Loading...
Publication
Comparação da modelação da digestão anaeróbica da ETAR do Choupal através de Redes Neuronais Artificiais e ADM1
| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 3.15 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
Resumo
O aumento dos últimos anos de Estações de Tratamento de Águas Residuais (ETAR) tem sido benéfico para o meio ambiente, pois as ETAR´s têm a função de receber e tratar as águas residuais de forma a serem devolvidas ao meio ambiente em condições ambientalmente seguras evitando assim a poluição.
Contudo as ETAR´ s acarretam muitos custos energéticos, posto isto, tem vindo a ser desenvolvidas soluções para colmatar estes custos.
Com este trabalho pretende-se otimizar o processo da Digestão Anaeróbia (DA) para se conseguir aumentar a produção de biogás que por sua vez é convertido em energia, contribuindo assim para a rentabilidade da exploração da ETAR. Nos últimos tempos as Redes Neuronais Artificiais (RNA) têm sido das soluções mais usadas neste tipo de processo da DA, pois são modelos matemáticos computacionais, sendo capazes de prever valores quando novos casos lhes são apresentados, têm características de desempenho inspirados no cérebro humano. Os modelos neuronais artificiais têm como fonte de inspiração as redes neuronais biológicas. Para o desenvolvimento das RNA foi utilizado o software NeuralTools da Palisade, esta consiste em quatro fases distintas, a preparação de dados, o treino, o teste e a previsão da rede. O estudo de caso incide nos dois digestores da ETAR do Choupal da AdCl, empresa onde foi realizado o estágio curricular que originou este trabalho. Para o presente estudo foram utilizadas variáveis do tratamento de lamas da ETAR, como % MS LM, a %MO LM, que entra no digestor I, Qentr.(m3/dia), o Qbiogás(m3/dia), a %MS, a %MO, a T(ºC), o pH, a alcalinidade, e os ácidos gordos voláteis de cada um dos digestores. Foram treinadas e testadas várias redes até se obter a melhor rede para de seguida ser utilizada para prever os valores de biogás. Houve dificuldades na definição das quantidades e disponibilidade de alguns dados prejudicando assim as redes, não conseguindo prever valores semelhantes dos valores reais do processo da DA, mas ainda assim foi possível obter em alguns valores essa igualdade. Verificou-se que as variáveis com maior impacto na modelação do processo anaeróbio são a MO(%)1, a MO(%)LM, a T(ºC)1, o Qentr(m3/dia). Conclui-se com este estudo, que usando o modelo RNA consegue-se obter resultados positivos no processo de DA permitindo assim otimizar a produção de biogás.
Abstract The increase in recent years of wastewater treatment plants (WWTP) has been a benefit to the environment, this is due to the fact that WWTP’s have the function of receiving and treating the waste water to be returned to the environment in environmentally safe conditions thus avoiding pollution. However the WWTP ́ s entails numerous energy costs. Due to that reason and to address these costs, solutions have been developed. This assignment intends to optimize the process of anaerobic digestion (AD) to increase the production of biogas which in turn is converted into energy, thus contributing to the profitability of exploration of the WWTP. In recent times the artificial neural networks (Ann) have been the most commonly used solutions in this type of process because they are mathematical models, capable to predict values when new cases are presented, they have performance characteristics inspired by the human brain. Artificial neural models have as a source of inspiration the biological neural networks. For the development of Ann the NeuralTools Palisade software was used, this consists of four distinct phases, the data preparation, training, testing and prediction of network. The case study focuses on the two digesters of WWTP of Choupal of AdCl, company where the curricular internship was held which led to this work. For this study we used variables of sludge treatment of WWTP to % MS LM,% MO LM, entering the digester I, Qentr.(m3/day), the Qbiogás (m3/day), the % MS,% MO, T (°C), pH, alkalinity and volatile fatty acids in each of the digesters. Several networks were trained and tested until the best network was found, to then be used to predict the values of biogas. There were difficulties in defining the quantity and availability of some data, hindering nets, failing to predict similar values of the real values of the process but it was still possible to obtain in some values that equality. It was found that the variables with the greatest impact in shaping the anaerobic process are the MO (%) 1, MO (%) LM, T (ºC) 1, Qentr (m3/day). This study concludes that using the RNA template can achieve positive results in the process, as a result, optimizing biogás production.
Abstract The increase in recent years of wastewater treatment plants (WWTP) has been a benefit to the environment, this is due to the fact that WWTP’s have the function of receiving and treating the waste water to be returned to the environment in environmentally safe conditions thus avoiding pollution. However the WWTP ́ s entails numerous energy costs. Due to that reason and to address these costs, solutions have been developed. This assignment intends to optimize the process of anaerobic digestion (AD) to increase the production of biogas which in turn is converted into energy, thus contributing to the profitability of exploration of the WWTP. In recent times the artificial neural networks (Ann) have been the most commonly used solutions in this type of process because they are mathematical models, capable to predict values when new cases are presented, they have performance characteristics inspired by the human brain. Artificial neural models have as a source of inspiration the biological neural networks. For the development of Ann the NeuralTools Palisade software was used, this consists of four distinct phases, the data preparation, training, testing and prediction of network. The case study focuses on the two digesters of WWTP of Choupal of AdCl, company where the curricular internship was held which led to this work. For this study we used variables of sludge treatment of WWTP to % MS LM,% MO LM, entering the digester I, Qentr.(m3/day), the Qbiogás (m3/day), the % MS,% MO, T (°C), pH, alkalinity and volatile fatty acids in each of the digesters. Several networks were trained and tested until the best network was found, to then be used to predict the values of biogas. There were difficulties in defining the quantity and availability of some data, hindering nets, failing to predict similar values of the real values of the process but it was still possible to obtain in some values that equality. It was found that the variables with the greatest impact in shaping the anaerobic process are the MO (%) 1, MO (%) LM, T (ºC) 1, Qentr (m3/day). This study concludes that using the RNA template can achieve positive results in the process, as a result, optimizing biogás production.
Description
Keywords
Biogás Estação de Tratamento de Águas Residuais