| The stability of heavily interconnected power systems is a primary concern in the power utility industry. Consequently, utilities are interested in real-time assessment of power system conditions. Accurate knowledge of the low-frequency electromechanical modes in power systems gives vital information about the stability of the system. Current techniques for estimating electromechanical modes are computationally intensive and rely on complex system models which are often inaccurate or incomplete. This research moves away from model-based approaches and uses measurement-based techniques. Current measurement-based techniques typically require a ringdown from a disturbance. This research applies signal analysis techniques, including block processing and adaptive filtering algorithms, to ambient power system data to estimate the frequency and damping ratio of the dominant electromechanical modes. This is a new approach in that the modes are estimated from measured ambient power system data without requiring a disturbance.;Programs are developed to implement block-processing and adaptive filtering algorithms, including the auto-regressive moving average (ARMA) and the least mean squares (LMS) techniques, on actual power system data from the Western United States and simulated data generated from 4 and 19-machine system models. The results show that given an adequate time interval of data, the dominant electromechanical modes are identified. There is more variability in the estimate of the damping ratio than the frequency. On data from the Western power grid, the results from ambient data compare well with Prony analysis of a ringdown immediately following the ambient data. |
