Performance of Stochastic Subspace Identification Methods in Presence of Forced Oscillations

Abstract

This paper evaluates the performance of data-and covariance- based Stochastic Subspace Identification (SSI) methods for simultaneous estimation of forced oscillations and system modes. Recent events in North American power grid point to resonant interactions of forced oscillations and electromechanical oscillatory modes of the system that can be problematic. In such cases, simultaneous estimation of both natural and forced oscillation characteristics is of great importance in understanding and analyzing the phenomena. In this paper, simulation cases are investigated to show the ability of different types of SSI methods in simultaneous estimation of forced oscillation estimates and natural mode estimates in the system when both kinds have their frequencies close to each other. Kundur two- area test system is used as the study system for simulations. In all the simulated cases, the frequency of the forced oscillation is very close to an inter-area or a local mode of the system. It is shown that even with the lowest possible model order, both system mode and forced oscillation are estimated by SSI methods and this is not related to the mode splitting phenomenon or high model orders. Furthermore, it is shown that although both SSI methods are capable of estimating the forced oscillation and system mode, bias in damping estimation of data- based SSI may be a challenge, especially for well- damped modes.