Speaker
Description
See full abstract here http://ocs.ciemat.es/EPS2019ABS/pdf/P5.1074.pdf
The most widely accepted models for plasma turbulent transport are based on gyrokinetic (GK) theory, but the computational times of full nonlinear GK codes prohibit their use for applications such as experimental scenario optimization and control-oriented modelling. Reduced formulations of these equations have resulted in quasilinear turbulent transport models, such as QuaLiKiz [1, 2], which has improved their speed by a factor of 106 compared to full nonlinear GK codes, but is still too slow for the desired application. Recent work has applied neural network (NN) regressions to emulate the reduced models, showing promising results in terms of bridging this gap [3, 4]. This study extends previous work done on the NN regression of QuaLiKiz by including additional input parameters, such as plasma rotation via Mtor and E × B shear, Shafranov shift via MHD, and a heavy impurity species to disentangle main ion dilution from Zeff. This is intended to capture a larger variety of plasma scenarios and improve the applicability of the NN predictions. In order to reduce the training set to a computationally viable size, experimental plasma profiles were extracted from the JET tokamak plasma device and
fitted using Gaussian Process Regression techniques [5], which rigourously accounts for experimental uncertainties. These techniques have also been applied to preparing integrated model inputs and improving their uncertainty quantification [6], further demonstrating their suitability for large-scale data extraction. The primary use of this database is to sufficiently populate a training set to perform exploratory modelling, while remaining focused to the capabilities of the
experimental device. This database can be extended using the developed workflows to include multiple plasma devices, further enhancing the capabilities of the NN regression.
References
[1] J. Citrin et al., Plasma Physics and Controlled Fusion 59, 12 (2017)
[2] C. Bourdelle et al., Plasma Physics and Controlled Fusion 58, 1 (2016)
[3] F. Felici et al, Nuclear Fusion 58, 9 (2018)
[4] J. Citrin et al., Nuclear Fusion 55, 9 (2015)
[5] M.A. Chilenski et al., Nuclear Fusion 55, 2 (2015)
[6] A. Ho et al., Nuclear Fusion, accepted (2019)