Deterministic and Stochastic Optimization of Acid Pretreatment for Lignocellulosic Ethanol Production

• Published in: Computer Aided Chemical Engineering Vol. 40; pp. 2149 - 2154
• Main Authors: Verma, Sumit Kumar, Shastri, Yogendra
• Format: Book Chapter
• Language: English
• Published: 01.01.2017
• Subjects: Acid pretreatment; multiobjective optimization; process optimization; stochastic programming; multiobjective optimization; process optimization; stochastic programming; Acid pretreatment
• ISBN: 9780444639653; 0444639659
• ISSN: 1570-7946
• DOI: 10.1016/B978-0-444-63965-3.50360-3

Pretreatment of lignocellulosic biomass is a critical and cost intensive step in ethanol production requiring optimization. Dilute acid pretreatment depolymerizes xylan into xylose and lignin into acid soluble lignin (ASL). Since ideal conditions for both depolymerization reactions are different, a multi-objective optimization problem (MOOP) has been formulated considering the batch mode of operation and reaction temperature as the decision variable. This study has solved two different bi-objective optimization problems, namely, the maximization of the weighted sum of xylose and ASL yields (problem-1), and minimization of the weighted variance of xylose and ASL yields around a fixed targeted value (problem-2). For both problems, the xylose and ASL yields at the end of the batch are considered. Initially, the deterministic MOOP for both problems was solved using the weighing method. The results showed a strong tradeoff between the xylose and ASL yield and the optimal temperatures varied between 144°C and 113°C depending on the weight. However, the acid pretreatment process is subject to several uncertainties including feedstock composition and kinetic parameters that govern the reactions. Therefore, expected values of both problems were optimized using stochastic MOOP. For MOOP with feedstock composition uncertainty, the results were similar to the deterministic MOOP due to the linear correlation of feedstock composition with yield. In contrast, the results of MOOP with kinetic parameter uncertainty were significantly different. The temperatures for Pareto optimal solutions for problem-1 varied relatively less (133-125°C) for different weights, indicating a conservative operation. Moreover, the temperatures for Pareto optimal solutions for problem-2 were more aggressive than the deterministic case due to highly sensitive xylose degradation kinetics. The use of problem-2 reduced the variability in the xylose yield by an average of 15% over problem-1, and the temperatures changed accordingly.