Abstract
This paper presents a method based on residual analysis and system identification techniques to localize the source of forced oscillations in power systems due to turbine governors in power plants. The method identifies the closed loop dynamics of power plants described by the swing equation. When forced oscillations are detected, they are located by finding the transfer functions that describe the behaviour of the corresponding plant the worst. The method presented in this paper locates the source of forced oscillations based only on local measurements at each plant. This may represent an inherent advantage since it may reduce the need of data measurement and communication. We investigate how the method performs under different process noise levels and also the common assumption that frequency measurements are a sufficient approximation of the machine’s rotational speed. The performance of the method is demonstrated with an hardware-in-the-loop approach on an experimental setup including a real time simulator and phasor measurement units. Sensitivities of the method to different forced oscillations and assumptions are analysed using numerical simulations.