Tokamak disruption prediction using different machine learning\n techniques

Disruption prediction and mitigation is of key importance in the development\nof sustainable tokamakreactors. Machine learning has become a key tool in this\nendeavour. In this paper multiple machinelearning models will be tested and\ncompared. A particular focus has been placed on their portability.This\ndescribes how easily the models can be used with data from new devices. The\nmethods used inthis paper are support vector machine, 2-tiered support vector\nmachine, random forest, gradient boostedtrees and long-short term memory. The\nresults show that the support vector machine performanceis marginally better\namong the standard models, while the gradient boosted trees performed the\nworst.The portable variant of each model had lower performance. Random forest\nobtained the highest portableperformance. Results also suggest that disruptions\ncan be detected as early as 600ms before the event.An analysis of the\ncomputational cost showed all models run in less than 1ms, allowing sufficient\ntimefor disruption mitigation.\n