Modeli̇ng short, medi̇um, and long-term energy producti̇on of wi̇nd power plants usi̇ng machi̇ne learni̇ng algori̇thms

Wind energy stands out as a prominent alternative among renewable energy sources due to its sustainability and environmentally friendly characteristics. However, high initial investment costs, the requirement for fixed infrastructure, and the limited availability of geographically suitable areas hinder its effective utilization. Moreover, the inherently variable and chaotic nature of wind makes accurate production forecasting a major challenge, which in turn affects energy trading and grid planning. Therefore, reliable forecasting of wind energy production is of critical importance for operational and strategic decision-making processes in the energy sector. This study presents wind energy production forecasts at hourly, daily, and monthly resolutions for the Söke Çatalbük Wind Power Plant located in Aydın, Turkey. Using meteorological and production data collected between 2018 and 2022, five different machine learning algorithms were implemented: Artificial Neural Networks (ANN), Random Forest (RF), Extreme Gradient Boosting (XGBoost), k-Nearest Neighbors Regression (KNN), and Multi-Layer Perceptron (MLP ANN). The data were preprocessed using both Standard Scaling and Min-Max Scaling methods prior to model training. The prediction performance was evaluated using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and the Coefficient of Determination (R²). Among the models, KNN achieved the highest accuracy in hourly predictions (R² = 0.7871), MLP ANN performed best in daily forecasts (R² = 0.8523), while Random Forest yielded the best results for monthly predictions (R² = 0.9187). Although scaling methods influenced certain error metrics, no consistent performance difference was observed overall. This study contributes to the literature by examining the temporal impact of different algorithms and preprocessing techniques on wind energy production forecasting.