A Robust Error Correction Strategy for High-Quality Day-Ahead Wind Power Interval Prediction in Transitional Weather

Weather transitions can cause rapid fluctuations in wind power and complicate the construction of prediction intervals. This study presents a robust error-correction framework that balances interval coverage and width. Multiple backward cloud transformation based on sampling with replacement (MBCT-SR) is combined with kernel-based fuzzy c-means (KFCM) clustering to represent and group error-related feature patterns. A temporal convolutional network and Transformer model (TCN-Transformer) predicts power and error sequences, and adaptive-bandwidth kernel density estimation (ABKDE) is used to construct prediction intervals. An improved multi-objective dung beetle optimizer (IMODBO) determines cluster-specific error-correction weights using coverage and width as competing objectives. The framework is evaluated using wind power data from China Longyuan Power Group Corporation Limited. The reported examples show reductions in interval width, with changes in coverage that depend on the test day and nominal coverage level.