Modelagem do percentual de metano em biogás produzido a partir de resíduos da pecuária

• Ano de defesa: 2022
• Autor(a) principal: Mandelli, Ana Paula
• 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: Resíduos agrícolas; Resíduos orgânicos; Energia - Fontes alternativas; Redes neurais (Computação); Agricultural wastes; Organic wastes; Renewable energy sources; Neural networks (Computer science); CNPQ::ENGENHARIAS::ENGENHARIA SANITARIA; Engenharia/Tecnologia/Gestão
• Link de acesso: http://repositorio.utfpr.edu.br/jspui/handle/1/31333

The agricultural activity is the basis of the economy in the western region of Paraná, and in recent decades, an expressive growth of this branch has been observed. However, this scenario has caused an increase in the amount of solid organic waste generated by agricultural processes, requiring the development of treatment technologies for this waste. In this context, anaerobic digestion is a viable option because, besides providing the appropriate treatment of the waste, it generates biogas as a co-product. The biogas, because it contains methane in the mixture, can be used in combustion systems for conversion of thermal or electrical energy. However, to evaluate the feasibility of biogas generation, especially the percentage of methane of a particular waste, laboratory tests of biodigestion are necessary. The disadvantage of estimating the percentage of methane is the time required for the analysis to be completed, between 30 and 60 days. Thus, the use of mathematical modeling to evaluate the methane percentage of a given waste, without the need for testing, would be of great value in project analysis of this kind. Furthermore, considering the inherent nonlinearity of biodigestion, since it involves bacteria, the use of Artificial Neural Networks (ANN) as a computational modeling tool can provide robustness to the adjusted model. In view of this, this study aimed at mathematical modeling of the methane percentage of livestock waste (pig, cattle, and poultry farming) using ANNs. As results, it was possible to verify that the methane percentages are similar to those of laboratory tests, leading to the conclusion that the ANN was able to predict the methane percentage in the biogas produced for most of the samples tested. However, for some samples, the ANN provided answers without physical meaning, and it was necessary to apply a filter to allow the ANN to provide answers only in those cases in which the methane concentration value was feasible.