摘要
现有方法忽略了分布式光伏时空关联性的动态变化,难以有效利用时空特征信息提升功率预测精度。考虑到分布式光伏出力的强波动特性与分布式光伏集群强时空关联性,提出一种基于时空关联动态表征与图卷积网络建模的分布式光伏超短期功率预测方法。首先,将各分布式光伏复杂出力序列分解为相对简单、波动较小的多个波动模态分量。然后,考虑到分布式光伏场站间时空关联性动态变化,利用数据驱动方式提取各类波动模态分量表征的各分布式光伏间深层次时空关联关系,并构建由各波动模态分量表征的多个动态时空图结构。在此基础上,建立考虑动态时空关联性的图卷积预测模型,针对不同模态下出力子序列分别预测,而后重构得到各场站功率进而获取区域分布式光伏总功率。最后,基于真实分布式光伏出力数据验证了所提方法的优越性。
Accurate distributed photovoltaic power forecasting provides important support for the safe and stable operation of active distribution networks.However,existing methods neglect the dynamic changes in the spatio-temporal correlations of distributed photovoltaics,making it difficult to effectively utilize spatio-temporal feature information to improve power forecasting accuracy.Considering the strong fluctuation characteristics of distributed photovoltaic output and the strong spatio-temporal correlation of distributed photovoltaic clusters,this paper proposes an ultra-short-term power forecasting method of distributed photovoltaics based on dynamic spatio-temporal correlation characterization and graph convolutional network modeling.First,the complex output sequences of each distributed photovoltaic are decomposed into multiple fluctuation mode components which are relatively simple and have smaller fluctuations.Then,considering the dynamic change of the spatio-temporal correlation between distributed photovoltaic sites,a data-driven method is used to extract the deep spatio-temporal correlation among distributed photovoltaic sites represented by various fluctuation mode components,and multiple dynamic spatio-temporal graph structures represented by each fluctuation mode component are constructed.On this basis,a graph convolutional forecasting model considering the dynamic spatiotemporal correlation is established.The output sub-sequences in different modes are forecasted,respectively,and then the power of each station is reconstructed to obtain the total regional distributed photovoltaic power.Finally,the superiority of the proposed method is verified based on real distributed photovoltaic output data.
作者
王玉庆
徐飞
刘志坚
甄钊
王飞
WANG Yuqing;XU Fei;LIU Zhijian;ZHEN Zhao;WANG Fei(Department of Electrical Engineering,North China Electric Power University,Baoding 071003,China;State Key Laboratory of Power System Operation and Control(Tsinghua University),Beijing 100084,China;Department of Power Engineering,North China Electric Power University,Baoding 071003,China;State Key Laboratory of Alternate Electrical Power System with Renewable Energy Sources(North China Electric Power University),Beijing 102206,China;Hebei Key Laboratory of Distributed Energy Storage and Microgrid(North China Electric Power University),Baoding 071003,China)
出处
《电力系统自动化》
EI
CSCD
北大核心
2023年第20期72-82,共11页
Automation of Electric Power Systems
基金
国家自然科学基金资助项目(52007092)
国家重点研发计划资助项目(2022YFB2403000)。
关键词
分布式光伏
超短期功率预测
波动性
时空关联性
分解
动态关联性
distributed photovoltaic
ultra-short-term power forecasting
volatility
spatio-temporal correlation
decomposition
dynamiccorrelation
