SUMMARYWhen a power system is subjected to a large disturbance, effective identification of vulnerable bus and generator groupings in the complex power grid contributes to the analysis of power system dynamic behavior. Also, computationally simple and reliable identification of clusters is of primary importance in the successful application of hybrid transient stability assessment methods. In the paper, a kernel-based clustering approach is proposed to finding communities of power grids and generators, using data clustering and multi-machine equivalent concept. The calculation of the cluster analysis is simple and fast, as only the network topology and generator dynamic parameters along the post-fault trajectory are required. Case studies using different scale test systems are given to illustrate the validity of the proposed method.