Artificial Neural Networks: a viable tool to design heat load smoothing strategies for the ITER Toroidal Field coils

In superconducting tokamaks, cryoplants provide the helium needed to cool the superconducting magnet systems. The evaluation of the heat load from the magnets to the cryoplant is fundamental for the design of the latter and the assessment of suitable strategies to smooth the heat load pulses induced by the pulsed plasma scenarios is crucial for the operation. Here, a simplified thermal-hydraulic model of an ITER Toroidal Field (TF) magnet, based on Artificial Neural Networks (ANNs), is developed and inserted into a detailed model of the ITER TF winding and casing cooling circuits based on the state-of-the-art 4C code, which also includes active controls. The low computational effort requested by such a model allows performing a fast parametric study, to identify the best smoothing strategy during standard plasma operation. The ANNs are trained using 4C simulations, and the predictive capabilities of the simplified model are assessed against 4C simulations, both with and without active smoothing, in terms of accuracy and computational time.