杨涛,孙志达,唐明,吴栋萁,王剑,李海波,雷一
YANG Tao,SUN Zhida,TANG Ming,WU Dongqi,WANG Jian,LI Haibo,LEI Yi

• 1:国网浙江省电力公司电力科学研究院
• 2:清华四川能源互联网研究院

摘要(Abstract)：

为了理清供电台区和用户之间的挂接关系,传统方法是依托人工或专用装置辨识,但存在效率低下和运行工况受限等问题。对此,在数据挖掘和机器学习技术基础之上,提出一种基于TensorFlow框架的改进BP(反向传播)户变关系识别方法。首先以Spearman相关系数作为K-means聚类的距离度量,对部分历史电压数据聚类,将聚类结果匹配所属台区进而构建训练样本,利用学习后的BP识别当前用户所对应的变压器及相别。为改善BP容易出现局部最优和收敛速度慢的问题,采用零均值化和Adam(自适应矩估计)对BP进行优化,并将BP部署在TensorFlow框架以进一步减小算法耗时。算例表明,所提算法能够有效提高户变关系识别准确率,提高辨识效率,具有良好的理论和应用价值。
To identify the connection relationship between transformers and users traditionally relies on manual work or special devices, which feature low efficiency and limited operating conditions. This paper, employing data mining and machine learning, presents an improved BP(backpropagation) method for transformer-user relationship identification based on the TensorFlow framework. Firstly, the Spearman correlation coefficient is used as the distance of K-means clustering. Some historical voltage data are clustered and the clustering results are matched with their transformers to construct training samples. Then the learned BP is used to identify the transformer and phase corresponding to the user voltage data. Meanwhile, to solve the problem of local optimization and slow convergence of BP, zero-mean and Adam(adaptive moment estimation), are used to optimize BP and deploy it in the TensorFlow framework to further reduce the time for computing. Examples show that the proposed algorithm can effectively improve the accuracy and efficiency of transformer-user relationship identification, and has good theoretical and application value.

关键词(KeyWords)： 户变关系;TensorFlow;BP;Spearman;K-means;零均值化;自适应矩估计
transformer-user relationship;TensorFlow;BP;spearman;K-means;zero-mean;Adam

Abstract:

Keywords:

基金项目(Foundation): 国网浙江省电力公司科技项目(5211DS19002U);; 清华大学电力系统及大型发电设备安全控制与仿真国家重点实验室开放课题面上项目(SKLD20KM03)

作者(Author): 杨涛,孙志达,唐明,吴栋萁,王剑,李海波,雷一
YANG Tao,SUN Zhida,TANG Ming,WU Dongqi,WANG Jian,LI Haibo,LEI Yi

DOI: 10.19585/j.zjdl.202108004

参考文献(References)：

• [1]杨晶伟.配电台区线损异常管理研究[J].自动化与仪器仪表,2017(8):55-57.
• [2]李建,赵汉昌.多功能低压台区识别设备的研制[J].电测与仪表,2014,51(13):107-111.
• [3]范荻,李绍山,李海涛,等.台区用户识别仪关键技术应用研究[J].华北电力技术,2010(7):27-30.
• [4]徐湛.台区用户识别仪应用分析[J].计量与测试技术,2009,36(11):26-28.
• [5]潘明明,田世明,吴博,等.基于智能电表数据的台区识别与窃电检测方法研究[J].智慧电力,2017,45(12):80-84.
• [6]阿辽沙·叶,顾君,张小秋.基于数据时空相关性的智能台区识别技术[J].现代电子技术,2018,41(4):30-33.
• [7]赵宇东.一种智能电能表台区归属自动判定方法[J].华北电力技术,2016(5):22-28.
• [8]周明,宋旭帆,涂京,等.基于非侵入式负荷监测的居民用电行为分析[J].电网技术,2018,42(10):3268-3276.
• [9]顾丹珍,艾芊,陈陈,等.自适应神经网络在负荷动态建模中的应用[J].中国电机工程学报,2007,27(16):31-36.
• [10]苏学能,刘天琪,曹鸿谦,等.基于Hadoop架构的多重分布式BP神经网络的短期负荷预测方法[J].中国电机工程学报,2017,37(17):4966-4973.
• [11]刘洋,许立雄.适用于海量负荷数据分类的高性能反向传播神经网络算法[J].电力系统自动化,2018,42(21):96-103.
• [12]李亚,刘丽平,李柏青,等.基于改进K-Means聚类和BP神经网络的台区线损率计算方法[J].中国电机工程学报,2016,36(17):4543-4552.
• [13]黎祚,周步祥,林楠.基于模糊聚类与改进BP算法的日负荷特性曲线分类与短期负荷预测[J].电力系统保护与控制,2012,40(3):56-60.
• [14]李亚,樊汝森,蒋伟,等.基于BP神经网络的智能台区识别方法研究[J].电测与仪表,2017,54(3):25-30.
• [15]XU W S,YU Q M,SUN Y L,et al.Research on improving training speed of LMBP algorithm and its simulation in application[C]//2007 International Conference on Computational Intelligence and Security(CIS 2007). Harbin,China:IEEE,2007:540-545.
• [16]吴云,雷建文,鲍丽山,等.基于改进灰色关联分析与蝙蝠优化神经网络的短期负荷预测[J].电力系统自动化,2018,42(20):67-72.
• [17]丁坚勇,朱炳翔,田世明,等.改进F-score特征选择的MPSO-BP神经网络短期负荷预测[J].电测与仪表,2018,55(15):36-41.
• [18]王亚琴,王耀力,王力波,等.一种改进果蝇算法优化神经网络短期负荷预测模型[J].电测与仪表,2018,55(22):13-18.
• [19]KINGMA D P,BA J,ADAM.A method for stochastic optimization[J/OL].[2021-03-05].Computer Science,2014.https://arxiv.org/abs/1412.6980.
• [20]LIU M L,GRANA D.Accelerating geostatistical seismic inversion using TensorFlow:a heterogeneous distributed deep learning framework[J].Computers&Geosciences,2019,124:37-45.
• [21]ABADI M,AGARWAL A,BARHAM P,et al.TensorFlow:large-scale machine learning on heterogeneous distribut ed systems[EB/OL].(2016-03-14).https://www.mendeley.com/catalogue/1a5efca7-0514-395b-a5e2-fc3805c7df53.
• [22]彭显刚,赖家文,陈奕.基于聚类分析的客户用电模式智能识别方法[J].电力系统保护与控制,2014,42(19):68-73.
• [23]胡锋,朱承治,汪志华.基于改进K-means算法的电力负荷分类研究[J].电子测量技术,2018,41(12):44-48.
• [24]张铁峰,顾明迪.电力用户负荷模式提取技术及应用综述[J].电网技术,2016,40(3):804-811.
• [25]邹云峰,梅飞,李悦,等.基于数据挖掘技术的台区合理线损预测模型研究[J].电力需求侧管理,2015,17(4):25-29.