Realtime Identification of Plasma Shape and Position on ITER Using an Equilibrium Database

Abstract

Realtime control of the plasma shape and position on ITER requires a fast and accurate determination of the equilibrium configuration. A suitable identification method must be robust in the presence of measurement noise and must also remain reliable when the plasma geometry makes possibly large deviations from the pre-programmed configuration. The realtime use of an iterative equilibrium code will, under benign conditions, yield the solution most closely satisfying a given set of input diagnostic measurements. However, apart from the relatively large computational requirement, the danger of uppredictable results due to faulty measurements and/or convergence to an incorrect solution by the minimization algorithm in such codes is always present. In contrast, a non-iterative identification of the equilibrium which naturally incorporates robustness against measurement noise is possible by methods based on offline analysis of a database of randomly generated equilibria chosen to span the entire parameter space accessible by ITER. We describe here the generation of one such database by the Garching Equilibrium Code1 using the ITER FDR PF coil set and spanning a cigar-like volume in a high-dimensional (16-D) parameter space whose axis is a line segment connecting Start of Burn (SoB) and End of Burn (EoB) fiducial ITER equilibria. We investigate the identification of a variety of plasma position, shape and current profile parameters from magnetic flux and field measurements prescribed by the most recent conceptual design of the magnetic sensor layout.2 We analyse a linear model, which is expected to be inadequate, and two leading identification techniques, namely Function Parameterization and Artificial Neural Networks.