Adaptive Identification of Wideband Oscillations in High-Voltage Power Systems: A WIFFT-ZoomFFT Fusion Method

Addressing the challenge of multi-modal identification of wide-band oscillations in “double-high” power systems, this paper introduces an adaptive wide-band oscillation mode identification method. This method achieves a frequency resolution of 0.2 Hz by dynamically fusing multiple algorithms (combining the wide-band processing of WIFFT with the local refinement of ZoomFFT, and adaptively switching analysis modes based on a main-lobe interference determination mechanism). It enhances the signal-to-noise ratio to the order of $10^{-11}$ by utilizing the side-lobe suppression characteristics of the window (side-lobe peak $\lt -92 \mathrm{~dB}$) and dynamic threshold filtering ($P_{t h}=\mu+K \sigma$). Furthermore, the introduction of three-spectrum line interpolation and window function phase compensation techniques controls the relative errors of amplitude and phase to the order of $10^{-7}$ and $10^{-10}$, respectively. Experimental results demonstrate that, under mixed noise scenarios (white noise, colored noise, and impulse noise), the frequency error is $\leq$ $1.1 \times 10^{-11} \mathrm{~Hz}$, the spectral leakage suppression ratio reaches -92 dB, and the minimum distinguishable frequency interval is 0.0001 Hz, which significantly outperforms traditional methods. The research findings provide high-precision parameter support for oscillation suppression strategies in “double-high” systems.