Wavelet-based electromechanical mode shape online identification from ambient data

The wavelet transform method for identifying a power system's electromechanical mode-shape properties from ambient data is presented. The relationship between wavelet power spectral density and modal eigenvectors is derived; then the wavelet-based coherence is presented to discriminate the mode shape of closely spaced modes. Simulation results with a two-area four-generator power system following stationary random excitation shows that the identified mode shape based on wavelet power spectrum agrees well with the eigenvector or Fourier-based mode shape. To overcome the limitations of Fourier transform on analyzing non-stationary signals, the output signals of the same system following non-stationary excitation is applied, and the wavelet spectral density is used to calculate its mode shape, which demonstrates that the wavelet-based method is applicable to identify mode shape from non-stationary signals.