Using Spectral Flatness to Detect and Label Power System Oscillations in the Presence of Intermittent Broadband Noise

With the ubiquitous deployment of phasor measurement units, the ability to automatically parse large-scale historical data sets to detect and label unwanted system dynamics is becoming increasingly important as it can lead to better informed decision making by power system operators. Most oscillation detection algorithms rely on pre-defined energy thresholds for discrete frequency bands to estimate the presence of oscillations, however, such techniques may not perform satisfactorily in systems with intermittent, high energy forced inputs from industrial equipment such as an electric arc furnace that increases the system's energy across the entire frequency spectrum. In this paper, we propose a method for detecting oscillations based on the flatness of the spectrum as opposed to the magnitude of its energy. It is shown that the proposed method is effective at detecting oscillations irrespective of a changing magnitude of input broadband noise.