Description
Predictions of the heat and particle transport in the present and future fusion devices plays crucial role in the modelling and analysis of the performance properties of the plasma scenarios. It is observed that the transport properties in the core and in the edge regions are coupled in H-mode plasmas, where the core pressure and the pedestal pressure are connected in loop via Shafranov shift and profile stiffness effects [1]. Therefore it is important to have accurate predictions in both core and edge regions. The core transport properties are modelled routinely using integrated modelling tools [2]. In contrary, the existing pedestal transport models based on the first principles do not cover all parameter spaces accurately, such as when resistive effects cannot be neglected, or when the pedestal density or separatrix density is not known. Therefore other approaches were developed to be able to model whole plasma domain in the integrated modelling workflows. The neural network prediction of the profiles in the edge region is one of the available approaches. The kinetic profiles are predicted in the edge region using available database of the limited set of (engineering) parameters and the values at the pedestal top (or slightly inwards into the plasma) are used as a boundary conditions for the core transport modelling. PENN model [3] utilizes such approach and is trained on the JET data using available database [4]. PENN model is coupled to the European Transport Simulator (ETS) integrated modelling workflow [5]. Recently the PENN model is updated to include isotope effect into the predictions. This work is concentrated on validation of the updated PENN model by modelling the core transport of the JET-ILW type I ELMy H-mode plasmas for different isotope mixtures with the ETS workflow using TGLF-SAT2 quasi-linear transport model [7] in the core and PENN predictions as a boundary conditions. The comparison with earlier obtained core modelling and experimental results [6] is performed.