Previsão do consumo de energia elétrica em um frigorífico: um estudo de caso utilizando regressão linear, redes neurais e máquinas de vetores de suporte

• Ano de defesa: 2021
• Autor(a) principal: Antoniolli, Eduarda Araújo
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Tecnológica Federal do Paraná; Medianeira; Brasil; Programa de Pós-Graduação em Tecnologias Computacionais para o Agronegócio; UTFPR
• Palavras-chave em Português: Agroindústria; Algorítmos computacionais; Redes neurais (Computação); Agricultural industries; Computer algorithms; Neural networks (Computer science); CNPQ::CIENCIAS EXATAS E DA TERRA; Engenharia/Tecnologia/Gestão
• Link de acesso: http://repositorio.utfpr.edu.br/jspui/handle/1/28877

This research aimed to forecast the electricity consumption of a poultry facility unit. For this, time series models were developed using learning algorithms, such as linear regression, neural networks and support vector machines, using WEKA software. The input was the unit’s consumption history, from November 2016 to December 2020. The hyperparameters of the models were chosen from the evaluation of the MAPE and MAE of the test set. For validation, a forecast of one semester ahead was performed, that is, for the period from January 2021 to June 2021. The data were compared with the actual consumption observed in the unit. The model with the best performance was the SVM, with a MAPE of 3.38%. The linear regression models and neural networks, despite presenting an error above the MAPE established for this work, of 5%, are also considered capable of estimating the future consumption of electricity for the unit in question.