文章基本信息

标题：Statistical Features Extraction and Performance Analysis of Supervised Classifiers for Non-Intrusive Load Monitoring
本地全文：下载
作者：Mehedi Hasan ; Dhiman Chowdhury ; Khalid Hasan 等
期刊名称：Engineering Letters
印刷版ISSN：1816-093X
电子版ISSN：1816-0948
出版年度：2019
卷号：27
期号：4
页码：776-782
出版社：Newswood Ltd
摘要：An approach to extract distinctive statistical featuresembedded in current and power signatures of differentelectrical appliances to substantiate efficacious classificationfor non-intrusive load monitoring (NILM) is presented in thisletter. Supervised classifiers - na¨ıve Bayes, multi-class supportvector machine (SVM), ensemble, binary decision tree (DT) anddiscriminant analysis are employed for performance evaluationbased on the extracted feature values. The testbed is COOLLNILM public dataset constituted by 42 devices of differentpower ratings. The training and testing accuracies along withcross-validation losses associated with each classification algorithmare determined. As a comparative analysis, binary DTclassifier produces the best results. Performance assessmentcorroborates the reliability of the proposed framework forNILM applications.
关键词：Current and power signatures; load disaggregation;NILM; statistical features; supervised classifiers