Identifying teleconnection predictors for medium-to-long term prediction of low wind power events

As wind power integration into the grid continues to increase, accurate medium-to-long term forecasting of low wind power events (LWPEs) becomes critical for ensuring grid stability and effective scheduling. Understanding the underlying mechanisms of LWPEs, particularly those driven by atmospheric intra-seasonal oscillations, is essential for improving forecast accuracy. This study proposes a method to identify LWPEs and attribute their teleconnection drivers based on causal analysis using the LightGBM algorithm. First, an LWPE identification model is developed using LightGBM and reanalysis data. A sample set of LWPEs is constructed from 20 years of historical reanalysis data. Then, causal inference is applied to identify key predictors with strong causal relationships, facilitating the attribution of teleconnection impact factors. The proposed method is validated using real-world data from a wind farm cluster in northern China.