Neural-Network Version of NUBEAM for Real-time Control and Scenario Optimization in DIII-D

M. Morosohk, M.D. Boyer, E. Schuster

Division of Plasma Physics (DPP) Annual Meeting of the American Physical Society (APS)

Portland, OR, USA, November 5-9, 2018

NUBEAM is a code developed to simulate with high accuracy the effects of neutral beams on the plasma. However, the code is computationally demanding due to its Monte Carlo nature. This makes it impractical for use in real-time control or in applications where a large number of simulation runs are required such as in model-based scenario optimization. Recently, a neural network model (NubeamNet) has been developed to closely approximate NUBEAM results for NSTX-U at drastically reduced run time [1]. In this work, a database of NUBEAM runs has been developed for DIII-D and used to train a neural network for model-based control and optimization applications. The trained neural network can match the NUBEAM data with high accuracy, and requires only microseconds to compute the outputs for a given set of inputs, making it a powerful tool for use in real-time control or in model-based scenario planning where a large number of simulation runs are required by the optimization algorithm. Simulation results illustrate the potential of the trained neural-network version of NUBEAM.

[1] M.D. Boyer et al., “Real-time capable neural network approximation of NUBEAM for use in the NSTX-U control system,” 45th EPS Conference on Plasma Physics, 2018.