Alternative optimization of multi-objective reservoirs with fuzzy optimum neural networks

Abstract

In the process of flood control operation of multi-objective reservoirs in flood season, it is necessary to select the best from feasible alternatives through comparison and filter. Because of the contradiction among objectives, which used to evaluate and select the satisfying alternative, the difficulty in assessing the weights of objectives, and the uncertainty of flood data, the selection is a multi-criteria and subjectivity problem under fuzzy environments. This paper applies fuzzy optimum neural networks to solve multi-criteria decision making problems and presents a new method to construct the training set for neural networks. First, the criterion set and its relative membership degree matrix are established for the evaluation of alternatives. Next, neural networks are trained with training set, which are constructed by the new method presented in this paper. And then substitute the candidate alternatives into the neural networks. The ranking of alternatives and the best one can be determined directly on the basis of the output of neural networks. The optimization process is simple and easy to use in practice. A case study shows that the method is reasonable and effective.