基于卷积神经网络的含分布式光伏配电网静态等值  

Static equivalent of distribution network with distributed photovoltaic based on convolutional neural network

作　　者：范海雯 郝旭东 赵康 邢法财 蒋哲 李常刚[1] FAN Haiwen;HAO Xudong;ZHAO Kang;XING Facai;JIANG Zhe;LI Changgang(School of Electrical Engineering,Shandong University,Jinan 250061,Shandong,China;Power Grid Technology Center,State Grid Shandong Electric Power Research Institute,Jinan 250003,Shandong,China)

机构地区：[1]山东大学电气工程学院,山东济南250061 [2]国网山东省电力公司电力科学研究院,山东济南250003

出　　处：《山东大学学报（工学版）》2023年第4期140-148,156,共10页Journal of Shandong University（Engineering Science）

基　　金：国家自然科学基金资助项目(52177096);国网山东省电力公司科技项目(SGSDDKOOWJJS2100208)。

摘　　要：为提升含分布式光伏配电网静态等值的运行方式适应性,提出一种基于卷积神经网络(convolutional neural networks,CNN)的含分布式光伏配电网静态等值方法。考虑源荷不确定性及相关性,基于核密度估计和Copula函数生成光伏、负荷功率场景并计算配网潮流。针对各单一运行方式下的等值问题,构造含分布式光伏配电网的等值模型,采用粒子群优化(particle swarm optimization,PSO)辨识变压器和线路参数。为提高模型参数计算效率,提出一种基于CNN的含分布式光伏配电网静态等值参数估计模型。在某省配电网算例下验证了所提方法的有效性。相较于其他方法,基于CNN的静态等值考虑了源荷功率的波动性及相关性,且提高了等值参数辨识效率,能够应用于静态等值参数的在线计算。This paper proposed a static equivalent method of the distribution network with distributed photovoltaic(PV)based on convolutional neural network(CNN)to adapt to the complex operation modes of distribution network.PV and load power scenarios were generated based on kernel density estimation and Copula function considering the correlation between fluctuation of PV and load,and the power flow of the distribution network was calculated.The static equivalent model of the distribution network with dis-tributed PV was constructed for each single operation mode,and particle swarm optimization(PSO)optimized the transformer and line parameters.A CNN-based static equivalent parameter estimation model of the distribution network with distributed PV was pro-posed to improve the efficiency of model parameter calculation.An example of a provincial distribution network verified the effec-tiveness of the proposed method.Compared with other methods,the static equivalent based on CNN considers the fluctuation and correlation of PV and load power.It improved the identification efficiency of equivalent parameters,which could be applied to the online calculation of static equivalent parameters.

关 键 词：分布式光伏 配电网 静态等值 卷积神经网络 粒子群优化 

分 类 号：TM71[电气工程—电力系统及自动化]