Topological analysis of cascading failures in Bay Area power grid

The power grid networks are vulnerable to various attacks which may lead to cascading failures. Severe cascading failures can cause large scale power outage and preventing them is believed to be one of the greatest challenges in power grid systems. In this paper, we study cascading failures using the Bay Area power grid data. We adopt two topology based models, which describe the definition of load and the reaction of the network upon substation failure or overloading. Traditionally, it is often believed that the attacker will knock down the node (i.e. substation) with the highest load, which represents the strongest attack strategy. However, our study shows that the effectiveness of the traditional load based attack strategy varies under different network models. Specifically, the load based strategy is not the strongest attack in the network model that assumes the overloaded substations completely fail to function. Under another network model in which the overloaded substations still function but with reduced efficiency in power delivery, attacking the node with the high load can cause severe cascading failure. In the worst case, attacking a single node with the second highest load can result in 44% loss of network efficiency.