Delft University of Technology

Production of ethanol fuel via syngas fermentation

Optimization of economic performance and energy efficiency
de Medeiros, Elisa M.; Noorman, Henk; Maciel Filho, Rubens; Posada, John A.

DOI
10.1016/j.cesx.2020.100056
Publication date
2020
Document Version
Final published version
Published in
Chemical Engineering Science: X

Citation (APA)
de Medeiros, E. M., Noorman, H., Maciel Filho, R., & Posada, J. A. (2020). Production of ethanol fuel via
syngas fermentation: Optimization of economic performance and energy efficiency. Chemical Engineering
Science: X, 5, [100056]. https://doi.org/10.1016/j.cesx.2020.100056
Important note
To cite this publication, please use the final published version (if applicable).
Please check the document version above.

Copyright
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent
of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Takedown policy
Please contact us and provide details if you believe this document breaches copyrights.
We will remove access to the work immediately and investigate your claim.

This work is downloaded from Delft University of Technology.

For technical reasons the number of authors shown on this cover page is limited to a maximum of 10.
 Chemical Engineering Science: X xxx (xxxx) xxx

Contents lists available at ScienceDirect

Chemical Engineering Science: X

journal homepage: www.elsevier.com/locate/cesx

Production of ethanol fuel via syngas fermentation: Optimization of

economic performance and energy efficiency
Elisa M. de Medeiros a,b,⇑, Henk Noorman a,c, Rubens Maciel Filho b, John A. Posada a
a
Department of Biotechnology, Delft University of Technology, van der Maasweg 9, Delft 2629 ZH, the Netherlands
b
School of Chemical Engineering, University of Campinas, Av. Albert Einstein 500, Campinas 13084-852, Brazil
c
DSM Biotechnology Center, PO Box 1, Delft 2600 MA, the Netherlands

a r t i c l e i n f o a b s t r a c t

Article history: In this work, a model was developed to predict the performance of a bubble column reactor for syngas
Received 13 September 2019 fermentation and the subsequent recovery of anhydrous ethanol. The model was embedded in an opti-
Received in revised form 16 January 2020 mization framework which employs surrogate models (artificial neural networks) and multi-objective
Accepted 21 January 2020
genetic algorithm to optimize different process conditions and design variables with objectives related
Available online xxxx
to investment, minimum selling price, energy efficiency and bioreactor productivity. The results indicate
the optimal trade-offs between these objectives while providing a range of solutions such that, if desired,
Keywords:
a single solution can be picked, depending on the priority conferred to different process targets. The
Bioethanol production
Renewable energy
Pareto-optimal values of the decision variables were discussed for different case studies with and without
Modeling the recovery unit. It was shown that enhancing the gas-liquid mass transfer coefficient is a key strategy
Multi-objective optimization toward sustainability improvement.
Syngas fermentation Ó 2020 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://
Artificial neural networks creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction constitutes a valorization pathway for works arising gases (WAG)

from steel production containing large amounts of CO, as demon-
Biofuels are one of the possible means to reduce CO2 emissions strated by the increasing number of commercial projects led by
in transportation, a sector responsible for roughly 25% of global LanzaTech and partners in the past years (ArcelorMittal, 2019;
greenhouse gas (GHG) emissions in 2016 (IEA, 2019). However, Biofuels Digest, 2019; Renewables Now, 2018).
the rapid expansion of 1st generation biofuels (i.e. those produced In order to upgrade the technology readiness level (TRL) and
from food crops) has also been associated with impacts that were achieve full commercialization, not only scaling-up technical
initially neglected or unforeseen, such as deforestation, indirect issues need to be resolved, but the process also needs to be boosted
land use change, and significant GHG emissions during agricultural at different levels, for example: improving gas-liquid mass transfer
stages (Goldemberg et al., 2008). Lignocellulosic or 2nd generation through better reactor design; genetically engineering bacteria;
biofuels have the potential to minimize these impacts by using adjusting the gasification process to deliver syngas with favorable
waste carbon materials such as agricultural residues, forestry composition; reducing energy use in the product recovery unit;
waste and marginal land (energy) crops. Yet, despite large efforts and tuning the process conditions at all units simultaneously to
towards commercialization, most of these production pathways approach overall optimal operation. Most published research about
face technical challenges that are inherent to emerging technolo- syngas fermentation has focused on running experiments to test
gies. An alternative path to biofuels and the object of this work is the capabilities of different strains and reactor configurations,
syngas fermentation, the microbial conversion of CO/H2/CO2 to and to study the effects of changing nutritional composition of
ethanol and potentially other chemicals using facultative auto- the liquid medium as well as the pH, such as done by Abubackar
trophic bacteria called acetogens. Since syngas can be produced et al. (2015). Many studies have also tried to elucidate the metabo-
via gasification of multiple types of feedstocks – including even lism of these microbes so that yield and selectivity of the desired
municipal solid waste –, this process is considered a promising product (usually ethanol) can be improved (Richter et al., 2016).
and flexible alternative to biochemical routes that rely on the lib- On the other hand, studies about modeling, simulation and opti-
eration of sugars from lignocellulosic biomass. Moreover, it also mization are still limited, despite being of paramount importance
for the evaluation of feasibility and comprehension of these sys-
⇑ Corresponding author. tems. A few recent works have discussed results obtained with
E-mail address: [email protected] (E.M. de Medeiros).
simulations using Aspen plus, such as de Medeiros et al. (2017),

https://doi.org/10.1016/j.cesx.2020.100056
2590-1400/Ó 2020 Published by Elsevier Ltd.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Please cite this article as: E. M. de Medeiros, H. Noorman, R. Maciel Filho et al., Production of ethanol fuel via syngas fermentation: Optimization of eco-
nomic performance and energy efficiency, Chemical Engineering Science: X, https://doi.org/10.1016/j.cesx.2020.100056
2 E.M. de Medeiros et al. / Chemical Engineering Science: X xxx (xxxx) xxx

Nomenclature

Greek symbols EtOH ethanol

c liquid phase activity coefficient FBM bare module cost factor
eG gas hold-up FST feed stage (T-01 or T-02)
g thermodynamic efficiency GRR gas recycle ratio (BCR)
l biomass growth rate [h
1] GRT gas residence time at the bottom of the reactor [min]
mi consumption or production rate of component i Hi Henry’s law constant of component i in water [Pa]
[molg
1h
1] HAc acetic acid
qL mass density of liquid [kgm
3] HMT high mass transfer (case study)
kd cell death rate [h
1]
Roman symbols kLa volumetric mass transfer coefficient [h
1]
a bubble column operation mode (1 for concurrent,
1 KT temperature adjustment coefficient for kLa
countercurrent) L length of vessel (BCR) [m]
AE anhydrous ethanol LHV lower heating value [MJkg
1 or MJkmol
1]
ANN artificial neural network ṁ mass flow rate [kgh
1 or kgyear
1]
AR vessel aspect ratio (BCR) MESP minimum ethanol selling price [$L
1]
BCR bubble column reactor MML million liters
CG,i, CL,i concentration of component i (gas or liquid phase) MML molar mass liquid phase [gmol
1]
[molm
3 or gL
1] MOGA multi-objective genetic algorithm
C*G,i, C*L,i Saturation concentration of component i (gas or liquid ṅMT mass transfer rate [molm
3h
1]
phase) [molm
3] NST number of stages (T-01 or T-02)
CL,X concentration of cells in the broth [gL
1] OF objective function
COL, CRM, CUT, CWT costs of operating labor, raw materials, utilities Psat,i vapor pressure of component i [Pa]
and waste treatment [$/year] QG, QL volumetric flow rate of gas or liquid in the BCR [m3h
1]
CBM, CTM mare module cost, total module cost [$] R gas constant = 8.314 Jmol
1K
1
COMd manufacturing costs excluding depreciation costs RR molar reflux ratio (T-01 or T-02)
C 0p purchase cost of equipment at base conditions S:F mass ratio side stream to feed stream
dC vessel diameter (BCR) [m] T temperature [°C or K]
D:F mass ratio distillate stream to feed stream uG, uL gas or liquid superficial velocity [mh
1]
Dfi diffusivity coefficient of component i in water [cm2s
1] VL volume of liquid (BCR) [m3]
DL axial dispersion coefficient [m2h
1] VR volume of reactor vessel (BCR) [m3]
Drate dilution rate in the BCR [h
1] Xi gas conversion (BCR), i = CO, H2
DV decision variable XP cell purge fraction (BCR)

Benalcázar et al. (2017), Pardo-Planas et al. (2017) and Roy et al. (iii) Development of a multi-objective optimization framework
(2015). Though these studies look at the whole process from feed- for the integrated process of anhydrous ethanol production
stock (biomass) to ethanol and provide meaningful estimates of from syngas via fermentation and distillation.
process performance in different domains (technical, economic (iv) Evaluation of Pareto-optimal solutions for the most impor-
and environmental), the process conditions were mostly fixed or tant input variables of the system, in terms of economic per-
in some cases changed through univariate sensitivity analysis, formance, thermodynamic efficiency and productivity.
and simulation of the bioreactor was based on strong assumptions
and simplifications with limited connection to the operating condi- 2. Methodology
tions. In a previous study (de Medeiros et al., 2019), we discussed
the contributions of more elaborate bioreactor models developed 2.1. Process overview
by other authors and presented the model of a dynamic continuous
stirred tank (CSTR), demonstrating its application for sensitivity Fig. 1 presents the conceptual process flowsheet to produce
analysis and technical optimization. In the present work, we anhydrous ethanol from syngas, which consists of three main unit
extended the CSTR model to a bubble column reactor model operations: (i) the syngas bioreactor (R-01); ethanol distillation (T-
(BCR) with distribution of key process variables in time and space, 01 and T-02) to achieve azeotropic composition (also called
and employed it in the optimization of operating (e.g. dilution rate) hydrous ethanol); (iii) ethanol dehydration in molecular sieves
and design (e.g. reactor size) variables considering different objec- (T-04 and T-05) to achieve 99.5% ethanol. There is also a vent
tives in parallel, both technical and economic. The BCR model was scrubber (T-03) to recover any ethanol present in CO2 off-gases.
first considered standalone and subsequently integrated with the Three main types of stream are recycled to the bioreactor: (i) water
distillation unit, whose input was concurrently optimized. The from the distillation bottoms and the scrubber; (ii) unconverted
contributions of this paper can be summarized as follows: syngas; and (iii) microbial cells. Sections 2.2 and 2.3 lay out the
methodology for process design and modeling of the two main
(i) Development of a spatial dynamic model of a bubble column blocks of the process: fermentation and product recovery.
reactor (BCR) for syngas fermentation considering kinetics of
cell growth and death, mixed product formation and acetic 2.2. Model of syngas fermentation in bubble column reactor (BCR)
acid re-assimilation.
(ii) Application of artificial neural networks as surrogate models The syngas bioreactor consists of a bubble column where syn-
for two types of intricate models: BCR model and distillation gas enters at the bottom and off-gas leaves from the top, a fraction
model (nonlinear system of MESH equations – material-equi of it being recycled back to the bottom after compression and cool-
librium-summation-enthalpy – solved in Aspen Plus). ing. Liquid flows continually in and out of the reactor and a fraction

Please cite this article as: E. M. de Medeiros, H. Noorman, R. Maciel Filho et al., Production of ethanol fuel via syngas fermentation: Optimization of eco-
nomic performance and energy efficiency, Chemical Engineering Science: X, https://doi.org/10.1016/j.cesx.2020.100056
 E.M. de Medeiros et al. / Chemical Engineering Science: X xxx (xxxx) xxx 3

Fig. 1. Simplified PFD of anhydrous ethanol production from syngas. C-01: gas compressor; E-01 to E-09: heat exchangers; MF-01: microfiltration membrane; R-01: syngas
bubble column reactor; T-01 and T-02: distillation columns; T-03: vent scrubber; T-04 and T-05: adsorption and regeneration columns.

of bacterial cells are recycled after separation from the products in the gas and in the liquid, i.e. CG,i and CL,i where i = CO, H2, CO2, etha-
a microfiltration membrane. In the reactor, cells are dispersed in nol (EtOH), acid acid (HAc), and H2O; and the concentration of cells
the liquid phase, where they consume CO, H2 and CO2 and produce in the liquid CL,X [gm
3]. Eqs. (1)–(3) summarize the governing
ethanol, acetic acid and CO2. The latter is both a product from the partial differential equations (PDEs) which are completed by the
oxidation of CO and a source of carbon in the reaction with H2, algebraic equations for the calculation of mass transfer rates (Eq.
therefore its consumption or production rate will depend on the (4)) and cell kinetics. For the latter, the variables mi (production/-
availability of the other two molecules. consumption rates), l (cell growth rate) and kd (cell death rate)
Similar models for this process with spatiotemporal distribu- are calculated or specified as explained in de Medeiros et al.
tion have been presented by Chen et al. (2018) and Li et al. (2019), at each discrete point of the system. The physical proper-
(2019), the main difference being the method to calculate the con- ties Hi, Psat,i and ci can also be found in de Medeiros et al. (2019).
sumption/production rates of components which in their works In Eqs. (2) and (3), a = 1 if the column is operating concurrently
was done via Flux Balance Analysis (FBA) while in the present and a =
1 otherwise. For the optimizations in the present work,
study we employ the microbial kinetics developed in our previous the latter is adopted.
work (de Medeiros et al., 2019). Moreover, the procedure to esti-
mate the gas velocity profile and hydrodynamic parameters (as- @C G;i uG @C G;i C G;i @uG n_ MT
¼


ð1Þ
sumed constant by Chen et al. (2018)) is different. FBA is a @t eG @x eG @x eG
powerful tool for the comprehension of cell metabolism, enabling
for example in silico simulations about the effects of gene deletion @C L;i auL @C L;i @ 2 C L;i n_ MT
and knockout; however, it also adds a significant amount of com- ¼
 þ DL þ þ mi C L;X ð2Þ
@t ð1
eG Þ @x @x2 ð1
e G Þ
plexity to the model with the inclusion of a linear programming
problem that must be solved during the integration of the ODEs.
Furthermore it requires the definition of an objective function @C L;X auL @C L;i @ 2 C L;i
¼
 þ DL þ ðl
kd ÞC L;X ð3Þ
(e.g. maximizing the biomass growth rate) and flux bounds which @t ð1
eG Þ @x @x2
are not always known and might not be constant during the pro-
cess. The kinetic model used in this work, on the other hand, is
 C G;i Hi  MML
based on global reactions for the consumption of gases and forma- n_ MT ¼ kL ai C L;i
C L;i ; C L;i ¼ ; mi ¼ ; i
mi R  T  qL
tion of products, embedding parameters that were previously esti-
mated with experimental data. ¼ CO; H2 ; CO2 ð4aÞ
The BCR was described here with the axial dispersion model of
Deckwer (1992), considering the following assumptions: (i)
 C L;i qL  R  T
isothermal operation at T = 37 °C; (ii) axial dispersion considered n_ MT ¼
kL ai C G;i
C G;i ; C G;i ¼ ; mi ¼ i
mi MML  ci  Psat;i
in the liquid phase but neglected in the gas phase; (iii) the pressure
¼ EtOH; HAc; H2 O ð4bÞ
profile is calculated from the liquid head; (iv) biochemical reac-
tions occur in the liquid phase with rates depending on the concen- The gas hold-up eG and the volumetric mass transfer coefficient
trations of the components; (v) gas velocity changes along the kLa [h
1] were calculated from the correlations by Heijnen and
column as the gas shrinks due to microbial conversion or expands van’t Riet (1984), adapted in Eqs. (5) and (6), where uG [mh
1] is
due to pressure reduction; (vi) hydrodynamic parameters are a the gas superficial velocity. In Eq. (6), the first term between brack-
function of the gas superficial velocity assuming heterogeneous ets is the kLa for air in water at 20 °C, which is adjusted to compo-
flow in air-water systems. In total 13 state variables are distributed nent i using the diffusivity coefficient Dfi, and to the reactor
in space: the concentrations [molm
3] of six chemical species in temperature by means of the coefficient KT = (1.024)(T
20) = 1.5

Please cite this article as: E. M. de Medeiros, H. Noorman, R. Maciel Filho et al., Production of ethanol fuel via syngas fermentation: Optimization of eco-
nomic performance and energy efficiency, Chemical Engineering Science: X, https://doi.org/10.1016/j.cesx.2020.100056
4 E.M. de Medeiros et al. / Chemical Engineering Science: X xxx (xxxx) xxx

(Heijnen and van’t Riet, 1984). The axial dispersion coefficient DL dance with the abovementioned design, though with minor modi-
[m2h
1] (Eq. (7)) is calculated with the correlation provided by fications: 25 stages, feed stream on the 5th stage from the top,
Deckwer et al. (1974). vapor-side stream removed from the 7th stage, top pressure of
2 bar, and molar reflux ratio (RR) of 3:1. Though in the design pre-
eG ¼ 0:6ðuG =3600Þ0:7 ð5Þ sented by Humbird et al. (2011) the fermentation broth had a
h
i
1=2 higher ethanol concentration (5.4% w/w), increasing RR beyond
kL a ¼ ð3600Þ 0:32  ðuG =3600Þ0:7  Df i =Df O2 ;air  KT ð6Þ 3:1, even for lower ethanol concentrations, was found to have mar-
ginal effects on the separation performance, especially if consider-

 ing the increased reboiler duty. The mass flow rate of the top
1:4
DL ¼ ð0:36Þ 2:7  ðdC  100Þ  ðuG =36Þ0:3 ð7Þ stream, which is rich in CO2, was defined so as to ensure that most
of the dissolved CO2 is removed, therefore it was fixed at 1.1 times
Finite differences were used to transform the system of partial
the mass flow rate of CO2 present in the feed stream.
differential algebraic equations (PDAE) to a system of ordinary dif-
The hot bottom stream from T-01, which contains most of the
ferential algebraic equations (DAE). This stiff problem was solved
acetic acid from the feed, is used to pre-heat the latter (in E-02),
in MATLAB R2015b using ode15s, which is usually the first recom-
whereas the CO2 stream, together with off-gas from the BCR, is
mended option before trying other stiff solvers such as ode23s. The
fed to the vent scrubber (T-03) to recover any ethanol that is pre-
state vector contained thus 13∙N variables where N = 15 is the
sent in these streams. The outlet liquid stream from T-03 is then
number of discretization points. The boundary conditions for both
mixed with bottoms from both distillation columns and part of
phases at the bottom (x = 0) and at the top (x = L), considering
the resulting stream is recycled to the bioreactor while the rest is
counter-current operation, are shown in Eqs. (8) and (9).
sent to wastewater treatment. The concentrated ethanol stream

 @C L;i  from T-01, which contains between 20 and 40% w/w depending
C G;i x¼0
C in
G;i ¼ 0 ; ¼0 ð8Þ
@x  x¼0
on the case, is sent to the rectification column (T-02) where it is
further concentrated to near-azeotropic composition (92.5% w/
  
@ C G;i  uG  
 in ð1
eG ÞDL @C L;i  w). The only fixed parameter in T-02 is the pressure (atmospheric);
 ¼ 0 ; C
C þ  ¼0 ð9Þ
@x x¼L all the other inputs are decision variables, as explained further in
L;i x¼L
@x x¼L
L;i
uL
Section 2.5. The distillate from T-02 is removed as saturated vapor
The gas velocity uG was calculated along the reactor by applying and superheated to 116 °C before being sent to dehydration with
a mole balance in the gas phase of each compartment starting from molecular sieves (columns T-04 and T-05). This unit was not mod-
the bottom. If the gas recycle ratio (GRR) is greater than zero elled but instead the outcomes were fixed considering two output
(where GRR is the ratio between the gas flow rate recycled to the streams: (i) the product (anhydrous ethanol, 99.5% w/w) which is
bottom of the bioreactor and the total gas flow rate at the top), then cooled to 38 °C (E-08); and (ii) a low-purity ethanol stream
then the flow rate and composition of the inlet are different from (72% w/w) produced during the regeneration step, which is recy-
those of the fresh syngas. In these cases an iterative procedure cled back to the rectifier.
was applied to find the correct properties of the inlet stream and The distillation columns were simulated in Aspen plus using the
the gas velocity profile: first an initial guess is assumed at the bot- RadFrac model which solves the system of MESH equations with
tom and used for the calculation of the velocity along the column, the Inside-Out algorithm. Four components were considered:
the new properties calculated at the top are then used to re- water, ethanol, acetic acid and CO2. The property methods were:
calculate the properties at the bottom, which replace the initial for the liquid phase, the activity coefficient model NRTL, due to
guess in the next iteration; the velocity profile is calculated again the non-ideal ethanol-water mixture; and for the gas phase,
and this procedure is repeated until the differences are negligible Hayden-O’Connell equation-of-state with Henry’s law, due to low
between two consecutive iterations. Though this subroutine makes pressure, presence of acetic acid which may cause vapor phase
the model more intricate, neglecting the change in gas velocity is association, and presence of dissolved gases (CO2).
an unrealistic assumption (Deckwer, 1976) and substantial differ-
ences (>50%) between the velocity at the inlet and outlet have been 2.4. Artificial neural network (ANN) surrogate models
observed not only with the model developed here but also by Li
et al. (2018). Surrogate modeling techniques have been applied by many
authors in the past years with the goal of reducing the complexity
2.3. Product recovery unit of computationally expensive models. Throughout this paper we
refer to the latter as ‘‘rigorous” models, although they also incorpo-
Anhydrous ethanol is recovered using distillation and molecular rate some level of simplification (e.g. equilibrium equations in the
sieve adsorption. In the present work, the design of these calculation of distillation columns).
operations was based on the flowsheet described by Humbird Recent works, such as Ibrahim et al. (2018) and Ye et al. (2019)
et al. (2011) for purification of ethanol produced via 2nd- have demonstrated the efficacy of this methodology for separation
generation biochemical route (dilute-acid pretreatment and enzy- processes, specifically for distillation and adsorption. In the present
matic hydrolysis). First, the dilute broth (1–4% w/w ethanol) is fed work, ANNs were trained with input from the two types of models
to a beer column (T-01), from which water is removed at the bot- laid out in Sections 2.2 and 2.3: (i) the PDAE system describing the
tom and CO2 at the top, while concentrated ethanol is removed as a BCR; and (ii) the RadFrac models in Aspen plus describing the dis-
vapor-side stream and fed to the rectification column (T-02). The tillation columns. In both cases, multi-layer feedforward networks
design choices adopted by Humbird et al. (2011) were considered were trained with Bayesian regularization backpropagation using
as starting point and a preliminary analysis was done to evaluate the ANN Toolbox in MATLAB R2015b. The networks contained
the effects of different parameters (results not shown here). In one input layer, one output layer and two hidden layers, with the
the first column it was observed that the mass ratio of side stream number of neurons per hidden layer being chosen after a few tests
to feed stream (S:F) had significantly stronger effects on the out- (in all cases between 5 and 25). For some responses, an ensemble
come (reboiler duty, ethanol recovery and concentration) than of networks was used to reduce the error of a single network by
other parameters, therefore this was chosen as the only decision averaging the predictions of the individual networks and removing
variable in this column. The other parameters were fixed in accor- outliers. For all trained ANNs the correlation coefficient (R value)

Please cite this article as: E. M. de Medeiros, H. Noorman, R. Maciel Filho et al., Production of ethanol fuel via syngas fermentation: Optimization of eco-
nomic performance and energy efficiency, Chemical Engineering Science: X, https://doi.org/10.1016/j.cesx.2020.100056
 E.M. de Medeiros et al. / Chemical Engineering Science: X xxx (xxxx) xxx 5

between network output and original response was higher than response is not straightforward because it depends on the gas
0.99. hold-up which is not fixed); (v) total cooling duty to keep the reac-
tor isothermal.
2.4.1. Bubble column reactor ANNs
Data was generated with the BCR model by varying the input 2.4.2. Distillation columns
vector and exporting the steady-state responses obtained after Data for the distillation columns was generated using Aspen
integration of the ODE system over a sufficiently long time span. plus and Aspen Simulation Workbook, which was used to launch
Around 2000 points were used to train the ANNs. The input vector the simulations automatically (roughly 2000 points for each
comprises in total 12 variables which are needed to define the DAE column). The main aspects of the simulations were detailed in
system, though not all of them are used as decision variables in the Section 2.3; and the following variables were considered inputs
optimization: (i) dilution rate (Drate), defined as the liquid volumet- for the ANNs, some of them being outputs of the bioreactor and
ric flow rate (QL) divided by the liquid volume (VL) in the bioreac- others being key parameters related to the design and operation
tor; (ii) the volume of liquid divided by the gas volumetric flowrate of the distillation column: (i) inlet mass fractions of ethanol, acetic
at the bottom of the reactor (VL/Qg,bot) which we refer to here as gas acid and CO2; (ii) number of stages (NST); (iii) feed stages (FST); (iv)
residence time (GRT); (iii) gas recycle ratio (GRR) defined in Sec- molar reflux ratio (RR); (v) mass ratio of side to feed stream (S:F),
tion 2.2; (iv) cell purge fraction (XP), which is the fraction of bio- only for T-01; (vi) mass ratio of distillate to feed stream (D:F), only
mass cells that are not recycled to the bioreactor; (v) vessel for T-02; (vii) pressure at the top; (viii) feed stream mass flow rate.
length (L); (vi) vessel aspect ratio (AR); (vii) and (viii) concentra- It should be noted that NST and FST are continuous variables when
tions of acetic acid and ethanol in the inlet liquid stream; (ix) to provided to the ANNs and for their training, but they are trans-
(xii) molar fractions of CO, H2, CO2 and H2O in the fresh syngas formed to integer values before running the Aspen simulations.
(note: for the optimization studies the syngas composition was FST is written as a fraction of the number of stages, such that
fixed at 50% CO and 50% H2). It should be noted that although FST = 0 when the feed stage is the first after the reboiler and
the liquid inlet composition (items vii and viii) is an input of the FST = 1 when it is the last before the condenser.
bioreactor, it is not an input of the integrated process (and even The following outputs were considered: (i) mass fractions of
less a decision variable in the optimization), because it depends ethanol and acetic acid at the bottom, at the top, and at the side
on the properties of the recycle streams from the distillation unit. stream (only for T-01); (ii) same for temperatures; (iii) column
The input matrix was generated using Latin hypercube sam- diameter; (iv) reboiler and condenser duties.
pling (LHS) to ensure good coverage of the input domain, which Multilayer feedforward neural networks were trained for both
considers realistic ranges of these variables (presented together pattern recognition and function approximation. In the first case,
with optimization results in Tables 1 and 2). ANNs were trained the ANNs were trained to classify the input vector as feasible or
to predict ten responses that are needed in the calculation of the infeasible according to the convergence status of the Aspen simu-
objective functions or as input for the distillation unit. These lations. The results of the successful simulations were then used to
responses belong to five categories: (i) concentrations of ethanol, train the ANNs as explained previously in Section 2.4.
acetic acid, cell biomass and dissolved CO2 at the liquid outlet;
(ii) gas superficial velocities at the bottom and at the top; (iii) 2.5. Optimization framework
gas conversions of CO and H2; (iv) pressure at the bottom (this
Multi-objective genetic algorithm (MOGA) was employed to
first optimize the standalone BCR and later the integrated process,
Table 1
with regard to the following objectives: (i) in the BCR, maximiza-
Multi-objective optimization of the bubble column reactor: ranges of decision
variables and process outcomes at the Pareto-optimal solutions. tion of ethanol productivity and minimization of CAPEX; and (ii)
in the whole process (shown in Fig. 1), minimization of CAPEX
Search space base HMT
per annual production of anhydrous ethanol (ṁAE), minimization
Drate [h
1] 0.01–0.15 0.088–0.14 0.11–0.15 of minimum ethanol selling price (MESP) and maximization of
GRT [min] 5–50 6.41–36.9 5.15–35.8
thermodynamic efficiency (g). MOGA aims at finding a set of
GRR 0–0.5 0.032–0.43 0.017–0.32
XP 0.05–0.2 0.035–0.049 0.025–0.056
non-dominated (i.e. optimal) solutions approaching the true Pareto
L [m] 25–50 35.6–47.3 35.5–42.5 front of the system, i.e. those feasible points for which one objec-
VR [m3] 500–900 502–572 500–548 tive function can only be reduced if other objective functions are
CL,EtOH [gL
1] – 6.21–34.1 12.5–39.5 increased. It’s worth noting that MOGA does not use weighted
CL,HAc [gL
1] – 1.35–2.75 2.17–2.98
functions to transform multiple objectives into one, therefore it
X CO – 0.56–0.74 0.81–0.93
X H2 – 0.42–0.66 0.56–0.87 ensures that objectives of different natures are considered fairly.
Though other methods exist for multi-objective optimization,

Table 2
Multi-objective and single-objective optimization of the integrated process: ranges of decision variables at the Pareto-optimal or optimal solutions.

Search space CAPEX  g CAPEX  g (HMT) MESP  g MESP  g (HMT) MESP MESP (HMT)
Drate (BCR) [h
1] 0.01–0.15 0.044–0.092 0.1022–0.1025 0.050–0.087 0.101–0.103 0.099 0.106
GRT (BCR) [min] 5–50 8.95–44.4 15.7–17.7 15.9–39.1 18.3–18.9 14.5 21.3
GRR (BCR) 0–0.5 0.062–0.43 0.489–0.493 0.092–0.45 0.478–0.496 0.0086 0.061
XP (BCR) 0.05–0.2 0.084–0.19 0.0553–0.0639 0.12–0.17 0.044–0.054 0.198 0.0902
L (BCR) [m] 25–50 44.9–49.9 41.3–43.4 48.3–49.9 42.6–42.8 49.9 47.2
VR (BCR) [m3] 500–900 710–853 630–645 713–811 645–654 633 595
S:F (T-01) 0.04–0.3 0.094–0.106 0.0796–0.0818 0.096–0.10 0.079–0.080 0.0963 0.0824
NST (T-02) 40–50 40–46 43–45 43–47 44 41 40
FST,V (T-02) 0.6–0.9 0.62–0.76 0.755–0.773 0.61– 0.80 0.76–0.78 0.67 0.724
FST,L (T-02) 0.2–0.5 0.26–0.47 0.336–0.359 0.297–0.496 0.34–0.36 0.22 0.456
RR (T-02) 3–6 4.4–5.80 3.78–3.89 5.14–5.83 3.95–4.02 5.25 4.17

Please cite this article as: E. M. de Medeiros, H. Noorman, R. Maciel Filho et al., Production of ethanol fuel via syngas fermentation: Optimization of eco-
nomic performance and energy efficiency, Chemical Engineering Science: X, https://doi.org/10.1016/j.cesx.2020.100056
6 E.M. de Medeiros et al. / Chemical Engineering Science: X xxx (xxxx) xxx

MOGA was chosen due to its robustness and easy implementation ditions (C 0p ), estimated for each equipment of capacity A using Eq.
using the gamultiobj function in MATLAB R2015b, which employs a (10). The total module cost (CTM), which considers total direct
variant of the NSGA-II method (non-dominated sorting genetic and indirect costs, as well as contingencies and fees, is calculated
algorithm). as shown in Eq. (11). In the present work, CAPEX is considered
For the standalone BCR, a population of 100 individuals was the same as CTM. The bare module cost of equipment i, CBM,i, is cal-
considered, while for the whole process populations ranging from
culated from C 0p;i applying bare module cost factors FBM,i which
250 to 1000 were tested at multiple runs.
account for pressure, materials and other items such as installation
The objectives were always defined as minimization functions,
and engineering. The coefficients K1 and K2 and the procedures to
hence in the first case (standalone BCR), the two objectives were:
calculate FBM for each type of equipment can be consulted in
OF 1 ¼
productiv ity ¼
Drate  C L;EtOH and OF 2 ¼ CAPEX BCR . In the
Turton et al. (2009).
second case (whole process), two pairs of objectives were evalu-
ated: (i) OF 1 ¼ CAPEX=m _ AE , OF 2 ¼
g; and (ii) OF 1 ¼ MESP, 2
log10 C 0p ¼ K 1 þ K 2 log10 ðAÞ þ K 2 ½log10 ðAÞ ð10Þ
OF 2 ¼
g. All the optimizations were performed for two case stud-
ies regarding mass transfer capacity: a base case and a high mass X
transfer case (HMT) in which the kLa calculated with Eq. (6) was C TM ¼ 1:18  C BM;i ; C BM;i ¼ C 0p;i  F BM;i ð11Þ
multiplied by a factor of 3, representing thus a hypothetical case i

where process intensification methods are employed to increase The manufacturing costs COMd (also called operating expenses or
mass transfer (this is further discussed in Section 3.3). A summary OPEX) comprise direct costs (e.g. raw materials [CRM], utilities
of the steps required to compute the objective functions is given [CUT], waste treatment [CWT], maintenance, operating labor [COL]),
next, along with the definition of the decision variables (DVs) (note fixed costs (e.g. overhead, depreciation, insurance) and general
that for the standalone BCR, the procedure stops after the 1st step): expenses (e.g. administration, R&D). The different components of
the manufacturing costs can be estimated using multiplication fac-
1. DVs related to the BCR (Drate, GRT, GRR, XP, L, VR) and fixed tors based on historical data; adding up all these costs leads to an
inputs (e.g. gas composition) are provided to the BCR model, expression such as Eq. (12), used in the present work, which con-
which employs ANNs to calculate multiple responses (broth siders midpoint values from the ranges reported in Turton et al.
composition, gas conversions, cooling requirements). The (2009).
design variables and some of the responses are also used to cal-
culate economic variables (investment, cost of raw materials COMd ¼ 0:18  C TM þ 2:73  C OL þ 1:23  ðC UT þ C WT þ C RM Þ ð12Þ
and utilities), as explained in Section 2.5.1. Pure water is consid-
The purchase price of syngas was considered 1 US$/kmol for the
ered as initial guess for the liquid feed of the bioreactor (this is
optimization runs, but its impact on MESP was later evaluated
later adjusted as explained in step 4). Power consumption and
along with other economic parameters. The assumption was based
cooling requirements associated with the gas compressor
on the ranges estimated with techno-economic modeling studies
(C-01) are also calculated in this step.
performed by Yao et al. (2018) and Martinez-Gomez et al. (2017)
2. Applicable responses from step 1 (e.g. broth composition), DVs
for syngas production from biomass and shale gas. Fuel for steam
related to T-01 (S:F) and fixed inputs are provided to the ANNs
generation was considered to be sugarcane bagasse in Brazil,
that calculate the outputs of T-01 (compositions of the streams,
assuming a price of 2 US$/GJ which is inside the reported range
reboiler and condenser duties, column diameter). Similarly,
from the last years (CanaOnline, 2016). The costs of utilities and
some of the outputs (e.g. reboiler duty) and inputs (e.g. number
waste treatment/disposal were calculated according to the
of stages) are used to calculate economic responses.
methodology of Ulrich and Vasudevan (2006), also assuming an
3. Applicable responses from step 2, DVs related to T-02 (NST, FST,V,
electricity price in accordance with the Brazilian context, at 0.12
FST,L and RR) and fixed inputs are provided to the model of T-02.
US$/kWh (CPFL Energia, 2019). The cost of operating labor (COL)
In this step, the distillate to feed ratio (D:F) in T-02 is determined
was assumed to be 3% of CAPEX.
through a minor optimization routine to guarantee an ethanol
With CAPEX (CTM) and OPEX, the discounted cash flows can be
mass fraction of 0.925 in the distillate. The ANNs are employed
calculated and the minimum ethanol selling price (MESP) can be
again with the complete input vector (including D:F) to generate
found with a subordinate optimization routine to achieve
the relevant responses; other outputs and inputs are used to
NPV = 0, where NPV (Net Present Value) is the cumulative dis-
compute economic results, and heating/cooling requirements
counted cash flow at the end of the project. For this calculation a
are also calculated for the adsorption unit.
few considerations were made: (i) 15 years of project life and
4. With the outputs from steps 2 and 3, the composition of the liq-
2 years of construction; (ii) land cost = 7.5% CTM; (iii) taxation
uid recycle stream is calculated and compared with the initial
rate = 35%; (iv) interest rate = 10%; (v) straight-line depreciation;
guess. If required, an iterative procedure is followed to match
(vi) project salvage value = 0; (vii) Chemical Engineering Plant Cost
both streams.
Index (CEPCI) = 615.9 as of December 2018 and CEPCI = 397 in base
5. If the ethanol mass fraction in the distillate from T-02 is 0.925
year (2001) (Chemical Engineering Magazine, 2019); and (viii)
and other requirements are met (no negative mass flows, ANN
350 days of operation per year.
inputs inside their training ranges, equipment capacities inside
feasible ranges), then the outputs from steps 1–4 are combined
to generate the objective functions, as further explained in 2.5.2. Computation of thermodynamic performance
Sections 2.5.1 and 2.5.2. The thermodynamic performance was evaluated with the
energy efficiency (g) as calculated in Eq. (13). The energy output
is given by the lower heating value of ethanol fuel (LHVAE = 26.5
2.5.1. Computation of economic performance MJkg
1) multiplied by its production rate, and the total energy
Economic performance was evaluated in terms of total capital input is given by the LHV of syngas containing 50% CO and 50%
investment (CAPEX) and minimum ethanol selling price (MESP). H2 (LHVsyngas = 252.4 MJkmol
1) plus heat provided by steam at
The total investment was estimated using the module costing tech- multiple parts of the process and the equivalent heat associated
nique described by Turton et al. (2009), in which all the costs with the production of electricity consumed in compressors and
derive from the individual purchase cost of equipment at base con- pumps. For the latter, work-related energy was translated to heat

Please cite this article as: E. M. de Medeiros, H. Noorman, R. Maciel Filho et al., Production of ethanol fuel via syngas fermentation: Optimization of eco-
nomic performance and energy efficiency, Chemical Engineering Science: X, https://doi.org/10.1016/j.cesx.2020.100056
 E.M. de Medeiros et al. / Chemical Engineering Science: X xxx (xxxx) xxx 7

by applying an electric generation efficiency of gCC = 60% for com- depicted in Fig. 2a for both case studies (normal and high mass
bined cycles operating with natural gas (Siemens AG, 2019). transfer – HMT), meaning that, as expected, higher productivities
can be achieved at the cost of higher investments. The Pareto fronts
LHV AE  m _ AE
g¼ ; E_ in also seem to comprise at least two convex sub-sets of solutions. For
LHV syngas  m_ syngas þ E_ in example, for the base case, one subset lies in the region with pro-
X X
¼ DH ^v  m _ steam þ W_ in =g ð13Þ ductivities 3.15, after which there is a clear change of pattern in
CC
the Pareto front, which nonetheless continues to be convex. It’s
worth recalling that the 1st objective function is the minimization
3. Results and discussion of –productivity, so it is not the objective function itself that is plot-
ted in the x-axis, but its opposite value. For both cases there is a
3.1. Techno-economic optimization of the bubble column reactor (BCR) threshold after which the capital costs continue to increase
without significant improvement in the productivity (e.g. after
Fig. 2 presents the main results of the optimization conducted 4.5 gL
1h
1 for the base case, as seen in Fig. 2a). These optimal
for maximization of ethanol productivity and minimization of cap- solutions are associated with larger reactor volumes (hence higher
ital cost considering the standalone BCR. The goal of this study was capital costs) obtained by decreasing the vessel aspect ratio (the
to analyze the trends in Pareto-optimal points before connecting length remains constant, as seen in Fig. 2c); but since the gains
the BCR to the purification unit. Firstly, there is a clear trade-off in productivity are only marginal, these are solutions without
between the two objectives, as shown with the Pareto fronts interest to the decision-making process. It is also verified that

Fig. 2. Bi-objective optimization of the BCR for maximization of ethanol productivity and minimization of capital costs: Pareto fronts (a); selected decision variables at the
Pareto-optimal solutions (dilution rate (b), reactor length (c), and gas residence time (d)); selected process outcomes at the Pareto-optimal solutions (ethanol concentration
(e), gas conversions (f)).

Please cite this article as: E. M. de Medeiros, H. Noorman, R. Maciel Filho et al., Production of ethanol fuel via syngas fermentation: Optimization of eco-
nomic performance and energy efficiency, Chemical Engineering Science: X, https://doi.org/10.1016/j.cesx.2020.100056
8 E.M. de Medeiros et al. / Chemical Engineering Science: X xxx (xxxx) xxx

increasing the mass transfer coefficient leads to higher productiv- rithms to improve this aspect is left as recommendation for a
ities with the same investment, at some points being almost twice future work.
the value of the base case. To support the interpretation of the Pareto fronts, the Pareto-
The different Pareto subsets are produced by different combina- optimal values of two selected decision variables are shown in
tions of the decision variables, as shown in Fig. 2b–d. Both case Fig. 4 along with process outcomes. In this figure, optimization A
studies show one subset of solutions characterized by large and (OptA, circles) refers to the objective pair (CAPEX,g) and B (OptB,
nearly constant values of dilution rate and reactor length. In this diamonds) to the pair (MESP,g). For both mass transfer cases, the
subset, the productivity continues to increase due to decreasing increase in g (accompanied by increase in CAPEX and MESP) was
GRT (Fig. 2d) and increasing gas recycle ratio (not shown). In the associated with a decrease in Drate (Fig. 4a) and increase in GRT
preceding subset, the increase in productivity is not only due to (Fig. 4b), though in the HMT case the range spanned by these vari-
decreasing GRT but also to increasing reactor length, in combina- ables was very narrow. Other DVs (not shown here) that had pre-
tion with changes in Drate and GRR. These variations, of course, also dominant increasing trends with g were the gas recycle ratio (GRR),
have effects on the capital costs (as seen in Fig. 2a): for example, the number of stages in T-02 and its reflux ratio. Fig. 4 also sug-
increasing the reactor length has a direct effect on the bioreactor gests that in the base case the OptB Pareto front (MESPxg) consti-
cost; on the other hand, decreasing GRT implies increasing the tutes a subset of OptA (CAPEXxg), with decision variables and
gas flow rate, which brings higher costs associated with larger process outcomes following similar trends and ranges. This also
gas compressor capacities. happens to a lesser extent with HMT.
As presented in Table 1, the decision variables at the optimal The minimum and maximum values of the optimal DVs are pre-
solutions are constrained to similar ranges for both study cases sented in Table 2, where it can be observed that for all case studies
(e.g. optimal vessel volumes are close to the lower bound of the and optimization pairs the ranges were significantly reduced from
stipulated search space), but the process outcomes are mostly the search space. For comparison, the ranges obtained by minimiz-
improved when mass transfer is enhanced. Higher conversions ing MESP as a single objective are also shown. The minimum values
are achieved for both CO and H2 (shown in Fig. 2f), as well as higher of MESP in this case were 0.958 US$/L (base) and 0.705 US$/L
ethanol concentrations (Fig. 2e), though in this case the HMT (HMT), but the energy efficiencies at these points were far from
results also show lower concentrations for a wide range of their highest values, at 0.31 and 0.50 respectively, indicating that
productivities. inefficiencies are compensated by low price of resources. An
intriguing outcome is that when optimizing a single objective
3.2. Thermo-economic optimization of the integrated process many DVs had optimal values outside the ranges of the Pareto-
optimal solutions; in fact, this was observed for all the DVs in
Integrating the BCR to the purification unit brought several dif- the HMT case. For the gas recycle ratio (GRR) specially, the
ficulties due to the high non-linearity of the MESH equations, single-objective optimum was close to zero while being almost
which often leads to infeasible solutions, and the inclusion of a at the search space upper bound for Pareto-optimal points with
recycle stream of distillate bottoms containing traces of ethanol higher g.
and acetic acid. This was observed by the much longer time This illustrates the importance of using multi-objective as
required to run the optimization (from 5 min to over 3 h), despite opposed to single-objective optimization: as technology progress
the use of reduced (surrogate) models. Another reason for the is driven by multiple values other than economic profit, it is inevi-
higher complexity of the integrated model is that the input vari- table that industries will seek to optimize conflicting objectives;
ables of the distillation columns must be set to guarantee an etha- and not only bi-objective but many-objective problems will need
nol mass fraction of 0.925 in the distillate stream of the second to be tackled. In fact, even if the goals are kept within the economic
column. sphere, it is still challenging to reduce them to a single variable,
The generated Pareto fronts are depicted in Fig. 3, where again since different indicators can be used to evaluate the profitability
the opposite of an objective function is shown: for all fronts in of a process: CAPEX, MESP, OPEX, NPV, payback time, ROI (return
Fig. 3, the 2nd objective function was to minimize
g, but g itself on investment), etc.
is plotted in the y-axis. As explained in Section 2.5, the optimiza- It’s worth noting that the optimization of the BCR alone is not
tion was performed for two mass transfer cases (base and HMT) enough for meaningful decision-making, since the operating and
and two pairs of objective functions. The difference between these design variables of the purification unit must be adapted to the
pairs is the formulation of the 1st objective function, which is outcomes of the bioreactor and optimized accordingly. When ana-
either to minimize CAPEX/ṁAE (Fig. 3a–b) or MESP (Fig. 3c–d). It’s lyzing the whole process, the possible choices for the operating
worth mentioning that the results presented here are the best conditions and design variables were optimized to give a range
found approximations of the Pareto fronts, which were selected of solutions from which a single one can be picked. If both energy
after multiple runs of the multi-objective genetic algorithm. efficiency and MESP are considered equally important, an illustra-
The first observation is that the thermodynamic efficiency was tive example for the base case is to take the solution at g  0.4 fol-
bounded to around 0.42 in the base model and 0.72 in the HMT lowing the notably steep increase in Fig. 3c, since after this point
case, also with lower CAPEX and MESP achieved in the latter. The the gains in efficiency are relatively slow with respect to MESP.
trade-off is visible for all cases, although the optimization of the This solution gives the following decision variables: Drate = 0.052 h
1,
g-MESP pair with HMT (Fig. 3d) led to a small set of non- GRT = 34.1 min, GRR = 0.37, XP = 0.13, L = 49.5 m, VR = 772 m3,
dominated solutions distributed between 0.707  MESP (US$/L) S:F = 0.10, NST = 45, FST,V = 0.65, FST,L = 0.4, and RR = 5.55. For the
 0.713. As explained by Goh and Tan (2009), finding a diverse high mass transfer case, a similar solution can be drawn out at
set of solutions is specially challenging in the presence of multi- g  0.714: Drate = 0.10 h
1, GRT = 18.6 min, GRR = 0.49, XP = 0.05,
modality (i.e. multiple local Pareto fronts), discontinuity and L = 42.7 m, VR = 652 m3, S:F = 0.08, NST = 44, FST,V = 0.76,
non-uniformity. Moreover, balancing diversity of solutions and FST,L = 0.34, and RR = 3.95.
proximity to the true Pareto front is a multi-objective problem in Relevant process outcomes are presented in Table 3. The aver-
itself, since these two qualities are both desirable and conflicting age ethanol concentration in the broth was around 29 gL
1 for
(Bosman and Thierens, 2003). Despite the lack of diversity in the the base case and 36 gL
1 for the HMT case, with the acetic acid
solutions presented in Fig. 3d, the results are deemed sufficient concentration staying under 4 gL
1. As consequence, lower energy
for the purposes of the present work, and testing different algo- consumption in the distillation unit (roughly 7.2 MJ/L ethanol)

Please cite this article as: E. M. de Medeiros, H. Noorman, R. Maciel Filho et al., Production of ethanol fuel via syngas fermentation: Optimization of eco-
nomic performance and energy efficiency, Chemical Engineering Science: X, https://doi.org/10.1016/j.cesx.2020.100056
 E.M. de Medeiros et al. / Chemical Engineering Science: X xxx (xxxx) xxx 9

Fig. 3. Bi-objective optimization of the integrated units: optimization A with base (a) and high mass transfer model (b); optimization B with base (c) and high mass transfer
model (d).

could be obtained with HMT, while the base case spans a range 3.3. Effects and prospects of mass transfer enhancement
from 7.4 to 9.4 MJ/L, mostly increasing with g. The results also
show that high CO conversions (Fig. 4c) of up to 97% can be It can also be observed in Fig. 4 and Table 2 that the optimal val-
achieved with increased mass transfer and 83% for the base case, ues of the decision variables differ significantly between the base
while the H2 conversions (Fig. 4d) were bounded to lower values case and the HMT case, demonstrating the importance of accurate
(74%) in both cases, due to the CO inhibition as considered in the kLa estimations for the effective application of model-based opti-
kinetic model. For comparison purposes, LanzaTech’s process is mization results. Several factors could compel the real kLa to devi-
able to achieve high and stable CO conversions above 90% and ate from the predictions of empirical correlations such as Eq. (6),
ethanol selectivity of 95% (LanzaTech, 2018). Though their process for example the presence of salts and alcohols in low concentra-
configuration might be similar to the one considered here, employ- tion, which have been shown to increase kLa by factors up to 2
ing gas-liquid column reactors, distillation to purify ethanol, and (van de Donk, 1981). Heijnen and van’t Riet (1984) have also noted
water recycle, the exact design and conditions of their process that, in non-coalescing media, fine bubble systems can present a
are not reported. Another point worth mentioning is that in the six-fold increase in kLa as compared to coarse bubble systems mea-
present work the gas composition was fixed at 50% CO and 50% sured at the same gas superficial velocity, although in strongly coa-
H2, but the proportions of CO/H2/CO2 are also expected to have sig- lescing media the difference is modest. Therefore the results
nificant impacts on the gas conversion and ethanol productivity in presented here suggest that great improvements can be achieved
the bioreactor (de Medeiros et al., 2019). Since the syngas compo- with ingenious tuning of kLa via enhancement techniques. Ulti-
sition is a result of design choices in upstream stages, which in turn mately, such ‘‘mass transfer enhancement” can be included as a
also affect prices and energy efficiency, another study is currently factor in the optimization framework in order to find the optimal
under development for the impact analysis and optimization of the amount of efforts and costs that should be spent into increasing
whole process including syngas production via gasification. mass transfer capacity in the bioreactor. Examples of process
The Pareto-optimal solutions of the integrated process span intensification (PI) strategies for this purpose are given next.
smaller ranges of ethanol productivity in the bioreactor than the Groen et al. (2005) presented a method to increase the oxygen
BCR optimization, indicating that higher productivities are not nec- transfer capacity in aerobic fermentation which consisted of inject-
essarily advantageous when taking into account the whole process, ing a second gas stream (pure oxygen) via a special nozzle config-
due to much higher capital costs (also corroborated by the steep uration to achieve supersonic velocities and a non-uniform bubble
increase in the Pareto fronts from Fig. 2a) and losses in gas conver- size distribution. With a ratio of 6:1 between the two gas flows
sion (thus lower energy efficiency). With regard to the annual pro- (air/pure oxygen) the authors were able to increase the mass trans-
duction of anhydrous ethanol (Fig. 4e), in the base case the trends fer rate 3.6 times. Along the lines of reducing bubble size to
are opposite to the energy efficiency, but with HMT there doesn’t increase the surface to volume ratio, the generation of micron size
seem to be a correlation between the two, with the optimal pro- (<1 mm) bubbles is generally considered an efficient way of
duction rate in this case being between 121 and 133 MML/year. enhancing gas-liquid mass transfer while requiring low power
However, if MESP is used as single objective both cases lead to sim- inputs: Bredwell and Worden (1998), for example, obtained a
ilar production rates of around 121–124 MML/year (Table 3). six-times higher kLa using a spinning-disk microbubble generator

Please cite this article as: E. M. de Medeiros, H. Noorman, R. Maciel Filho et al., Production of ethanol fuel via syngas fermentation: Optimization of eco-
nomic performance and energy efficiency, Chemical Engineering Science: X, https://doi.org/10.1016/j.cesx.2020.100056
10 E.M. de Medeiros et al. / Chemical Engineering Science: X xxx (xxxx) xxx

Fig. 4. Bi-objective optimization of the integrated units: selected decision variables at the Pareto-optimal solutions (dilution rate (a) and gas residence time (b)); selected
process outcomes at the Pareto-optimal solutions (CO conversion (c), H2 conversion (d), production rate (e), and energy consumption (f)).

Table 3
Multi-objective optimization of the integrated process: ranges of process outcomes at the Pareto-optimal solutions.

CAPEX  g CAPEX  g (HMT) MESP  g MESP  g (HMT) MESP MESP (HMT)

CL,EtOH (BCR) [gL
1] 26.3–34.7 36.6–38.3 26.0–30.1 35.5–36.7 28.4 34.6
CL,HAc (BCR) [gL
1] 1.61–3.74 2.33–2.43 1.97–3.44 2.33–2.46 4.20 3.0
Heat consumption [MJL
1 AE] 7.42–9.35 6.93–7.11 8.41–9.41 7.18–7.30 8.71 7.45
Electricity consumption [kWhL
1 AE] 0.40–0.58 0.37–0.41 0.40–0.53 0.36–0.37 0.43 0.25
X CO (BCR) 0.63–0.82 0.95–0.96 0.71–0.83 0.97 0.68 0.97
X H2 (BCR) 0.40–0.74 0.52–0.69 0.50–0.71 0.70–0.74 0.40 0.78
ṁAE [MMLyear
1] 63.5–158 123–133 68.8–146 125–127 124 121
EtOH Productivity (BCR) [gL
1h
1] 1.17–3.19 3.74–3.93 1.33–2.63 3.66–3.73 2.81 3.67

as compared to the conventional sparging system in a stirred tank microbubbles, relying on different aspects: with or without accom-
for syngas fermentation. Moreover, while micro-porous diffusers panying liquid flows, using polymers or with very low power con-
can be used to generate microbubbles in a passing gas stream, sev- sumption such as microchannel techniques and ultrasonic systems
eral works have demonstrated even further size reduction by oscil- (Parmar and Majumder, 2013).
lating this stream, with the oscillation frequency in this case being Finally, nanoparticles have also been shown to improve mass
a central optimization variable (Brittle et al., 2015). A myriad of transfer, though only limited work has been developed for syngas
other methods have also been reported for the formation of fermentation. Kim and Lee (2016), for example, observed substan-

Please cite this article as: E. M. de Medeiros, H. Noorman, R. Maciel Filho et al., Production of ethanol fuel via syngas fermentation: Optimization of eco-
nomic performance and energy efficiency, Chemical Engineering Science: X, https://doi.org/10.1016/j.cesx.2020.100056
 E.M. de Medeiros et al. / Chemical Engineering Science: X xxx (xxxx) xxx 11

Fig. 5. Sensitivity of MESP to economic parameters: base (a) and high mass transfer model (b).

tial increase in the dissolved concentrations of CO, H2 and CO2 increase with g, which nonetheless was restricted to maximum
when using functionalized nanoparticles, and they also reported values of 0.42 (base case) and 0.72 (HMT). The trends of the deci-
that adding a magnetic layer to the particles enabled their reuse sion variables along the Pareto fronts were discussed and it was
up to five times, thereby improving their economic feasibility. seen that optimization results are greatly affected by the mass
transfer calculations, thus corroborating that kLa enhancement is
3.4. Sensitivity of MESP a promising strategy for global process improvement. The results
presented here provide information about key process conditions
The Pareto-optimal MESP was in the range 1.0–1.3 US$/L (base) and design variables, as well as sustainability targets and limita-
and 0.707–0.713 US$/L (HMT), while in a previous work by our tions in syngas fermentation. This work can be further extended
group the estimated MESP for a process including biomass gasifica- to evaluate and optimize the whole process from biomass waste
tion and heat/power generation was in the range 0.63–0.93 US$/L to ethanol or other chemicals, and its reliability can be increased
hydrous ethanol (i.e. before dehydration) (de Medeiros et al., with the development of new models for syngas fermentation.
2017). In another study about syngas fermentation, even higher
values between 1.58 and 1.93 US$/L were found (Benalcázar CRediT authorship contribution statement
et al., 2017). Not only technical considerations and process model-
ing methods lead to these disparities, but also economic consider- Elisa M. de Medeiros: Conceptualization, Methodology,
ations such as the prices of utilities and raw materials. With that in Software, Formal analysis, Investigation, Writing - original draft,
mind, Fig. 5 shows the sensitivity of MESP to several economic Writing - review & editing, Visualization. Henk Noorman: Concep-
parameters, where the midpoint is one of the Pareto-optimal solu- tualization, Resources, Writing - review & editing, Supervision,
tions from Fig. 3c and 3d. In this analysis the parameters were var- Funding acquisition. Rubens Maciel Filho: Conceptualization,
ied each at a time, with the other parameters being fixed at the Resources, Supervision. John A. Posada: Conceptualization,
values used in the optimization. For both cases the highest impacts Resources, Writing - review & editing, Supervision.
on MESP (around 15% of its original value) were caused by a 30%
change in the capital investment, which might also explain why Declaration of Competing Interest
the optimizations with CAPEX and MESP have similar results. A
forthcoming work by our group suggests that in other cases MESP The authors declare that they have no known competing finan-
can decrease with higher g, probably because the contributions of cial interests or personal relationships that could have appeared
raw materials and utilities are more significant than in the present to influence the work reported in this paper.
study.
Acknowledgments
4. Conclusions
The authors thank DSM (Netherlands) and BE-Basic Foundation
Different process systems engineering (PSE) tools (modeling, for the financial support provided in the form of a Ph.D. scholarship
simulation, neural networks, genetic algorithm) were applied in for E. M. de Medeiros. The research was also supported by São
this work to develop an optimization framework through which Paulo Research Foundation (FAPESP-Brazil), grant #2015/20630-
ranges of operating conditions and design variables can be selected 4. This work is part of a Dual Degree Ph.D. project under the agree-
for optimal production of ethanol fuel from syngas via fermenta- ment between UNICAMP (Brazil) and TU-Delft (Netherlands).
tion, taking into account objectives of distinct natures. When opti-
mizing a standalone bioreactor (BCR) in terms of ethanol References
productivity and capital costs, it was observed that Pareto-
Abubackar, H.N., Veiga, M.C., Kennes, C., 2015. Carbon monoxide fermentation to
optimal values of productivity ranged between 0.75–4.5 gLh
1 ethanol by Clostridium autoethanogenum in a bioreactor with no accumulation
for the base model and 1.5–5.4 gL1h
1 for an optimistic case with of acetic acid. Bioresour. Technol. 186, 122–127.
higher mass transfer capacity (HMT), with the productivity ArcelorMittal, 2019. World first: ArcelorMittal Belgium starts construction on two
pioneering projects to further reduce carbon emissions. https://belgium.
increasing at the cost of higher investment due to larger reactor
arcelormittal.com/en/two-pioneering-projects-to-further-reduce-carbon-
vessel and gas compressor. In the second part of this work, the emissions/ (accessed 22-07-2019).
BCR model was coupled to the purification unit including water Benalcázar, E.A., Deynoot, B.G., Noorman, H., Osseweijer, P., Posada, J.A., 2017.
recycle, and the integrated process was optimized in terms of eco- Production of bulk chemicals from lignocellulosic biomass via thermochemical
conversion and syngas fermentation: a comparative techno-economic and
nomic variables (CAPEX and MESP) and energy efficiency (g). It was environmental assessment of different site-specific supply chain configurations.
observed that, with the assumptions used here, MESP and CAPEX Biofuels. Bioprod. Bioref. 11, 861–886.

Please cite this article as: E. M. de Medeiros, H. Noorman, R. Maciel Filho et al., Production of ethanol fuel via syngas fermentation: Optimization of eco-
nomic performance and energy efficiency, Chemical Engineering Science: X, https://doi.org/10.1016/j.cesx.2020.100056
12 E.M. de Medeiros et al. / Chemical Engineering Science: X xxx (xxxx) xxx

Biofuels Digest, 2019. South Africa gives green light to country’s first LanzaTech economics for biochemical conversion of lignocellulosic biomass to ethanol:
ethanol project. http://www.biofuelsdigest.com/bdigest/2018/10/23/south- dilute-acid pretreatment and enzymatic hydrolysis of corn stover. Technical
africa-gives-green-light-to-countrys-first-lanzatech-ethanol-project/ (accessed Report, NREL/TP-5100-47764 (May 2011).
22-07-2019). Ibrahim, D., Jobson, M., Li, J., Guillén-Gosálbez, G., 2018. Optimization-based design
Bosman, P.A.N., Thierens, D., 2003. The balance between proximity and diversity in of crude oil distillation units using surrogate column models and a support
multiobjective evolutionary algorithms. IEEE Trans. Evolut. Comput. 7 (2), 174– vector machine. Chem. Eng. Res. Des. 134, 212–225.
188. IEA, 2019. https://www.iea.org/statistics/co2emissions/ (accessed 26/06/2019).
Bredwell, M.D., Worden, R.M., 1998. Mass-transfer properties of microbubbles. 1. Kim, Y.-K., Lee, H., 2016. Use of magnetic nanoparticles to enhance bioethanol
Experimental studies. Biotechnol. Prog. 14, 31–38. production in syngas fermentation. Bioresour. Technol. 204, 139–144.
Brittle, S., Desai, P., Ng, W.C., Dunbar, A., Howell, R., Tesař, V., Zimmerman, W.B., LanzaTech, 2018. LanzaTech – No such thing as waste. https://ccnet-nibb.co.uk/wp-
2015. Minimising microbubble size through oscillation frequency control. content/uploads/2018/04/simpson.pdf (accessed 08/08/2019).
Chem. Eng. Res. Des. 104, 357–366. Li, X., Griffin, D., Li, X., Henson, M.A., 2019. Incorporating hydrodynamics into
CanaOnline, 2016. Há um ano, bagaço de cana era vendido por R$150,00 a tonelada, spatiotemporal metabolic models of bubble column gas fermentation.
agora chega a R$25,00. http://www.canaonline.com.br/conteudo/ha-um-ano- Biotechnol. Bioeng. 116, 28–40.
bagaco-de-cana-era-vendido-por-r15000-a-tonelada-agora-chega-a-r2500. Martinez-Gomez, J., Nápoles-Rivera, F., Ponce-Ortega, J.M., El-Halwagi, M.M., 2017.
html (accessed 19/03/2019). Optimization of the production of syngas from shale gas with economic and
Chemical Engineering Magazine, 2019. https://www.chemengonline.com/pci safety considerations. Appl. Therm. Eng. 110, 678–685.
(accessed 20/03/2019). Pardo-Planas, O., Atiyeh, H.K., Phillips, J.R., Aichele, C.P., Mohammad, S., 2017.
Chen, J., Daniell, J., Griffin, D., Li, X., Henson, M.A., 2018. Experimental testing of a Process simulation of ethanol production from biomass gasification and syngas
spatiotemporal metabolic model for carbon monoxide fermentation with fermentation. Bioresour. Technol. 245, 925–932.
Clostridium autoethanogenum. Biochem. Eng. J. 129 (2018), 64–73. Parmar, R., Majumder, S.K., 2013. Microbubble generation and microbubble-aided
CPFL Energia, 2019. https://www.cpflempresas.com.br/institucional/tarifas.aspx? transport process intensification – a state-of-the-art report. Chem. Eng. Process.
emp=CPFL (accessed 19/03/2019). 64, 79–97.
de Medeiros, E.M., Posada, J.A., Noorman, H., Osseweijer, P., Filho, R.M., 2017. Renewables Now, 2018. LanzaTech, Shougang open sustainable ethanol refinery in
Hydrous bioethanol production from sugarcane bagasse via energy self- China. https://renewablesnow.com/news/lanzatech-shougang-open-
sufficient gasification-fermentation hybrid route: simulation and financial sustainable-ethanol-refinery-in-china-615844/ (accessed 22-07-2019).
analysis. J. Clean. Prod. 168, 1625–1635. Richter, H., Molitor, B., Wei, H., Chen, W., Aristilde, L., Angenent, L.T., 2016. Ethanol
de Medeiros, E.M., Posada, J.A., Noorman, H., Filho, R.M., 2019. Dynamic modeling of production in syngas-fermenting Clostridium ljungdahlii is controlled by
syngas fermentation in a continuous stirred-tank reactor: multi-response thermodynamics rather than by enzyme expression. Energy Environ. Sci. 9,
parameter estimation and process optimization. Biotechnol. Bioeng. 2019, 1– 2392–2399.
15. https://doi.org/10.1002/bit.27108. Roy, P., Dutta, A., Deen, B., 2015. Greenhouse gas emissions and production cost of
Deckwer, W.-D., 1976. Non-isobaric bubble columns with variable gas velocity. ethanol produced from biosyngas fermentation process. Bioresour. Technol.
Chem. Eng. Sci. 31, 309–317. 192, 185–191.
Deckwer, W.-D., 1992. Bubble Column Reactors (V. Cottrell, Trans.). John Wiley & Siemens AG, 2019. Gas Turbine Portfolio Brochure. https://
Sons Ltd, Chichester. siemens.com/gasturbines (accessed 17-07-2019).
Deckwer, W.-D., Burckhart, R., Zoll, G., 1974. Mixing and mass transfer in tall bubble Turton, R., Bailie, R.C., Whithing, W.B., Shaeiwitz, J.A., 2009. Analysis, Synthesis, and
columns. Chem. Eng. Sci. 29, 2177–2188. Design of Chemical Processes. Prentice Hall, New Jersey.
Goh, C.-K., Tan, K.C., 2009. Evolutionary Multi-objective Optimization in Uncertain Ulrich, G.D., Vasudevan, P.T., 2006. How to estimate utility costs. Chem. Eng. (April
Environments: Issues and Algorithms. Springer-Verlag, Berlin Heidelberg. 2006), 66–69.
Goldemberg, J., Coelho, S.T., Guardabassi, P., 2008. The sustainability of ethanol Van de Donk, J., 1981. Water Aeration with Plunging Jets PhD thesis. Delft
production from sugarcane. Energy Policy 36, 2086–2097. University of Technology.
Groen, D.J., Noorman, H.J., Stankiewicz, A., 2005. Improved method for aerobic Yao, Z., You, S., Ge, T., Wang, C.-H., 2018. Biomass gasification for syngas and biochar
fermentation intensification. In: Jansens, P.J., Stankiewicz, A., Green, A. (Eds.), co-production: energy application and economic evaluation. Appl. Energy 209,
Proceedings of Sustainable (Bio)Chemical Process Technology Incorporating 6th 43–55.
Int. Conf. Process Intensification. BHR Group Ltd., Cranfield, pp. 105–112. Ye, F., Ma, S., Tong, L., Xiao, J., Bénard, P., Chahine, R., 2019. Artificial neural network
Heijnen, J.J., Vant Riet, K., 1984. Mass transfer, mixing and heat transfer phenomena based optimization for hydrogen purification performance of pressure swing
in low viscosity bubble column reactors. Chem. Eng. J. 28, B21–B42. adsorption. Int. J. Hydrogen Energy 44, 5334–5344.
Humbird, D., Davis, R., Tao, L., Kinchin, C., Hsu, D., Aden, A., Schoen, P., Lukas, J.,
Olthof, B., Worley, M., Sexton, D., Dudgeon, D., 2011. Process design and

Please cite this article as: E. M. de Medeiros, H. Noorman, R. Maciel Filho et al., Production of ethanol fuel via syngas fermentation: Optimization of eco-
nomic performance and energy efficiency, Chemical Engineering Science: X, https://doi.org/10.1016/j.cesx.2020.100056