Machine Learning-Based Wind Energy Forecasting Using Weather Parameters: The Example of Yalova

In this study, various machine learning algorithms were evaluated for estimating wind energy production using hourly meteorological data of Yalova province in 2018. The input parameters were input parameters of weather parameters such as temperature, relative humidity, air pressure, wind direction, and wind speed. In the analysis performed on a total of 50530 data points, methods such as Gradient Boosting (GB), Random Forests (RF), k-nearest neighbor (kNN), and Stochastic gradient descent (GBD) were compared. Model performances were evaluated according to Mean Absolute Error (MAE), Mean Square Error (MSE), Root Mean Square Error (RMSE), MAPE, and R2 criteria. According to the results, the best-performing algorithm was RF with an MSE value of 0.039, RMSE value of 0.197, MAE value of 0.081, MAPE value of 0.377, and R² score of 0.961. On the other hand, the SGD model showed the lowest performance with an MSE value of 0.175, RMSE value of 0.418, MAE value of 0.303, MAPE value of 0.581, and R² score of 0.822. These findings show that machine learning models, supported by selecting the correct weather parameters, can provide high accuracy in estimating wind energy production and contribute to energy management policies in this direction.