Harmonic Time–Frequency Analysis and Detection Method Based on Improved MSST

This paper proposes a method for harmonic time–frequency analysis and detection based on an improved multi-synchrosqueezing transform (MSST). The aim is to address the significant endpoint problem of the synchrosqueezing transform (SST) in power harmonic analysis. This approach initially employs the Burg method to estimate the parameters of the auto-regressive (AR) model for the harmonic signal. Subsequently, it conducts multiple iterative computations on the SST results of the extended harmonic signal, further compressing the time–frequency spectrum energy to obtain a more precise time–frequency spectrum of the harmonic signal. Additionally, it utilizes the robust reconstruction capability of MSST to decompose the harmonic signal and obtain a series of intrinsic mode functions (IMF) with different frequencies. Finally, the Hilbert Transform is applied to identify the harmonic parameters of each IMF component and accomplish harmonic detection. The simulation experiments and measured data results demonstrate that the proposed method outperforms the Hilbert-Huang Transform (HHT) and SST methods in achieving more accurate time–frequency analysis and detection of harmonic signals. It also reveals the time–frequency characteristics and variation patterns of power grid harmonics, making it of great significance for harmonic control.