Problems involving DC Optimal Power Flow have been solved by several neural approaches proposed in the literature. The approaches based on Hopfield models are those more used for mapping these types of problems. However, most of these models ignore the transmission system, which leave of taking into account important active power constraints. This fact can lead to solutions obtained by the networks to improper dispatch policies. Moreover, problems involving convergence speed and difficulty of guaranteeing feasible solutions are also found in some approaches. This paper presents a modified Hopfield network to solve DC Optimal Power Flow in an efficient way. The transmission system in this model is represented by linear load flow equations and constraints on active power flows. The internal parameters of the modified Hopfield network are computed using the valid-subspace technique, which guarantees the obtainment of feasible solutions (equilibrium points) by the developed network. Simulation results and a sensitive analisys involving IEEE test systems are presented to illustrate the efficiency of the proposed approach.
DC optimal power flow; artificial neural networks; systems optimization; Hopfield networks; power systems
