Redes neurais artificiais e relação ZR aplicadas à estimativa de chuva
Resumo
Resumo: Por meio deste trabalho apresenta-se a utilização de redes neurais artificiais, em particular as rede neurais com funções de base radial e perceptron de múltiplas camadas, além de uma terceira técnica que busca por uma nova relação ZR. O objetivo deste trabalho é comparar o desempenho das técnicas ao converter dados de refletividade de radar em taxa precipitação de chuva. As metodologias propostas são comparadas com a relação ZR de Marshall-Palmer. Todos os dados meteorológicos utilizados são provenientes do SIMEPAR através do radar meteorológico e uma rede de pluviômetros. Os resultados apresentam-se satisfatórios após alguns índices de comparação serem testados, como a raiz do erro médio quadrático que enquanto para a relação de Marshall-Palmer fornece o valor 6.139 para os dados de treinamento, a rede RBF retorna 2.789, a rede MLP 3.289 e a nova relação ZR 2.895. Abstract:Through this work we present the use of artificial neural networks, in particular the neural network with radial basis functions and multilayer perceptron, and a third technique that searches for a new ZR relationship. All with the aim of comparing their ability to convert radar reflectivity data in rainfall. It also compares the proposed methodologies with the Marshall-Palmer ZR relationship. All meteorological data is provided by SIMEPAR through weather radar and rain gauge network. The results are found satisfactory after some comparison indices were tested, as the root mean square error for the relationship of Marshall-Palmer gives the value of 6.139 for the training data, the RBF network returns 2.789, the MLP network 3.289 and the new ZR relationship 2.895.
Collections
- Teses & Dissertações [9979]