Implementation of matrix compression in the coupling of JOREK to realistic 3D conducting wall structures

TL;DR

This paper presents a method to compress response matrices in JOREK simulations of fusion plasmas using SVD, enabling more efficient modeling of complex 3D conducting wall structures like ITER.

Contribution

The study introduces the application of SVD-based matrix compression to improve memory efficiency in JOREK's coupling with detailed 3D conducting wall models.

Findings

Successful application of SVD compression to response matrices

Enhanced memory efficiency in plasma instability simulations

Potential for simulating large-scale fusion devices like ITER

Abstract

JOREK is an advanced non-linear simulation code for studying MHD instabilities in magnetically confined fusion plasmas and their control and/or mitigation. A free-boundary and resistive wall extension was introduced via coupling to the STARWALL and CARIDDI codes, both able to provide dense response matrices describing the electromagnetic interactions between plasma and conducting structures. For detailed CAD representations of the conducting structures and high resolutions for the plasma region, memory and computing time limitations restrict the possibility of simulating the ITER tokamak. In the present work, the Singular Value Decomposition provided by routines from the ScaLAPACK library has been successfully applied to compress some of the dense response matrices and thus optimize memory usage. This is demonstrated for simulations of Tearing Mode and Vertical Displacement Event instabilities. An outlook to future applications on large production cases and further extensions of the method are discussed.