A New Approach for Nontechnical Losses Detection Based on Optimum-Path Forest

dc.contributor.authorOba Ramos, Caio Cesar [UNESP]
dc.contributor.authorde Sousa, Andra Nunes
dc.contributor.authorPapa, João Paulo [UNESP]
dc.contributor.authorFalcao, Alexandre Xavier
dc.contributor.institutionUniversidade Estadual Paulista (Unesp)
dc.contributor.institutionUniversidade Estadual de Campinas (UNICAMP)
dc.date.accessioned2014-05-20T13:25:59Z
dc.date.available2014-05-20T13:25:59Z
dc.date.issued2011-02-01
dc.description.abstractNowadays, fraud detection is important to avoid nontechnical energy losses. Various electric companies around the world have been faced with such losses, mainly from industrial and commercial consumers. This problem has traditionally been dealt with using artificial intelligence techniques, although their use can result in difficulties such as a high computational burden in the training phase and problems with parameter optimization. A recently-developed pattern recognition technique called optimum-path forest (OPF), however, has been shown to be superior to state-of-the-art artificial intelligence techniques. In this paper, we proposed to use OPF for nontechnical losses detection, as well as to apply its learning and pruning algorithms to this purpose. Comparisons against neural networks and other techniques demonstrated the robustness of the OPF with respect to commercial losses automatic identification.en
dc.description.affiliationSão Paulo State Univ, Dept Elect Engn, Intelligent Tech & Power Syst Lab, São Paulo, Brazil
dc.description.affiliationSão Paulo State Univ, Dept Comp, São Paulo, Brazil
dc.description.affiliationUniv Estadual Campinas, Inst Comp, São Paulo, Brazil
dc.description.affiliationSão Paulo State Univ, Dept Comp Sci, São Paulo, Brazil
dc.description.affiliationUnespSão Paulo State Univ, Dept Elect Engn, Intelligent Tech & Power Syst Lab, São Paulo, Brazil
dc.description.affiliationUnespSão Paulo State Univ, Dept Comp, São Paulo, Brazil
dc.description.affiliationUnespSão Paulo State Univ, Dept Comp Sci, São Paulo, Brazil
dc.format.extent181-189
dc.identifierhttp://dx.doi.org/10.1109/TPWRS.2010.2051823
dc.identifier.citationIEEE Transactions on Power Systems. Piscataway: IEEE-Inst Electrical Electronics Engineers Inc, v. 26, n. 1, p. 181-189, 2011.
dc.identifier.doi10.1109/TPWRS.2010.2051823
dc.identifier.issn0885-8950
dc.identifier.lattes9039182932747194
dc.identifier.urihttp://hdl.handle.net/11449/8303
dc.identifier.wosWOS:000286516100021
dc.language.isoeng
dc.publisherInstitute of Electrical and Electronics Engineers (IEEE)
dc.relation.ispartofIEEE Transactions on Power Systems
dc.relation.ispartofjcr5.255
dc.relation.ispartofsjr2,742
dc.rights.accessRightsAcesso restrito
dc.sourceWeb of Science
dc.subjectNontechnical lossesen
dc.subjectoptimum-path foresten
dc.subjectpattern recognitionen
dc.titleA New Approach for Nontechnical Losses Detection Based on Optimum-Path Foresten
dc.typeArtigo
dcterms.licensehttp://www.ieee.org/publications_standards/publications/rights/rights_policies.html
dcterms.rightsHolderIEEE-Inst Electrical Electronics Engineers Inc
unesp.author.lattes9039182932747194
unesp.author.orcid0000-0002-6494-7514[3]
unesp.campusUniversidade Estadual Paulista (Unesp), Faculdade de Ciências, Baurupt
unesp.departmentComputação - FCpt

Arquivos

Licença do Pacote

Agora exibindo 1 - 2 de 2
Nenhuma Miniatura disponível
Nome:
license.txt
Tamanho:
1.71 KB
Formato:
Item-specific license agreed upon to submission
Descrição:
Nenhuma Miniatura disponível
Nome:
license.txt
Tamanho:
1.71 KB
Formato:
Item-specific license agreed upon to submission
Descrição: