NubeamNet: Accelerated predictive modeling of NSTX-U beam deposition for optimization and control

Model-based control and scenario development will be critical for safely and efficiently reaching optimal performance of present-day and future fusion devices. The model-based approach relies on a hierarchy of models of varying fidelity and speed, tailored to different roles in the design process. To enable high-fidelity beam deposition calculations for real-time optimization and control applications on NSTX-U, a neural network model, NubeamNet, has been developed based on NUBEAM calculations. The model evaluates heating, torque, and current drive profiles from equilibrium parameters and measured profiles. The database was generated from interpretive TRANSP analysis of shots from the 2016 NSTX-U campaign. Predictions made for the testing data demonstrate the ability of the model to generalize and accurately reproduce profiles and scalar quantities. Hardware-in-the-loop simulations of the model in the NSTX-U plasma control system demonstrate the suitability of the model for real-time applications. Applications of the model, including estimation of Zeff and anomalous fast ion diffusivity to match measured neutron rates, will be presented.