The U.S. Power Grid is a national resource of vital importance -- a severe disruption would cause severe economic damage; would affect critical infrastructure resources, such as hospitals, transportation networks, and the Internet; and would have significant impact on national security. In August, 2003, a large power blackout affected many states, principally in the Northeast, with large economic impact. Had the blackout laster longer the impact would likely have been far more severe. Similar blackouts have affected other parts of the world in recent years. It has been fairly clearly established that these events were not caused by a lack of generation power. Rather, they were caused by poor system fault tolerance in the transmission network. To put it differently, transmission capacity was tight; this fact, combined with poor understanding of the network dynamics -- in particular, the underlying combinatorial complexity-- and inadequate real-time techniques, conspired to generate fully cascading events.
The central focus of our work will be to use techniques from Operations Research and Computer Science to develop high-performance computational implementations of algorithms that address critical problems in the operation and management of large-scale power grids. The work, which will include both methodology and and experimentation, will tackle long-term planning models, where the focus is on reinforcing a power grid in an economic manner so as to make it robust against potential faults; and ``online'' models, where a fast algorithm allows a network to appropriately react when facing a developing cascade.
