Identification of actuator model in an electrical machine by prediction error method and cultural particle swarm optimization

Abstract

This paper discusses identification of an actuator model, which has been built inside a two-pole induction motor in order to control rotor vibrations. The methods used for identification are prediction error method and cultural particle swarm optimization with mutation. The first-mentioned method produces a black box model with correspondence to input-output measurements. The second method is used to identify parameters of a linear time-invariant state-space model, which is based on electromechanical equations. The results are compared in time domain and in frequency domain.