A new controlled islanding algorithm based on spectral clustering

When islanding operation of system is unavoidable, controlled islanding need determine proper splitting strategies and split the entire interconnected transmission network into islands ensuring generation/load balance and generator coherency. For a large-scale power system, the controlled islanding problem is very complex in general because a combinatorial explosion of strategy space happens. This paper mainly studies how to find proper splitting strategies of large-scale power systems using a two-step spectral clustering based method. In the first step, the machine nodes (generators and dynamic loads) will be grouped by normalized spectral clustering with dynamic models to satisfy the dynamic constraint, and the machine grouping results will be served as prior knowledge of the next step. In the second step, all the nodes will be grouped by constrained spectral clustering with power flow data to satisfy the static constraint and get the minimal cut set solution. Simulation results on IEEE 9- and 118-bus networks show that this method is efficient for controlled islanding of large-scale power systems.