Seasonal Flow Forecast for the Orós Dam (Ceará, Brazil) Using Neural Networks and the Resampling Technique of K-neighbors

• Main Authors: Araújo, Carla Beatriz Costa De, Filho, Francisco De Assis De Souza, Luiz Martins De Araújo Júnior, Cleiton Da Silva Silveira
• Format: Dataset
• Language: unknown
• Published: SciELO journals 2021
• Subjects: 40107 Meteorology; FOS Earth and related environmental sciences; Nash; Pacific; North Atlantic
• Online Access: https://dx.doi.org/10.6084/m9.figshare.14282000; https://scielo.figshare.com/articles/dataset/Seasonal_Flow_Forecast_for_the_Or_s_Dam_Cear_Brazil_Using_Neural_Networks_and_the_Resampling_Technique_of_K-neighbors/14282000

Abstract The objective of this work is to perform a comparative flow forecast for the Orós basin (Ceará, Brazil) using artificial neural networks (RNA) and k-neighbors re-sampling technique. The models were developed from the historical series of 100 years of hydrometeorological data (sea surface temperature and flows). Both use as predictors the temperatures of the North Atlantic, South Atlantic and Equatorial Pacific oceans, and forecast July in the next year’s rainy season (January to June). The k-neighbors model was elaborated from the identification of the closest neighbor years for the resampling of the approximation, since the RNA model was formulated from the synaptic and bias weights obtained in the training phase of the network. The Nash-Suttcliffe (E) efficiency coefficient, the coefficient of determination (R²), the Taylor diagram (2001) and the coefficient of determination (R²) were used for the validation step. maximum likelihood ratio. For all comparative variables, the neural model presented better values, indicating that this represents more efficiently the behavior of the flows to the reservoir.