Abstract Details
Abstracts
Author: Orso Meneghini
Requested Type: Consider for Invited
Submitted: 2024-03-21 23:51:32
Co-authors: T. Slendebroek, B.C. Lyons, J. Mcclenaghan, L. Stagner, J. Harvey, T.F. Neiser, A. Ghiozzi, G. Dose, J. Guterl, T. Cote, N. Shi, D. Weisberg, S.P. Smith, B.A. Grierson
Contact Info:
General Atomics
3550 General Atomics Court
San Diego, San Diego 92121-1
United States of Ame
Abstract Text:
FUSE [Meneghini, IAEA 2023] is a framework that integrates first-principle, machine-learning, and reduced models to enable comprehensive high-fidelity self-consistent simulations for fusion power plant design. FUSE was developed from the ground up exclusively in the Julia programming language, with a focus on computational efficiency and scalability. By leveraging Julia's performance, incorporating machine learning strategically, and developing solvers with innovative formulations (eg. for equilibrium and transport), FUSE achieves a remarkable 1000-fold improvement in speed over previous integrated modeling frameworks.
A multi-objective optimization approach is used to efficiently explore designs that meet the stringent requirements of fusion energy generation but also adhere to economic and engineering constraints. This process by its very nature can seamlessly capture the tradeoffs of different requirements, leading stakeholders to make a more informed decision, and to designs that enjoy broader acceptance and support.
FUSE can run both stationary and time-dependent simulations. Dynamics are efficiently modeled by coupling physics models that take into consideration the separation of time and spatial scales that naturally occur in a fusion plasma, as well as in the plant thermal conversion system. The time-dependent and multi-objective optimization capabilities of FUSE are exploited to optimize the time-traces of different actuators to achieve a desired state (ie. trajectory-optimization).
At its core FUSE leverages the ITER IMAS data structure standard to facilitate effective data exchange among its various models. This not only enables the framework to be modular, allowing for the execution of models of varying fidelity across a broad range of domains, but also ensures its direct applicability to ITER, as well as numerous other fusion experiments that have adopted the IMAS standard.
Work supported by General Atomics corporate funding.
Comments: