Application of evolving neural network to unit commitment

Abstract

This paper reports the initial research results of applying an evolving neural network to unit commitment. In this technique, a genetic algorithm is evolved to intelligently decide the initial weights and the connection in the artificial network to solve the unit commitment problem. By using the proposed approach, any stagnation during NN training can be prevented. Besides, the proposed NN converges into a global minimum for a given range of space. The NN would not be trapped into an undesirable local minimum as with the case of backpropagation algorithms. Also, the evolving NN with weight or topology options have lower training error when compared to NN with random initial weights.