Modelo de previsões de vazões para o Nordeste setentrional utilizando análise das componentes principais

Abstract

The article presents the results of a model that contains predictions of seasonal and interannual flow to a system of reservoirs in northeastern North, using historical series from 1931 to 1998, with their respective securities. The methodology presented uses the analysis of principal components (PCA) system of 18 reservoirs and under these components is done a linear regression performed on the climatic indices. Using the parameters of the regression is applied the methodology of the k-nearest neighbors to calculate the probability distribution of flows of tributaries. The application of the k-neighbors is given to three reservoirs: Armando Ribeiro, Coremas and Oros. The values for the correlation between the values observed were lower than those obtained from the model to predict the reservoir Oros in isolation, letting space for improvement and better results.