An improved ramp events forecasting of wind generating resources using ensemble learning of numerical weather prediction: The case of Jeju Island's wind farms

Wind energy is an extensively used renewable energy resource. However, its variability and intermittency challenge grid reliability and stability. Grid stability issues caused by ramp events can be mitigated using accurate ramp forecasting methods. In this study, focusing wind farms in Jeju Island, we developed a ramp event forecasting model based on a Light Gradient Boosting Machine (LGBM) and incorporated the output of an ensemble numerical weather prediction (NWP) model to account for prediction uncertainties. Ensemble NWP is a specialized forecasting method that considers the uncertainties inherent in weather prediction. Multiple wind speed scenarios from the ensemble NWP were applied to the wind power forecasting model as input data. The proposed model provided probabilistic ramp event forecasts. To validate the effectiveness of the proposed methodology, the model results were applied to the empirical system of Jeju Island. The results of the model can aid stakeholders in deciding ramp rate control methods, such as storage-system usage, and thus, contributing to the increasing the integration of wind energy.