A Dynamic Cluster-Aware Wind Power Forecasting Framework for Sustainable Renewable Energy Integration

Wind power plays an increasingly important role in the global energy transition. However, its power output exhibits significant uncertainty due to rapid variations in meteorological conditions. Existing forecasting methods still face challenges in large-scale wind farm cluster scenarios. In such cases, spatial heterogeneity and temporal asynchrony among wind farms cannot be fully characterized, which limits the overall prediction accuracy. To address these issues, this study proposes a novel hierarchical and adaptive collaborative forecasting framework for wind farm clusters by integrating meteorology-driven dynamic clustering with deep learning-based prediction. First, a multidimensional feature system is constructed by jointly considering static wind farm attributes and dynamic meteorological variation trends. Based on a sliding time window, real-time meteorological similarity among wind turbines is evaluated, allowing meteorological data to actively drive the formation and continuous evolution of adaptive subcluster structures. Subsequently, a deep learning model is developed to perform short-term power forecasting at the dynamic subcluster level. This approach enables the framework to flexibly capture spatio-temporal heterogeneity while maintaining robust prediction capability under varying cluster structures. Experimental results based on real-world wind farm cluster data demonstrate that the proposed method achieves superior accuracy and robustness compared with conventional whole-farm forecasting and static clustering approaches. The proposed framework enhances forecasting reliability, thereby supporting renewable energy integration and sustainable low-carbon power systems.