Scenario adaptive disruption prediction study for next generation burning-plasma tokamaks

Next generation High Performance (HP) tokamaks risk damage from unmitigated disruptions at high current and power. Achieving reliable disruption prediction for a device's HP operation based on its Low (LP) data is key to success. In this letter, through explorative analysis dedicated numerical experiments multiple existing tokamaks, we demonstrate how the operational regimes of can affect power trained predictor. First, our results suggest data-driven predictors abundant LP discharges work poorly regime same tokamak, which consequence distinct distributions tightly correlated signals related in these two regimes. Second, find that matching parameters among strongly improves cross-machine accuracy implies model learns underlying scalings dimensionless physics like q95, βp confirms importance cross machine domain perspective. Finally, show absence target devices, best predictivity be achieved by combining with other machines. These provide possible predictor development strategy next such as ITER SPARC, highlight developing baseline scenario future machines collect more relevant disruptive data.