The effect of meteorological factors on wind power plants

Wind energy is a renewable and sustainable energy source which gains inreasingly more importance around the world. However, the instability of wind speed creates uncertainties in wind energy production and this leads to significant problems in terms of grid integration and energy planning. The aim of this study is to comparatively examine the predictive performances of K-Nearest Neighbors (KNN), Extreme Gradient Boosting (XGBoost), Random Forest (RF) and Light Gradient Boosting Machine (LightGBM) algorithms in wind power generation using meteorological data. Historical generation data of a wind power plant were used in the study. The input variables used in the forecast models are wind speed, blade angle, temperature, rotor speed, and nacelle direction; The output variable is wind power generation. In order to increase the prediction performance of each model, hyperparameter optimization was performed using the Optuna library based on the Bayesian optimization method, and at this stage, the 5-fold cross validation technique was applied. Prediction accuracy of models was evaluated with statistical performance criteria such as Root Mean Square Error (RMSE), Mean Absolute Error (MAE) and Coefficient of Determination (R²).