多分辨率动态模态分解在宽频谐波不稳定识别中的应用

Mode identification of harmonic instability in modern power electronics-dominated power grids facilitates oscillation detection, stability assessment and further damping control. However, complex oscillation components and possible uncharacteristic transient variations of the signal pose challenges to robust and accurate mode identification of oscillation signals. In this article, an identification method based on multiresolution dynamic mode decomposition (MR-DMD) algorithm is leveraged to extract dominant system modes with time-frequency information at multiple frequency decomposition levels and temporal subsets, thereby improving robustness in handling missing data. In addition, key algorithm parameters are defined and derived to reveal their impact on the algorithm's performance, while a framework of parameter design is further developed to optimize the identification results. The proposed method and framework are verified and interpreted under diverse oscillation scenarios using real measurement data from experimental conditions.