Sistema inteligente baseado em árvore de decisão, para apoio ao combate às perdas comerciais na distribuição de energia elétrica

• Ano de defesa: 2006
• Autor(a) principal: Reis Filho, José
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Uberlândia; BR; Programa de Pós-graduação em Engenharia Elétrica; Engenharias; UFU
• Palavras-chave em Português: Árvore de decisão; Identificação de fraudes; Energia elétrica; Inteligência artificial; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA
• Link de acesso: https://repositorio.ufu.br/handle/123456789/14493

The increase in commercial losses in electric utility companies has been a reason of great concern for these companies. The main motives of the increase in these losses are two: fraud practiced by the consumers; and problems in the energy meters. Nowadays, to identify one of the two problems mentioned above, in-site inspections are required. However, due to the high number of consumer unities, such inspections are done without any previous analysis of the consumer behavior, which results in a low rate of problem identification. On the other hand, electric utility companies have a database with much information about their consumers. So, this information can be used to identify the behavior profile of those consumers that are likely to be frauding or having problems with their energy meters. However, due to high quantity of data, it is demanding the use of an automatic process for identification of such behavior profiles. The goal of this work is to develop a decision support system to combat commercial losses in distribution power systems. Such system is based on Knowledge Discovery in Database KDD, which refers to discovering of knowledge in database, which may increase the rate of successful in-site inspections. The tool used to do the data mining stage of the KDD is Decision Tree. This is an artificial intelligence technique that tries to emulate human abilities in a computer system, and it learns from data and it is used for classification type of problems.