基于改进Informer模型的风电预测  

Wind power prediction based on improved Informer model

作　　者：张振亚 杜春梅[1] ZHANG Zhenya;DU Chunmei(Hebei University of Architecture,Zhangjiakou,Hebei 075132,China)

机构地区：[1]河北建筑工程学院,河北张家口075132

出　　处：《计算机应用文摘》2024年第18期62-65,69,共5页

摘　　要：针对风电预测中的长时序和精确预测困难等问题,风电预测在平衡能源供需方面具有重要意义。文章提出了一种基于Informer网络结合空洞卷积的多输入多输出预测模型--CCN-Informer。该模型通过改进输入序列的全局信息提取能力,从而提高预测准确性。实验中,空洞卷积用于特征提取,进一步提高了预测精度。模型基于实际风电场数据集进行了训练和测试。大量实验结果表明,所提方法显著提高了平均风电功率的预测精度。Wind power forecasting is of great significance in balancing energy supply and demand,as it addresses the challenges of long-term and accurate forecasting in wind power forecasting.The article proposes a multi input multi output prediction model based on Informer network combined with dilated convolution—CCN—Informer.This model improves prediction accuracy by enhancing the global information extraction capability of the input sequence.In the experiment,dilated convolution was used for feature extraction,further improving the prediction accuracy.The model was trained and tested based on an actual wind farm dataset.Numerous experimental results have shown that the proposed method significantly improves the prediction accuracy of average wind power.

关 键 词：风电功率预测 长序列输入预测 空洞卷积 注意力机制 

分 类 号：TM614[电气工程—电力系统及自动化]