| Abstract: | Large interconnected power systems are usually subjected to natural oscillation (NO) and forced oscillation (FO). NO
occurs due to system transient response and is characterized by several oscillation modes, while FO occurs due to
external perturbations driving generation sources. Compared to NO, FO is considered a more severe threat to the
safe and reliable operation of power systems. Therefore, it is important to locate the source of FO so corrective
actions can be taken to ensure stable power system operation. In this paper, a novel approach based on two-step
signal processing is proposed to characterize FO in terms of its frequency components, duration, nature, and the
location of the source. Data recorded by the Phasor Measurement Units (PMUs) in a Wide Area Monitoring System
(WAMS) is utilized for analysis. As PMU data usually contains white noise and appears as multi-frequency oscillatory
signal, the first step is to de-noise the raw PMU data by decomposing it into a series of intrinsic mode functions
(IMF) using Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (ICEEMDAN)
technique. The most appropriate IMF containing the vital information is selected using the correlation technique.
The second step involves various signal processing and statistical analysis tools such as segmented Power Spectrum
Density (PSD), excess kurtosis, cross PSD etc. to achieve the desired objectives. The analysis performed on the
simulated two-area four-machine system, reduced WECC-179 bus 29 machine system, and the real-time power
system PMU data set from ISO New England, demonstrates the accuracy of the proposed method. The proposed
approach is independent of complex network topologies and their characteristics, and is also robust against
measurement noise usually contained in PMU data. |
