Optimal extractive distillation process for bioethanol dehydration

Abstract

The large-scale production of bioethanol fuel requires energy demanding distillation steps to concentrate the diluted streams from the fermentation step and to overcome the presence of the ethanol-water azeotrope. The conventional separation sequence consists of three distillation columns performing several tasks with high energy penalties: pre-concentration of ethanol (PDC), extractive distillation (EDC) and solvent recovery (SRC). It is remarkable that almost all papers on this topic focus on the azeotropic separation only, while neglecting the pre-concentration step. Usually, ethanol concentration in the first distillate stream is arbitrarily considered close to the azeotropic composition. While the energy usage in the PDC increases as the distillate composition gets closer to the azeotrope, the energy requirements in the EDC-SRC units decreases as the feed to EDC becomes richer in ethanol-and the other way around. This paper addresses this key trade-off of the distillate composition-a fundamental issue that was not studied before. Aspen Plus simulations were used to investigate how this parameter affects the energy usage and investment costs of the complete system. This issue applies in any other methods using a pre-concentration column (e.g. extractive and azeotropic distillation). The optimal economics is reached at a distillate concentration of 91 %wt (or 80% mol) ethanol, where the energy use is only 2.11 kWh (7596 kJ) per kg ethanol.

Full text

Optimal extractive distillation process for bioethanol dehydration
Radu M. Ignat,1 A.A. Kiss,2 Costin S. Bildea1
1University Politehnica of Bucharest, Bucharest/Romania;
2AkzoNobel Research, Development & Innovation, Deventer/Netherlands;
Abstract
The large-scale production of bioethanol fuel requires energy
demanding distillation steps to concentrate the diluted streams from
the fermentation step and to overcome the presence of the ethanol-
water azeotrope. The conventional separation sequence consists of
three distillation columns performing several tasks with high energy
penalties: pre-concentration of ethanol (PDC), extractive distillation
(EDC) and solvent recovery (SRC). It is remarkable that almost all
papers on this topic focus on the azeotropic separation only, while
neglecting the pre-concentration step. Usually, ethanol concentration
in the first distillate stream is arbitrarily considered close to the
azeotropic composition. While the energy usage in the PDC
increases as the distillate composition gets closer to the azeotrope,
the energy requirements in the EDC-SRC units decreases as the
feed to EDC becomes richer in ethanol – and the other way around.
This paper addresses this key trade-off of the distillate composition –
a fundamental issue that was not studied before. Aspen Plus
simulations were used to investigate how this parameter affects the
energy usage and investment costs of the complete system. This
issue applies in any other methods using a pre-concentration column
(e.g. extractive and azeotropic distillation). The optimal economics is
reached at a distillate concentration of 91 %wt (or 80% mol) ethanol,
where the energy use is only 2.11 kWh (7596 kJ) per kg ethanol.
Keywords
Extractive distillation, economic optimum, process optimization
1. Introduction
Bioethanol is one of the most promising alternative and sustainable biofuel. The
bioethanol production at industrial scale relies on several processes, such as: corn-
to-ethanol, sugarcane-to-ethanol, basic and integrated lignocellulosic biomass-to-
ethanol. After the initial pre-treatment steps, the raw materials enter the fermentation
stage where ethanol is produced (Vane, 2008). A common feature of all these
technologies is the production of diluted bioethanol – about 5-12 %wt ethanol – that
needs to be further concentrated to a maximum allowed water content of 0.2 %vol
(EU), 0.4 %vol (BR) or 1.0 %vol (US) according to various standards.
Several energy demanding separation steps are required to reach high purities,
mainly due to the presence of the binary azeotrope ethanol-water (95.63 %wt
ethanol). The first step is carried out in a pre-concentration distillation column (PDC)
that concentrates ethanol from 5-12% up to near azeotropic compositions (Frolkova,
2012). The second step is the ethanol dehydration up to higher concentrations above
the azeotropic composition, which is more complex and of greater research interest.
867

Several alternatives are also available and well described in the literature:
pervaporation, adsorption, pressure-swing distillation, extractive distillation (ED),
azeotropic distillation (AD), as well as hybrid methods (Vane, 2008; Frolkova, 2012).
Extractive distillation (ED) remains the option of choice in case of large scale
production of bioethanol fuel, and it involves an extractive distillation column (EDC)
and a solvent recovery column (SRC) for the ethanol dehydration – see Figure 1.
PDC
EDC
PDC-TOP
Ethanol
EDC-BTM
Feed
Water
Solvent
SRC
Water
Solvent
PDC –pre-concentration distillation column
EDC –extractive distillation column
SRC –solvent recovery column
(Ethylene glycol)
(recycle)
Figure 1. Flowsheet for bioethanol pre-concentration and dehydration by extractive distillation
Almost all reports focus only on the separation of ethanol-water azeotrope,
neglecting the pre-concentration step. Typically, the ethanol concentration in the first
distillate stream is arbitrarily considered close to the azeotropic composition. Though
the energy usage in the PDC increases as the distillate composition approaches the
azeotrope, the energy requirements in the EDC and SRC units decrease
correspondingly as the feed to EDC becomes richer in ethanol. This paper addresses
this key trade-off of the distillate composition – a fundamental issue that was not
studied before. A mixture of 10 %wt (4.2 %mol) ethanol is concentrated and
dehydrated using ethylene glycol as solvent. Rigorous Aspen Plus simulations were
used to investigate how this parameter affects the energy usage and investment
costs of the complete system. Note that this important issue applies in any other
dehydration methods using a pre-concentration column.
2. Problem statement
The composition of the distillate from the PDC unit is a key design optimization
variable that was so far neglected in the optimal design of extractive distillation
systems for ethanol dehydration. For example, Ryan and Doherty (1989) assumed a
composition of 94.9 %wt (88 %mol) ethanol which is rather close to the azeotropic
composition, while other authors (Kiss and Ignat, 2012; Kiss and Suszwalak, 2012; Li
and Bai, 2012) selected more practical compositions of about 93.5 %wt (85 %mol).
The problem is how to select this key design parameter such that the energy
requirements and the capital cost of the two sections of the process (pre-
concentration and dehydration of ethanol) are economically balanced to minimize the
overall costs. To solve this problem, we investigate here the effect of the PDC
distillate composition and prove that the optimal value is lower than what was
considered so far in the literature.
868

3. Results and discussion
Extractive distillation performs the separation of close boiling components or
azeotropes in the presence of a miscible, high boiling, relatively non-volatile
component that forms no azeotrope with the other components in the mixture. For
the ethanol-water mixture, ethylene glycol remains the most common entrainer used
in extractive distillation processes. However, the use of ethylene glycol could become
restricted in the future due to its toxicity. For this reason novel solvents are currently
explored, as for example: glycerol, hyperbranched polymers and ionic liquids.
Aspen Plus simulations were performed using the rigorous RADFRAC unit for
distillation. NRTL (non-random two-liquid) was used as the most adequate property
method, due to the presence of a non-ideal mixture containing polar components.
The ternary mixture ethanol-water-glycol presents a single binary azeotrope and no
liquid phase splitting – as shown in previous work (Kiss and Suszwalak, 2012). The
feed used here is the diluted bioethanol stream (10 %wt or 4.2 %mol ethanol)
obtained by fermentation. This is distilled to a composition below the azeotropic one,
and then dehydrated to a purity of over 99.8 %wt ethanol, to comply with all the
bioethanol standards. The production rate considered in this work is 100 ktpy.
The conventional sequence presented in Figure 1 consists of three distillation units:
pre-concentration distillation column (PDC), extractive distillation column (EDC) and
solvent recovery column (SRC). The first column (PDC) in the sequence has the
function to separate water as bottom stream and a near-azeotropic composition
mixture as distillate – sent afterward to the second column (EDC). In the EDC unit,
ethylene glycol – used as a high boiling solvent – is added on a stage higher than the
feed stage of the ethanol-water mixture. Due to the presence of the solvent the
relative volatility of ethanol-water is changed such that the separation becomes
possible. High purity ethanol is collected as top distillate product of the EDC, while
the bottom product contains only solvent and water. The solvent is then completely
recovered in the bottom of the third column (SRC), cooled in a heat recovery system,
and then recycled back to the extractive distillation column. An additional water
stream is obtained as distillate of the SRC unit. The bottom product of the SRC unit
constitutes the solvent recycle stream.
The SQP optimization method and the effective sensitivity analysis tool from Aspen
Plus® were used in the optimization procedure of all processes. Backed by a solid
theoretical and computational foundation, the sequential quadratic programming
method has become one of the most successful methods for solving nonlinearly
constrained optimization problems (Murray, 1969; Bartholomew-Biggs, 2008). The
objective of the optimization is to find the optimal trade-off between the energy
requirements and the equipment cost, both translated into the total annual cost
(TAC). The objective function used approximates very well the minimum of total
annualized cost of a conventional distillation column. The procedure was described in
detail in our recent work (Kiss and Ignat, 2012; Kiss and Suszwalak, 2012).
min NT (RR+1) = f (NT,i, NF,i, SFR, RRi, Vi) (1)
Subject to
mm xy  ≥
where i is the distillation column (PDC, EDC, SRC), NT is the total number of stages,
NF is the feed stage, SFR is the solvent-to-feed ratio, RR is the reflux ratio, V is the
boilup rate for each of the three columns, while ym and xm are vectors of the obtained
and required purities for the m products.
869

In order to perform a fair comparison between all process alternatives, the total
investment costs (TIC), total operating costs (TOC) and total annual costs (TAC)
were calculated, as described in our previous studies (Kiss and Ignat, 2012; Kiss and
Suszwalak, 2012). The equipment costs are estimated using correlations from the
Douglas textbook, updated to the level of 2010. The Marshall & Swift equipment cost
index (M&S) considered in this work has a value of 1468.6. Moreover, a price of 600
US $/m2 was used for calculating the cost of the sieve trays, and the following utility
costs were considered: US $0.03/t cooling water and US $13/t steam. For the TAC
calculations, a total plant lifetime of 10 years was considered (Kiss and Ignat, 2012).
The composition of the pre-concentrated ethanol stream was varied in the range 75-
93.5 %wt (54-85 %mol) ethanol and for each value considered the process flowsheet
was optimized. Table 1 shows the main results of the sensitivity analysis, including
the total investment costs (TIC), total operating costs (TOC) and the total annual cost
(TAC) as well as the total reboiler duty and the specific energy use per kg product.
Table 1. Results of the sensitivity analysis: key performance indicators (KPI) as function of
the composition of the pre-concentrated ethanol stream
Pre-
concentrated
EtOH (wt%)
Total
investment
cost (TIC)
Total
operating
cost (TOC)
Total
annual
cost (TAC)
Reboiler duties:
PDC, EDC, SRC (kW)
Energy use
(kW/kg
EtOH)
75.0
$4,299,460
$6,003,454
$6,433,400
18135 / 6658 / 4025
2.31
80.0
$4,197,003
$5,842,719
$6,262,419
18427 / 6347 / 3292
2.25
85.0
$4,138,478
$5,684,488
$6,098,336
18487 / 6259 / 2578
2.19
87.0
$4,054,603
$5,590,383
$5,995,843
18547 / 6021 / 2315
2.15
89.0
$3,983,370
$5,506,929
$5,905,266
18608 / 5833 / 2051
2.12
90.0
$3,951,436
$5,493,809
$5,888,952
18680 / 5823 / 1927
2.12
91.0
$3,915,109
$5,475,770
$5,867,281
18847 / 5673 / 1829
2.11
91.5
$3,969,593
$5,542,080
$5,939,039
19208 / 5658 / 1793
2.13
92.0
$3,994,262
$5,624,435
$6,023,861
19777 / 5589 / 1679
2.16
93.0
$4,199,949
$6,042,605
$6,462,600
21885 / 5577 / 1542
2.32
93.5
$4,409,534
$6,445,864
$6,886,817
23865 / 5574 / 1453
2.47
In addition, Figure 2 shows the optimal composition value of the pre-concentrated
ethanol stream for minimal specific energy use and lowest total annual cost (TAC). It
is worth noting that when the pre-concentrated ethanol stream has a composition
below the optimal value, the duty of the PDC decreases with the ethanol
concentration in the distillate, while the duties of the EDC and SRC units increases
since more effort is needed to remove the higher amount of remaining water.
Similarly, for pre-concentrated compositions higher than the optimal value, the duties
of the EDC and SRC units is lower since less effort is needed to remove the smaller
amount of remaining water. However, the duty of the PDC unit has a very steep
increase due to approaching the azeotropic composition. Balancing these two effects
lead to the optimal value of 91 %wt ethanol in the pre-concentrated stream. Just by
changing this key parameter, over 15% energy savings are possible in existing plants
that still use a pre-concentrated stream of near azeotropic composition. It is worth
mentioning that a similar value for the trade-off concentration (80% mol ethanol in the
beer-still distillate) was recently reported for the case of a heterogeneous azeotropic
distillation process for ethanol dehydration, using a more concentrated ethanol feed
(5% mol) and benzene or cyclohexane as light entrainers (Luyben, 2012).
870

2.10
2.15
2.20
2.25
2.30
2.35
2.40
2.45
2.50
74 76 78 80 82 84 86 88 90 92 94
Pre-concentrated ethanol / [wt%]
Specific energy use / [kWh/kg]
$5,800
$6,000
$6,200
$6,400
$6,600
$6,800
$7,000
74 76 78 80 82 84 86 88 90 92 94
Pre-concentrated ethanol / [wt%]
TAC/ [US k$]
Figure 2. Specific energy use per kg of ethanol product (left) and total annual cost (right), as
function of the composition of the pre-concentrated ethanol stream
Table 2 lists the key design and process parameters of the optimized flowsheet. Note
that in case of the non-optimal configurations, the number of stages varies within ±
20% more or less stages depending on the separation difficulty. The effect of the pre-
concentrated ethanol composition on the equipment design can be summarized as
follows: the number of stages and the diameter of the PDC column increases with the
ethanol concentration in the pre-concentrated stream, due to the more difficult
separation and higher reflux required. However, for the EDC and SRC columns the
variation of the required number of stages is rather minor due to the insignificant
change in the separation difficulty, while the column diameters are increasing at
lower pre-concentrated composition since more water is present in the feed.
Table 2. Design parameters of an optimal conventional sequence for bioethanol dehydration
by extractive distillation
Design parameters
PDC
EDC
SRC
Unit
Total number of stages
30
17
16
–
Feed stage number
19
11
8
–
Feed stage of extractive solvent
–
4
–
–
Column diameter
2.9
1.5
1
m
Operating pressure
1
1
1
bar
Feed composition (mass fraction)
Ethanol : water
Water : solvent
0.1 : 0.9
–
0.91 : 0.09
–
–
0.055 : 0.945
kg/kg
kg/kg
Feed flowrate (mass)
Ethanol
Water
Solvent
12500
112500
0
12494
1236
20793
0.625
1215
20788
kg/hr
kg/hr
kg/hr
Reflux ratio
1.31
0.24
0.45
kg/kg
Reboiler duty
18847
5673
1829
kW
Condenser duty
-8600
-3652
-1112
kW
Ethanol recovery
–
99.96
–
%
Water recovery
99.98
–
99.98
%
Solvent (EG) recovery
–
–
99.91
%
Purity of bioethanol product
–
99.80
–
%wt
Purity of water by-product
99.99
–
99.99
%wt
Purity of ethylene glycol recycle
–
–
99.99
%wt
In order to assess the controllability of the optimal design, a dynamic simulation
model was built using Aspen Dynamics. For all columns, the pressure is controlled by
condenser duty, while the distillate and bottoms flow rates are used to control the
871

levels in the reflux drums and column sumps, respectively. The pre-concentration
column is operated at constant reflux ratio, while the temperature in the stripping
section (stage 25) is controlled by the reboiler duty. Similarly, the EDC unit is
operated at constant reflux, constant solvent to feed ratio, the temperature in the
lower part (stage 15) being controlled by the reboiler duty. Dual temperature control
(stages 4 and 13), by means of reflux rate and reboiler duty, is employed for SRC.
Figure 3 presents results of dynamic simulation which prove the controllability of the
optimal design. Starting from the steady state, the feed flow rate is increased by
10%, from 125 to 137 ton/h (Figure 3, left). The transitory regime lasts for about 2
hours, new values for the product flows being established. The water and ethanol
purity remain very close to the initial value. In a second simulation (Figure 3, right),
the concentration of the raw material is reduced from 10 to 8 %wt ethanol. The new
values of the product flow rates are achieved in about 2 hours, with minor deviations
of the product purities.
0.995
0.996
0.997
0.998
0.999
1
0 2 4 6 8 10
Time / [h]
Mass fraction
0
0.02
0.04
0.06
0.08
0.1
0.12
PDC Water
EDC Ethanol
SRC Water
PDC Feed
0.995
0.996
0.997
0.998
0.999
1
0 2 4 6 8 10
Time / [h]
Mass fraction
0
0.02
0.04
0.06
0.08
0.1
0.12
PDC Water
EDC Ethanol
SRC Water
PDC Feed
Figure 3. Dynamic simulations for +10% feed flowrate disturbance (left) and a reduction from
10 to 8 %wt of ethanol concentration in the feed (right)
4. Conclusions
A key contribution of this study is creating awareness that the composition of the pre-
concentrated ethanol stream is an important design optimization variable that was
neglected so far in many articles, as well as calculating the optimal value of this
parameter in order to obtain minimum total annual costs.
Rigorous Aspen Plus simulations were successfully used to investigate how the
trade-off of the distillate composition affects the energy usage and the investment
costs of the complete system for ethanol dehydration by extractive distillation, using
ethylene glycol as a mass separation agent. In addition, Aspen Dynamics simulations
were employed to prove the controllability of the optimized process. The economical
optimum was found at a distillate concentration of 91 %wt (or ~80% mol) ethanol,
where the energy use is 2.11 kWh (7596 kJ) per kg ethanol (Kiss and Ignat, 2013).
References
Bartholomew-Biggs M., Nonlinear optimization with engineering applications, Springer 2008.
Frolkova A. K., Raeva V. M., 2012, Theor. Found. Chem. Eng., 44, 545-566.
Kiss A. A., Ignat R. M., 2012, Sep. Purif. Technol., 98, 290-297.
Kiss A. A., Suszwalak D. J-P. C., 2012, Sep. Purif. Technol., 86, 70-78.
Kiss A. A., Ignat R. M., 2013, Energy Technol., 1, 166-170.
Li G., Bai P., 2012, Ind. Eng. Chem. Res., 51, 2723-2729.
Luyben W. L., 2012, Ind. Eng. Chem. Res., 51, 16427-16432.
Murray W., in Optimization (Ed. Fletcher R.), Academic Press, New York 1969.
Ryan P. J., Doherty M. F., 1989, AIChE J., 35, 1592-1601.
Vane L. M., 2008, Biofuels Bioprod. Bioref., 2, 553-588.
872