Evaluation of AR and ARMA models for estimation of core stability in nuclear reactors

High-amplitude power oscillations are undesirable in a nuclear boiling water reactor, since they are a disturbance to reactor operation. The core stability is estimated from passive measurements of the neutron flux, using AR or ARMA models, to verify the stability within the operating domain. Our main concern is to estimate the core stability as accurately as possible. The aim of this paper is to study how this estimation should be carried out. More specifically, pretreatment of data, filtering and decimation, is discussed as means to focus the identification to the interesting frequency domain and reduce the mean square error.