Since 18 of December 2019 conferences.iaea.org uses Nucleus credentials. Visit our help pages for information on how to Register and Sign-in using Nucleus.

3–6 Sept 2024
ITER Headquarters
Europe/Vienna timezone

Machine Learning model for real-time SPARC vertical stability observers

4 Sept 2024, 09:25
25m
Council Room (ITER Headquarters)

Council Room

ITER Headquarters

Contributed Oral Prediction and Avoidance Prediction & Avoidance

Speaker

Arunav Kumar

Description

Author: A. Kumar,
Co-authors: C. Clauser, T. Golfinopoulos, D.T. Garnier, J. Wai, D. Boyer, A. Saperstein, R. Granetz, C. Rea

Given the demanding requirements of the SPARC high-field tokamak (B_{0}=12.2 T) and its operation with high elongated plasma (\kappa_{sep}=1.97), robust real-time-compatible vertical stability observers are paramount. In this work, we present the fast surrogate-based modeling approach for observers (such as energy functional, VDE n=0 growth rate, stability margins ​​(m_s), inductive stability margins (m_i), max-Z, and frequency components), integrating advanced 2D electro-mechanical circuit and dynamic plasma response models [1]. These surrogate observers employ transformer-based machine learning techniques; trained to replicate and predict the results of the filamentary semi-rigid body MEQ-RZIp and deformable free boundary MEQ-FGE (and its linearized version FGElin) code suite, as detailed in Carpanese et al. [2]. The training dataset incorporates simulated SPARC primary reference discharge scenarios and the Alcator C-Mod (hot VDEs) 2012-2016 disruption warning database. To enhance robustness, these observers will also be trained over a range of simulated L- and H-mode SPARC plasma scenarios, including periods without and with ELM triggering (via artificial vertical kicks [2]). We will report on the assessment of observers’ sensitivity to the underlying RZIp & FGE models and their proximity to stability boundaries; thereby supporting disruption prediction and avoidance.

Acknowledgements
Work funded by Commonwealth Fusion Systems under grant number RPP2023.

References
1. M. L. Walker & D. A. Humphreys (2006), 50:4,473-489, DOI: 10.13182/FST06-A1271
2. Carpanese et al. 2020 EPFL PhD thesis no. 7914
3. Sartori F. et al Proc. 35th EPS Conf. on Plasma Physics (Hersonissos, Greece, 9–13 June 2008) vol 32D P5.045

Speaker's title Mr
Speaker's email address arunavk@psfc.mit.edu
Speaker's Affiliation MIT-PSFC, Cambridge, MA, USA
Member State or IGO United States of America

Primary author

Co-authors

Dr Cesar Clauser (Massachusetts Institute of Technology) Theodore Golfinopoulos (Plasma Science and Fusion Center, Massachusetts Institute of Technology) Darren Garnier (MIT Plasma Science and Fusion Center) Dr J Wai (Commonwealth Fusion Systen) Dan Boyer (Commonwealth Fusion Systems) Alex Saperstein (MIT - PSFC) Robert Granetz (MIT) Cristina Rea (Massachusetts Institute of Technology)

Presentation materials