基于V-I轨迹颜色编码的非侵入式负荷识别方法  

Non-intrusive Load Monitoring Method Based on V-I Trajectory Color Coding

作　　者：解洋 梅飞[2] 郑建勇 高昂[3] 李轩 沙浩源[3] XIE Yang;MEI Fei;ZHENG Jianyong;GAO Ang;LI Xuan;SHA Haoyuan(School of Cyber Science and Engineering,Southeast University,Nanjing 211102,China;College of Energy and Electrical Engineering,Hohai University,Nanjing 211100,China;School of Electrical Engineering,Southeast University,Nanjing 210096,China)

机构地区：[1]东南大学网络空间安全学院,江苏省南京市211102 [2]河海大学能源与电气学院,江苏省南京市211100 [3]东南大学电气工程学院,江苏省南京市210096

出　　处：《电力系统自动化》2022年第4期93-102,共10页Automation of Electric Power Systems

基　　金：国家重点研发计划资助项目(2018YFB1500800)。

摘　　要：在非侵入式负荷识别中基于原始电压-电流(V-I)轨迹特征的识别方法,难以对相似轨迹特征的负荷做出有效辨识。因此,提出了一种基于V-I轨迹特征的颜色编码方法,并利用K-means聚类算法和AlexNet神经网络进行负荷特征的辨识。首先,运用K-means聚类算法对负荷的有功和无功功率特征进行初步分类。然后,对未分类成功的负荷进行V-I轨迹构建和颜色编码处理,生成带有颜色特征的V-I轨迹。最后,运用AlexNet神经网络对负荷进行训练和分类,达到快速精细化的分类效果。针对公共数据集PLAID和WHITED,运用原始V-I轨迹特征和进行颜色编码后V-I轨迹的识别效果做对比分析,可知所提方法在节省计算时间的同时也提高了识别的准确度,提升效果明显。In the non-intrusive load monitoring, the monitoring method based on the original(voltage-current) V-I trajectory characteristics is difficult to effectively identify the loads with similar trajectory characteristics. Therefore, a color coding method based on V-I trajectory characteristics is proposed, and the K-means clustering algorithm and Alex Net neural network are used to identify load characteristics. First, the K-means clustering algorithm is used to preliminarily classify the active power and reactive power characteristics of the load. Then, construction and color coding of the V-I trajectory are performed on the unclassified loads to generate V-I trajectories with color characteristics. Finally, the Alex Net neural network is used to train and classify the load to achieve a fast and refined classification effect. For the public data sets PLAID and WHITED, the monitoring effect of original V-I trajectory characteristics and color-coded V-I trajectories are compared and analyzed. It is known that the proposed method not only saves calculation time but also improves the monitoring accuracy, and the improvement effect is obvious.

关 键 词：非侵入式负荷识别 V-I轨迹特征 颜色编码 AlexNet神经网络 K-MEANS聚类算法 

分 类 号：TM714[电气工程—电力系统及自动化] TP183[自动化与计算机技术—控制理论与控制工程]