Hybrid deep learning architecture for general disruption prediction across tokamaks

In this letter, we present a new disruption prediction algorithm based on Deep Learning that effectively allows knowledge transfer from existing devices to new ones, while predicting disruptions using very limited disruptive data from the new devices. Future fusion reactors will need to run disruption-free or with very few unmitigated disruptions. The algorithm presented in this letter achieves high predictive accuracy on C-Mod, DIII-D and EAST tokamaks with limited hyperparameter tuning. Through numerical experiments, we show that good accuracy (AUC=0.959) is achieved on EAST predictions by including a small number of disruptive discharges, thousands of non-disruptive discharges from EAST, and combining this with more than a thousand discharges from DIII-D and C-Mod. This holds true for all permutations of the three devices. This cross-machine data-driven study finds that non-disruptive data is machine-specific while disruptions are machine-independent.