Stabilization of burn conditions in an ITER FEAT like tokamak with uncertainities in the helium ash confinement time
Résumé
In this work we demostrate using a two-temperature volume average 0D model that robust stabilization, with regard the hellium ash confinement time , of the burn conditions of a tokamak reactor with the ITER FEAT design parameters can be achieved using Radial Basis Neural Networks (RBNN). Alpha particle thermalization time delay is taken into account in this model. The control actions implemented by means of a RBNN, include the modulation of the DT refueling rate, a neutral He-4 injection beam and auxiliary heating powers to ions and to electrons; all of them constrained to lie within allowable range values. Here we assume that the tokamak follows the IPB98(y,2) scaling for the energy confinement time, while the helium ash confinement time is assumed to be independently estimated on-line. The DT and helium ash confinement times are assumed to keep a constant relationship at all times. An on-line noisy estimation of the helium ash confinement time is simulated by corrupting it with pseudo Gaussian noise.
Loading...