Eventos Anais de eventos
ENCIT 2020
18th Brazilian Congress of Thermal Sciences and Engineering
Methodology for Estimating Daily Solar Radiation Using Neural Networks
Submission Author:
Vinícius Laguardia de Castro Oliveira , MG
Co-Authors:
Vinícius Laguardia de Castro Oliveira, Ronei Avelar Soares, Mateus de Medeiros Ramos, Felipe Venancio Mitkiewicz Silva, Cristiana Brasil Maia
Presenter: Vinícius Laguardia de Castro Oliveira
doi://10.26678/ABCM.ENCIT2020.CIT20-0525
Abstract
This work aims to establish a methodology to estimate the incident global solar radiation combining an analytical method with artificial neural networks (ANN), using ten cities in the state of São Paulo as reference. The analytical method depends on meteorological parameters, among them the clearness index (Kt), which might not be available. Therefore, to overcome this issue, two different ANN were used to obtain the global solar radiation likewise the Kt, outputs to the networks. For the first network (Network I), day, latitude, longitude and altitude of the location were used as input parameters. For the second one (Network II), in addition to the first network’s variables, the average local daily temperature and humidity were also used. The results indicated that the approximation factors of Network I were similar to the performance of the analytical method, while Network II had higher approximation factors. For cities known by the networks, an average approximation factor of 53% was obtained for Network I and 80% for Network II. For nearby cities not present in the training database, the average approximation factor was 57% for Network I and 54% for Network II. Even with limitations, it was possible to estimate global solar radiation using ANN.
Keywords
Solar Energy, solar radiation, neural networks, Artificial Intelligence
DOWNLOAD PDF VIEW PRESENTATION
