STOCHASTIC MODELING OF MONTHLY RIVER FLOWS BY SELF-ORGANIZING MAPS
Keywords:Artificial neural networks, method of fragments, water resources
AbstractExtreme hydrological conditions and increasing water demands observed in semiarid Brazil have generated conflicts regarding to the best use of existing water resources. Synthetic generation models of river flows are often used as support for the definition of water system operating rules, which allow the establishment of rationing policies before water scarcity spells. This work aims at verifying the applicability of models based on self-organizing maps (SOM) for stochastic modeling of monthly river flows. The basic principle of the study consisted of using SOM models in order to define the deterministic component of river flow series and a density probability function (stochastic component) to represent the resulting residuals. During calibration of all networks, values of NASH were above 0.9989 for the applications. The results were promising, indicating that the established models are capable of producing synthetic series of inflows with excellent performance.
Download data is not yet available.