基于动态图注意力的风电场组合预测模型

Wind Farm Combination Forecasting Model Based on Dynamic Graph Attention

为了实现风电场用能管理的高效调度,充分提取多站点间时空特征的潜在联系,提出一种基于动态图卷积和图注意力的多站点短期风电功率时空组合预测模型。使用图卷积实现多站点间时序特征的邻居聚合,并使用图注意力机制加强其对空间特征的提取能力。同时,针对传统模型无法处理图节点关联性实时变化的问题,先在图卷积过程中依据站点间的相关系数和距离动态构建邻接矩阵,再使用门控循环单元处理动态图卷积输出的上下文信息,最后完成风电功率预测。实验结果表明,所提出的组合模型在预测精度、稳定性和多步预测性能方面均最优。

In order to realize efficient scheduling of wind farm energy use management and fully extract the potential relationship between spa⁃tial and temporal characteristics of multiple sites,a multi-site short-term wind power spatio-temporal combination prediction model based on dynamic graph convolution and graph attention is proposed.Firstly,graph convolution is used to realize neighbor aggregation of temporal fea⁃tures among multiple sites,and graph attention mechanism is used to enhance its ability to extract spatial features.At the same time,in view of the problem that the traditional model cannot handle the real-time changes of the graph node correlation,firstly,the adjacency matrix is dy⁃namically constructed according to the correlation coefficient and distance between the sites during the graph convolution process;secondly,the gated cycle unit is used to process the context information of the output of the dynamic graph convolution;finally,the wind power predic⁃tion is completed.The experimental results show that the proposed combined model is optimal in terms of prediction accuracy,stability and multi-step prediction performance.