Since 18 of December 2019 conferences.iaea.org uses Nucleus credentials. Visit our help pages for information on how to Register and Sign-in using Nucleus.

17–22 Oct 2016
Kyoto International Conference Center
Japan timezone

Big Data Machine Learning for Disruption Predictions

20 Oct 2016, 14:00
4h 45m
Kyoto International Conference Center

Kyoto International Conference Center

Takaragaike, Sakyo-ku, Kyoto 606-0001 Japan
Poster EXS - Magnetic Confinement Experiments: Stability Poster 6

Speaker

Dr William Tang (Princeton Plasma Physics Laboratory)

Description

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 big-data-driven statistical methods featuring machine learning (ML). An especially time-urgent and very challenging problem facing the development of a fusion energy reactor today is the need to reliably mitigate and avoid 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. These major macroscopic events lead to rapid termination of plasma discharges including severe impulsive heat loads damaging material components. Since they can damage the surfaces of the machine, which in turn can cost hundreds of millions of dollars to remediate, it is critical that the international fusion mission engage in multiple avenues to accelerate progress toward achieving the capability to reliably avoid such events with better than 95% predictive capability. Accordingly, this paper will present results from the development and testing of ML-based-methodologies – an exciting R&D approach that is increasingly deployed in many scientific and industrial domains -- to help provide much-needed guidance for disruption avoidance in JET. Working on this repository of the most important and largest (nearly a half petabyte and growing) data base of fusion-grade plasmas, JET statistical scientists have successfully deployed ML software interfaced with the large JET data base over the course of the past 6 years. This has produced encouraging results involving primarily the application of the support vector machine (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 portability of the associated software beyond JET to other current tokamak systems and to ITER in the future. This will necessitate more realistic multi-dimensional, time-dependent, and much larger complex data instead of the simpler zero-dimensional, temporal data considered at present in all of the JET ML studies. Current analysis (via MDS+ tree) of the JET disruption data base has been enabled by formal approval of EUROfusion/JET.
Country or International Organization USA
Paper Number EX/P6-47

Primary author

Dr William Tang (Princeton Plasma Physics Laboratory)

Co-authors

Dr Andrea Murari (EUROfusion Consortium, JET) Dr Augusto Pereira (EUROfusion Consortium, JET) Dr Eliot Feibush (Princeton Plasma Physics Laboratory) Dr Isabel Nunes (EUROfusion Consortium, JET) Dr Jesus Vega (EUROfusion Consortium, JET) Dr Jong Choi (Oak Ridge National Lab) Mr Matthew Parsons (Princeton Plasma Physics Laboratory) Dr Tahsin Kurc (Stony Brook University)

Presentation materials