中文摘要：

随着WAMS系统的广泛应用,功角量测数据为基于轨迹的同调机组分群研究提供了可靠的数据支持。利用聚类方法人为因素干扰小、自适应能力强的优点,提出了一种基于EMD的聚类树分群方法。该方法建立在系统聚类分析的基础上,以各机受扰轨迹之差的信号能量最小为聚类准则,可以实现多机系统的动态分群。为解决电力系统受扰后动态轨迹非平稳、非线性问题,采用EMD预处理方法,实现原始数据的高频滤波和平稳化处理。EPRI-36节点系统和新英格兰10机39节点系统的算例分析证明,在网络拓扑结构基本不变的情况下,不同扰动方式下聚类分群结果总体上一致,从而佐证了该方法的有效性。

英文摘要：

　With the application of the Wide Area Measure System（WAMS）,the mass data of phase-swing curves is sufficiently supporting the study on coherency identification of power generation sets.Because of the ability of anti-disturbance and self-adaptive,the clustering-tree method for coherency identification based on Empirical Mode Decomposition（EMD） is put forward.The method can realize dynamic coherency identification for multiple sets on the basis of system clustering analysis,taking the clustering principle of minimum signal energy of disturbance trace.In order to solve the problems of non-stable and non-linear of electric system after disturbance,a new approach to data pretreatment which is called EMD is used to realize high-frequency filter and stabilization of original data.Case studies of New England 10-machine 39-bus system and EPRI 36-bus system show that the coherency identification results are consistent with different disturbance modes for the same topology structure,which indicates that the method is effective.