Utilização de técnicas de inteligência computacional na predição de dados meteorológicos

Abstract

This paper presents a computational approach for the one step ahead prediction in meteorological data series belonging to regions of Paty do Alferes and Paracambi, both located in the state of Rio de Janeiro (RJ). Two models of Artificial Neural Networks (ANN) were used: Multilayer Perceptron (MLP) and Radial Basis Function (RBF). To confirm the performance of the models were performed the prediction of hourly and monthly variables. These data were compared with results obtained by Multiple Linear Regression models (RLM), but also with the data registered by meteorological stations and analyzed by statistical techniques. It showed a favorable result, reaching between 91% and 96% of correct predictions for all cases. Moreover, the predictions also showed a strong linear correlation with the actual data, keeping it from 0.61 to 0.94. As a result, the RNAs can stand out as a strong tool for prediction of meteorological data analyzed.