Parallel processing implementation of the unit commitment problem with full AC power flow constraints

The authors describe a parallel implementation of the Lagrangian Relaxation Algorithm with variable duplication for the thermal unit commitment problem. The formulation was previously reported by the authors and allows inclusion of the full nonlinear AC network power flow model, which permits addressing voltage limits, as well as more realistic branch flow limits than is possible with a linear DC flow model. Thus, potential VAr production can be used as another criterion for commitment of otherwise expensive generators in strategic locations. The algorithm is highly parallelizable, and the authors have taken advantage of this in a version currently being developed for the Cornell Theory Center's Velocity AC3 NT cluster.