Multi-objective optimization of operation of lignocellulosic acetone-butanol-ethanol fermentation with ex situ butanol recovery (ESBR)
Introduction
Biobutanol produced from acetone-butanol-ethanol (ABE) fermentation by Clostridium is a promising alternative to established biofuels (e.g. ethanol and methanol), given its advantages like high energy density, low corrosiveness, and easy blending with gasoline (Dürre, 2007). However, there are major hurdles to overcome for commercial scale biobutanol production, including low productivity and yield caused by the toxicity of produced butanol, and high cost of the feedstock, which is glucose typically (Wang et al., 2017). To address these limitations, the possibility of using cheaper feedstocks (e.g., woody biomass), the development of a strain tolerant to butanol toxicity, and of in-situ or ex-situ butanol recovery techniques have been studied (Gottumukkala et al., 2017; Ibrahim et al., 2017).
Lignocellulosic biomass, one of nonedible crops, has extensively been studied as a cheap and sustainable feedstock for biofuel production. The representative hexose and pentose contained in the lignocellulosic biomass are glucose and xylose, respectively. In general, the wild-type Clostridium, a commonly used strain for biobutanol production, prefers glucose over xylose such that it hardly consumes xylose in the presence of glucose, due to a mechanism known as carbon catabolite repression (CCR) (Ren et al., 2010). In order to effectively utilize the lignocellulosic biomass in the ABE fermentation, many researchers have developed genetically modified strains having not only a high butanol tolerance but also the capability to consume glucose and xylose simultaneously (Gu et al., 2010; Ren et al., 2010). Along with these efforts, the possibility of removing butanol during on-going fermentation has been investigated by integrating separation techniques like gas stripping (Ezeji et al., 2004), liquid-liquid extraction (Karcher et al., 2005), perstraction (Karcher et al., 2005), and adsorption (Oudshoorn et al., 2009). Among them, adsorption has been evaluated as the simplest and most energy-efficient (Oudshoorn et al., 2009).
In an effort to combine adsorption with ABE fermentation to enhance the fermentation performance, Eom and coworkers (Eom et al., 2015) suggested a novel continuous fermentation system for butanol production called “ex situ butanol recovery by adsorption system” (referred to as “ESBR-by-adsorption system” hereafter). The system consists of one fermenter and two adsorption columns and the saturated adsorption column is periodically replaced with a fresh one from the regeneration process. This way, the produced butanol is continuously recovered through the adsorption to keep the butanol concentration in the fermenter below its threshold level, thus alleviating the inhibitory effect of butanol on the cell growth and ABE production. The ESBR-by-adsorption system was shown to give a substantially improved butanol productivity, by as much as 5.5 and 3.7 times higher compared to the batch and fed-batch fermentation with in situ butanol recovery (Kim et al., 2017). More recently, this system was tested with the feed of glucose/xylose mixtures using a genetically engineered Clostridium acetobutylicum (Lim et al., 2019b). With the mixture feed, the sugar feed cost could be significantly reduced, by as much as 40% compared to the glucose-based ESBR-by-adsorption system.
Although the glucose/xylose-based ESBR-by-adsorption system can provide a significant improvement in butanol productivity and economic efficiency, it presents certain operational challenges, which arise from the integration of highly nonlinear kinetics of the co-fermentation and the adsorption as well as the cyclic dynamics due to the periodic switching of adsorption columns. In addition to this, multiple, competing objectives typical of fermentation processes, such as maximizing productivity and yield vs. minimizing sugar loss, conflict with each other. Moreover, the productivity and yield of the co-fermentation of glucose and xylose vary significantly with their feed ratio. Therefore, a systematic model-based optimization of the glucose/xylose-based ESBR-by-adsorption system is essential to better understand its dynamic behavior and to optimize its performance.
Several model-based optimization studies have been performed for lignocellulosic butanol fermentation processes integrated with vacuum evaporation (Mariano et al., 2010; Mariano et al., 2009), pervaporation (Sharif Rohani et al., 2015), and gas stripping (Sharif Rohani et al., 2015), but an optimization study of the lignocellulosic ESBR-by-adsorption system is currently missing in literature. In our previous study (Lim et al., 2019b), we developed a mathematical model of the lignocellulosic ESBR-by-adsorption system and conducted a simple cyclic steady state (CSS) optimization to compare performances of the glucose-based and glucose/xylose-based systems. The optimization was performed with respect to a single objective function represented as a weighted-sum of butanol productivity and sugar loss with fixed weights, and therefore did not give insights into how the optimal operation condition would change as the weights of the different objectives are varied. This is important as priorities of the different objectives (e.g., maximizing productivity or yield, or minimizing substrate loss) may change as the feedstock cost changes. Moreover, butanol yield, which is an important performance indicator, was not considered in the previous work. Finally, the feed concentration, a crucial operating degree of freedom, was not considered as a decision variable as only the feeding rate and circulation rate were varied.
This work aims to provide a more comprehensive optimization study of the lignocellulosic ESBR-by-adsorption system. An optimal operation strategy of the system is suggested through a model-based optimization addressing multiple objectives and feedstock ratios. Multi-objective optimization is performed to provide insight into the relationship between major operating variables and operational objectives of the system and to suggest optimal operating conditions, which can improve the system's performance and economic viability. Based on the model developed in our previous work (Lim et al., 2019b), key operating variables, feed concentration, feeding rate, and circulation rate, are optimized for three different objective functions of the biobutanol production, i.e., butanol productivity maximization, butanol yield maximization, and sugar loss minimization. Since the system operates in a periodic manner, the objective functions and constraints are evaluated at the system's CSS, which is calculated through a sequential approach. In addition, as the feedstock source has a considerable influence on the performance of lignocellulosic-based ABE fermentation due to the varying glucose-to-xylose ratio in the lignocellulosic hydrolysates, the optimal operating conditions calculated from the multi-objective optimization are compared for two different lignocellulosic biomass types, empty fruit brunch (EFB) and pine tree. Finally, an open-loop sensitivity analysis study is conducted to identify the most influential input parameters on the system's performance and to identify conditions under which the system's performance remains stable despite uncertainty.
Section snippets
Dynamic Model
In the ESBR-by-adsorption system, the fermenter is connected with the external adsorption column filled with adsorbent resins, and the fermentation broth is circulated between the fermenter and the adsorption column continuously (Fig. 1). Before a cyclic operation, the fermentation proceeds without feeding and circulation until the butanol concentration in fermenter reaches a pre-set level (6-8 g/L). When the butanol concentration reaches the pre-set level, the circulation begins and feeding
Multi-objective optimization
For the operation of fed-batch or continuous fermentation processes, it is important to find an optimal feeding strategy (Mears et al., 2017). In the ESBR-by-adsorption system, the circulation rate affects the speed of adsorption and thus the cycle time (Kim et al., 2017). Thus, for the CSS optimization of the lignocellulosic ESBR-by-adsorption system, the feeding rate, feed concentration and circulation rate are considered as the major operating variables. They are optimized with respect to
Open-loop sensitivity analysis
In order to analyze the effects of disturbances in the operating variables and uncertainties in the model parameters, such as adsorbent degradation and cell variability, on the system outputs at CSS, an open-loop sensitivity analysis is performed. At the optimal CSS operating conditions obtained from the optimization, we investigate how the objective function values, key process outputs (e.g., the concentrations of cell, glucose, xylose, and butanol) and the cycle time are affected by ±10%
Open-loop sensitivity analysis at CSS
The open-loop sensitivity at CSS can be measured by how the end state and objective function values at a new CSS reached after a parameter perturbation differ from those at the reference CSS. Here, the reference CSS is the CSS under an optimal operating condition obtained from the optimization in Section 3, and the new CSS is the one obtained through the successive substitution starting from the reference CSS with a given perturbed parameter. It is assumed that there is no feedback action to
5. Conclusions
A model-based cyclic steady state (CSS) optimization of lignocellulosic ABE fermentation with ex situ butanol recovery by adsorption was performed with respect to multiple operational objectives, which are maximizing productivity, maximizing yield, and minimizing sugar loss. As a result, Pareto optimal solutions were obtained and the relationships between each objective and operating conditions were analyzed. It was identified that butanol productivity can be significantly increased through a
CRediT authorship contribution statement
Ha-Eun Byun: Conceptualization, Methodology, Software, Writing - original draft. Boeun Kim: Conceptualization, Writing - original draft. Jongkoo Lim: Investigation, Resources. Jay H. Lee: Writing - review & editing, Supervision.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgment
This work was supported by the Wastes to Energy Technology Development Program funded by the Ministry of Environment, Republic of Korea (Project No.: 2013001580001) and the National Research Foundation of Korea (NRF) grant funded by the Korean government (MSIT) (NRF-2015M1A8A1076118).
References (27)
- et al.
Periodic states of adsorption cycles—I. Direct determination and stability
Chem. Eng. Sci.
(1994) - et al.
Trends and advances in conversion of lignocellulosic biomass to biobutanol: microbes, bioprocesses and industrial viability
Renew. Sustain. Energy Rev.
(2017) - et al.
Cellulosic biobutanol by clostridia: challenges and improvements
Renew. Sustain. Energy Rev.
(2017) Effective implementation of the ε-constraint method in multi-objective mathematical programming problems
Appl. Math. Comput.
(2009)- et al.
A review of control strategies for manipulating the feed rate in fed-batch fermentation processes
J. Biotechnol.
(2017) - et al.
Multi-objective dynamic optimization study of fed-batch bio-reactor
Chem. Eng. Res. Des.
(2017) - et al.
Identification and inactivation of pleiotropic regulator CcpA to eliminate glucose repression of xylose utilization in clostridium acetobutylicum
Metabolic Eng.
(2010) - et al.
Current advances on fermentative biobutanol production using third generation feedstock
Biotechnol. Adv.
(2017) - et al.
On-constraint based methods for the generation of Pareto frontiers
J. Mech. Eng. Autom.
(2013) - et al.
A fast and elitist multiobjective genetic algorithm: NSGA-II
IEEE Trans. Evol. Comput.
(2002)