Neural networks for generation scheduling in power systems

A neural network for generation scheduling in power systems is presented. The network consists of two levels which correspond to different types of variables in the generation scheduling problem. The first level is a neural net which solves economic dispatch, a sub-problem in the generation scheduling problem. Its outputs indicate the power generation of generating units. The second level is a Boltzmann machine, a stochastic neural net which determines the off/on status of units. Simulation on a 20 generating-unit system shows that fast and optimal solutions can be obtained using the proposed neural network.<<ETX>>