Logotipo do repositório
 

Publicação:
A NILM Dataset for Cognitive Meters Based on Conservative Power Theory and Pattern Recognition Techniques

dc.contributor.authorSouza, Wesley A.
dc.contributor.authorMarafão, Fernando P. [UNESP]
dc.contributor.authorLiberado, Eduardo V. [UNESP]
dc.contributor.authorSimões, Marcelo G.
dc.contributor.authorDa Silva, Luiz C. P.
dc.contributor.institutionUniversidade Estadual de Campinas (UNICAMP)
dc.contributor.institutionUniversidade Estadual Paulista (Unesp)
dc.contributor.institutionColorado School of Mines
dc.date.accessioned2019-10-06T16:54:30Z
dc.date.available2019-10-06T16:54:30Z
dc.date.issued2018-12-15
dc.description.abstractThis paper presents a novel dataset capable of classifying and disaggregating residential appliances for the development of smart or cognitive power meters. This novel dataset uses power indicators (also denoted as conformity factors) from the conservative power theory (CPT), which are calculated from measured voltage and current waveforms during the operation of residential loads. The association of CPT power indicators with suitable pattern recognition algorithms (PRA) and a power signature state machine provides proper identification of each appliance. So, the paper also presents a detailed evaluation of possible PRA for this application, especially the SVM—support vector machine, OPF—optimum-path forest, MLP—multilayer perceptron, KNN—K-nearest neighbor and DT—decision tree. All these algorithms have been compared regarding accuracy and computational time. Validation results point out that KNN would be the best choice for dealing with the proposed dataset, leading to an accuracy higher than 98%.en
dc.description.affiliationDepartment of Energy and Systems (DSE) School of Electrical and Computer Engineering (FEEC) University of Campinas (UNICAMP), Av. Albert Einstein, 400
dc.description.affiliationInstitute of Science and Technology of Sorocaba (ICTS) Univ. Estadual Paulista (UNESP), Av. Três de Março, 511
dc.description.affiliationDepartment of Electrical Engineering Colorado School of Mines
dc.description.affiliationUnespInstitute of Science and Technology of Sorocaba (ICTS) Univ. Estadual Paulista (UNESP), Av. Três de Março, 511
dc.description.sponsorshipFundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)
dc.description.sponsorshipIdFAPESP: 2012/19375-1
dc.format.extent742-755
dc.identifierhttp://dx.doi.org/10.1007/s40313-018-0417-4
dc.identifier.citationJournal of Control, Automation and Electrical Systems, v. 29, n. 6, p. 742-755, 2018.
dc.identifier.doi10.1007/s40313-018-0417-4
dc.identifier.issn2195-3899
dc.identifier.issn2195-3880
dc.identifier.scopus2-s2.0-85056083377
dc.identifier.urihttp://hdl.handle.net/11449/189859
dc.language.isoeng
dc.relation.ispartofJournal of Control, Automation and Electrical Systems
dc.rights.accessRightsAcesso aberto
dc.sourceScopus
dc.subjectCognitive meter
dc.subjectConservative power theory
dc.subjectPattern recognition algorithms
dc.subjectResidential appliance recognition dataset
dc.subjectSmart meter
dc.titleA NILM Dataset for Cognitive Meters Based on Conservative Power Theory and Pattern Recognition Techniquesen
dc.typeArtigo
dspace.entity.typePublication
unesp.author.orcid0000-0002-3431-6359[1]
unesp.campusUniversidade Estadual Paulista (UNESP), Instituto de Ciência e Tecnologia, Sorocabapt
unesp.departmentEngenharia de Controle e Automação - ICTSpt

Arquivos