基于mRMR-R-GCN的配电网低压台区线损率预测模型  

A Power-Loss Prediction Model of Line Loss Rate for Low-Voltage Areas in Distribution Network Based on mRMR-R-GCN

作　　者：赵雅婷 林顺富 姜恩宇 沈运帷 ZHAO Yating;LIN Shunfu;JIANG Enyu;SHEN Yunwei(Shanghai University of Electric Power,Shanghai 200090,China;State Grid Shanghai Urban Electric Power Supply Company,Shanghai 200080,China)

机构地区：[1]上海电力大学,上海200090 [2]国网上海市区供电公司,上海200080

出　　处：《上海电力大学学报》2025年第2期120-125,133,共7页Journal of Shanghai University of Electric Power

摘　　要：针对低压台区线损率预测涉及海量异构多元数据的统计分析这一问题,提出了一种基于最大相关最小冗余(mRMR)算法和关系图卷积神经网络(R-GCN)模型的低压台区线损率预测模型。首先,针对供售电量、供电半径、负载率、光伏出力等电气特征指标进行清洗和归一化处理。其次,采用mRMR算法计算电气特征指标和低压台区线损率之间的相关度和冗余度,构建低压台区线损特征指标体系,进而利用K-Means++聚类算法对低压台区线损特征指标历史数据进行聚类,并采用变异系数(CoV)法对低压台区线损特征指标进行权重赋值。最后,采用R-GCN模型分析赋权低压台区线损特征指标和线损率的内部关联,建立低压台区线损率预测模型。以某低压台区实际数据对所提线损率预测模型进行验证,证明了所提模型的有效性。The power-loss prediction of low-voltage platform area involves the statistical analysis of massive heterogeneous multivariate data.A prediction of line loss rate of low-voltage distribution transformers in areas based on mRMR-R-GCN model is proposed.Firstly,electrical indicators such as power supply and sales,power supply radius,load rate,and photovoltaic output in low-voltage substation area are cleaned and normalized.Secondly,mRMR algorithm is used to calculate the correlation and redundancy between the multivariate electrical indexes and the power-loss rate in substation area,and the characteristic index system of power-loss in substation area is constructed.Furthermore,the K-Means++clustering algorithm is used to cluster the historical data of power-loss characteristic indexes in substation area,and the CoV method is used to assign the weight of power-loss characteristic indexes.Finally,the R-GCN model is used to analyze the internal correlation between the weighted line loss characteristic index and the power-loss rate,and the power-loss prediction model in the substation area is established.The actual data of a low-voltage substation area are used to verify the proposed power-loss prediction model,and the effectiveness of the proposed method is proved.

关 键 词：低压台区线损率 最大相关最小冗余算法 K-Means++聚类算法 关系图卷积神经网络 

分 类 号：TM77[电气工程—电力系统及自动化]