基于EEMD与CNN模型的多标签负荷识别方法  

Multi-label load identification method based on EEMD and CNN model

作　　者：程志友[1,2] 程安然 李悦 姜帅 CHENG Zhiyou;CHENG Anran;LI Yue;JIANG Shuai(Power Quality Engineering Research Center of the Ministry of Education,Anhui University,Hefei 230601,China;School of Internet,Anhui University,Hefei 230039,China)

机构地区：[1]教育部电能质量工程研究中心,安徽大学,安徽合肥230601 [2]安徽大学互联网学院,安徽合肥230039

出　　处：《电工电能新技术》2023年第2期88-96,共9页Advanced Technology of Electrical Engineering and Energy

基　　金：国家自然科学基金项目(61672032);安徽省科技重大专项(18030901018);安徽省自然科学基金项目(2108085QE237)。

摘　　要：识别用户用电负荷组成与用电行为是智能电网技术发展的重要研究内容之一。本文提出了一种基于集合经验模态分解(EEMD)结合卷积神经网络(CNN)的多标签负荷识别方法,实现对用户负荷有效的非侵入式监测。首先从检测到事件的聚合测量数据中提取单周期电流波形,应用集合经验模态分解将电流分解为两种模态分量,接着应用欧氏距离相似度函数将分解后的电流转化为二维矩阵表示,通过CNN多标签分类器自动提取矩阵的有效特征,最后利用公开数据集对所提出的方法进行了实验验证。结果表明,基于EEMD处理后的负荷识别准确率高,能够有效地实现多标签负荷识别。To identify the composition and behavior of users’power load is one of the important research contents in the development of smart grid technology.In this paper,a multi-label load identification method based on ensemble empirical mode decomposition(EEMD)and convolutional neural network(CNN)is proposed to realize effective non-invasive monitoring of user load.Firstly,the single cycle current waveform is extracted from the aggregated measurement data of the detected event,and the current is decomposed into two modal components by ensemble empirical mode decomposition.Secondly,the decomposed current is transformed into two-dimensional matrix representation by Euclidean distance similarity function,and the effective features of the matrix are automatically extracted by CNN multi-label classifier.Finally,the proposed method is verified by experiments using public data sets.The results show that the load identification based on EEMD has high accuracy and can effectively realize multi-label load identification.

关 键 词：集合经验模态分解 卷积神经网络 欧式距离相似度函数 多标签 负荷识别 

分 类 号：TM721[电气工程—电力系统及自动化]