基于分位数时空图神经网络的分布式光伏聚合不确定性表征方法  

A Distributed Photovoltaic Aggregation Uncertainty CharacterizationMethod Based on Quantile Spatial-temporal Graph Neural Network

作　　者：曾锃 肖茂然 夏元轶 张震 殷俊杰 余益团 张瑞 窦春霞[2] ZENG Zeng;XIAO Mao-ran;XIA Yuan-yi;ZHANG Zhen;YIN Jun-jie;YU Yi-tuan;ZHANG Rui;DOU Chun-xia(Information&Telecommunication Branch,State Grid Jiangsu Electric Power Co.,Ltd.,Nanjing 210024,China;Institute of Carbon Neutralization Advanced Technology,Nanjing University of Posts and Telecommunications,Nanjing 210023,China)

机构地区：[1]国网江苏省电力有限公司信息通信分公司,南京210024 [2]南京邮电大学碳中和先进技术研究院,南京210023

出　　处：《科学技术与工程》2024年第34期14726-14733,共8页Science Technology and Engineering

基　　金：国网江苏省电力有限公司科技项目(5210ED230009)。

摘　　要：聚合分布式光伏发电功率的不确定性量化对于电力系统的决策控制是至关重要的。提出了一种数据驱动的分布式光伏发电功率不确定性聚合量化方法。首先,考虑分布式光伏的时空分布特性,提出了一个基于时空图卷积神经网络的不确定性聚合模型,该模型可以有效地挖掘出数据的时空特征;其次,将时空图卷积神经网络与分位数回归模型相结合,在不需要假设聚合不确定性的概率分布的情况下,实现了聚合不确定性的量化与表征;另外考虑到分位数的交叉问题,设计了一种分位数映射方法,规避了聚合不确定性量化结果的交叉问题。最后,基于IEEE-33节点系统进行了验证,结果验证了所提出方法的有效性。The quantification of uncertainty in aggregated distributed photovoltaic power generation is crucial for decision control in power systems.A data-driven aggregation quantification method for power uncertainty in distributed photovoltaic power generation was proposed.Firstly,considering the spatial-temporal distribution characteristics of distributed photovoltaics,an uncertainty aggregation model based on spatial-temporal graph convolutional neural networks was proposed,which can effectively mine the spatial-temporal characteristics of data.Secondly,aggregated uncertainty was quantified and characterized without assuming its probability distribution through the combination of spatial-temporal graph convolutional neural networks with quantile regression models.In addition,considering the cross problem of quantiles,a quantile mapping method to avoid the cross problem of quantized results with aggregated uncertainty was proposed.Finally,the effectiveness of the proposed method was verified based on the IEEE-33 bus system.

关 键 词：分布式光伏 不确定性 图神经网络 分位数回归 数据驱动 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]