Towards In-Between-Discharges Model-Based Scenario Planning in NSTX-U Via Fast Nonlinear Optimization

Abstract

The realization of advanced scenarios in tokamaks is achieved by carefully selecting the actuator trajectory waveforms, which defines a feedforward control problem. As an alternative to the usual ``trial-and-error’ approach, a more systematic approach to scenario planning via model-based optimization has been proposed. By parameterizing the actuator trajectories, the feedforward control inputs are determined by minimizing a cost function measuring the distance between actual and desired plasma state. This arbitrary cost function, which can weigh different properties of the desired plasma state, is minimized subject to plasma-dynamics, actuator, and state constraints by using Sequential Quadratic Programming. To avoid spending time in numerically computing the gradients of the cost function with respect to the to-be-optimized parameters, analytical expressions of these gradients are pre-calculated in this work. These expressions require the integration of a plasma transport model for NSTX-U, which is provided in this case by the Control Oriented Transport SIMulator (COTSIM). This fast feedforward-control optimizer has the potential of being used routinely for in-between-discharges scenario planning at NSTX-U.

Sai Tej Paruchuri

Postdoctoral Research Associate in Plasma Control

My research interests include plasma control, dynamics and controls, vibrations and adaptive structures, data-driven modeling.