A coupled gradient network for the solution of the temporal unit commitment problem in power systems planning

This paper deals with a neural network solution for a problem specific to the power industry. In particular, the unit commitment problem is addressed. This paper is an extension of the results given in Watta and Hassona (1994), where the static unit commitment problem was discussed. The temporal problem involves the scheduling of unit commitment over a planning horizon of more than one hour. Using a cascade and interconnect strategy, the coupled gradient network given in the above paper and Watta (1994) is extended to handle the temporal case. As before, though, the dynamics of the network are defined by gradient descent on a global energy function, and hence the network converges to locally optimal solutions. After formulating the model, the parameter selection involved in actually simulating the network is discussed. Finally, this paper concludes with a discussion of some simulation results