Prediction of Wind Speed by Using Machine Learning

Due to the depletion of fossil fuel resources and environmental concerns caused by traditional fuel systems in recent years, the share of renewable energy sources in current energy production has been increasing. Among these energy sources, wind and solar energy stand out compared to other sources. Wind energy is a clean, sustainable and low-cost energy source. Wind and solar energies vary considerably according to the stochastic environment of meteorological conditions. Solar and wind energy variability and uncontrollability lead to power quality, generation-consumption balance and reliability problems of solar and wind energy systems. For this reason, it is important to know and predict the wind speed and solar radiation characteristics of the regions where the systems are installed. In this study, meteorological data of Antalya Serik Region were analyzed using statistical methods and wavelet transform. Thus, the potentials of wind and solar energies in the study area and large and small-scale events affecting these potentials were determined. In addition, a short-term estimation study was made for wind intensity and solar radiation using the time series of meteorological data. Besides SARMA, SARMAX and NAR models, Wavelet-NARX, SARMAX-NAR and NAR-SARMAX hybrid models are employed. Hybrid models are successfully produced better results than component forecasts.