一种基于多分类器耦合的非侵入式负荷辨识方法  

A Non-intrusive Load Identification Method Based on Coupling Weak Multiple Classifiers

作　　者：易仕琪 孔政敏[1] 王帅 霍梓航 YI Shiqi;KONG Zhengmin;WANG Shuai;HUO Zihang(School of Electrical Engineering and Automation,Wuhan University,Wuhan,Hubei 430072,China;Guangdong Provincial Key Laboratory of New Technology for Smart Grid,China Southern Power Grid Technology Co.,Ltd.,Guangzhou,Guangdong 510080,China)

机构地区：[1]武汉大学电气与自动化学院,湖北武汉430072 [2]广东省智能电网新技术企业重点实验室(南方电网电力科技股份有限公司),广东广州510080

出　　处：《广东电力》2023年第8期89-96,共8页Guangdong Electric Power

基　　金：广东省智能电网新技术企业重点实验室项目(2020B1212070025);国家自然科学基金项目(62173256);国家重点研发计划项目(2021ZD0112702)。

摘　　要：新一代智能电表实现了电压、电流数据高频采样,使得用户电力负荷的辨识特征具有更多的可选性,同时多类别的负荷特征输入对于增强负荷辨识算法性能提出了更高的要求。为此,提出一种基于弱分类器耦合的非侵入式负荷辨识方法。首先,提取家用电器的电量和非电量等多种负荷特征,并将这些特征作为弱分类器的输入;然后,将各个弱分类器的输出进行耦合,采用经典的Adaboost架构提高分类器的性能,实现非侵入式电力负荷辨识。最后,在AMPds公用数据集上进行测试,结果表明所提方法能够达到较高的负荷辨识准确率。A new generation of intelligent meters realizes high-frequency sampling of voltage and current data,leading to more choosability of identification features of users' load.Meanwhile,the input of classified load features puts forward higher requirement for increasing the performance of load identification algorithms.Therefore,this paper proposes a kind of non-intrusive load identification method based on weak classifier coupling.Firstly,the multiple types of load features are extracted from the load as the input of the weak classifiers,including the electrical and non-electrical features.Then,the output of the weak classifiers method is coupled and the performance of the classifiers are improved by the classical Adaboost architecture so as to realize non-intrusive power load identification.Finally,the results of test on the AMPds public dataset show that the proposed method has higher load identification accuracy.

关 键 词：非侵入式负荷辨识 弱分类器 ADABOOST算法 

分 类 号：TM714[电气工程—电力系统及自动化] TM933.4