关键词: 转折性天气, 日前风功率预测, 逆向云变换, 误差类型聚类, 修正权重优化, 核密度估计

Abstract:

[Objectives] In recent years, transitional weather has occurred frequently, and the randomness and volatility of wind power generation have intensified. This paper proposes a robust error correction strategy for high-quality day-ahead wind power interval prediction to address the problem that it is difficult for existing day-ahead wind power prediction solutions to take into account the interval coverage rate and interval width of the power prediction interval. [Methods] First, kernel-based fuzzy c means (KFCM) clustering algorithm combined with multiple backward cloud transformation based on sampling with replacement (MBCT-SR) is proposed to accurately cluster error types in complex transitional weather. Then, a point prediction model temporal convolutional network transformer (TCN-transformer) and an improved kernel density estimation day-ahead wind power interval prediction method are established, which improves the fitting accuracy of power interval. Finally, improved multi objective dung beetle optimizer (IMODBO) algorithm is designed to perform robust error correction on different clusters, which effectively reduces the day-ahead wind power prediction interval width while ensuring high interval coverage. [Results] At the 95% confidence level, the PICP increased by a maximum of 5.884%, and the PINAW increased by a maximum of 35.01%. The effectiveness of the proposed method in this paper has been validated. [Conclusions] This approach significantly improves the fitting accuracy of error distribution, avoids the issue of poor interval quality caused by local over-correction or under-correction, greatly enhances the efficiency of the algorithm search, and achieves higher-quality wind power interval prediction.

Key words:

"> transitional weather, day-ahead wind power prediction, backward cloud transformation, error type clustering, correction weight optimization, kernel density estimation