@Article{LimaMarPerLorHei:2016:FoSuSo,
author = "Lima, Francisco Jos{\'e} Lopes de and Martins, Fernando Ramos and
Pereira, Enio Bueno and Lorenz, Elke and Heinemann, Detlev",
affiliation = "{Instituto Nacional de Pesquisas Espaciais (INPE)} and
{Universidade Federal de S{\~a}o Paulo (UNIFESP)} and {Instituto
Nacional de Pesquisas Espaciais (INPE)} and {University of
Oldenburg} and {Universidade Federal de S{\~a}o Paulo
(UNIFESP)}",
title = "Forecast for surface solar irradiance at the Brazilian
Northeastern region using NWP model and artificial neural
networks",
journal = "Renewable Energy",
year = "2016",
volume = "87",
pages = "807--818",
month = "Mar.",
keywords = "Artificial neural network, Solar energy forecast, Solar
irradiance, WRF model.",
abstract = "There has been a growing demand on energy sector for short-term
predictions of energy resources to support the planning and
management of electricity generation and distribution systems. The
purpose of this work is establishing a methodology to produce
solar irradiation forecasts for the Brazilian Northeastern region
by using Weather Research and Forecasting Model (WRF) combined
with a statistical post-processing method. The 24 h solar
irradiance forecasts were obtained using the WRF model. In order
to reduce uncertainties, a cluster analysis technique was employed
to select areas presenting similar climate features. Comparison
analysis between WRF model outputs and observational data were
performed to evaluate the model skill in forecasting surface solar
irradiance. Next, model-derived short-term solar irradiance
forecasts from the WRF outputs were refined by using an artificial
neural networks (ANNs) technique. The output variables of the WRF
model representing the forecasted atmospheric conditions were used
as predictors by ANNs, adjusted to calculate the solar radiation
incident for the entire Brazilian Northeastern (NEB) (which was
divided into four homogeneous regions, defined by the Ward
method). The data used in this study was from rainy and dry
seasons between 2009 and 2011. Several predictors were tested to
adjust and simulate the ANNs. We found the best ANN architecture
and a group of 10 predictors, in which a deeper analyzes were
carried out, including performance evaluation for Fall and Spring
of 2011 (rainy and dry season in NEB, mainly in the northern
section). There was a significant improvement of the WRF model
forecasts when adjusted by the ANNs, yielding lower bias and RMSE,
and an increase in the correlation coefficient.",
doi = "10.1016/j.renene.2015.11.005",
url = "http://dx.doi.org/10.1016/j.renene.2015.11.005",
issn = "0960-1481",
language = "en",
targetfile = "Lima_forecast.pdf",
urlaccessdate = "27 abr. 2024"
}