To succesfully suppress rotating magnetic islands in large tokamaks such as ASDEX-U and ITER
(which is currently under construction), the rotation frequency and the phase of the islands need
to be tracked accurately in realtime. In previous experiments on the TEXTOR tokamak, this was
achieved using a phase-locked loop (PLL). However, the required frequency range in which the
islands need to be tracked for island control on ASDEX-U is too large for a PLL.
In this report an adaptive extended Kalman filter (AEKF) and adaptive unscented Kalman filter
(AUKF) are presented and applied to the problem of tracking periodic signals with time-varying
frequency and amplitude.
Also a high-gain observer is considered, which appeared to be unsuited for this specific problem.
The performance of the AEKF and AUKF using several signal models is evaluated using simulations.
From these simulations it is concluded that the AEKF and AUKF are able to track the frequency
and phase of periodic signals in a much larger range than a PLL. Tests on real data from the
TEXTOR plant revealed that the AEKF is not always able to track magnetic islands, the AUKF is
much more robust and also performed well on real data, which shows that the unscented Kalman
filter is a good candidate for tracking the frequency and phase of magnetic islands.