基于谱聚类算法的电力系统关键断面自动识别

Automatic Discovery of Critical Section Based on Spectral Clustering

为加强对关键断面的监控,提高电力系统安全稳定运行水平,提出了一种基于谱聚类的系统关键断面自动识别方法,即以节点导纳Laplace矩阵参数中的物理信息反映节点之间拓扑关系和耦合程度,通过引入谱聚类算法,利用数据挖掘技术对节点导纳Laplace矩阵的特征向量进行聚类,形成层次分明的分区方案,进而通过分区自动识别系统关键断面,并以分区效果系数和相关系数反映关键断面识别效果。最后以IEEERTS 79系统为例验证了该算法的可行性,为电力系统关键断面的识别提供了一种新思路。

Based on spectral clustering,a new method of automatic identification of system critical sections is proposed to strengthen monitoring critical section and improve security and stability operation of power systems.The parameters of nodal conductance Laplace matrix are used to reflect topological and couple relationships of two nodes.And then data mining technology is adopted to cluster the eigenvectors of nodal conductance Laplace matrix and the coherent partition scheme is formatted.Finally,the system critical section is automatically discovered with partition.The identification effect is evaluated with the partition coefficient and correlation coefficient.Taking IEEE-RTS 79 system as an example,the feasibility of the method is verified.So,it provides a new idea for automatic discovery of the critical section of power systems.