Predict first: turbulent transport validation within integrated modeling on JET and ASDEX Upgrade

Integrated modelling is extensively carried out on JET and ASDEX Upgrade pulses applying the first-principle-based quasilinear turbulent transport model QuaLiKiz [www.qualikiz.com, Citrin PPCF2017] within JINTRAC [Romanelli PFR2014].

For the first time, the evolution of density, rotation and temperature profiles for electrons and multiple ions (including Tungsten) is modelled within a flux-driven transport code over multiple confinement time along with the self-consistent prediction of the current diffusion, heat sources, radiation and magnetic equilibrium, revealing the complex interactions and multiple nonlinearities at play.

Agreement between the predicted profiles and the measured ones is obtained in all channels, including the time evolution of W 2D profiles, for various hybrid and baseline H modes in both JET [Breton NF 2018] and ASDEX Upgrade. The transition between the LOC and the SOC regimes is also reproduced. In JET, the core W accumulation is shown to be reinforced by NBI central particle source and NBI torque. Applying ICRH can prevent W accumulation. The impact of the boundary conditions, at the pedestal top or at the LCFS, is investigated using uncertainties produced by Gaussian process regression techniques. Based on this successful integrated modelling, extrapolations to higher power and longer pulses are carried out to prepare the upcoming JET DT campaign. In ASDEX Upgrade, the modeling is used to assess the impact of neoclassical, turbulent and MHD driven transport on avoidance of central W accumulation.

A surrogate model of QuaLiKiz, 4 orders of magnitude faster than the original model, has been produced through a neural network regression of 3x10⁸ flux calculations over 9 input dimensions. Presently, a proof-of-principle 4D version is implemented in the control-oriented fast tokamak simulator RAPTOR for simultaneous heat and particle transport [Felici NF 2018].