Towards fully-predictive transport modelling in ASDEX Upgrade H-modes

Abstract The common way to predict energy confinement in future devices such as ITER is use scaling laws, based on parameters regression of large cross-machine databases. However, this approach limited: the variables are not purely engineering parameters, physics quantities plasma density n e also input; power regressions fail capture important regime transitions; profile effects T i / or reverse magnetic shear retained. As a consequence, scatter large, but even some dependences known be limited validity, P IPB98(y, 2) scaling. Dimensionless physics, if validated, provides reliable basis for predicting device. there no experimental data boundary condition, nor empirical knowledge that Also, full gyro-kinetic codes with all relevant ingredients computationally unfeasible full-radius modelling. Integrated modelling recently developed IMEP workflow allows simulate from separatrix centre, without making any input. Although physics-based models used core transport, transport shown predicted less accuracy than pedestal energy. Therefore, we validate most established quasi-linear TGLF and QuaLiKiZ different regimes conditions, assessing their strengths limits. accurately variety including dominant ion electron heating, respectively. well-known shortcoming near top solved, while new collision operator model improves temperature simulation significantly. Also Neural Network version QuaLiKiZ, very fast computation time, appears well applicable heated H-modes.