ESTUDO PARA O DESENVOLVIMENTO DE UM PREVISOR DE DESCARGAS ELÉTRICAS ATMOSFÉRICAS APLICADO À REGIÃO COSTEIRA DO ESTADO DO RIO DE JANEIRO

AUTOR(ES)
DATA DE PUBLICAÇÃO

2005

RESUMO

The atmospheric dynamics evidently is very complex. There are many macro and micron scales processes and meteorological variables involved in the atmospheric physicalphenomena. The storms with electrical discharges are distinguished, among these phenomena, by the damage consequences to the human beings, directly or indirectly. Many researchers have pursued the possibility of forecasting the occurrence of a storm with electrical discharges, principally in the last three decades. However, there are not improvements in forecast performance, mainly due to phenomenon complexity. The main objective of the present dissertation was to accomplish a study to determine the viability or not of constructing a forecast system of atmospheric electrical discharges from artificial intelligence techniques, specifically artificial neural networks (NN). The base of the system was constituted of numerical simulations results of the atmospheric dynamics obtained from the mesoscale model MM5. It was identified meteorological variables (outputs of MM5) that would have some correlation with the electrical discharges. These variables act as input in the NN, which generate the forecast, consisting in the number of electrical discharges that will reach the ground some posterior time to the relative time of the simulated atmospheric fields. The region chosen for this study includes the state of Rio de Janeiro, the south of Espírito Santo, the southeast of Minas Gerais and the Atlantic Ocean. Besides possessing a detection system of electrical discharges, this region contains the most important concentration of oil platforms of Brazil, being strategic to better know the atmospheric behavior in this place. Before developing the forecast system based on NN, some comparative tests were made using the MM5 simulation results and GOES infrared imagery, in order to survey the model prognostic capability. The forecast system showed reasonable results, indicating that the NN application may be a promising way to the electrical discharge forecast. However, it is necessary a better investigation, mainly with relation to the accomplishment of others tests with a bigger set of electrical discharges real data.

ASSUNTO(S)

numerical simulation meteorologia sistema de previsão simulação numérica modelo de mesoescala mm5 redes neurais artificias forecast system mesoscale model mm5 artificial neural networks descargas elétricas atmosféricas atmospheric electrical discharges

ACESSO AO ARTIGO

Documentos Relacionados