Short-Term Wind Power Forecasting under Data Loss Conditions: A Case Study in Vietnam

The effective utilization of wind energy strongly relies on accurate power forecasting, in which short-term prediction plays a crucial role in grid operation.In practice, measurement data are often incomplete due to data loss.Existing studies typically address this issue by either focusing on data imputation, developing forecasting models based on deep learning or machine learning (DL/ML), or integrating numerical weather prediction (NWP) models and data security.However, only a limited number of approaches effectively combine robust imputation with powerful time-series forecasting models, while also ensuring comprehensive evaluation under various data loss scenarios and maintaining both interpretability and practical applicability.Moreover, regional characteristics significantly influence forecasting methods and wind power management strategies.To address these challenges, this paper proposes a hybrid XGBoost-GRU model for short-term wind power forecasting, with a case study in Southern Vietnam.Experimental results demonstrate that the proposed model outperforms the baseline GRU model by achieving higher predictive accuracy and maintaining stable performance under different data conditions.