer Predicts Tokamak Global Plasma Parameters
6.1
Chenguang Wan et al. present a transformer-based model (submitted Feb 22, 2026) that predicts six global plasma parameters on the WEST tokamak using only pre-discharge signals. Trained on 550 discharges, it achieves average MSE 0.026, R^2 0.94 and inference times around 0.1 seconds. The results suggest data-driven surrogates can aid discharge planning, scenario evaluation, and real-time control of tokamak plasmas.
Key Points
- 1Develops transformer model predicting six global plasma parameters using pre-discharge signals.
- 2Demonstrates high accuracy (average MSE 0.026, R^2 0.94) across 550 WEST discharges.
- 3Enables fast inference (~0.1s), supporting rapid scenario design and potential real-time control.
Scoring Rationale
Solid, directly usable surrogate model with high accuracy, limited by single-venue arXiv preprint and niche tokamak scope.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
