April 4-6

Abstract Details

files Add files

Abstracts

Author: Benjamin J Faber
Requested Type: Poster
Submitted: 2022-03-04 16:05:54

Co-authors: J.M. Duff, A. Bader, C.C. Hegna

Contact Info:
University of Wisconsin-Madison
2221 Sherman Ave, Apt 401
Madison, WI   53704
United States

Abstract Text:
Numerous recent advances in the theory and computation of plasma confinement properties in 3D magnetic fields has reestablished the stellarator as a competitive fusion concept. These advances share a common thread, where the confining external magnetic field is manipulated to generate new stellarator equilibria with better confinement properties. Performing this procedure within a numerical optimization loop to efficiently find new configurations presents significant challenges. The numerical codes used in stellarator optimization are in general not written to couple easily together, due to obstacles such as differing programming languages and resource requirements, separate input and output formats, or incompatible magnetic field representations. To alleviate these issues, a new suite of stellarator physics and optimization packages have been written in the dynamic, high-performance Julia programming language, centered around the StellaratorOptimization.jl package. These packages provide common interfaces between different magnetic coordinates systems, equilibrium solvers, and physics analysis codes that allows for users to easily add new methods and procedures to optimization problems without needing to understand the underlying optimization internals. Importantly, by leveraging these interfaces, methods for computing physics targets can be implemented independently of the equilibrium solver. Furthermore, by using Julia package manager, different optimization algorithms can be explored seamlessly without any need for pre-compilation. The first results from StellaratorOptimization.jl will be presented demonstrating the dynamic use of both local and global MHD equilibrium solvers with different combinations of optimization algorithms and physics targets of varying complexity, such as quasisymmetry, collisionless fast particle confinement, and turbulent confinement.

Comments: