Home // International Journal On Advances in Networks and Services, volume 5, numbers 1 and 2, 2012 // View article
Authors:
Aubai Alkhatib
Siegfried Heier
Melih Kurt
Keywords: Artificial Neural Networks; Wind Speed; Mean root square error; Training functions; Short term prediction; Long term prediction
Abstract:
Wind speed forecasting is an essential prerequisite for the planning, operation, and maintenance works associated with wind energy engineering. This paper attempts to forecast fluctuations based only on observed wind data using the data-driven artificial neural network approach. Wind fluctuations with varying lead times ranging from a half year to a full year are predicted at Al-Hijana, Syria with the pre-preparation for the available data. Two layers of feed-forward back-propagation networks were used along with the conjugate gradient algorithm and other tested training functions. The results show that artificial neural network models perform extremely well as low values of errors resulting between the measured and predicted data are obtained. The present work contributes to previous work in the field of wind energy independent power producer market and may be of significant value to Syria, considering that the country is currently in the process of transitioning into a free energy market. It is likely that this modeling approach will become a useful tool to enable power producer companies to better forecast or supplement wind speed data. Two main types of wind speed prediction tool is discussed in this paper. One prediction tool with no time shift and the other prediction tool with time shift, where in the second type two different time periods were used to show the different between long term prediction and short term prediction.
Pages: 149 to 158
Copyright: Copyright (c) to authors, 2012. Used with permission.
Publication date: June 30, 2012
Published in: journal
ISSN: 1942-2644