Identification of Plasma Equilibria in ITER from Magnetic Measurements Via Functional Parameterization and Neural Networks

Real-time control of the plasma shape in the International Thermonuclear Experimental Reactor (ITER) calls for a fast and accurate identification of the equilibrium starting from magnetic measurements. The technique proposed for ITER interpolates the actual equilibrium within a previously generated dataset where each parameter is given a sufficiently wide range of variation. The properties of the artificial neural networks (ANNs) are shown to be well suited for this task. The satisfactory comparison with the functional parameterization, which is currently adopted for the feedback control in ASDEX-Upgrade, makes the proposed technique well linked to the experience available in current experiments. The ANN technique also provides an algorithm for the selection of the number and location of the magnetic sensors, which is an important issue for the ITER design. A preliminary analysis of the effects of the eddy currents flowing in the structure is also included. Numerical results presented refer to the so-called TAC-4 ITER geometry; extrapolation to update geometries with a close poloidal field concept is straightforward.