摘要
传统的水电用户异常检测分别在用水和用电系统中进行,容易造成检测结果不准确或漏检。如何有效检测水电异常用户,对电力系统和用水系统提出了新的考验。以实际的水电用户数据为样本,对比了水电比聚类分析法、水量电量二维向量聚类分析法和基于SVM的异常检测法,并充分结合水量、电量数据,进行异常水电用户检测。结果表明,对于城市居民用户,相比其它两种算法,基于SVM的异常检测法可以更加准确、高效的检测出异常水电用户。
Traditional water/electricity user anomaly are detected respectively in the water and electricity system,vulnerable to inaccurate results or misdetection.So how to effectively detect abnormal water/electricity users is a challenge for the power system and water system.With the real water/electricity user data as sample,this paper compares the three methods：the water/electricity ratio clustering analysis method,water/electricity quantity two-dimensional vector clustering analysis method and anomaly test method based on SVM.And fully combined with the data of water/electricity quantity,abnormal water and electricity users are detected.The results show that,for urban users,anomaly test method based on SVM is the most accurate and effective in detecting the abnormal water and electricity users compared with other two methods.
出处
《电力与能源》
2015年第3期355-359,共5页
Power & Energy
关键词
水电用户
异常检测
聚类分析
SVM
water/electricity users
anomaly detection
clustering analysis
SVM
作者简介
罗飞鹏（1969），男，高级经济师，长期从事电力营销管理工作。
