Initialisation of the augmented Hopfield network for improved generator scheduling

An artificial neural network algorithm for generator scheduling is proposed. The algorithm employs an unfeasible Lagrangian dual maximum solution to initialise the neurons of an augmented Hopfield network. The proposed algorithm produces cheaper solutions when compared with Lagrangian relaxation or a randomly initialised augmented Hopfield network. The algorithm also has shorter convergence times than the augmented Hopfield network, but is not as fast to converge as Lagrangian relaxation.