Stochastic Fuzzy Neural Network: Case Study of Optimal Reservoir Operation

Abstract

Reservoirs play an important role within the water resources management framework. Among the available techniques for reservoir optimal operation, the most well known one is stochastic dynamic programming (SDP). In recent years, artificial intelligence techniques such as genetic algorithms (GA) and artificial neural networks have arisen as an alternative to overcome some of the limitations of traditional methods. Some of these limitations are related to the difficulty in combining SDP with other simulation and prediction models, the curse of dimensionality due to the increase in the number of decision and state variables, and the error resulting from the rough discretization of these variables. Here, we introduce a new approach for system optimization and operation, named stochastic fuzzy neural network (SFNN), which can be defined as a neuro-fuzzy system that is stochastically trained (optimized) by a GA model to represent the system operational strategy. Moreover, to deal with imprecision originated by the discretization of inflow intervals (events) in calculating the transition probabilities, we applied the method based on the conditional probability of a fuzzy event. To investigate the applicability and efficiency of the proposed method, the Barra Bonita Reservoir, Brazil, is stochastically optimized and operated. The results found by the SFNN method were compared to the results of other available dynamic programming models, showing success in developing and applying the proposed method to optimal reservoir operation.