EFIT-Prime: Probabilistic and physics-constrained reduced-order neural network model for equilibrium reconstruction in DIII-D
Physics of Plasmas · 01.09.2024 · 31 (9)
DOI: 10.1063/5.0213609 ↗Kurz gefasst
Fachartikel in der Zeitschrift Physics of Plasmas, online erschienen am 06.09.2024. Verfasst von S. Madireddy und 10 weiteren Autor:innen. Die Arbeit wurde bisher 7-mal zitiert (Crossref) und nennt 45 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Fusion Energy Sciences, Argonne National Laboratory.
Abstract (Originalsprache)
We introduce EFIT-Prime, a novel machine learning surrogate model for EFIT (Equilibrium FIT) that integrates probabilistic and physics-informed methodologies to overcome typical limitations associated with deterministic and ad hoc neural network architectures. EFIT-Prime utilizes a neural architecture search-based deep ensemble for robust uncertainty quantification, providing scalable and efficient neural architectures that comprehensively quantify both data and model uncertainties. Physically informed by the Grad–Shafranov equation, EFIT-Prime applies a constraint on the current density Jtor and a smoothness constraint on the first derivative of the poloidal flux, ensuring physically plausible solutions. Furthermore, the spatial location of the diagnostics is explicitly incorporated in the inputs to account for their spatial correlation. Extensive evaluations demonstrate EFIT-Prime's accuracy and robustness across diverse scenarios, most notably showing good generalization on negative-triangularity discharges that were excluded from training. Timing studies indicate an ensemble inference time of 15 ms for predicting a new equilibrium, offering the possibility of plasma control in real-time, if the model is optimized for speed.
Förderung
- Fusion Energy Sciences
- Argonne National Laboratory
Zitieren
APA
Madireddy, S., Akçay, C., Kruger, S. E., Amara, T. B., Sun, X., McClenaghan, J., Koo, J., Samaddar, A., Liu, Y., Balaprakash, P., & Lao, L. L. (2024). EFIT-Prime: Probabilistic and physics-constrained reduced-order neural network model for equilibrium reconstruction in DIII-D. Physics of Plasmas, 31(9). https://doi.org/10.1063/5.0213609
Vancouver
Madireddy S, Akçay C, Kruger SE, Amara TB, Sun X, McClenaghan J, et al. EFIT-Prime: Probabilistic and physics-constrained reduced-order neural network model for equilibrium reconstruction in DIII-D. Physics of Plasmas. 2024;31(9). doi:10.1063/5.0213609
BibTeX
@article{Madireddy2024,
author = {Madireddy, S. and Akçay, C. and Kruger, S. E. and Amara, T. Bechtel and Sun, X. and McClenaghan, J. and Koo, J. and Samaddar, A. and Liu, Y. and Balaprakash, P. and Lao, L. L.},
title = {EFIT-Prime: Probabilistic and physics-constrained reduced-order neural network model for equilibrium reconstruction in DIII-D},
journal = {Physics of Plasmas},
year = {2024},
volume = {31},
number = {9},
doi = {10.1063/5.0213609}
}
EUVIMED