“The research team proposed a robust error correction strategy, established a point prediction model in the time domain convolutional Transformer network, and designed an improved kernel density estimation method for predicting wind power intervals before the day, providing a solution to improve the accuracy of wind power prediction under transitional weather conditions.”