A Two-Stage Decomposition Method for Harmonic Detection Using UWPT and Spectral Envelope Prior-Assisted VMD

Accurate and rapid harmonic detection technologies can effectively monitor and safeguard the healthy and secure operation of power systems. The variational mode decomposition (VMD) and wavelet transform (WT)-based methods are extensively employed in harmonic analysis. However, these methods exhibit higher efficacy in analyzing structurally simple or simulated signals. The random distribution of complex interference in actual nonlinear, time-varying distorted signals increases detection time complexity while also aggravating boundary artifacts of the VMD-based methods causing harmonic overdecomposition and spectral leakage of the WT-based methods responsible for intensified aliasing and noise. Consequently, a two-stage decomposition (TSD) method for harmonic detection is proposed. The first stage applies the time-efficient low-dimensional undecimated wavelet packet transform (UWPT) to isolate the fundamental component representing the signal’s principal energy. The frequency band components of other harmonics are reconstructed from jointly adjacent wavelet packet coefficients to mitigate spectral leakage. To extract individual harmonics via VMD of the reconstructed subband signals in the second stage, a method for estimating the decomposition order of VMD using the prior of local maxima in the signal’s spectral envelope is presented. Meanwhile, distorted signals collected from real-world power systems with 6- and 24-pulse rectifiers are used for validation. Experimental results show that the proposed TSD not only effectively detects the fundamental component but also significantly improves the frequency resolution of other individual harmonics and suppresses noise.