Experiment data-driven modeling of tokamak discharge in EAST

Abstract

Abstract A neural network model of tokamak discharge is developed based on the experimental dataset of a superconducting long-pulse tokamak (EAST) campaign 2016–2018. The purpose is to reproduce the response of diagnostic signals to actuator signals without introducing additional physical models. In the present work, the discharge curves of electron density n e, stored energy W mhd, and loop voltage V loop were reproduced from a series of actuator signals. For n e and W mhd, the average similarity between the modeling results and the experimental data achieve 89% and 97%, respectively. The promising results demonstrate that the data-driven methodology provides an alternative to the physical-driven methodology for tokamak discharge modeling. The method presented in the manuscript has the potential of being used for validating the tokamak’s experimental proposals, which could advance and optimize experimental planning and validation.