Sergio Velázquez
University of Las Palmas de Gran Canaria, Las Palmas, Canary Islands, Spain
José A. Carta
University of Las Palmas de Gran Canaria, Las Palmas, Canary Islands, Spain
José Matías
University of Vigo, Vigo, Spain
Ladda ner artikel
http://dx.doi.org/10.3384/ecp11057834Ingår i: World Renewable Energy Congress - Sweden; 8-13 May; 2011; Linköping; Sweden
Linköping Electronic Conference Proceedings 57:12, s. 834-841
Publicerad: 2011-11-03
ISBN: 978-91-7393-070-3
ISSN: 1650-3686 (tryckt), 1650-3740 (online)
In order to estimate the amount of electricity that can be produced by a potential wind farm it is important to know how the wind resource performs at the site where it is to be installed. Of fundamental importance in an analysis of the wind resource is the wind speed parameter. Understanding how this parameter behaves over periods of time that cover ten or more years (long-term) is vital for an accurate estimation that will span the working life of the wind installation. However; in most cases there is insufficient data available about the candidate site to enable a long-term study.
In this work; the long-term wind power density at a candidate site is estimated through the use of a Measure-Correlate-Predict (MCP) algorithm and an Artificial Neural Network model (ANN). To evaluate the accuracy of the estimations different metrics are used; with a comparison of the results obtained for each of them.
The mean hourly wind speeds and directions obtained from twenty-two weather stations located on different islands in the Canary Archipelago (Spain) are used for this study.
Among the conclusions that are reached is that the use of one or another metric (or combination of metrics) in the wind power density estimation process can lead to differing interpretations and/or conclusions. For this reason; it is important that the most appropriate metric (or set of metrics) is chosen at each moment for the study that is being carried out.
Wind Power Density; Short-Term estimation; Long-Term estimation; Artificial Neural Networks; Measure Correlate Predict
Inga referenser tillgängliga
