Skip to Main Content
Large-scale power blackouts caused by cascading failure are inflicting enormous socioeconomic costs. We study the problem of cascading link failures in power networks modelled by random geometric graphs from a percolation-based viewpoint. To reflect the fact that links fail according to the amount of power flow going through them, we introduce a model where links fail according to a probability which depends on the number of neighboring links. We devise a mapping which maps links in a random geometric graph to nodes in a corresponding dual covering graph. This mapping enables us to obtain the first-known analytical conditions on the existence and non-existence of a large component of operational links after degree-dependent link failures. Finally, we present a simple but descriptive model for cascading link failure, and use the degree-dependent link failure results to obtain the first-known analytical conditions on the existence and non-existence of cascading link failures.