Speaker
Aaron Ho
Description
See the full Abstract at http://ocs.ciemat.es/EPS2018ABS/pdf/P4.1088.pdf
Turbulent transport model validation at JET using integrated modelling
enhanced by Gaussian process regression
A. Ho1 , J. Citrin1 , F. J. Casson2 , F. Auriemma3 , C. Bourdelle4 , P. Manas5 , G. Szepesi2 ,
H. Weisen6 , and JET Contributors∗
1
DIFFER, De Zaale 20, 5612 AJ, Eindhoven, The Netherlands
2 EURATOM-CCFE Fusion Association, Culham Science Centre, Abingdon, OX14 3DB, UK
3 Consorzio RFX, Associazione EURATOM-ENEA sulla Fusione, Padova, Italy
4 CEA, IRFM, F-13108 Saint-Paul-lez-Durance, France
5 Max-Planck-Institut für Plasmaphysik, D-85748 Garching, Germany
6 Swiss Plasma Center, EPFL, 1015 Lausanne, Switzerland
∗ See the author list of “X. Litaudon et al 2017 Nucl. Fusion 57 102001"
Due to increasing complexity and costs of experimental fusion plasma devices, more empha-
sis is being placed on plasma models to assist in the design process. To have confidence in these
model predictions, a self-consistent connection between the predictions and experimental mea-
surements must be ensured via model validation. However, the high sensitivity and non-linear
nature of plasma models demand a more rigourous uncertainty treatment in order to determine
the significance of any reported agreement between model and experiment. By using Gaus-
sian Process Regression (GPR) techniques [1, 2] on the measurement data, which can provide
both fit and fit gradient envelopes while maintaining tractability for large-scale data processing,
validation and sensitivity studies can be performed with increased statistical rigour.
This study outlines the application of GPR techniques to profile fitting for use in tokamak
turbulence transport model validation within integrated modelling. With properly tuned opti-
mizers, the developed profile fitting tool can process a single time window in ∼2 min., allowing
the processing of measurements from an entire discharge in reasonable time. The advantages
of this approach were demonstrated through a JETTO integrated modelling simulation [3, 4]
of the JET ITER-like-wall discharge #92436 with the QuaLiKiz quasilinear turbulent transport
model [5, 6]. Excellent agreement was achieved between the fitted and simulated profiles for
ne , Te and Ωtor simultaneously but the simulation underpredicts Ti for this discharge. This un-
derprediction is suspected to be from known physics which is currently being included in the
transport model. The fit envelopes have allowed for more rigourous error propagation through
the model, such as Monte Carlo studies of transport model boundary conditions within the fit
uncertainties, and the definition of a figure-of-merit to assess the quality of this agreement.
References
[1] M.A. Chilenski et al., Nuclear Fusion 55, 2 (2015)
[2] C.E. Rasmussen, C.K.I. Williams, Gaussian Processes for Machine Learning, (2006)
[3] M. Romanelli et al., Plasma and Fusion Research 9, 01 (2014)
[4] G. Cenacchi, A. Taroni, JETTO: A free boundary plasma transport code, JET-IR (1988)
[5] J. Citrin et al., Plasma Physics and Controlled Fusion 59, 12 (2017)
[6] C. Bourdelle et al., Plasma Physics and Controlled Fusion 58, 1 (2016)
