Large-Scale Cascading Failure Mitigation in Power Systems via Typed-Graphlets Partitioning

Large-scale cascading power failures impact nations economically and socially. Current literature is lacking effective methods and tools to prevent failures from globally propagating in the system. In this paper, we bridge this gap by proposing a partitioning method that exploits the heterogeneity of the power system by identifying the most vulnerable typed-graphlets. We formulate the partitioning problem as a large-scale mixed integer program, which we solve using Benders' method. We show through simulations that i) when typed-graphlets are used, a much larger number of failing power nodes is required to reach a complete power blackout compared to the case without typed-graphlets, and ii) the overall damage can be reduced by an average of 61% compared to the case without partitioning.