Monday, April 30, 2007
6-48
Bioethanol production optimization: A thermodynamic analysis
Alvarez Victor, Ccopa Rivera Elmer, Carvalho da Costa Aline, Aznar Martin, and Maciel Filho Rubens. School of Chemical Engineering, State University of Campinas, Cidade Universitária Zeferino Vaz, CEP: 13081-970, Campinas - SP, Brazil
The difficulty in modeling biotechnological processes is essentially on the precise description of the kinetics and robust modeling can only be achieved by incorporating reliable procedures to easily update the model when there are changes in operational conditions. An accurate model should also consider the influence of thermodynamic variables.
In this work, Artificial Neural Networks (ANN) has been used in order to reduce the optimization complexity of a model for bioethanol production. The ANN has been developed to estimate temperature, thermodynamic variables and biomass, substrate and product concentrations. Subsequently, the ANN model was optimized using Genetic Algorithms and Successive Quadratic Programming to determine the optimal operational conditions. An extractive alcoholic fermentation process was considered. The process consists of four interlinked units: fermentor, centrifuge, cell treatment unit and an ethanol-water separation unit (vacuum flash vessel).
The phase equilibrium in binary mixtures found in the flash vessel has been modeled using the predictive Soave-Redlich-Kwong, with original and modified molecular parameters. The presence of polar substances found in mixture and the many components (others different from ethanol and water) makes it difficult to model these mixtures. The case studied considered five binary water mixtures: acetic acid, acetaldehyde, furfural, methanol and 1-pentanol. These are the substances that are considered to be legal compounds by the Brazilian legislation.
A simulator of this system was developed to explore the possible operational strategies for a large-scale system. The impact of each strategy on the process behavior as well as assessment of possible implementation difficulties are carefully considered and discussed.
In this work, Artificial Neural Networks (ANN) has been used in order to reduce the optimization complexity of a model for bioethanol production. The ANN has been developed to estimate temperature, thermodynamic variables and biomass, substrate and product concentrations. Subsequently, the ANN model was optimized using Genetic Algorithms and Successive Quadratic Programming to determine the optimal operational conditions. An extractive alcoholic fermentation process was considered. The process consists of four interlinked units: fermentor, centrifuge, cell treatment unit and an ethanol-water separation unit (vacuum flash vessel).
The phase equilibrium in binary mixtures found in the flash vessel has been modeled using the predictive Soave-Redlich-Kwong, with original and modified molecular parameters. The presence of polar substances found in mixture and the many components (others different from ethanol and water) makes it difficult to model these mixtures. The case studied considered five binary water mixtures: acetic acid, acetaldehyde, furfural, methanol and 1-pentanol. These are the substances that are considered to be legal compounds by the Brazilian legislation.
A simulator of this system was developed to explore the possible operational strategies for a large-scale system. The impact of each strategy on the process behavior as well as assessment of possible implementation difficulties are carefully considered and discussed.
See more of Poster Session 2
See more of General Submissions
See more of The 29th Symposium on Biotechnology for Fuels and Chemicals (April 29 - May 2, 2007)
See more of General Submissions
See more of The 29th Symposium on Biotechnology for Fuels and Chemicals (April 29 - May 2, 2007)