Prediction of wind speed using machine learning and deep learning

In today's world, the energy sector is undergoing significant changes due to the acceleration of technological advancements and the impact of environmental factors. In this process of transformation, renewable energy sources, particularly wind energy, are becoming increasingly important. Wind energy plays a crucial role in sustainable energy portfolios, and wind speed forecasting is a critical factor in improving the performance of wind energy farms and optimizing energy production processes. This research focuses on wind speed prediction in the provinces of Bozcaada, Gökçeada, Ezine, and Ayvacık in Çanakkale province, aiming to effectively assess the wind energy potential. Accurate prediction of wind speed has the potential to increase the efficiency of wind energy farms, minimizing fluctuations in energy production. This study examines the performance of various machine learning and deep learning algorithms to enhance wind speed prediction. Among the applied algorithms are Long Short-Term Memory (LSTM), Extreme Gradient Boosting (XGBoost), k-Nearest Neighbors, Advanced Artificial Neural Networks, and Convolutional Neural Networks. The results generally indicate that LSTM and XGBoost models achieve high accuracy in wind speed prediction compared to other models. Specifically, the XGBoost model demonstrates the highest accuracy with an R-squared value of 0,9438 in Ayvacık, while showing similar high R-squared and low error metrics in Bozcada and Ezine. Conversely, the LSTM model achieves high accuracy with an R-squared value of 0,964 in Gökçeada, although it exhibits disadvantages in certain error metrics like MAPE. These findings suggest that LSTM generally provides high accuracy in wind speed prediction, with other models also proving suitable in specific scenarios. The study emphasizes the potential to improve prediction model accuracy through enhanced data quality and the use of different algorithms. Future research should focus on further enhancing prediction models and improving operational efficiency in the wind energy sector.