摘要
针对用电侧异常行为自动识别中存在的错识和漏识问题,提出融合离群点检测与K-means的用电侧异常行为自动识别方法。通过对用电侧行为数据进行填补及标准化处理,实现对原始数据的预处理;通过对用电侧行为进行离群点检测,深入挖掘数据中的潜在规律,提取离散特征的数据点。利用K-means算法对检测出的时间离群点序列进行聚类,识别序列中的异常行为,实现融合离群点检测与K-means的用电侧异常行为自动识别。实验证明,所设计方法的错识率不超过1.5%,漏识率不超过1%,可实现对用电侧异常行为的自动识别。
To address the problems of misidentification and missed identification in the automatic identification of abnormal behaviors on the power consumption side,an automatic identification method for abnormal behaviors on the power consumption side fusing outlier detection and K-means is proposed.Preprocessing of raw data is realized by imputing and standardizing the behavioral data on the power consumption side.Potential rules in the data are deeply explored and discrete feature data points are extracted by conducting outlier detection on the behaviors on the power consumption side.The K-means algorithm is used to cluster the detected temporal outlier sequences,identify abnormal behaviors in the sequences,and realize the automatic identification of abnormal behaviors on the power consumption side fusing outlier detection and K-means.Experiments show that the misidentification rate of the designed method does not exceed 1.5%and the missed identification rate does not exceed 1%,which can realize the automatic identification of abnormal behaviors on the power consumption side.
作者
陈普
刘仲
刘元强
CHEN Pu;LIU Zhong;LIU Yuanqiang(Integrated Electronic Systems Lab Co.,Ltd.,Jinan,Shandong 250010,China)
出处
《自动化应用》
2026年第2期170-172,共3页
Automation Application
关键词
离群点检测
K-MEANS
用电侧
异常行为
标准化
outlier detection
K-means
power consumption side
abnormal behavior
standardization
