Data-Driven Models for NSTX-U Vertical Stability Control and Prospects for Negative Triangularity Plasmas

A new analysis of the initial NSTX-U experiments reveals that data-driven models of the plasma's vertical position can be iteratively tuned on a shot-by-shot basis to reduce vertically unstable behavior while developing new operating scenarios. In anticipation of the NSTX-U tokamak returning to operation next year, this work has examined potential updates to the vertical stability controller and how these updates would enable control of negative triangularity plasmas. Advanced regression techniques were used to select which real-time magnetic diagnostics to use as input signals and to optimize the vertical position model to make the controller robust to signal errors. The open-source time-dependent Grad-Shafranov solver TokaMaker has been used to simulate vertical position control in NSTX-U and demonstrate the potential to iteratively tune the vertical position model parameters on a shot-by-shot basis to reduce the time that it takes to develop new scenarios, including scenarios with negative triangularity plasmas. The TokaMaker code has also been deployed to test how effectively the existing shape controller can scan from shapes with a positive triangularity to those with a negative triangularity.