A model-based approach for filtering magnetic pitch angles obtained by the Motional Stark Effect diagnostic

摘要

Introduction One of the challenges in fusion research is the optimisation of the operation regime of current and future tokamaks. An important figure of merit for plasma stability and performance is the safety factor (q). It is inversely proportional to the current density and defined as the ratio of the number of m poloidal turns a field line has to complete to do n toroidal turns in the tokamak. However, no direct measurement of the full q-profile is available. It can be calculated by equilibrium reconstruction codes which solve the Grad-Shafranov equation (=force balance in the plasma), constrained by polarisation angle measurements obtained by the Motional Stark Effect diagnostic (MSE). The MSE diagnostic measures the polarisation direction of light emitted by neutral particles injected into the plasma from which the local direction of the magnetic field can be derived. A challenge for the MSE diagnostic is the high accuracy required for equilibrium reconstruction, typically a few tenths of degrees. To enable MSE-constraint equilibrium reconstruction even in situations where high accuracy measurements are unavailable, such as the high-density discharges on the ASDEX-U tokamak (AUG) [2], we propose an observer-based approach to filter the measurements. An observer is based on a model of the system taking the underlying physics into account from which predictions about the expected measurement can be made. Combining the predicted and real measurement can generally lead to better results than would be obtainable by using only one of the two. As a model of our system, we use RAPTOR [3], a faster than real-time transport simulator which solves the coupled ID poloidal flux diffusion and electron heat diffusion equations. From the poloidal flux Ψ and electron temperature profile many otherwise unavailable parameters of the system can be reconstructed. We propose to use RAPTOR to filter the measured polarisation angles with an Extended Kalman Filter (EKF). The EKF has