基于组合模型的电力用户用电行为分层分类方法  

Hierarchical Classification Method for Power Consumption Behavior of Power Users Based on Combination Model

作　　者：肖庆追 李捷 陈鹤峰[2] 魏彩娥 XIAO Qingzhui;LI Jie;CHEN Hefeng;WEI Cai’e(School of Internet Finance and Information Engineering,Guangdong University of Finance,Guangzhou 510520,China;School of Applied Mathematics,Guangdong University of Technology,Guangzhou 510520,China;School of Entrepreneurship Education,Guangdong University of Finance,Guangzhou 510520,China)

机构地区：[1]广东金融学院互联网金融与信息工程学院,广州510520 [2]广东工业大学应用数学学院,广州510520 [3]广东金融学院创业教育学院,广州510520

出　　处：《电力系统及其自动化学报》2023年第5期82-88,94,共8页Proceedings of the CSU-EPSA

基　　金：广东省高等教育教学研究和改革项目(2019B2685)。

摘　　要：针对电网企业对电力用户进行精细化管理需求,提出一种组合模型实现对“正常”、“窃电”、“漏电”和“计量异常”4种用电行为的分层分类。首先,利用电压不平衡度等3维特征对高维用电数据进行降维表征,然后提出磷虾算法优化的一类支持向量机KH-OC-SVM(krill herd optimized one-class support vector machine)分类器将特征向量自动划分为“正常”和“异常”2类,最后利用所提基于密度的K-均值(K-means)聚类算法对“异常”数据进行分析,将其自动划分为“窃电”“漏电”和“计量异常”3种异常用电行为。算例结果表明,所提方法能够有效实现电力用户用电行为的自动分类。According to the demand of power grid enterprises for the fine management of power users,a combination model is proposed to realize the hierarchical classification of four kinds of power consumption behaviors,i.e.,“nor⁃mal”,“power theft”,“leakage”and“metering abnormality”.First,the three-dimensional characteristics including voltage unbalance are used to reduce the dimension of high-dimensional power consumption data.Then,a krill herd op⁃timized one-class support vector machine(KH-OC-SVM)classifier is proposed to automatically divide the feature vec⁃tors into“normal”and“abnormal”.Finally,the proposed K-means clustering algorithm based on density is used to ana⁃lyze the“abnormal”data,which is automatically divided into three kinds of abnormal power consumption behaviors,i.e.,“power theft”,“leakage”and“metering abnormality”.The results of an example show that the proposed method can ef⁃fectively realize the automatic classification of power users’power consumption behavior.

关 键 词：异常用电 分层分类 特征提取 基于密度的K-均值聚类 磷虾算法 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]