中国电机工程学报2025，Vol.45Issue(12)：4693-4706,中插13,15.DOI:10.13334/j.0258-8013.pcsee.232642
基于尾流关联的动态超图风电功率超短期预测方法
Dynamic Hypergraph Wind Power Ultra Short Term Prediction Method Based on Wake Correlation
摘要
Abstract
Accurate wind power prediction is of great practical significance for the safe and stable operation of the power system.Due to factors such as wake effects,there are complex correlation characteristics between wind turbines in a wind farm.Existing research has overlooked the dynamic changes in spatiotemporal correlation characteristics,and traditional graph based binary relationship representation methods are also difficult to accurately characterize the complex and diverse spatiotemporal relationships between wind turbines,resulting in difficulty in accurately capturing the spatiotemporal characteristics between fans.Considering the poor solvability of deep learning models and the impact of wake effects on wind power,this paper proposes a dynamic hypergraph wind power ultra short-term prediction method based on wake correlation.First,each wind turbine is regarded as a node,with the historical power of each wind turbine as the feature input,and the spatial position and complex relationships of the wind turbine as the hyperedges.A dynamic hypergraph representation structure of the wind turbine is constructed along the time dimension.Then,combining the wind direction data and wind turbine information at each moment,based on the Jensen wake model principle,a dynamic hypergraph based on wake correlation is constructed in the form of ray method.On this basis,for the special data structure of dynamic hypergraphs,a spatiotemporal aggregation feature extraction module based on dynamic hypergraph convolution and a spatiotemporal feature fitting module based on bidirectional long short-term memory(BiLSTM)are constructed to extract dynamic spatiotemporal features and achieve accurate prediction.Finally,experimental analysis is conducted based on real wind power data to verify the superiority of this method from multiple dimensions.关键词
风功率预测/超图/动态特性/时空特性/尾流效应/动态超图卷积Key words
wind power prediction/hypergraph/dynamic characteristics/spatio-temporal characteristics/wake effect/dynamic hypergraph convolution分类
动力与电气工程引用本文复制引用
钟吴君,李培强,涂春鸣..基于尾流关联的动态超图风电功率超短期预测方法[J].中国电机工程学报,2025,45(12):4693-4706,中插13,15.基金项目
国家重点研发计划项目(2021YFB2601504) （2021YFB2601504）
国家自然科学基金项目(52377097). National Key R&D Program of China(2021YFB2601504) （52377097）
Project Supported by National Natural Science Foundation of China(52377097). （52377097）
