Prony analysis method can detect oscillation frequency, damping, phase and amplitude directly from time series data. Several applications based on the method have been developed in power systems, but it is said that it gives less accurate results for noisy data. In this paper, a power swing mode detection method is presented which is suitable for noisy measured data in power systems.
First, the analysis parameters such as sampling interval, number of data and number of orders are investigated. For detection of low frequency inter-area modes, longer sampling interval (0.2 to 0.4 s) is preferable. The importance of each detected mode can be evaluated by an index based on integration of each waveform. And for data containing large noise, the accuracy of analysis can be improved by applying a low-pass filter. By modifying amplitudes and phases based on the filter characteristics, the original waveform eliminating noise can be estimated precisely.
Effectiveness of the method were verified through examinations using active power and bus voltage data actually measured in a power system.