Parameter estimation for a nonlinear control-oriented tokamak profile evolution model

Abstract

A control-oriented tokamak profile evolution model is crucial for the development and testing of control schemes for a fusion plasma. The RAPTOR (RApid Plasma Transport simulatOR) code was developed with this aim in mind (Felici 2011 Nucl. Fusion 51 083052). The performance of the control system strongly depends on the quality of the control-oriented model predictions. In RAPTOR a semi-empirical transport model is used, instead of a first-principles physics model, to describe the electron heat diffusivity in view of computational speed. The structure of the empirical model is given by the physics knowledge, and only some unknown physics of , which is more complicated and less well understood, is captured in its model parameters. Additionally, time-averaged sawtooth behavior is modeled by an ad hoc addition to the neoclassical conductivity and electron heat diffusivity. As a result, RAPTOR contains parameters that need to be estimated for a tokamak plasma to make reliable predictions. In this paper a generic parameter estimation method, based on the nonlinear least-squares theory, was developed to estimate these model parameters. For the TCV tokamak, interpretative transport simulations that used measured profiles were performed and it was shown that the developed method is capable of finding the model parameters such that RAPTOR’s predictions agree within ten percent with the simulated q profile and twenty percent with the measured profile. The newly developed model-parameter estimation procedure now results in a better description of a fusion plasma and allows for a less ad hoc and more automated method to implement RAPTOR on a variety of tokamaks.