1 EX / P 647 Big Data Machine Learning for Disruption Predictions

Building the scientific foundations needed to develop fusion power in a timely way can be facilitated not only by familiar “hypothesis-driven”/ first principles approaches but also by engaging modern bigdata-driven statistical methods featuring machine learning (ML) -an exciting R&D approach that is increasingly deployed in many scientific and industrial domains. An especially time-urgent and very challenging problem facing the development of a fusion energy reactor today is the need to deal reliably with large-scale major disruptions in magnetically-confined tokamak systems such as the Joint European Torus (JET) today and the burning plasma ITER device in the near future. Significantly improved methods of prediction with better than 95% predictive capability are required to provide sufficient advanced warning for disruption avoidance or mitigation strategies to be effectively applied before critical damage is done to the machine. The supervised machine learning classification technique featured in Support Vector Machines (SVM’s) has been further advanced to this end and will be presented in this paper together with early results from ML studies utilizing multi-dimensional signal data to initiate the development of cross-machine-portable predictors. Working on it’s repository of the most important and largest (nearly a half petabyte and growing) data base of fusion-grade plasmas, JET’s statistical scientists have successfully deployed ML software interfaced with the large JET data base over the course of the past 7 years. This has produced encouraging results involving primarily the application of the SVM approach. The goals for the present investigations are to: (i) achieve greater predictive reliability by improving the physics fidelity of the classifiers within the “supervised” ML workflow; and (ii) establishing the cross-machine portability of the associated software beyond JET to other current tokamak systems and to ITER in the future. In order to so, it will be necessary to address the more realistic multi-dimensional, time-dependent, and much larger complex data instead of the simpler zerodimensional, temporal data considered at present in all of the JET ML studies.