Abstract Details
Abstracts
Author: Christopher J Hansen
Requested Type: Poster
Submitted: 2024-04-12 16:11:36
Co-authors: S. Joung, A. O. Nelson, S. Madrieddy, A. Samaddar, D. Smith, F. Ebrahimi
Contact Info:
Columbia University
500 W 120th St
New York, New York 10027
United States
Abstract Text:
Commercial fusion reactors based on the tokamak are expected to require operation with high-performance while avoiding the ELMs that typify standard H-mode operation in present machines. Several such ELM-free regimes are currently under investigation for this purpose, including: RMP-suppression of ELMs, QH-mode and related naturally ELM-free H-mode variants, and negative triangularity shaping. While experimentally accessible in current devices, their operational boundaries (where ELMs reappear) are generally not well understood due to complex nonlinear mechanisms, such as saturation and hysteresis. The SciDAC-5 Center for Edge of Tokamak Optimization (CETOP) is focused on improving our ability to predict ELM-free regimes from design-level information (eg. equilibria and modeling) to enable optimization of future reactor designs for these states. As part of this work Machine Learning (ML) will be applied in two areas: 1) The development of ML models from experimental data to aid in the validation of high-fidelity simulations of ELMy and ELM-free discharges. Application to prediction of detailed single and multi-ELM dynamics and time-delay between ELMs are being explored to enable validation of long-time dynamics with finite-time simulations. 2) The development of ML models for prediction of ELMy or ELM-free operation from design-level information to enable optimization of future device designs. These models will be trained on a hybrid database of simulations and experimental data with additional ML/AI techniques applied to guide the expansion of the database through additional high-fidelity simulations or experiments in an optimized way. Progress and plans on this work will be presented. This research program is supported by US Department of Energy SciDAC program under award DE-AC02-09CH11466.
Comments:
Please group with other CETOP posters.