Real-time capable first principle based modelling of tokamak turbulent ă transport
Résumé
A real-time capable core turbulence tokamak transport model is ă developed. This model is constructed from the regularized nonlinear ă regression of quasilinear gyrokinetic transport code output. The ă regression is performed with a multilayer perceptron neural network. The ă transport code input for the neural network training set consists of ă five dimensions, and is limited to adiabatic electrons. The neural ă network model successfully reproduces transport fluxes predicted by the ă original quasilinear model, while gaining five orders of magnitude in ă computation time. The model is implemented in a real-time capable ă tokamak simulator, and simulates a 300 s ITER discharge in 10 s. This ă proof-of-principle for regression based transport models anticipates a ă significant widening of input space dimensionality and physics realism ă for future training sets. This aims to provide unprecedented ă computational speed coupled with first-principle based physics for ă real-time control and integrated modelling applications.