Previsão das vazões afluentes diárias por rede neural para projeção de PLD

• Ano de defesa: 2019
• Autor(a) principal: Mendes, Natalia Dias Sardinha
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Rio de Janeiro; Brasil; Instituto Alberto Luiz Coimbra de Pós-Graduação e Pesquisa de Engenharia; Programa de Pós-Graduação em Engenharia Elétrica; UFRJ
• Palavras-chave em Português: Engenharia elétrica; Previsão das vazões afluentes diárias; Rede neural artificial; Projeção de PLD; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA
• Link de acesso: http://hdl.handle.net/11422/12287

The prediction of affluent flows is one of the main inputs in the definition of the Monthly Operating Cost - CMO of the Brazilian electrical system, as well as, the Market Clearing Price – PLD. Therefore the prediction of affluent flows is a challenging area in relation to the complexity of problems, besides the possibility of generating high financial profits for the energy market agents. This work aimed to calibrate a daily affluent flow forecasting model at the Três Marias hydroelectric plant in the São Francisco basin, 7 days ahead, using the neural networks technique, considering the natural affluent flows from the plant's own reservoir, and pluviometric and fluviometric information from hydrometric stations upstream of the study region. The prediction model of this paper using the neural network technique produced better results than the models used by the electrical system, such as the PREVIVAZH traditional statistician and even the Neuro3M neural network model, which was used until the beginning of this year in the same reservoir.