Predictions of wind speed and wind power potential using artificial neural networks

In this study, wind power density in the southern and southwestern region of Turkey was determined by using the Weibull and Rayleigh probability density functions and the Wind Atlas Analysis and Application Program (WAsP). Moreover, wind speed of a target station was predicted using wind speeds of reference stations and other meteorological parameters with the artificial neural networks (ANNs) method. ANNs were also applied to predict the long term monthly temperature and rainfall at any specific point of Turkey based on the use of the neighboring measuring stations data. The wind data taken with an interval of one hour were measured by the General Directorate of Electrical Power Resources Survey Administration (EIE) at 9 measuring stations such as Akhisar, Belen, Datça, Gelendost, Söke, Gökçeada, Foca, Gelibolu and Bababurnu. The long-term wind data, containing hourly wind speeds and directions, cover the period from the year 1998 to 2003. Wind speeds of reference stations and other meteorological parameters were used as an input neurons in the ANN architecture, while the target station wind speed was used as an output neuron in the ANN architecture. In MATLAB, different artificial neural network learning algorithms have been used to establish the ANN models. The results obtained with these models were compared with the actual data. Comparisons have shown that there was a good agreement between predicted and measured data. Errors obtained in these models are well within acceptable limits. Keywords: Artificial neural networks, Meteorological parameters, Reference station, Target station, Wind speed.