An ensemble method for short-term wind power prediction considering error correction strategy

With a high proportion of renewable energy injected into the power grid, the accurate wind power prediction of wind farms is key to improving power quality and ensuring the stable operation of the power grid. In this paper, an ensemble learning prediction model considering the rolling error correction strategy used for short-term wind power prediction is proposed. Firstly, the ensemble learning prediction model integrating multiple Gradient Boosting Trees (GBDTs) based on Bayesian optimization is established, and the characteristics of prediction errors are analyzed. To further improve the prediction accuracy, a daily rolling error correction strategy based on the Swing Window segmentation method, Spearman correlation coefficient and Quantile Regression is established to obtain the optimal compensation amount. Finally, the corrected short-term wind power prediction results are obtained. The two-year dataset collected from a regional wind farm in China is used as the benchmark test. Different prediction models and error correction strategies are compared, and the effectiveness of the proposed method is comprehensively evaluated. The results from different seasons show that the proposed prediction method has good stability, and the proposed error correction strategy has good generalization ability.