Predição da performance de um reator UASB para o tratamento de vinhaça usando identificação e redes neuronais

• Autor(a) principal: Friderichs, André
• Data de Publicação: 2013
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Manancial - Repositório Digital da UFSM
• dARK ID: ark:/26339/001300000bcm7
• Texto Completo: http://repositorio.ufsm.br/handle/1/7965
• topic: Etanol; Vinhaça; UASB; Identificação; Redes neuronais; Ethanol; Stillage; UASB; Identification; Neural network; CNPQ::ENGENHARIAS::ENGENHARIA DE PRODUCAO

Brazil is a tropical country with a huge amount of natural energy resources. In view of the growth that the country is experiencing, such resources exploitation becomes increasingly attractive. Among many resources alternatives, the biomass is one of the most notable mainly due to its applicability in farms and agro-industries around the country. The use of biomass to ethanol production, even on a small scale, results in a considerable production of stillage waste production that presents high organic matter content and that is seen as a highly polluting effluent. The anaerobic digestion of stillage in Upflow Anaerobic Sludge Blanket (UASB) reactors is an efficient alternative to stillage treatment as well as to biogas production. This work presents the study of empirical modeling, using tools such as artificial neural networks (ANN) and parametric identification, of an UASB reactor operation which treats the distillation stillage of ethanol produced from two different biomasses: raw starch (potato) and saccharide (sugar cane). The inputs used in the models were chosen by statistical methods according to a series of parameters that are monitored during the experimental reactor operation, where it is evident the importance of the initial Chemical Oxygen Demand, temperatures and the period of operation of the reactor with the same charge. The results were promising for the use of such tools in performance estimation of highly complex biological systems such as the anaerobic digestion, chosen as case study in this work, being achieved in the best cases a correlation of 0,98841 for potato stillage, and a correlation of 0,99738 for the stillage of sugar cane using neural networks.