Speaker
Description
The safe and controlled exhaust of heat from magnetically confined fusion plasmas requires having dynamic information about the system. On present-day reactors, this information is obtained using system identification experiments, which involve perturbing the gas injection rate and observing the Scrape-Off-Layer (SOL) response in the frequency domain. However, this strategy is cumbersome for future reactors, as accidental disruptions or reattachment of the SOL plasma could lead to catastrophic damage to the device. Therefore, numerical dynamic models of the exhaust plasma, validated against experiments, are needed. To this end, we present a multi-sine gas injection perturbation approach that uses high-fidelity time-dependent SOLPS-ITER simulations, with a grid extending to the vessel wall, to quantify the plasma response. Comparison to an existing set of experiments on MAST-U is done to validate this approach. The simulation grid was generated using the Grid Optimization and Adaptation Toolbox (GOAT), and both the Advanced Fluid Neutral (AFN) model and the kinetic neutral model Eirene available in SOLPS-ITER are used. New boundary condition types were implemented in the code to enable multi-sine gas injection perturbations in SOLPS-ITER. In steady state, qualitatively similar behaviour to previous SOLPS-ITER simulations with impurities and kinetic neutrals is observed, allowing comparison between our model and experimental data at similar operating conditions. The comparison in the frequency domain shows that the AFN SOLPS-ITER simulations predict much faster response times to divertor fuelling perturbations than the experiments, around a factor of 10 at 15 Hz, likely in part due to the fluid approximation used for the neutrals. Another possible cause of the discrepancy is the effect of remote vessel regions in MAST-U, some of which are not included in the computational grid. These regions can act as neutral gas reservoirs which introduce additional timescales to the system. Therefore, an additional boundary condition type was implemented, which models these reservoirs as 0D particle balance equations. When including the reservoir models for remote vessel regions, the response time increases, but is still too low compared to experiments by a factor of 4 at 15 Hz. Simulations employing Eirene for kinetic treatment of neutrals show better agreement with the experiments, i.e. within a factor of 2 at 15 Hz. In conclusion, this work presents a first step towards using the SOLPS-ITER code package for synthetic system identification using multi-sine perturbations.