Kalman filter methods for real-time frequency and mode number estimation of MHD activity in tokamak plasmas

Abstract

Magnetohydrodynamic (MHD) activity in magnetically confined fusion experiments is often associated with detrimental effects such as increased radial transport and consequent loss of confinement. In particular, the (2,1) neoclassical tearing mode (NTM), when proceeding to mode-locking, is a potentially disruptive instability hence with the potential to compromise the mechanical integrity of the machine. It is therefore quite significant to be able to characterize in real-time the most virulent and performance limiting instabilities such that adequate mitigation or complete stabilization using feedback control methods are employed during the plasma discharge. This work proposes a Kalman filter (KF) based mechanism for providing, in real-time, the amplitude and phase evolution of instabilities within a predefined set of mode numbers. The method relies on two KF implementations: a non-linear KF isolating the non-stationary dominant signal component of a sensor measurement and subsequently a linear KF which projects the former, for a collection of sensors, onto a predefined set of mode numbers. A basic overview of algorithms commonly used for real-time mode number analysis is also presented along with applications of the proposed algorithm to recently recorded data of the Joint European Torus (JET) tokamak.