Speaker
Onnie On-Ying Luk
Description
See the full Abstract at http://ocs.ciemat.es/EPS2018ABS/pdf/P1.1102.pdf
Multiscale Fusion Plasma Simulations of Varied Tokamak Scenarios
within the ComPat Framework
O.O. Luk1 , O. Hoenen1 , A. Bottino1 , B.D. Scott1 , D.P. Coster1
1 Max-Planck-Institut für Plasmaphysik, Boltzmannstr. 2, D-85748, Garching, Germany
Fully simulating the impact of turbulence on the performance of fusion devices such as ITER
is challenging, especially when fusion plasmas exhibit highly disparate spatio-temporal scales.
Currently, there are single-scale models developed to study turbulence (gyrokinetic models) and
transport (large-scale simplified models) separately. To go beyond single-scale simulations, the
Computing Patterns for High Performance Multiscale Computing (ComPat) project [1] takes the
component based approach to construct multiscale simulations by connecting existing single-
scale models (submodels) together into a workflow. This approach has simpler algorithm and
codebase, therefore each submodel is easier to validate, verify, maintain and optimize. In addi-
tion, ComPat incorporates the concept of Multiscale Computing Patterns [2] into its framework,
so that applications can run efficiently when one or several submodels require computing capa-
bilities at the petascale.
In this work, the ComPat framework is taken to build a multiscale fusion application that
brings equilibrium (fixed boundary codes), turbulence (either local flux-tube or global gyroki-
netic code) and core transport (1D code) models together. Using both the unified datastructure
developed in EUROfusion and the Multiscale Coupling Library and Environment (MUSCLE2),
one can set up a reliable multiscale fusion plasma simulation based on existing single scale
codebases. A significant advantage to such approach, is the ease with which individual sub-
model can be replaced by another that offers similar functionality. However, there are major
challenges arise to such framework. Several of these challenges will be addressed, including
time bridging between turbulence and transport models, defining quasi-steady state of core
plasma, implementing global gyrokinetic code into current simulation framework, and opti-
mizing the overall simulation runtime. All the simulation results presented here are based on
initial conditions from ASDEX Upgrade- and JET-sized cases.
References
[1] http://compat-project.eu
[2] S. Alowayyed, D. Groen, P.V. Coveney, and A.G. Hoekstra, Journal of Computational Science 22, 15-25
(2017)
