Development of surrogate-optimization techniques to accelerate transport predictions
Development of surrogate-optimization techniques to accelerate transport predictions
Simulation
Plasma Turbulence

Development of surrogate-optimization techniques to accelerate transport predictions

Explore our latest research projects and innovations in the field.

Principal Investigator
Ronald Ballinger
Professor Emeritus
and
Nuclear Science and Engineering
Team
Earl Marmar
Earl Marmar
Johan Frenje
Johan Frenje
01
Importance of research

In this project, we continue the development of the in-house, open-source PORTALS framework. By employing machine-learning toolsets, solutions to core turbulent transport coupled to the macroscopic plasma evolution can be found at much reduced cost than with standard techniques. Predictions of burning plasmas (SPARC, ITER) and current experiments (DIII-D, ASDEX Upgrade, C-Mod, JET) of unprecedented fidelity are performed with PORTALS.

Importance of Research

For a long time, predictions of tokamak core performance accounting for nonlinear turbulence physics were not possible, with only a couple of examples in the literature, and with limited fidelity. This has changed thanks to the introduction of machine learning techniques in PORTALS and the integration of high-performance codes like CGYRO in GPU systems. Now, we are capable of running simulations of core performance with self-consistently nonlinear turbulent transport routinely, which will help design better and more optimized burning plasma experiments and fusion pilot plants.

02
Methods

This project is heavily based on data science and numerical methods. We employ python-based, object-oriented, machine-learning-ready workflows that couple to high-performance simulation codes.

03
Project Outputs
Publications

P. Rodriguez-Fernandez, N.T. Howard, A. Saltzman, S. Kantamneni, J. Candy, C. Holland, M. Balandat, S. Ament and A.E. White, “Enhancing predictive capabilities in fusion burning plasmas through surrogate-based optimization in core transport solvers“, arXiv:2312.12610 [physics.plasm-ph] (2023). https://arxiv.org/abs/2312.12610

P. Rodriguez-Fernandez, N.T. Howard and J. Candy, “Nonlinear gyrokinetic predictions of SPARC burning plasma profiles enabled by surrogate modeling“, Nucl. Fusion 62, 076036 (2022). https://iopscience.iop.org/article/10.1088/1741-4326/ac64b2

04
Funding Acknowledgements

This project has been funded by several awards, including the Commonwealth Fusion Systems’ RPP020, the MFE Cooperative Agreement, and more recently the SMARTS (Surrogate Models for Accurate and Rapid Transport Solutions) SciDac award in collaboration with Dr. Christopher Holland (University of California San Diego). We also collaborate informally with the Adaptive Experimentation team at Meta.