Defending Electrical Power Grids

Abstract : This thesis considers the problem of protecting an electrical power grid against a potential attack on its physical infrastructure. We develop a mathematical model, called Defense of Known Interdictions (DKI), that identifies the optimal set of components to defend in an electrical power grid given limited defensive resources. For a small test network, we show that defending fewer than 10% of the buses reduces the possible disruption from an attack by over 20%. Previous research has developed optimization models, called I-DCOPF, to find optimal or near optimal interdiction plans for electrical power grids. DKI solution time is determined by I-DCOPF solution time. We develop a model, called the Network Dual Relaxation (NDR), to replace I-DCOPF and reduce solution times. NDR approximates electrical power grid behavior as a minimum cost network flow and uses this approximation to quickly estimate a lower bound for the exact interdiction model. We test NDR on a portion of the North American power grid with a computational limit of 6000 seconds. Results with ten buses defended show that NDR finds solutions that are, on average, 40% better than those of the exact I-DCOPF model with a significant reduction in computational time.