Calculating and calibrating a persistence measure for use in monitoring power system vulnerability

Wide-area monitoring using phasor measurement units (PMUs) offers the possibility of detecting conditions during which the power system may be vulnerable to destabilizing events. Such events are often characterized by undamped oscillations and are preceded by conditions of very low damping. Current techniques for estimating damping from low SNR ambient data are less accurate than desired, although fairly accurate for estimating frequencies of oscillation. We introduce a robust persistence measure that can be used in lieu of damping estimates to assess conditions for vulnerability to dynamic instability. The measure relies on the calculation of the energies of normalized sample autocorrelations of select signals. In this paper we focus on the technical details associated with spectral smoothing and temporal isolation in the calculation of the normalized autocorrelations, to permit effective calibration of the metric for both stable and unstable systems. An example is provided using measurements from the field.