Probability Density Forecasting of Wind Power Based on Transformer Network with Expectile Regression and Kernel Density Estimation

A comprehensive and accurate wind power forecast assists in reducing the operational risk of generation, improves safety stability system, maintains balance generation. Herein, a hybrid probabilistic density forecasting approach based on transformer network combined with expectile regression kernel estimation (Transformer-ER-KDE) is methodically established. The prediction results various levels are exploited as input estimation, optimal bandwidth achieved by employing leave-one-out cross-validation to arrive at complete probability curve. In order more assess predicted results, two sets evaluation criteria constructed, including metrics for point interval prediction. generation dataset from official website Belgian grid company Elia employed validate proposed approach. experimental reveal that Transformer-ER-KDE method outperforms mainstream recurrent neural models terms error. Further, suggested capable accurately capturing uncertainty through construction intervals curves.