A graph-algebraic approach for detecting islands in power system

Network islanding is a critical event which is often caused by power system contingencies. In order to safeguard system operation and control, it is imperative to timely detect existing or emerging islands beforehand. This paper develops a new and effective solution for online detection of islanding transition. This approach can largely facilitate secure operation and control of power systems based on complex network and graph theories. In this paper, different graph partitioning methods based on complex network are firstly compared. Then a novel graph-algebraic model based on spectral analysis is developed. In this model, Laplacian matrix of power system network (PSN) is established, which contains significant topological information of power grids. Subsequently, spectral patterns analysis can assist to identify emerging grid islands rapidly and accurately. To evaluate the performance of this approach, IEEE New England 39-bus system is used in our case study.