TOWARDS IMPROVED ETHANOL

PRODUCTION FROM
LIGNOCELLULOSIC BIOMASS

Cristian Fernando Triana

November 2016

A thesis submitted for the degree of

Doctor of Philosophy of the University College London

Department of Chemical Engineering

University College London

2
I, Cristian F. Triana, confirm that the work presented in this thesis is my own. Where
information has been derived from other sources, I confirm that this has been
indicated in the thesis.

________________
Cristian F Triana C

3
 ABSTRACT

Ethanol from biological feedstock has emerged as a promising alternative for the
generation of energy from renewable sources in order to mitigate the damages caused
by the gas emissions associated to the consumption of fossil fuels. In many countries,
ethanol is already being produced at industrial scale from different biological raw
materials. However, there are some technical issues related to this process that need
to be addressed and one of the major problems is the high heat requirements which
makes this process less competitive against well-established fuels.

This work proposes an optimisation methodology based on a dynamic approach to

improve the overall efficiency of the process by considering new configurations and
designs that allow the reduction of operating costs, usage of utilities, the size of units,
etc. The work initially provides an introduction to the concept of fuels and the
current global scenario regarding their production and consumption. Next, a general
review of biofuels is given, in particular the production of ethanol from corn stover
and the different units involved in this process. Additionally, mathematical
formulations of the different units in the process are presented including detailed
kinetic models and dynamic mass and energy balances. These models are first
validated and then used in the construction of an overall model of the entire process
which so far has not been available in open literature. This thesis also presents the
development of an empirical mathematical model of an organophilic membrane for
ethanol removal from aqueous solutions to increase the separation rates of ethanol in
the process.

Finally, this works presents the optimisation of the ethanol production process
considering the implementation of heat storage units to reduce the consumption of
utilities such as steam and cooling water by reducing the Total Annualised Cost
(TAC). The results obtained show that the implementation of heat integration in the
process can achieve a reduction of 7 % in the TAC and 10 % in the total energy
consumption. These results indicate that ethanol production from corn stover with
the use of energy storage is a viable alternative for energy generation that can
become part of the main market of the production of green technologies.

4
 ACKNOWLEDGEMENTS

I would like to start this off by saying thanks to my family: My mom, my sister and
my nephew. You have been the strength and the motivation to keep on working and
doing what I am doing. You have shared my achievements and failures and never
stopped supporting me and believing in me.

I would like to express my gratitude to my supervisors who supported, encouraged

and contributed greatly in the development of this work and to my academic
progress. First, to Professor Eric Fraga who helped me a lot during my application
process to UCL and who always challenged me to be more confident about my work
and how to present my results. And to Professor Eva Sorensen for her great support
in many aspects related to education, especially, for introducing me to different
scenarios in which I met great people, I saw great places and I was allowed to present
my work to different audiences.

Special thanks to Professor Andrzej Górak and to Dr Philip Lutze for giving me the
opportunity of working with them for 4 months at the Laboratory of Fluid
Separations (FVT) at TU Dortmund. I would also like to thank Ms Kathrin Kissing
and Dr Johannes Holtbrügge for the technical support provided during the
development of my experimental work in Dortmund.

I would like to acknowledge the financial support of the Colombian Institute of

Science and Innovation (COLCIENCIAS) and the Engineering and Physical
Sciences Research Council (EPSRC) that made this research possible. Conference
funding from UCL Graduate School and IChemE Fluid Separation Subject Group is
also greatly acknowledged.

Finally, to my friends: Andres Calderon, Masha Koleva, Carlos Perez, Alexandros

Adam, Adrian Aguirre, Joakim Beck, Harry Christodoulou, Ben Dercks, Alexa
Duarte, Juan Carlos Uribe, Anna Kunze, Davide Bascone, David Lorenzo, Juan
Pablo Mariscal, Sebastian Siegert, Dr. Javier Fontalvo, Euan Corrall and all the
funky crew from the Elephants Head and the IoE.

5
Table of contents

Abstract 4
Acknowledgements 5
List of tables 10
List of figures 14
Chapter 1 – Ethanol production process from lignocellulosic biomass 19
1.1 Introduction 19
1.2 Economics 20
1.3 Biofuels 23
1.4 Ethanol 25
1.5 Lignocellulosic biomass 26
1.6 Current outlook for biofuels 28
1.7 Summary 31
1.8 Aims 31
1.9 Organisation 31
Chapter 2 – Literature review 33
2.1 Introduction 33
2.2 General aspects of the ethanol production process from
34
lignocellulosic biomass
2.3 Pretreatment 36
2.4 Detoxification 41
2.5 Enzymatic hydrolysis 42
2.5.1 Enzymatic activities 42
2.5.2 Enzymatic mechanism 43
2.6 Fermentation 45
2.7 Separation processes 50
2.8 Heat integration and heat storage 51
2.9 Summary and research statement 57
Chapter 3 – Mathematical models and model validation 60
3.1 Introduction 60
3.2 Process description 62
3.3 Dilute acid pretreatment 62
6
 3.4 Evaporation 67
3.5 Detoxification 67
3.6 Simultaneous saccharification and co-fermentation (SSCF) 69
3.6.1 Enzymatic hydrolysis 69
3.6.2 Co-fermentation 71
3.7 Separation 74
3.7.1 Distillation 74
3.7.2 Membrane-based operation 74
3.8 Results 75
3.9 Pretreatment stage 76
3.10 Detoxification stage or Overliming 77
3.11 Enzymatic hydrolysis stage 77
3.12 Co-fermentation stage 80
3.13 Distillation stage 81
3.14 Pervaporation stage 81
3.15 Conclusions 81
Chapter 4 – Ethanol recovery from aqueous solutions using a
83
PERVAPTM 4060 organophilic membrane
4.1 Introduction 83
4.2 Experimental materials and methods 85
4.2.1 Materials 86
4.2.2 Membrane and module 87
4.2.3 Gas chromatography 91
4.3 Experimental procedure 92
4.4 Modelling 94
4.4.1 Approach 94
4.4.2 Pure component and mixture property models 95
4.4.3 Models for permeance 95
4.5 Results 97
4.5.1 Model discrimination 98
4.6 Conclusions 111
Chapter 5 – Optimisation of the separation section of the ethanol
113
production process

7
 5.1 Introduction 113
5.2 Design of distillation columns 114
5.2.1 Shortcut methods 115
5.2.2 Group methods 115
5.2.3 Rigorous tray-by-tray optimisation models 116
5.3 Optimisation of a distillation system 116
5.3.1 Specifications of the distillation columns 123
5.3.2 Results of the optimisation of the distillation system 125
5.4 Optimisation of the pervaporation system using a PAN-B5
127
hydrophilic membrane
5.4.1 Specifications of the pervaporation network using a
131
hydrophilic membrane
5.4.2 Results of the optimisation of the pervaporation network using
132
a hydrophilic membrane
5.5 Optimisation of the pervaporation system using a PERVAPTM 4060
138
organophilic membrane
5.5.1 Specifications for the pervaporation network using an
140
organophilic membrane
5.5.2 Results of the optimisation of the pervaporation network using
141
an organophilic membrane
5.6 Optimisation of the complete separation section – Case study 150
5.6.1 Specifications of the separation section 150
5.6.2 Results of the separation sections 151
5.7 Sensitivity analysis 155
5.8 Conclusions 163
Chapter 6 – Heat integration across an ethanol production process 166
6.1 Introduction 166
6.1.1 Direct heat integration (DHI) 167
6.1.2 Indirect heat integration (IHI) 168
6.1.3 Mixed direct-indirect heat integration (MDIHI) 169
6.2 Heat transfer fluids 169
6.3 Simulation procedure 172
6.3.1 Configuration of the ethanol process without heat storage 172

8
 6.3.2 First superstructure of the ethanol production process with
179
heat storage
6.3.3 Second superstructure of the ethanol production process with
183
heat storage
6.4 Results 184
6.4.1 Specifications 184
6.4.2 Optimisation results – pretreatment stages 185
6.4.3 Optimisation results – separation section 190
6.4.4 Optimisation results – energy demand and economic analysis 195
6.5 Conclusions 196
Chapter 7 – Conclusions and future work 200
7.1 Conclusions 200
7.2 Future work 203
List of publications and events 206
REFERENCES 207
NOMENCLATURE 232
Appendix A – Kinetic models, mass and energy balances for the
235
production of ethanol from corn stover
A.1 Pretreatment 235
A.2 Detoxification 237
A.3 Evaporation 238
A.4 Simultaneous saccharification and co-fermentation (SSCF) 239
A.4.1 Saccharification 239
A.4.2 Simultaneous saccharification and co-fermentation 240
A.5 Distillation column 244
A.5.1 Tray 244
A.5.2 Condenser 246
A.5.3 Drum 247
A.5.4 Reboiler 248
A.5.5 Tray section 249
A.6 Membrane system 251
A.6.1 Fibre side 251
A.6.2 Shell side 252

9
List of tables

Table 1.1: Energy densities for some biofuels (Drapcho et al., 2008) 25
Table 1.2: Cellulose, hemicellulose and lignin content in agricultural 29
residues (% w/w dry basis)
Table 1.3: Cellulose, hemicellulose and lignin content in by-products of 29
agricultural processes (% w/w dry basis)
Table 1.4: Cellulose, hemicellulose and lignin content in crops for energy 30
generation (% w/w dry basis)
Table 2.1: Pretreatment methods for lignocellulosic materials 38
Table 2.2: Literature review on the pretreatment of lignocellulosic biomass 39
for ethanol production
Table 2.3: Enzymes used in conversion of cellulose into glucose 44
Table 2.4: Use of enzymatic hydrolysis in the production of glucose and 46
other reducing sugars from cellulose
Table 2.5: Microorganisms used in ethanol production 47
Table 2.6: Recent experimental works related to the implementation of the 48
SSCF process in the production of ethanol from lignocellulosic biomass
Table 2.7: Separation methods used in the dehydration of ethanol 52
Table 3.1: Chemical characterisation of corn stover (dry basis % w/w) 65
reported by Bhandari et al., (1984); Čuček et al., (2011); Esteghlalian et al.,
(1997) and Liu and Chen, (2016)
Table 3.2: Estimated ethanol purity for an ethanol feed concentration of 94% 80
(w/w)
Table 4.1: Operating conditions used in the experiments 87
Table 4.2: List of components produced during the fermentation of 89
lignocellulosic hydrolysates
Table 4.3: Specifications for the optimal operation of the PERVAPTM 4060 91
membrane (Sulzer Chemtech, 2014)
Table 4.4: Experimental results for the removal of ethanol from aqueous 100
solutions using the PERVAPTM 4060 membrane
Table 4.5: Partial fluxes of ethanol and water under the influence of different 101
o
impurities at 40 C and a permeate pressure of 20 mbar

10
Table 4.6: Parameters of the mathematical model of the organophilic 102
membrane
Table 4.7: Model validation using additional experimental data 105
Table 4.8: Lack-of-fit test for the parameter estimation of the membrane 105
model
Table 4.9: Parameter estimation problem for the permeance of ethanol and 108
water using Eq. 4.3
Table 4.10: Parameter estimation problem for the permeance of ethanol and 109
water using Eq. 4.4
Table 4.11: Parameter estimation problem for the permeance of ethanol and 110
water using Eq. 4.5
Table 5.1: Input variables used in the optimisation of the distillation systems 124
Table 5.2: Initial guesses and bounds used in the optimisation of the 126
distillation systems
Table 5.3: Results of the optimisation of the distillation systems 126
Table 5.4: Input variables used in the optimisation of the pervaporation 131
system using the PAN-B5 hydrophilic membrane
Table 5.5: Initial guesses and bounds used in the optimisation of the PAN- 132
B5 hydrophilic membrane
Table 5.6: Optimal parameters for a minimum TAC in the PAN-B5 135
hydrophilic membrane system
Table 5.7: Input variables used in the optimisation of the pervaporation 140
system using the PAN-B5 hydrophilic membrane
Table 5.8: Initial guesses and bounds used in the optimisation of the PAN- 140
B5 hydrophilic membrane
Table 5.9: Optimal parameters for a minimum TAC in the PERVAP 142
4060TM organophilic membrane system
Table 5.10: Input variables used in the optimisation of the distillation 143
column linked to the organophilic membrane system
Table 5.11: Results of the optimisation of the distillation systems using the 144
permeate stream of the pervaporation system shown in Figure 5.15 as feed
Table 5.12: Initial guesses and bounds used in the optimisation of the PAN- 146
B5 hydrophilic membrane in the configuration with the organophilic

11
membrane
Table 5.13: Initial guesses and bounds used in the optimisation of the PAN- 147
B5 hydrophilic membrane
Table 5.14: Optimal parameters for a minimum TAC in the pervaporation 147
network using the distillate stream of the single distillation column shown in
Figure 5.13
Table 5.15: Input variables used in the optimisation of the separation section 150
with and without the organophilic membrane
Table 5.16: Results of the optimisation of the superstructures of the 154
separation stages of the ethanol production process
Table 5.17: Case studies considered in the optimisation of the separation 156
section
Table 5.18: Sensitivity analysis of the separation section 158
Table 6.1: List of heat transfer fluids (Vignarooban et al., 2015) 170
Table 6.2: List properties of Therminol 66 (Eastman, 2016) 173
Table 6.3: Initial guesses used in the optimisation of the process with and 183
without heat integration
Table 6.4: Operating conditions for the reactors and evaporator 183
Table 6.5: Optimal results of the pretreatment stages for all the 185
configurations
Table 6.6: Results of the optimisation of the pretreatment stages and 188
composition of the process streams (% w/w)
Table 6.7: Results of the optimisation of the separation stages without heat 191
storage (membrane area of 500 m²)
Table 6.8: Results of the optimisation of the separation stages in 191
Superstructure 1 (membrane area of 500 m²)
Table 6.9: Results of the optimisation of the separation stages in 192
Superstructure 2 (membrane area of 500 m²)
Table 6.10: Energy requirements for the different configurations of the 193
ethanol process (kW)
Table 6.11: Economic analysis for equipment used in the ethanol production 194
process 1x103 (US$/year)
Table 6.12: Economic analysis for utilities and services used in the ethanol 194

12
production process 1x103 (US$/year)
Table 6.13: Area and cost for heat exchangers in all configurations 195

13
List of Figures

Figure 1.1: Global oil consumption since 2004 (U.S. Department of Energy,
20
2015)
Figure 1.2: Spot crude prices since 2004 (BP, 2016a) 20
Figure 1.3: Global consumption and production of natural gas since 2004
22
(U.S. Department of Energy, 2016)
Figure 1.4: Global consumption and production of coal since 2004 (BP,
22
2016a, 2012; U.S. Department of Energy, 2015)
Figure 2.1: Standard configuration for ethanol production 35
Figure 2.2: Configuration of the ethanol production process from
37
lignocellulosic biomass considering two fermentation stages
Figure 2.3: Configuration of the ethanol production process from
lignocellulosic biomass considering co-fermentation of both pentoses and 37
hexoses
Figure 2.4: Configuration of the ethanol production process from
lignocellulosic biomass considering simultaneous saccharification and co- 37
fermentation (SSCF)
Figure 2.5: Liquid-Vapour Equilibrium (VLE) calculated using ChemCAD
50
for the mixture ethanol/water at 1 atm
Figure 3.1: Flowsheet of the ethanol production process from corn stover 64
Figure 3.2: Xylose remaining in the solid fraction in the hydrolysis of
hemicellulose. The left side represents the results obtained with gPROMS.
78
The right side is the experimental data presented by Esteghlalian et al.,
(1997)
Figure 3.3: Model validation for the detoxification of acid hydrolysates. The
left side shows the results using gPROMS. The right side shows the 78
experimental data obtained by Purwadi et al., (2004)
Figure 3.4: Validation of the kinetic model for enzymatic saccharification.
The left side presents the results using gPROMS. The right side presents the 79
experimental data presented by Kadam et al., (2004)
Figure 3.5: Validation of the fermentation model proposed by Leksawasdi et
79
al., (2001). The left side presents the results using gPROMS. The right side

14
presents the experimental data
Figure 3.6: Validation profile for liquid weight fraction of binary mixture
80
ethanol/water using gPROMS and ChemCAD
Figure 4.1: Experimental design for the evaluation of the parameters of the
mathematical model of the PERVAPTM 4060 membrane (C = feed 87
concentration, T = feed temperature, P = permeate pressure)
Figure 4.2: Simplified flowsheet of the laboratory-scale set-up for
pervaporation experiments. (1) Feed tank, (2) Heat exchanger, (3)
88
Membrane module, (4) Thermal oven, (5) Cooling traps, (6) Vacuum pump,
(7) Exhaust, (8) Heating cabinet and (9) Back-pressure regulator
Figure 4.3: Pervaporation plant 90
Figure 4.4: Membrane module 90
Figure 4.5: Thermal oven 90
Figure 4.6: Jacketed tank 90
Figure 4.7: Cooling traps 90
Figure 4.8: Partial flux for ethanol and water using the different
105
formulations proposed in Equations 4.3 to 4.5
Figure 4.9: Parity plots with the fitted parameters for ethanol (top) and water
(bottom) using: (a) Eq. 4.3, (b) Eq. 4.4, (c) Eq. 4.5, (d) Eq. 4.3, (e), Eq. 4.4, 107
(f) Eq. 4.5
Figure 5.1: Sections of the distillation column used in the rigorous tray-by-
117
tray optimisation
Figure 5.2: a) Rectification section with side reflux stream and b) stripping
117
section with side boilup stream
Figure 5.3: Configurations for the distillation system considered in this
work: a) single distillation column and b) double distillation system with 119
mixed waste
Figure 5.4: Outer Approximation (OA) algorithm for the solution of the
124
MINLP problem using gPROMS (Process System Enterprises, 2015)
Figure 5.5: Streams and concentrations in a single distillation column using
128
the results of the optimisation presented in Table 5.3
Figure 5.6: Streams and concentrations in a double-column distillation
128
system using the results of the optimisation presented in Table 5.3

15
Figure 5.7: Pervaporation network design using PAN-B5 hydrophilic
129
membrane
Figure 5.8: Optimisation of the hydrophilic pervaporation system at different
membrane areas for the system consisting of a single distillation column 133
shown in Figure 5.5
Figure 5.9: Comparison of total annualised cost vs. energy consumption for
134
three and four in-series pervaporation stages shown in Figure 5.8
Figure 5.10: Results of the optimisation of the pervaporation network using
PAN-B5 hydrophilic membrane and a membrane area of 250 m2 for the 137
distillation system shown in Figure 5.5
Figure 5.11: Pervaporation network design using PERVAPTM 4060
139
organophilic membrane
Figure 5.12: Results of the optimisation of the pervaporation network using
PERVAPTM 4060 organophilic membrane and a membrane area of 20 m2 143
and 19 in-parallel modules
Figure 5.13: Results of the optimisation of the distillation column under the
145
specifications presented in Table 5.10
Figure 5.14: Optimisation of the hydrophilic pervaporation system at
different membrane areas for the system consisting of a single distillation 145
column shown in Figure 5.13
Figure 5.15: Results of the optimisation of the pervaporation network using
PAN-B5 hydrophilic membrane and a membrane area of 150 m2 for the 149
distillation system shown in Figure 5.14
Figure 5.16: Superstructure of the separation section of the ethanol
production process: a) single distillation and pervaporation network with
PAN-B5 hydrophilic and b) pervaporation network with PERVAPTM 149
organophilic membrane linked to a single distillation column and a
pervaporation network with PAN-B5 hydrophilic membrane
Figure 5.17: Results of the optimisation of the separation section with a
152
PAN-B5 hydrophilic membrane area of 250 m2
Figure 5.18: Results of the optimisation of the separation section including a
PERVAPTM 4060 organophilic membrane with 19 in-parallel membrane 153
modules with an area of 20 m2 and a PAN-B membrane with a membrane

16
area of 150 m2
Figure 5.19: Case 1 - Results of the optimisation of the separation section of
the ethanol production process with the PERVAPTM 4060 organophilic 160
membrane considering a feed concentration of ethanol of 4 % (w/w)
Figure 5.20: Case 2 - Results of the optimisation of the separation section of
the ethanol production process with the PERVAPTM 4060 organophilic 161
membrane considering a feed concentration of ethanol of 6 % (w/w)
Figure 5.21: Case 3 - Results of the optimisation of the separation section of
the ethanol production process with the PERVAPTM 4060 organophilic
162
membrane considering a concentration of ethanol in the distillate stream of
90 % (w/w)
Figure 5.22: Evaluation of the minimum TAC of the separation section with
163
and without the organophilic membrane considering different cases studies
Figure 6.1: Flowsheet of the ethanol production process from corn stover
175
without heat integration
Figure 6.2: Superstructure 1 of the ethanol production process from corn
stover with heat storage (using the bottom stream from the distillation 182
column to heat the stream of Therminol 66)
Figure 6.3: Superstructure 2 of the ethanol production process from corn
stover with heat storage using the heat released in the condenser to heat the 186
stream of Therminol 66
Figure 6.4: Results of the simulation and optimisation of the pretreatment
187
stages of the ethanol production process
Figure 6.5: Results of the optimisation of the separation stages without heat
189
storage 182
Figure 6.6: Results of the simulation and optimisation of the separation
189
stages with heat storage (Superstructure 1)
Figure 6.7: Results of the simulation and optimisation of the separation
190
stages with heat storage (Superstructure 2)
Figure 6.8: Temperatures of heat transfer fluid Therminol 66 across
198
Superstructure 1
Figure 6.9: Temperatures of heat transfer fluid Therminol 66 across
199
Superstructure 2

17
Figure 7.1: Flowsheet of possible scenarios for the optimisation of the
204
separation section
Figure A.1: Scheme of a jacketed reactor used in the formulation of the mass
236
and energy balances
Figure A.2: Scheme of a jacketed reactor used in the formulation of the mass
241
and energy balances
Figure A.3: Schematic of a distillation column tray 245
Figure A.4: Flowsheet of the top of the distillation column 246
Figure A.5: Schematic of a Kettle reboiler 247
Figure A.6: Schematic of the membrane module (Marriott and Sorensen,
251
2003; Marriott et al., 2001; Tsuyumoto et al., 1997)

18
 Chapter 1 – Ethanol production process from
lignocellulosic biomass

Abstract

This chapter presents a report on the production and consumption of fuels around
the world. This information shows an increase in the consumption of traditional fuels
in the last years by the countries of the Convention on the Organization for
Economic Co-operation and Development (OECD). The elevated consumption of
sources such as oil and coal has not only impacted global economy but also, the
environment. This chapter presents a breakdown of the different biofuels that are
currently produced around the world and the different raw materials used in their
production, focusing mainly on the production of ethanol from lignocellulosic
biomass.

1.1 Introduction
The population in this planet keeps on increasing at elevated rates and this only
represents a challenge for world’s leaders and their governments in the search for
efficient paths to provide food, healthcare and services to their people. However, the
ever-increasing populace not only denotes a problem in terms of food and healthcare
but also in terms of the energy demand that is necessary to maintain the gears of
development and industrialisation well-greased. Energy has become a priority in
every country due to its great influence in the creation of jobs, the modernisation of
infrastructure, military, politics, etc. (Cardona and Sánchez, 2007; Drapcho et al.,
2008; Hoekman, 2009). Therefore, a summary of the current state of fuels is
presented in this chapter and the trends in their prices and consumption throughout
the years. This chapter will also include the production of biofuels as an alternative
to conventional energy sources leading to the selection of ethanol as the fuel of
choice for the development of this thesis. Finally, this chapter will cover the
different raw materials used in the production of ethanol via fermentation and the
different units involved therein (e.g. reactors, evaporators, distillation columns, etc.).

19
 92

90

Global oil consumption

(million barrels daily)
88

86

84

82

80
2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Year

Figure 1.1: Global oil consumption since 2004 (U.S. Department of Energy, 2015)

120 Dubai
Brent
Nigerian forcados
100 West Texas intermediate
US$ per barrel of oil

80

60

40

20

0
2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Year

Figure 1.2: Spot crude prices since 2004 (BP, 2016a)

1.2 Economics
In emerging economies, the net energy consumption in 2015 increased significantly
since 2004 (see Figure 1.1) with China alone accounting for 71% of global energy
consumption growth. The consumption in the countries of the Convention on the
Organization for Economic Co-operation and Development (OECD), which include
EU countries, Canada, Turkey, Japan and the US, declined, led by a sharp drop in
Japan which in volumetric terms was the world's largest decline (BP, 2012; U.S.

20
Department of Energy, 2015). The data reported by both BP and the U.S. Department
of Energy suggest that the growth in global CO2 emissions from energy use
continued in 2015, but at a slower rate than back in 2010 (BP, 2016a; U.S.
Department of Energy, 2015). Crude oil prices peaked in April 2011 following the
loss of Libyan supplies and decreases of 9.7% were reported from 2013 to 2014. In
2015, the prices of oil plummeted approximately 47 % (BP, 2016a) (see Figure 1.2).

Natural gas prices in Europe and Asia – including spot markets and those indexed to
oil – presented a noticeable increased along with oil prices, although movements
within the year varied widely. North American prices reached record discounts to
both crude oil and international gas markets due to continued robust regional
production growth (Demirbas, 2007; Hoekman, 2009). Countries outside the OECD
once again accounted for all of the net growth in global consumption. Chinese
consumption growth, for instance, was below average but still recorded the largest
increment to global oil consumption with 390,000 barrels/day (an increment of 2.1 %
from 2013 to 2015). The global consumption of natural gas in 2015 increased only
0.4 % from 2013 and with an average price reduction in some OECD countries of 8.7
%. Figure 1.3 shows the global consumption and production of natural gas reported
by the U.S. Energy Information Administration.

Global coal consumption fell by 1.8% in 2015, well below the 10-year average
annual growth of 2.1% and the largest reported decline thus far (see Figure 1.4). The
entire net decline was accounted for by the US (-12.7%, the world’s largest
volumetric decline) and China (-1.5%), partially offset by modest increases in India
(+4.8%) and Indonesia (+15%). Coal’s share of global primary energy consumption
fell to 29.2%, the lowest share since 2005, (BP, 2016a)

21
 3600
Consumption
3500 Production
3400
Billions of cubic meters 3300
of natural gas 3200
3100
3000
2900
2800
2700
2600
2500
2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Year
Figure 1.3: Global consumption and production of natural gas since 2004 (U.S. Department of
Energy, 2016)

4000
Consumption
3900
3800 Production
3700
3600
equivalent of coal
Million tonnes oil

3500
3400
3300
3200
3100
3000
2900
2800
2700
2600
2500
2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Year
Figure 1.4: Global consumption and production of coal since 2004 (BP, 2016a, 2012; U.S.
Department of Energy, 2015)

These statistics suggest that traditional energy sources (i.e. oil, natural gas, coal, etc.)
are changing in unpredictable ways which directly impact several aspects of society
(economics, foreign politics, culture, etc.). Also, the means to obtain these fuels and
their consumption are believed to be detrimental to the environment. Several
alternatives for the reduction of toxic gas emissions to the atmosphere have been
proposed, but yet, fossil fuels are currently the most used energy sources in the world
22
(Escobar et al., 2009; Hoekman, 2009; Rašković et al., 2010). However, BP also
refers to a global increase of 7% in the production of renewable sources in their
report of June 2016 and an increase in the total world consumption of 11%, a trend
that has persisted for the last 10 years (BP, 2016a; Koizumi, 2015; U.S. Department
of Energy, 2015).

1.3 Biofuels
The production of biofuels has been proposed as a sustainable alternative for energy
generation in order to reduce the usage of fossil fuels whose emissions are believed
to be the main cause of global warming and the so-called “greenhouse effect” (Balat
et al., 2008; Cardona et al., 2010; Demirbas, 2007). Biofuels make reference to the
compounds whose origins are a manifestation of the capture and storage of solar
energy through photosynthetic reactions (Raman et al., 2015). In the case of
vegetable forms such as plants and algae, the oils are products of photosynthesis
(Drapcho et al., 2008; Karlsson et al., 2014; Sánchez and Cardona, 2012). There are
several forms of biofuels such as biodiesel which are a direct product of chemically
transesterified oils, alcohols and alkanes which can be produced from organic
substrates via anaerobic fermentation and biofuels like hydrogen that can be
produced using chemical and biological routes (Cardona et al., 2010; Demirbas,
2007; Drapcho et al., 2008; Hoekman, 2009).

Biofuels are classified according to their source and type: primary and secondary
biofuels. Primary biofuels are usually used in an unprocessed form, mainly for
heating, cooking or electricity generation. These primary biofuels include hard and
soft wood, wood chips and pellets, etc. (Drapcho et al., 2008; Escobar et al., 2009).
They are often found as by-products in processes such as deforestation, agriculture,
fishery products, municipal wastes, food industry and food services (Cardona et al.,
2010; Drapcho et al., 2008; Felix and Tilley, 2009).

Secondary biofuels are the result of the conversion of biopolymers found in the raw
material through biological paths such as anaerobic metabolism or fermentation
(Siqueira et al., 2008; Sreenath et al., 2001; Sreenath and Jeffries, 2000; Triana et al.,
2011). Secondary biofuels are further divided into first, second and third generation
on the basis of raw material and technology used for their production (Bai et al.,
23
2008; Balat et al., 2008; Binod et al., 2010; Drapcho et al., 2008). The first
generation refers to the biofuels obtained from seeds, grains and sugars via
fermentation (e.g. ethanol from molasses or starch, and biodiesel from transesterified
seed oils such as soybean oil). The second generation of secondary biofuels are
obtained from lignocellulosic biomass (raw material mainly composed by cellulose,
hemicellulose and lignin) to produce either ethanol via enzymatic hydrolysis and
fermentation, or methane via anaerobic digestion (Balat et al., 2008; Petersson et al.,
2007). Finally, the third generation comprises all the biofuels obtained from algae
and sea weeds (Nigam, 2001).

Biofuels can be solids such as fuel wood, charcoal, and wood pellets; or liquid, such
as ethanol, biodiesel and pyrolysis oils; or gaseous, such as biogas and hydrogen
(Cardona and Sánchez, 2007; Drapcho et al., 2008; Nigam, 2001). Table 1.1 shows a
comparative study between the different forms of biofuels in terms of the energy
density. The information presented in Table 1.1 allows a better understanding about
the energy potential that these components possess and how relevant their production
is for industrial purposes and for social and economic growth. Hydrogen has the
highest energy density of common fuels expressed on a mass basis. Other fuels such
as gasoline and biodiesel have energy densities ranging from 40–46 kJ/g. Alcohols,
on the other hand, present energy densities in the 20–30 kJ/g range but their
combustion is complete (lower concentrations of toxic emissions) (Hoekman, 2009;
Reijnders, 2006).

Liquid biofuels are primarily used to fuel vehicles, engines and energy cells for
electricity generation. There are several reasons for biofuels to be considered as
relevant technologies by both developing and industrialised countries (Hoekman,
2009; Koizumi, 2015; Su et al., 2015). For instance, biofuels represent a potential
solution to energy security, environmental concerns, foreign exchange savings, and
socioeconomic issues related to the rural sector (Demirbas, 2007; Koizumi, 2015).
Additionally, all biofuels have very low sulphur levels and many of them have low
nitrogen content which means that no SOx and very low concentrations of NOx gases
are produced during combustion (Demirbas, 2007; Hoekman, 2009; Nigam, 2001).

24
 Table 1.1: Energy densities for some biofuels (Drapcho et al., 2008)
Energy
Fuel source
density (kJ/g)
Hydrogen 143.0
Methane 54.0
Biodiesel 46.0
Gasoline 44.0
Soybean oil 40.2
Coal 35.0
Ethanol 29.6
Methanol 22.3
Soft wood 20.4
Hard wood 18.4
Bagasse 17.5

1.4 Ethanol
Ethanol will be considered in this study since it has a wide application in industry
(e.g. pharmaceutics, food and chemical industry). Table 1.1 presents the energy
density of some biofuels of industrial application. Ethanol, for instance, shows an
energy density of 29.6 kJ/g, which suggests that this biofuel has applicability and
potential for a well-established fuel for engines, electricity generation, etc. One of the
routes to produce ethanol is the hydration of ethylene in which excess of this gas gets
in contact with high pressure and high temperature steam to produce ethanol in a
reversible reaction catalysed by Phosphoric Acid (V), also known as synthetic
ethanol (Roberts and Caseiro, 1977).

𝐻3 𝑃𝑂4
𝐶2 𝐻4 (𝑔) + 𝐻2 𝑂 ⇔ 𝐶2 𝐻5 𝑂𝐻(𝑣𝑎𝑝)

Another route, currently out of circulation for being considered obsolete due to its
high energy consumption and low conversion of reactants is the hydration of
ethylene using sulphuric acid (Roberts and Caseiro, 1977; Streitwieser and
Heathcock, 1976).

𝐶2 𝐻4 (𝑔) + 𝐻2 𝑆𝑂4 → 𝐶2 𝐻5 𝑆𝑂4 𝐻

25
𝐶2 𝐻5 𝑆𝑂4 𝐻 + 𝐻2 𝑂 → 𝐶2 𝐻5 𝑂𝐻(𝑣𝑎𝑝) + 𝐻2 𝑆𝑂4

Both of these routes are expensive and require high levels of energy (i.e. temperature
300 oC and pressure 60 – 70 atm) and product recovery (the removal of the catalyst
and the outlet streams with low pH). Another route and the one that is currently
receiving most of the attention is the production of ethanol via fermentation. This
method allows the production of ethanol from substrates that mainly consist of
reducing sugars such as hexoses and pentoses.

The importance of bioethanol as an alternative source of energy lies on the concept

of clean energy which basically means that the production of ethanol should come
from renewable sources that allow this technology to be sustainable and
environmentally friendly (Fan et al., 2014; Karlsson et al., 2014; Raman et al., 2015;
Su et al., 2015). To this date, many countries are producing biofuels. The main
producers are USA with approximately 18.3 billion litres followed by Brazil with
17.5 billion litres and China with 1 billion litres on a yearly basis (Baeyens et al.,
2015; García et al., 2013; Petrobras, 2016).

The current trend of these countries, and others, is an increasing interest in the use of
particular renewable sources for energy generation purposes. The main raw materials
for the ethanol production in these countries are sugar cane, molasses and corn, but
there are also many other sources available (Hasunuma and Kondo, 2012; Kravanja
et al., 2013; Triana et al., 2015a). In this work, the raw material considered for the
ethanol production process is lignocellulosic biomass. The variety of lignocellulosic
materials is vast, ranging from solid urban waste, agricultural residues, paper pulp,
wood etc. However, none of these materials have a direct application in the
production of high-added-value products, such as ethanol or other metabolites of
industrial application. This work therefore aims to address the use of these materials
as potential feedstock for the ethanol production process.

1.5 Lignocellulosic biomass

Traditionally, lignocellulosic biomass has been utilised for direct combustion.
However, burning biomass still produces pollutants such as dust and acid rain gases

26
(e.g. sulphur dioxide and nitrogen oxides). Nevertheless, the amount of sulphur
dioxide, for instance, produced during the combustion of biomass is 90% less than
the amount produced by burning coal, and the concentration of atmospheric
pollutants produced during this process are insignificant in comparison to other
pollution sources (Chen and Zhang, 2015; Raman et al., 2015; Saxena et al., 2009).
Lignocellulosic biomass comprises all the materials with high content of cellulose
and hemicellulose trapped in the cellular walls by lignin. These materials are
classified as follows (Drapcho et al., 2008; Escobar et al., 2009; Hahn-Hagerdal et
al., 2006; Hoekman, 2009):

 Agricultural residues
 By-products of agricultural processes
 Crops for energy generation

Agricultural residues are the type of lignocellulosic biomass that is generated as a

result of processes such as fruit harvesting, seeds, animal wastes and more. These
materials are generated in large amounts, are easily accessible and also, fairly
inexpensive than other sources of carbon such as molasses and starch (Ballesteros et
al., 2004; Kadam et al., 2000; Palmqvist and Hahn-Hagerdal, 2000a). Some
examples of these materials and their chemical composition are shown in Table 1.2.

By-products of agricultural processes are the result of processes carried out in order
to harvest crops or to renew soil. Sugar cane bagasse is one the most used
agricultural residues in ethanol production via fermentation, especially in South
America where is being produced in large amounts (Aguilar et al., 2002; Avci et al.,
2013). Some other materials are listed in Table 1.3. Crops for energy generation
make reference to the crops that are meant to be used only for energy generation and
not for human consumption. Table 1.4 shows some examples of these materials and
their organic content. In general, the usage of biomass as a source of energy is of
interest in the production of biofuels due to the following benefits:

 Lignocellulosic biomass is a renewable, potentially sustainable and relatively

environmentally friendly source of energy due to its high organic content

27
 (Abedinifar et al., 2009; Bai et al., 2008; Balat et al., 2008; Ballesteros et al.,
2004)
 An increased use of biomass would extend the lifetime of diminishing crude
oil supplies (Cardona et al., 2010; Cardona and Sánchez, 2007; Hahn-
Hagerdal et al., 2006)
 Lignocellulosic materials are an energy source that could improve economies
and energy security (Sanchez and Cardona, 2005; Saxena et al., 2009; Triana
et al., 2011)
 The use of lignocellulosic biomass could lead to a reduction on the
production and accumulation of carbon dioxide in the atmosphere (Cardona
and Sánchez, 2007; Demirbas, 2007; Drapcho et al., 2008)

1.6 Current outlook for biofuels

The ethanol production from lignocellulosic biomass is a process that is still in an
early stage of development and requires a great deal of investigation. When
considering the implementation of the production of ethanol from lignocellulosic
biomass, several aspects need to be taken into account. For example, one of these
aspects is the availability of the raw material. Although many countries have
functional plants of bioethanol from lignocellulosic biomass, not every country has
access to the raw material and this can have a negative impact on the design of the
plant as costs related to importation and transportation can increase (Mabee et al.,
2011). Another relevant aspect in the design of an efficient plant of ethanol
production is the chemical composition of the raw material. Differences in the
chemical composition of the lignocellulosic material can affect the yield and the
amount of substrate produced during the pretreatment stage, the size of the
equipment and the energy requirements which later translates into lower yields of
ethanol during fermentation (Hasunuma and Kondo, 2012; Triana et al., 2011).

When considering the applicability of the ethanol production process in industry,

another aspect that should be considered is the operating conditions; in order to
obtain high yields of ethanol, the enzymes require long periods of time
(approximately 60 - 80 hours) to produce the necessary amount of reducing sugars
for the subsequent fermentation (Avci et al., 2013; Cara et al., 2007; X. Zhang et al.,
2009). Additionally, the fermentation stage also requires long residence times to
28
reach the desired concentration of the final product. In addition to the residence time,
these two stages require pH and temperature conditions that guarantee the survival of
the biological agent (Morales-Rodriguez et al., 2012, 2011). In the case of the
pretreatment stage, this is one of the most problematic parts of the process, as large
amounts of energy are needed to break down the lignin to release the cellulose and
hemicellulose (Behera et al., 2010; Davis et al., 2005; Golias et al., 2002; Hahn-
Hagerdal et al., 2006; Patle and Lal, 2008; Zhu and Pan, 2010).

Table 1.2: Cellulose, hemicellulose and lignin content in agricultural residues

(% w/w dry basis)
Material Cellulose Hemicellulose Lignin Reference
Barley straw 36 – 37 17 – 19 12 – 13 (Freer and Detroy, 1983)
(Drapcho et al., 2008;
Corn stover 36 – 39 21 – 22 19 – 20
Mielenz et al., 2009)
(Binod et al., 2010;
Rice straw 32 – 47 19 – 27 5 – 24 Drapcho et al., 2008;
Karimi et al., 2006)
(Drapcho et al., 2008;
Wheat straw 15 – 40 25 – 35 <8 Mielenz et al., 2009;
Saha et al., 2005)

Table 1.3: Cellulose, hemicellulose and lignin content in by-products of agricultural processes
(% w/w dry basis)
Material Cellulose Hemicellulose Lignin Reference
Sugar cane (Cardona et al., 2010;
40 – 45 30 – 35 20 – 30
bagasse Carrasco et al., 2011)
(Noureddini and Byun,
Corn fibres 13 – 18 35 – 40 7–8 2010;
Rasmussen et al., 2010)
(Saha & Cotta 2007;
Rice husk 15 – 36 12 – 35 8 – 16
Saha & Cotta 2008)
Soy husk 20 – 51 10 – 20 1–4 (Mielenz et al., 2009)

29
 Table 1.4: Cellulose, hemicellulose and lignin content in crops for energy generation
(% w/w dry basis)
Material Cellulose Hemicellulose Lignin Reference
(Drapcho et al., 2008;
Hybrid
36 – 37 17 – 19 12 – 13 H.-J. Huang et al.,
poplar
2009)
(Drapcho et al., 2008;
H.-J. Huang et al., 2009;
Willow 36 – 39 21 – 22 19 – 20
Sassner et al., 2008,
2006)
(Drapcho et al., 2008;
Hay 32 – 47 19 – 27 5 – 24
Xu et al., 2010)

Alternative configurations based on new technologies have arisen to improve the

efficiency of the process. Several authors have suggested a configuration that
comprises two saccharification stages (i.e. one to produce xylose from hemicellulose
and another one to produce glucose from cellulose), whereas some other authors
have considered one unit that combines the fermentation and the enzymatic
saccharification (Agbogbo et al., 2006; Ballesteros et al., 2004). Current research is
also focusing on the design of new and genetically modified strains, able to grow in
the hydrolysates obtained on the stages of saccharification and new separation
methods that allow to reduce energy consumption and increase product purity (Bai et
al., 2008; Ballesteros et al., 2004; Cardona and Sánchez, 2007; Saxena et al., 2009).

Last but not least, energy consumption within the process has become a relevant
aspect. As previously mentioned, most of the units of the process require large
amounts of energy to guarantee either a high yield of substrate or a high
concentration of the key product in the separation and purification stages. Energy
reduction is therefore the main motivation behind this work. The aim is to determine
the optimal configuration of the process which will distribute more effectively the
required heat within the plant with minimum energy losses, and therefore, with
minimum operating costs.

30
 1.7 Summary
This chapter presented an introduction to the concept of energy generation and the
different sources used for this purpose. The chapter also conveyed the idea that fossil
fuels have become an intrinsic part of society and that their use is directly linked to
development and industrialisation. However, as many studies have shown in the last
decades, the usage of fossil fuels and the gas emissions associated to their
consumption are leading to harmful effects on the ozone layer, which can be
observed in the rise of the level of the seas, the melting of glaciers in the poles,
droughts, among others. The objective of this chapter was to present a sustainable
alternative for energy generation to contribute to the reduction in the utilisation of
traditional fossil fuels. The production of ethanol from agricultural residues stands as
an interesting proposal for clean energy since the combustion of this alcohol is
complete and the concentration of toxic gas emissions significantly lower than in the
cases of oil, carbon and natural gas.

1.8 Aims
The main objective of this thesis is to conduct a more comprehensive study based on
rigorous optimisation methods to determine the optimal distribution of the heat
sources within the ethanol production process considering aspects such as size of the
plant, design of the different units and costs. This is a novel approach to reduce
energy consumption which considers robust kinetic models, dynamic mass and
energy balances, equations of state, activity models, etc. The deliverable of this work
can be seen in Chapters 4 through 6.

1.9 Organisation
This thesis consists of seven chapters and an appendix and their organisation is the
following:

 Chapter 1: Introduction to the concept of fuels and the current global scenario
regarding production and consumption. Chapter 1 also presents a general
review of biofuels leading to production of ethanol from lignocellulosic
biomass.
 Chapter 2: Updated state-of-the-art for the some works related to the
production of ethanol and the different units involved in the process. This
31
 literature survey includes new technologies, methods of production,
pretreatment techniques, new strains, etc.
 Chapter 3: Presentation of the mathematical models used in this work and
their respective validation.
 Chapter 4: The development of an empirical mathematical model of an
organophilic membrane for ethanol removal from aqueous solutions.
 Chapter 5: Optimisation of the separation stages (i.e. distillation and
pervaporation) for the mixture ethanol/water.
 Chapter 6: Optimal distribution of the heat sources throughout the process to
improve energy efficiency. This chapter presents heat integration of the units
as a method to reduce the usage of utilities and therefore the reduction in the
overall energy consumption within the process.
 Chapter 7: General conclusions of the thesis emphasising on the objectives
achieved. Recommendations for future work are proposed in this chapter
which have to show that the PhD work can have continuity and applicability.
 Appendix A: Detailed kinetic models applied to mass and energy balances of
the units in the process.

32
 Chapter 2 – Literature review

Abstract

Chapter 2 presents the main stages of the ethanol process and covers some of the
most recent publications regarding pretreatment, overliming, enzymatic hydrolysis,
fermentation and separation methods. Furthermore, a review of the different
approaches on heat integration and optimisation of the ethanol production process
will be introduced in order to establish what has already been explored and what is
open for investigation.

2.1 Introduction
This chapter intends to convey the importance of a thorough literature review and
how it can be useful to set the ground and the direction of modern applications and
methodologies for the production of ethanol from lignocellulosic biomass. In this
chapter, recent developments and new methods in the pretreatment of lignocellulosic
biomass as well as the recent applications for detoxification, enzymatic hydrolysis
and fermentation are presented. However, many of these works do not present an
analysis of how this specific unit or units affect the entire plant or if they are
economically feasible or not. The mathematical models for the reactors as well as for
the hydrophilic membrane are taken from literature and modelled and simulated
using gPROMS. The models of the distillation column, evaporator and heat
exchangers are formulated by the author based on mass and energy balances with
their respective correlations for heat transfer and VLE.

The standard configuration for ethanol production consists of four main stages:
Pretreatment, enzymatic hydrolysis, fermentation and separation (Drapcho et al.,
2008; Sanchez and Cardona, 2005; Sánchez and Cardona, 2012). This work also
includes the review of hybrid processes and heat integration. Hybrid processes have
emerged as an alternative to improve the efficiency of a plant and reduce costs.

33
The literature review conducted on this subject shows great advances and interesting
results in terms of energy consumption, design and operating conditions in both
simulation and experimentation. On the other hand, heat integration for the ethanol
process has not been widely studied and few researchers have really focused their
effort on this matter in order to improve not only the efficiency of the plant but also
the design thereof. The different studies will be further discussed throughout this
chapter. This chapter will conclude with the presentation of the research statement
which will be useful to stablish the deliverables of this thesis and its novelty.

2.2 General aspects of the ethanol production process from lignocellulosic

biomass
Figure 2.1 shows the general series of steps in the ethanol production process for any
raw material. In the case of ethanol production from lignocellulosic biomass, the raw
material consists mainly of cellulose, hemicellulose and lignin. Lignin is a
biopolymer whose bonds keep cellulose and hemicellulose trapped in the cellular
walls of the material. The pretreatment stage (which usually operates at high
temperatures and pressures) basically breaks said bonds and releases hemicellulose
and part of cellulose making both biopolymers more susceptible to the action of
chemical agents (e.g. acids, bases or enzymes) (Agbogbo et al., 2006; Esteghlalian et
al., 1997). The content of the pretreatment reactor is then passed through a filter and
divided into two streams: a liquid fraction (rich in reducing sugars) and a solid
fraction (mainly unreacted cellulose and other biopolymers).

The solid fraction is then treated with specialised enzymes (e.g. cellulases and β-
glucosidases) in order to degrade cellulose into cellobiose, glucose and other hexoses
(Bezerra and Dias, 2005; Cara et al., 2007). In some cases and depending on the
fermenting strain, other configurations for the ethanol production process can be
considered.

34
 Figure 2.1: Standard configuration for ethanol production

Figure 2.2 shows an arrangement where two fermenters, one for hexoses and one for
pentoses, are required (Agbogbo et al., 2006; Krishnan et al., 1999; Olofsson et al.,
2008). In Figure 2.2, the fermenting microorganism is not necessarily the same in
both fermenters since most natural strains are not capable of assimilating both
hexoses and pentoses, simultaneously (Morales-Rodriguez et al., 2011; J. Zhang et
al., 2009). Figure 2.3 shows an example of a configuration where specialised strains
are utilised in the conversion of both reducing sugars and is known as separate
hydrolysis and fermentation (SHF) (Abedinifar et al., 2009).

The ethanol production process from lignocellulosic biomass can also be approached
by considering a configuration where enzymatic hydrolysis and co-fermentation are
combined into a single unit in order to increase the yield of ethanol by reducing the
inhibitory effects of substrate over the enzymes and product over the fermenting
strains (Ballesteros et al., 2004; Kadar et al., 2004; Morales-Rodriguez et al., 2011).
Figure 2.4 shows the configuration for the simultaneous saccharification and co-
fermentation process (SSCF) (Jin et al., 2012; Su et al., 2012). This configuration is
the one to be considered in this work since it combines two processes which already
guarantees fewer units in the plant and also contributes to the reduction in the usage
of utilities. Other units are also shown in Figure 2.4, including an evaporator which is
used to concentrate xylose in the liquid fraction and a detoxification stage where the
toxic by-products obtained during the pretreatment are degraded into insoluble salts
to reduce their inhibitory effect on the fermenting strains (Purwadi et al., 2004).

35
Companies such as Petrobras (Brazil), INCAUCA (Colombia), BP (Brazil) use
molasses and sugar cane bagasse to produce ethanol at continuous operation of the
four main stages (i.e. including pretreatment, fermentation using yeast, two
distillation columns and a dehydration section) (BP, 2016b; Incauca, 2016; Petrobras,
2016). The configurations presented in Figures 2.2 to 2.4 are proposed in several
works but their implementation in industry still requires further investigation. These
configurations also require an environmental analysis which determines the viability
of this process and how its residual streams can be treated without causing further
pollution.

2.3 Pretreatment
Every lignocellulosic material used in the production of ethanol has to be pretreated
in order to reach high levels of conversion of the degradable biopolymers. The
purpose of the pretreatment stage is to produce reducing sugars (e.g. glucose and
xylose) from cellulose and hemicellulose. The pretreatment of the raw material has to
meet the following requirements to guarantee high yields (Abedinifar et al., 2009;
Drapcho et al., 2008; Freer and Detroy, 1982):

 Release of the main biopolymers such as cellulose and hemicellulose by

degrading lignin
 Production of reducing sugars that can be assimilated by fermenting
microorganisms
 Minimisation on the degradation or loss of usable carbohydrates

During the pretreatment, the cellular matrix of the lignocellulosic material is broken,
releasing cellulose and hemicellulose. Additionally, the crystallinity of cellulose
decreases, making it more susceptible to a further enzymatic attack and, at the same
time, the hemicellulose is partially hydrolysed into pentoses (mainly xylose). Table
2.1 shows some of the most employed methods in the pretreatment of lignocellulosic
biomass and the production of reducing sugars. The raw material considered in this
work is corn stover, which is mainly composed of cellulose, hemicellulose, lignin
and some other compounds such as protein and ashes. This raw material is very
representative due to its organic content (i.e. the content of degradable biopolymers)
and its availability in many countries around the world.
36
Figure 2.2: Configuration of the ethanol production process from lignocellulosic biomass
considering two fermentation stages

Figure 2.3: Configuration of the ethanol production process from lignocellulosic biomass
considering co-fermentation of both pentoses and hexoses

Figure 2.4: Configuration of the ethanol production process from lignocellulosic biomass
considering simultaneous saccharification and co-fermentation (SSCF)

37
 Table 2.1: Pretreatment methods for lignocellulosic materials
Pretreatment Features Reference
Concentrated The raw material is treated with concentrated (Abedinifar et al. 2009;
acid (CA) sulphuric acid at 121 °C for 4 hours. Saxena et al. 2009)
A 2 % (w/w) solution of H2SO4 or HCl is mixed (Mohagheghi et al. 2006;
Dilute acid (DA) with the raw material and heated at 121°C for 3 Saxena et al. 2009;
to 4 hours. Sukumaran et al. 2009)
(Abedinifar et al. 2009;
Alkaline A 2 % (w/w) solution of NaOH or KCl is mixed
Saxena et al. 2009;
hydrolysis (AH) with the solid and heated at 121°C for 3 hours.
Sukumaran et al. 2009)
The raw material gets in contact with ammonia
Ammonia fibre
in a ratio of 2 kg of NH3 per kg of solid at 90 °C (Balat et al., 2008)
explosion (AME)
for 30 min.
The raw material is submitted to a reaction in (Kim et al. 2008;
Ozonolysis (O) presence of ozone at ambient temperature and Sanchez & Cardona
pressure. 2005)

Liquid The raw material is heated in water in a ratio of

(Prasad et al., 2007)
hot water (LHW) 1:10 at 220 °C for 2 min.
The raw material gets in contact with high
(Balat et al. 2008;
pressure steam (5 – 10 bar) for residence times
Steam Sanchez & Cardona
from 5 to 30 min. It is quickly depressurised so
explosion (SE) 2005;
the expansion of the vapour can cause the
Sassner et al. 2006)
break-up of the cellular matrix.

All the techniques listed in Table 2.1 have shown promising results at laboratory
scale. Pretreatment methods such as dilute-acid (DA), liquid hot water (LHW) and
steam explosion (SE) are the most commonly used for lignocellulosic biomass.
However, kinetic models for most of these pretreatments are not available in
literature and a comparison of different methods is therefore not possible. For this
work, the pretreatment technique considered is dilute-acid (DA) which is widely
used in industry in the production of biofuels from molasses and sugar cane bagasse.
The kinetic parameters for the mathematical model for corn stover are reported in
literature by Esteghlalian et al., (1997). All kinetic models and parameters will be
discussed later in Chapter 3.

38
 Table 2.2: Literature review on the pretreatment of lignocellulosic biomass for ethanol
production
Description of the
Work Characteristics Reference
pretreatment
Steam explosion,
organosolv, and sulfite Advantages for woody
Three different pretreatment
pretreatment to biomass conversion
techniques for (Zhu and Pan,
overcome especially for softwood
delignification and 2010)
lignocelluloses species; 30 min
conversion of hemicellulose
recalcitrance reaction time at 180 °C
(SPORL)
High temperature and
Production of reducing
SPORL and dilute- pressure operation for
sugars with SPORL is (Shuai et al.,
acid (DA) SPORL and DA
87.9 %, compared to 2010)
pretreatments pretreatments in the
56.7 % with dilute acid
delignification of spruce
The conversion of
hemicellulose is 90 %.
Dilute mixed-acid Combination of sulphuric
The solid ratio 1:10 (Jackson de
pretreatment of and acetic acid to improve
showed the highest Moraes Rocha
sugarcane bagasse for the conversion of
conversion and the et al., 2011)
ethanol production hemicellulose into xylose
lowest furfural
concentration
Xylose (91.4% w/w)
Dilute acid Dilute H3PO4 (0.0–2.0%,
was obtained from (Avci et al.,
(DA) pretreatment v/v) is used to pre-treat corn
corn stover pretreated 2013)
with phosphoric acid stover (10%, w/w)
with H3PO4 1% (v/v)

Pretreatment is one of the key stages of the ethanol production process from
lignocellulosic biomass and many techniques have been investigated in order to
improve the production of xylose by reducing the crystallinity of lignin and degrade
its very intertwined structure (Avci et al., 2013; Jung et al., 2013). Table 2.2 shows
some of the research on the pretreatment of lignocellulosic biomass and the key
features and the results found therein.

All the methods presented in Table 2.2 not only seek to increase the production of
monomeric sugars from the raw material but also to reduce the production of by-
products such as furfural, HMF and organic acids which reduce pH in the growth

39
medium and raise the osmotic pressure of the cells during the fermentation (Diaz et
al., 2015; Elgharbawy et al., 2016; Kumar et al., 2009; Liguori et al., 2015; Öhgren et
al., 2007a).

Table 2.2: Literature review on the pretreatment of lignocellulosic biomass for ethanol
production (cont.)
Description of the
Work Characteristics Reference
pretreatment
Optimised operating Increase of ethanol yield
Dilute sulfuric acid conditions for the from 52.5 % to 87.5 %
pretreatment of oil pretreatment of oil palm to when pretreated EFB (Jung et al.,
palm empty fruit increase the production of slurry was treated with 2013)
bunches (EFB) ethanol in a microwave activated carbon before
digester subjecting to fermentation
Conversions of cellulose
Pilot-scale steam Comparative study between
pulp of 90 % and 79 %
explosion and diluted steam explosion (SE) and (Rocha et
were obtained from SE
sulfuric acid dilute acid (DA) in al., 2015)
and DA pretreatment,
pretreatments sugarcane bagasse
respectively
The moderate temperature
Pretreatment of corn stover pretreatment with
(Swain and
Utilisation of and rice husk using ammonia is suitable for
Krishnan,
aqueous ammonia aqueous solutions of high recovery of hexose
2015)
ammonia and pentose during
hydrolysis
Microwave irradiation is
applied to enhance enzyme 23 % of lignin is removed
Microwave-assisted
hydrolysis of corn straw when the material is (Diaz et al.,
pretreatment of
and rice husk immersed in pretreated in glycerol at 2015)
lignocellulosic biomass
water, aqueous glycerol or 180 °C for 2 min
alkaline glycerol
The operation goes from
IL’s have been increasingly
temperatures between 100-
Ionic liquid exploited as solvents and/or (Elgharbawy
120 °C and times between
pretreatment reagents in many et al., 2016)
1-5 h, depending on the
applications
raw material

40
 2.4 Detoxification
The detoxification stage is used for the reduction on the concentration of toxic
compounds produced during the pretreatment stage (i.e. furfural,
hydroxymethylfurfural and some organic acids). Regardless of the pretreatment
technique used to break down lignin and to degrade hemicellulose into reducing
sugars, these toxic compounds are formed (Mohagheghi et al., 2006).

The detoxification or overliming stage consists of a reactor in which the liquid

fraction coming out of the pretreatment stage is treated with a chemical agent to
neutralise the acids in the hydrolysate solution and to convert the furfural and HMF
into insoluble salts that can be separated by filtration. The operating conditions for
this stage are usually between 30 – 60 oC, pH between 9 – 12 and residence time of
30 min. The most used agent for the detoxification of the acid hydrolysates produced
during pretreatment is Ca(OH)2 showing good results in terms of conversion of
furfural and organic acids and good applicability in industry (Chandel et al., 2007;
Klasson et al., 2013; A. D. Moreno et al., 2013; Purwadi et al., 2004; Xie et al.,
2015).

Purwadi et al., (2004) presented a very detailed experimental work for the
detoxification of acid hydrolysates from Swedish spruce and the degradation of
reducing sugars during this stage. Purwadi et al., (2004) also proposed a
mathematical model to describe the reactions that would be more likely to take place
during this stage. However, not many authors have decided to evaluate these
parameters or try to obtain other kinetic models. Telli-Okur and Eken-Saracoglu,
(2008) presented a similar approach as Purwadi et al., (2004) for the neutralisation
and detoxification of the hydrolysates from sunflower seeds using Ca(OH)2 obtaining
high rates of furan removal (41 – 68 %).

Klasson et al., (2013) presented an experimental work in which they performed a

simultaneous detoxification, enzymatic hydrolysis and fermentation from milled
switchgrass using biochar, seeking to reduce the inhibition of furfural and other by-
products present in the hydrolysates. Similarly to Klasson et al., Xie et al., (2015)
presented the experimental procedure to reduce the concentration of toxic by-
products using amino acids. Five amino acids are considered in this work: cysteine,
41
histidine, lysine, tryptophan and asparagine. The results of this research showed a
total conversion of furfural and other by-products. However, the authors have yet to
estimate costs related to the utilisation of amino acids. In general, the overliming
process applied to ethanol production has been very useful for the degradation of
toxic compounds and its outcomes and operating conditions often go hand in hand
with the performances in the pretreatment stage and the fermentation stage.

2.5 Enzymatic hydrolysis

The pretreatment stage produces two fractions: A liquid fraction rich in xylose and a
solid fraction which is rich in unreacted cellulose and lignin. Different pretreatment
techniques available, such as dilute-acid pretreatment, ozonolysis, steam explosion,
among others, have high hemicellulose conversion during this stage (approx. 18 – 30
g/L of xylose) (Saha et al., 2005; Sreenath and Jeffries, 2000; Telli-Okur and Eken-
Saracoglu, 2008). The remaining solid material, rich in cellulose, is then directed to a
further stage of pretreatment in which the biopolymer is broken down into its basic
structural units (mainly glucose and cellobiose) by the action of enzymes known as
cellulases (Bansal et al., 2009; Bezerra and Dias, 2005; Kadam et al., 2004). This
pretreatment stage is used in the production of substrate for the fermenting
microorganisms, which is an important part in the ethanol production process and is
also known as enzymatic hydrolysis. This reaction is widely used in industry
reporting high yields in glucose production (approx. 15 – 30 g/L of glucose) (Bansal
et al., 2009; Cara et al., 2007; J. Zhang et al., 2009).

2.5.1 Enzymatic activities

The mechanism to convert cellulose into glucose and cellobiose can make use of two
different major groups of enzymes (Nigam, 2001; Sreenath et al., 2001):

1. Endoglucanases (1,4-β -D-glucanohydrolases).

2. Exoglucanases:

 Cellodextrinases (1,4-β-D-glucan glucanohydrolases)

 Cellobiohydrolases (1- β -D-glucan cellobiohydrolases)
 β-Glucosidases (β -glucoside glucohydrolases)

42
Endoglucanases cut randomly in the amorphous sites of the polysaccharide chains of
cellulose producing olygosaccharides of several lengths and consequently new ends
of the chain. Exoglucanases act gradually over the reducing or non-reducing ends of
the chains of cellulose releasing glucose (by the action of glucanhydrolases) and
cellobiose (by the action cellobiohydrolases) as products in major proportion. β-
Glucosidases degrade the cellobiose into glucose in a parallel reaction (Mohagheghi
et al., 2006; Telli-Okur and Eken-Saracoglu, 2008).

2.5.2 Enzymatic mechanism

The enzymatic hydrolysis takes place on the surface of cellulose and it has a complex
nature which involves many more stages that any other reaction catalysed by
enzymes. The different action mechanisms over the substrate are the following:

1. Adsorption of cellulase in the substrate (Jervis et al., 1997)

2. Location of a bond susceptible to hydrolysis on the surface of the substrate
(formation of chain ends in the case of carbohydrolases and bond division in
the case of Endogluconases) (Jervis et al., 1997)
3. Formation of the enzyme-substrate complex (Divne et al., 1998; Mulakala
and Reilly, 2005)
4. Hydrolysis of the β-glucosidic bond followed by the simultaneous sliding of
the enzyme along the chain of cellulose (Divne et al., 1998; Mulakala and
Reilly, 2005)
5. Desorption of the cellulases from the substrate (Divne et al., 1998; Mulakala
and Reilly, 2005)
6. Hydrolysis of cellobiose by the action of β-glucosidases and the subsequent
inhibition of the enzyme by this product (Bezerra and Dias, 2005; Holtzapple
et al., 1990; Xiao et al., 2004)

Enzymes commonly used in glucose production from cellulose are shown in Table
2.3. The most commonly used enzyme in industry, given its easy acquisition and low
prices, is Celluclast 1,5 L.

43
 Table 2.3: Enzymes used in conversion of cellulose into glucose
Yield (g of sugars
Cellulase Reference
/g of solid residues)
(Cara et al. 2007;
Celluclast 1,5L 0.11 – 0.26
Zhang et al. 2009)

Mycotech 0.13 – 0.25 (Cara et al., 2007)

Genencor 0.17 – 0.30 (Cara et al., 2007)

(Bansal et al. 2009;
Cellubrix 0.20 – 0.40
Petersson et al. 2007)

The enzymes reported in Table 2.3 require specific conditions of temperature and pH
in order to ensure a high yield of glucose production and to avoid the formation of
microbial agents that can cause inhibition. Some authors recommend supplementing
the action of cellulases with β-glucosidases to further increase the production of
glucose by degrading cellobiose (Beck et al., 1990; Palmarola-Adrados et al., 2005).
Depending on the raw material and the cellulose crystallinity level, a cocktail of
several enzymes is implemented in the production of reducing sugars as a way to
increase the conversion of cellulose and other polymers that can be reduce to their
monomeric units.

Table 2.4 shows the different works where enzymatic hydrolysis has been used in the
production of hexoses from lignocellulosic biomass. The works presented in Table
2.4 are some of the most recent investigations on enzymatic hydrolysis of cellulose.
Although the results reported in these works are promising in terms of the level of
conversion and production of glucose, few publications have touched on the
modelling and mathematical formulation.

Kadam et al., (2004) presented a model based on the Michaelis-Menten kinetics for
the production of glucose, xylose and cellobiose taking into consideration inhibitory
effects and loss of enzyme activity. This approach, unlike other publications, focuses
on the different possible routes that can be measured during the enzymatic hydrolysis
and the inhibitory effects of both substrate and product over the activity of the
enzyme.

44
This work has been cited by over 90 authors, some of which include Kumar and
Wyman, (2008), Engel et al., (2012), Gaykawad et al., (2012), Morales-Rodriguez et
al., (2012), Cheng et al., (2015), Liguori et al., (2015), Niu et al., (2016), among
others. The work of Morales has also been included in this thesis, as explained in the
following chapters, since it merges both the kinetic model of the enzymatic
hydrolysis with the model of co-fermentation of xylose and glucose using bacteria.

2.6 Fermentation
During the production of the reducing sugars (i.e. pentoses and hexoses), some by-
products are also produced with inhibitory effects on the fermenting microorganisms
(Chandel et al., 2007; Sreenath et al., 2001; Triana et al., 2011; Zhu et al., 2015). The
fermentation stage is the step of the process in which the lignocellulosic hydrolysates
are converted into ethanol via anaerobic metabolism. This is an important step in the
process and most of the current research is focused on finding the optimal conditions
of aspects such as new technologies and methodologies for growth, operating
conditions and the design of new fermenting strains in order to achieve high ethanol
yields by using all the available substrate (Baeyens et al., 2015; Carrasco et al.,
2011; Liguori et al., 2015; Swain and Krishnan, 2015). Table 2.5 shows some of the
most commonly employed strains in industry along with the ethanol yields.

The yields listed in Table 2.5 are calculated as experimental ethanol yield/theoretical
ethanol yield (theoretical ethanol yield is 0.511 g of ethanol/g of available substrate)
(Drapcho et al., 2008). Some researchers have developed new alternative
configurations for this stage. Some of these alternatives include the recycling of the
cells or the substrate, continuous fermentation with simultaneous saccharification or
fermentation of only glucose, presenting good results. However, in the case of
lignocellulosic biomass, most of the hydrolysates not only present hexoses but also
pentoses. Some of the microorganisms in Table 2.5 are therefore considered when
multiple-substrate fermentations are required to ensure higher yields of ethanol (Cho
et al., 2010; Dien et al., 1998; Zhao et al., 2008).

45
Table 2.4: Use of enzymatic hydrolysis in the production of glucose and other reducing sugars
from cellulose
Work Enzyme Characteristics Reference
Simultaneous
saccharification and The raw material is pretreated
(Ballesteros
fermentation (SSF) Celluclast 1.5 L with steam. SSF is carried out at
et al., 2004)
from woody and 42 oC for 72 hours
herbaceous materials
Enzymatic hydrolysis at 50 °C
Hydrolysis of Celluclast 1.5 L and (Kumar et
and 150 rpm for 72 h and at 2%
Japanese red pine Novozyme 188 al., 2009)
(w/v) substrate concentration
Estimation of the kinetic
Enzymatic hydrolysis
CPN cellulase and β- parameters for a mathematical (Kadam et
of corn stover for
glucosidase Novo 188 model that describes al., 2004)
ethanol production
saccharification of cellulose
SSF of lignocellulose Production of ethanol using (Hasunuma
to ethanol with coupled fermentation and and
Celluclast 1.5 L o
thermotolerant yeast enzymatic hydrolysis at 48 C for Kondo,
strains 24 hours 2012)
Optimisation of
Lignocellulose rich sweet
saccharification of
Celluclast 1.5 L + β- sorghum bagasse was used as a
sweet sorghum (Saini et
glucosidase Novozyme substrate for cellulolytic
bagasse using al., 2013)
188 hydrolysis at 28 ± 2 °C for
response surface
8 days
methodology
T. reesei cellulolytic
Production of enzymes from
Use of Cellulases cocktail for the
Trichoderma reesei as means to (Berrin et
from Trichoderma saccharification of
optimise the saccharification of al., 2014)
reesei lignocellulosic
lignocellulosic biomass
feedstocks
Biological β-glucosidase from
pretreatments to almonds (49,290),
Production of ethanol from
increase the efficiency hemicellulase from A. (Larran et
switchgrass with high content of
of the niger (H2125), al., 2015)
C4 using Fungal supernatants.
saccharification from cellulase from A. niger
Spartina argentinensis (C1184)
SSCF for improving
Steam-exploded corn stover for (Liu and
xylose conversion
Cellulase Cellic CTec2 ethanol production using SSCF Chen,
from steam-exploded
at 50 oC at high solid loading 2016)
corn stover

46
 Table 2.5: Microorganisms used in ethanol production

Microorganism Substrates Yield (%) Reference

Glucose, Galactose, (Behera et al., 2010;

Saccharomyces cerevisiae 43 – 87
Maltose Siqueira et al., 2008)
Glucose, Xylose, (Drapcho et al., 2008;
Pichia stipites 83 – 92
Arabinose, Mannose Nigam, 2001)
Glucose, Xylose, (Chandel et al. 2007;
Candida shehatae 75 – 90
Arabinose, Mannose Huang et al. 2009)
Glucose, Xylose, (Freer & Detroy 1982;
Candida lusitaniae 70 – 85
Sucrose, Cellobiose Sreenath et al. 2001)
Glucose, Xylose, (Jamai et al. 2007;
Candida tropicalis 76 – 85
Sucrose, Cellobiose Rattanachomsri et al. 2009)
(Garcia et al.,2010;
Escherichia coli Glucose, Glycerol 70 – 90
O’Brien et al., 2004)
Glucose, Xylose, (Behera et al., 2010;
Zymomonas mobilis 80 – 92
Fructose, Sucrose Leksawasdi et al., 2001)
Glucose, Fructose, ( Ji et al. 2009;
Klebsiella oxytoca 80 – 90
Sucrose, Maltose Sanchez & Cardona 2005)
Glucose, Xylose, (Romero et al., 2007;
Pachysolen tannophilus 75 – 80
Glycerol Zhao et al., 2008)
Clostridium thermocellum Glucose, Xylose 60 – 80 (Balusu et al., 2005)
Kluyveromyces marxianus Glucose 70 – 80 (Kadar et al., 2004)

Publications related to this stage have explored different aspects such as operating
conditions, growth media, fermenting microorganisms, and design of the reactor
(Gaykawad et al., 2013; Kim et al., 2008; Öhgren et al., 2007a; Sanchez et al., 2004;
Sasaki et al., 2015; Sreenath et al., 2001). However, current research is more
focused on the fermenting microorganisms (e.g. yeast, bacteria or fungi) and how to
obtained higher concentrations of ethanol with lower production of cells in a shorter
residence time (Golias et al., 2002; Tang et al., 2008; Tian et al., 2009; Zhu et al.,
2015). In terms of the design of the reactor, research is also looking to couple other
processes with the fermentation stage in order to increase ethanol yield and reduce
the inhibitory effects of substrate, product and by-products over the performance of
the microorganism (Kumar and Wyman, 2008; Palmqvist and Hahn-Hagerdal,
2000b; Ranjan et al., 2009; Xiao et al., 2004).

47
Some of these combinations include simultaneous saccharification and fermentation
(SSF), simultaneous saccharification and co-fermentation (SSCF), removal of
ethanol using selective membranes, genetically modified strains for specific
substrates, etc. (Gaykawad et al., 2013, 2013; Huang et al., 2006; Niemisto et al.,
2013; Tusel and Brüschke, 1985). Table 2.6 shows some of the works related to the
production of ethanol from lignocellulosic hydrolysates.

Table 2.6: Recent experimental works related to the implementation of the SSCF process in the
production of ethanol from lignocellulosic biomass
Work Strain and enzyme Characteristics Reference
Two-step SSCF to
convert AFEX-treated
Ammonia fibre explosion
switchgrass to ethanol
S. cerevisiae 424A(LNH- is used to pretreat
using commercial (Jin et al.,
ST) + Spezyme CP and switchgrass and SSCF is
enzymes and 2010)
Novozyme 188 carried out at 35 °C for
Saccharomyces
120 h
cerevisiae 424A(LNH-
ST)
SSCF of AFEXTM Saccharomyces
pretreated corn stover cerevisiae 424A(LNH-
for ethanol production ST) + Spezyme CP (for Conversion of glucose and
using commercial cellulose conversion into xylose into ethanol from (Jin et al.,
enzymes and glucose and cellobiose) + pretreated corn stover at 2012)
Saccharomyces Novozyme 188 (for 45 °C for 168 h
cerevisiae 424A(LNH- cellobiose conversion
ST) into glucose)
In situ laccase treatment
of steam-exploded S. cerevisiae F12 + (A. D.
SSCF of laccase at 35 °C
wheat straw in SSCF Cellic CTec2 and Cellic Moreno
and pH 5.5 for 144 h
processes at high dry Htec2 et al., 2013)
matter consistencies

48
 Table 2.6: Recent experimental works related to the implementation of the SSCF process in
the production of ethanol from lignocellulosic biomass (cont.)
Work Strain and enzyme Characteristics Reference
High-Solid SSCF of The raw material is
Zymomonas mobilis CP4
alkaline-pretreated alkaline-pretreated
+ cellulase (GC220) with (Su et al.,
corncob using Corncob. The operating
30 FPU/g cellulose and 2012)
recombinant conditions are pH = 5.5
Multi-effect xylanase
Zymomonas mobilis CP4 and 30 °C for 96 h
Fed-batch SSCF using
steam-exploded wheat All the experiments were
straw xylose-fermenting S. cerevisiae KE6-12 + run at 35°C and 180 rpm
(A. Moreno
Saccharomyces Cellic CTec2 and Cellic for 144 hours (batch
et al., 2013)
cerevisiae strain: effect HTec2 SSCF) or 168 hours (fed-
of laccase batch SSCF)
supplementation
S. cerevisiae TMB3400 +
Steam pretreated wheat.
SSCF of whole wheat in Cellulase (Celluclast 1.5
The operating conditions (Erdei et al.,
integrated ethanol L)
of the SSCF process are at 2013)
production and b-glucosidase
35 °C for 120 h
(Novozym 188)
Genetically modified S.
Metabolic engineering Saccharomyces
cerevisiae for SSCF at 30
of Saccharomyces cerevisiae ethanol strains (Romaní et
°C for 140 hours obtaining
cerevisiae PE-2 and PE-2 and CAT-1 al., 2015)
92% of the theoretical
CAT-1 + Cellic Ctec2
yield
SSCF for pretreated rice
SSCF by using xylose- Saccharomyces
straw at 35 °C for 48 h
fermenting cerevisiae (Sasaki et al.,
reaching ethanol yield up
Saccharomyces MN8140X/TF-TF + 2015)
to 74 % of the theoretical
cerevisiae Cellic CTec2
yield
SSCF of dry diluted Genetically modified S.
S. cerevisiae SyBE005 +
acid pretreated corn cerevisiae for SSCF with (Zhu et al.,
Accellerase 1500 and
stover at high dry 65% yield with 25% solid 2015)
Novozyme 188
matter loading loading
SSCF at different
SSCF for steam- Saccharomyces
glucose/xylose ratios 30 (Liu and
exploded corn stover at cerevisiae IPE003 +
°C, and 96 h with pH Chen, 2016)
high solid loading Cellic CTec2
ranges from 4 to 5

49
 375 1.0
Liquid
Vapour 0.9

370 0.8

Vapour mole fraction

0.7
Temperature (K)
365 0.6

0.5

360 0.4

0.3

355 0.2

0.1

350 0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Mole fraction Liquid mole fraction

Figure 2.5: Liquid-Vapour Equilibrium (VLE) calculated using ChemCAD for the mixture
ethanol/water at 1 atm

2.7 Separation processes

In the ethanol production process, the separation section or product recovery
represents a challenge given the presence of a minimum boiling point azeotrope in
the binary mixture ethanol/water (see Figure 2.5) using the NRTL method for the
activity coefficients and Equation of State PSRK for the acentric factor. Since a high
concentration of ethanol is required (approx. 99.5 % w/w) for fuel applications in
engines or in any other machinery, high levels of energy are necessary to obtain a
concentration close to the azeotrope (approx. 94% w/w) (Bai et al., 2008; Drapcho et
al., 2008).

Separation methods such as distillation and adsorption have shown high rates of
product recovery but with the energy-related costs in order to obtain concentrations
of ethanol around 94 % (w/w), which is still not suitable for biofuel purposes (Balat
et al., 2008; Lee et al., 2012). Alternative configurations to simple distillation
include: extractive distillation, swing-pressure distillation, double-effect distillation,
dividing wall distillation, etc. However, their implementation often entails additional
expenses related to controllability and operation and, in the case of extractive
distillation, solvent recovery (Kiss and Olujic, 2014; Le et al., 2015; Okoli and
Adams, 2015).

Another alternative for ethanol dehydration is membrane-assisted distillation, in

which membrane-based techniques are coupled with distillation systems to remove a

50
specific component or components from the feed mixture even beyond the barriers of
an azeotrope with lower energy consumption (Koch and Gorak, 2014; Koltuniewicz,
2010; Kookos, 2003). In most membrane separations, the feed stream is split into
two product streams: permeate and retentate. From which side the key product is
removed depends highly on the nature of the membrane (Chovau et al., 2011; Fan et
al., 2014; O’Brien et al., 2004; Valentinyi et al., 2013). The separation of the key
product is mainly caused by the concentration or pressure gradient between both
sides of the membrane. Since the membrane is highly selective, the key component
permeates throughout the membrane from the feed stream and is then removed by the
injection of an inert gas or by simply applying vacuum on the permeate side. Some
cases of membrane-based separations include osmosis, pervaporation and vapour
permeation (Koch and Gorak, 2014; Niemisto et al., 2013; Toth and Mizsey, 2015;
Valentinyi et al., 2013). Table 2.7 presents some of the most recent works related to
the purification of ethanol for fuel applications using different separation techniques.

2.8 Heat integration and heat storage

The optimisation of a process that includes heat exchange is an important step to
improve the productivity and efficiency of a plant by reducing energy consumption
and operating costs. It is for this reason that mathematical and thermodynamic
methods are used to solve the optimisation problems by re-distributing the energy
sources (i.e. utilities, process streams, etc.) (Baños et al., 2011; Biegler et al., 1999;
Bryson and Ho, 1975; Rašković et al., 2010; Stuart and El-Halwagi, 2012a).
Conventional methods for the optimisation of a heat exchange network can be
classified into three general categories: Pinch point analysis, mathematic methods
and metaheuristic methods (Hahn-Hagerdal et al., 2006):

1. Pinch point analysis: This analysis is a widely used method to determine the
minimal demand for hot and cold duties supplied by utilities and to maximise
heat recovery. The heat content of all the streams in the process and the initial
and final temperatures must be known and then a heat exchange network
(HEN) can be designed with minimum energy demand (Rašković et al., 2010;
Stuart and El-Halwagi, 2012a). The process data is represented as a set of
energy flows, or streams, as a function of heat load (kW) against temperature
(oC). These data are combined for all the streams in the plant to give
51
 composite curves, one for all hot streams (releasing heat) and one for all cold
streams (requiring heat) (Baeyens et al., 2015; Fornell and Berntsson, 2012;
Kravanja et al., 2013). The point of closest approach between the hot and
cold composite curves is the pinch point (or just pinch) with a hot stream
pinch temperature and a cold stream pinch temperature. This is where the
design is most constrained.

Table 2.7: Separation methods used in the dehydration of ethanol

Work Process Characteristics Approach Reference
Double distillation
Extractive
system. First column
Control of an extractive distillation process
for the removal of
distillation process to to produce Simulation using (Gil et al.,
ethanol (99.7 %
dehydrate ethanol using anhydrous ethanol ASPEN Plus 2012)
w/w) and the second
glycerol as entrainer using glycerol as
column for the
entrainer
recovery of glycerol
The membrane is
The use of
placed inside the
A new hybrid distillation and
distillation column in
distillation– pervaporation for
order to remove Simulation using (Haelssig et
pervaporation process ethanol separation
ethanol beyond the ASPEN Plus al., 2012)
are combined in a single considering
azeotrope at different
unit Membrane
ratios and permeate
Dephlegmation
rates
Several
configurations were
Innovative single step Extractive Simulation and
taking into
bioethanol dehydration dividing-wall optimisation (Kiss and
consideration
in an extractive dividing- distillation using using ASPEN Ignat, 2012)
obtaining
wall column glycerol Plus
concentrations of
99.8 % (w/w)
Bioethanol
dehydration, using
Enhanced bioethanol dividing-wall (ED) and (AD) used
dehydration by columns (DWC) in DWC processes Simulation and (Kiss and
extractive and azeotropic extended to reach purities of optimisation in Suszwalak,
distillation in dividing- extractive bioethanol over 99.8 ASPEN plus 2012)
wall columns distillation (ED) wt.%
and azeotropic
distillation

52
 Table 2.7: Separation methods used in the dehydration of ethanol (cont.)
Work Process Characteristics Approach Reference
Ethanol dehydration
Process analysis and Combination of Simulation in
process obtaining
optimisation of hybrid distillation, vapour ASPEN Plus (Roth et al.,
concentrations of
processes for the permeation and supported with 2013)
99.6 % (w/w) with
dehydration of ethanol adsorption experiments
lower capital costs
Distillate is sent to a
Ethanol recovery from membrane module to
fermentation broth with Membrane-assisted dehydrate ethanol in
(Vane et al.,
integrated vapour stripping the retentate. Experimentation
2013)
distillation−membrane process Permeate is recycled
process to the column. Purity
of ethanol 99 %
The hybrid process
produces conc. of
Azeotropic distillation Distillation column ethanol >99.4%wt.
and hybrid system for followed by The hybrid system (Kunnakorn
Experimentation
water–ethanol hydrophilic NaA presents energy et al., 2013)
separation zeolite membrane savings of 52.4%
against azeotropic
distillation
Optimisation of
hybrid processes Reduction of energy
Logic hybrid simulation- with a hydrophilic consumption for
Simulation using (Caballero,
optimisation algorithm membranes using a process where two or
MatLab 2015)
for distillation design tray-by-tray more components are
optimisation separated
approach
Distillate is sent to a
Ethanol recovery from membrane module to
fermentation broth with Membrane-assisted dehydrate ethanol in
(Vane et al.,
integrated vapour stripping the retentate. Experimentation
2013)
distillation−membrane process Permeate is recycled
process to the column. Purity
of ethanol 99 %

Hence, by finding this point and starting the design there, the energy targets
can be achieved using heat exchangers to recover heat between hot and cold
streams in two separate systems, one for temperatures above pinch
temperatures and one for temperatures below pinch temperatures (Baños et
al., 2011; Biegler et al., 1999; Bryson and Ho, 1975; Rašković et al., 2010;
Stuart and El-Halwagi, 2012a; Yee et al., 1990a, 1990b; Yee and Grossmann,
53
 1990). In practice, during the pinch analysis of an existing design, often
cross-pinch exchanges of heat are found between a hot stream with its
temperature above the pinch and a cold stream at temperatures below the
pinch. Removal of those exchangers by alternative matching makes the
process reach its energy target (Gorji-Bandpy et al., 2011). This method is
flexible and provides a wide perspective to the designer in terms of
configuration of the units and the solution. However, it is not possible to
simultaneously minimise the price of the network and the amount of heat
exchangers with this method (Gorji-Bandpy et al., 2011).

2. Mathematical methods: These methods make reference to the mathematical

formulations which describe all the units and the system as a whole. This set
of differential and algebraic equations includes the mass, momentum and
energy balances and the solution of these models is usually carried out by
using numerical methods (Gorji-Bandpy et al., 2011; Yee et al., 1990a,
1990b; Yee and Grossmann, 1990). For the optimisation of the process and
the heat exchange network, some authors suggest methods such as:

 Linear programming (LP): linear programming is a technique for the

optimisation of a linear objective function, subject to linear equality and
linear inequality constraints. Its feasible region is a convex polyhedron,
which is defined as the intersection of finitely many half spaces, each of
which is defined by a linear inequality. Its objective function is a real-
valued affine function defined on said polyhedron. A linear programming
algorithm finds a point in the polyhedron where this function has the
smallest (or largest) value if such a point exists (Biegler et al., 1999;
Bryson and Ho, 1975; Vanderbei, 2010).
 Nonlinear programming (NLP): The problem is called a nonlinear
programming problem if the objective function is nonlinear and/or the
feasible region is determined by nonlinear constraints (Biegler et al.,
1999; Yee et al., 1990a, 1990b; Yee and Grossmann, 1990).
 Mixed-integer linear programming (MILP): In a linear programming
method, all the decision variables x are continuous variables which may

54
 take on any positive value Ax > b and x > 0. In MILP, the decision
variables are integer values (Stuart and El-Halwagi, 2012b).
 Mixed-integer nonlinear programming (MINLP): It refers to
mathematical programming method with continuous and discrete
variables and nonlinearities in the objective function and constraints. The
use of MINLP is a natural approach of formulating problems where it is
necessary to simultaneously optimise the system structure (discrete) and
parameters (continuous) (Bryson and Ho, 1975; Stuart and El-Halwagi,
2012a; Yee et al., 1990a, 1990b; Yee and Grossmann, 1990).

3. Metaheuristics methods: Metaheuristics are general algorithmic

frameworks, often nature-inspired, intended to solve complex optimisation
problems. In recent years, metaheuristic are emerging as useful alternatives to
more classical approaches also for solving optimisation problems that include
in their mathematical formulation uncertain, stochastic, and dynamic
information, (Blum et al., 2005; Gendreau and Potvin, 2010). The
metaheuristic algorithms have several advantages over the mathematical
solvers. These methods are preferably used for complex spaces and since they
do not use step sizes for solving, the objective function can be non-linear and
non-continuous (Baños et al., 2011; Blum et al., 2005; Gendreau and Potvin,
2010). Some Metaheuristic methods include Simulated Annealing, Genetic
Algorithm, Tabu Search Method, among others:

 Simulated Annealing: This is a multivariable combinational optimisation

technique, based on an analogy with statistical mechanics. The basic
principle of this method resemblances the annealing of solids in which the
atoms are reorganized in order to obtain a minimum internal energy. For a
given configuration, a random move is carried out by randomly picking a
molecule and moving it in a random direction for a random distance (AN
et al., 2008; Liu et al., 2013).
 Genetic algorithms: In a genetic algorithm, a population of candidate
solutions (called individuals, creatures, or phenotypes) is evolved towards
better solutions. Each candidate solution has a set of properties (its
chromosomes or genotype) which can be mutated. Traditionally, solutions
55
 are represented in binary as strings of 0s and 1s, but other encodings are
also possible (Dieterle et al., 2003). The evolution usually starts from a
population of randomly generated individuals and is an iterative process.
The population in each iteration is also known as generation. In each
generation, the fitness of every individual in the population is evaluated;
the fitness is usually the value of the objective function in the
optimisation problem being solved. The fittest individuals are
stochastically selected from the current population, and each individual's
genome is modified (recombined and possibly randomly mutated) to form
a new generation. The new generation of candidate solutions is then used
in the next iteration of the algorithm. Commonly, the algorithm
terminates when either a maximum number of generations has been
produced, or a satisfactory fitness level has been reached for the
population (Rivera et al., 2006).
 Tabu Search: This is a heuristic approach for solving combinatorial
optimisation problems by using a guided, local search procedure to
explore the entire solution space without becoming easily trapped in local
optimal. It has been shown to be effective for scheduling problems,
traveling salesman problems, constraint satisfaction problems, as well as
many engineering optimisation problems, such as general zero-one mixed
integer programming problems (Teh and Rangaiah, 2003).

The main objective of this work is the integration of the heat sources within the
ethanol production process in order to reduce the overall energy consumption. In
order to achieve this goal, several aspects of the process need to be addressed. For
instance, the design of the plant, the optimal operating conditions of the different
units in the process, the evaluation of different alternatives that can improve the
efficiency of the process, sustainability, economic feasibility, etc.

This project wants to show that the production of biofuels is an appealing alternative
to the utilisation of fossil fuels and that they can also be implemented around the
world. Several authors have explored different configurations and technologies in
heat integration in the production of bioethanol from lignocellulosic biomass. Some

56
recent works include: Fornell and Berntsson, (2012), Grisales et al., (2008), Kravanja
et al., (2013), Modarresi et al., (2012), Rašković et al., (2010), Čuček et al., (2011).
These works are mainly focused on the implementation of the Pinch Analysis in the
production of ethanol from biomass and savings of energy (Fornell and Berntsson,
2012; Grisales et al., 2008; Kiran and Jana, 2015; Rašković et al., 2010; Yee et al.,
1990a, 1990b; Yee and Grossmann, 1990).

2.9 Summary and research statement

The literature review conducted in this chapter summarises the work done on design,
experimentation, modelling and optimisation of the different units involved in the
production of ethanol from lignocellulosic biomass. The main aspects and
characteristics of the studies were highlighted. It can be seen that for the
pretreatment, enzymatic hydrolysis and fermentation stages, current research is
mainly dedicated to the improvement of the operating conditions and techniques used
in the laboratory in order to achieve higher conversions. There is no economic
analysis behind most of these studies and there is no implementation in industrial
cases. However, this gap in literature can be advantageous and interesting to explore,
especially the application of detailed mathematical models into more robust
optimisation problems.

Mathematical models presented by Esteghlalian et al., (1997), Purwadi et al., (2004),

Kadam et al., (2004) and Leksawasdi et al., (2001) are some of the most relevant and
widely used kinetic models for pretreatment, detoxification, enzymatic hydrolysis
and co-fermentation, respectively. These models take into account several aspects
including inhibition by product, inhibition by by-product, inhibition by substrate, loss
of activity, the influence of the operating conditions, etc. These kinetic models will
be used in the simulation of the ethanol plant in gPROMS as described in the next
chapters.

Most of the work that has been done on hybrid processes is focused on the design of
systems with emphasis on simulation and optimisation. The majority of the work
presented in literature contained some form of comparative analysis to conventional
distillation processes (Kiran and Jana, 2015; Kiss and Olujic, 2014; Le et al., 2015;
Sudhoff et al., 2015). These comparative studies highlighted the benefits of these
57
new technologies and sold the idea that their implementation can result in significant
reductions in energy consumption and costs.

While previous works on hybrid processes have shown savings in costs, it should be
noted that the extent of said savings may vary significantly between studies. Works
by Kiss and collaborators have shown the benefits of different configurations for
dividing wall and membrane-based operations using modelling, which allows the
understanding of the theoretical flexibility of the system, as well as limitation of
these processes compared to existing technologies (Kiss and Ignat, 2012; Kiss and
Olujic, 2014; Kiss and Suszwalak, 2012). Publications by Górak and collaborators
include simulation work in different separation systems, which have been validated
against experimental studies in a pilot plant (Koch et al., 2013; Koch and Gorak,
2014; Roth et al., 2013; Sudhoff et al., 2015).

Research related to heat integration only made reference to the use of the Pinch
Analysis, which is very helpful and widely recognised. However, this approach does
not consider aspects such as the design of the plant and the configuration thereof or
costs related to operation and equipment. The research group of Chemical, Catalytic
and Biotechnological Processes (PQCB) led by Dr. Carlos Ariel Cardona has studied
the production of bioethanol from different raw materials found in Colombia’s rural
areas. Their approach consists of using ASPEN Plus to simulate the process and
evaluate the costs using the Economics tool known as Icarus developed by Aspen
Tech. They also include in all of their publications an analysis for the environmental
impact using the WAR algorithm which evaluates the toxicity of the outlet streams of
the process based on their chemical composition (Grisales et al., 2008; Sanchez and
Cardona, 2005; Sánchez and Cardona, 2012; Triana et al., 2011). Similarly,
researchers such as Dr. Ricardo Morales-Rodriguez and collaborators from the
University of Guanajuato in Mexico, have been working in the development of new
configurations of the process using ASPEN Plus to reduce the production costs
(Morales-Rodriguez et al., 2012, 2011).

The deliverable of this thesis is a practical methodology for the improvement of the
ethanol production process from lignocellulosic biomass, considering aspects such as
design and size of each unit within the plant, yield of ethanol, concentration of
58
ethanol at the end of the process, among other aspects. Other studies have used other
packages to simulate the entire plant. However, most of these works only consider
stoichiometric approaches for the reactors and none of them have focused on the
optimisation of the distillation section and the pervaporation network as well as the
minimisation of the total annualised costs for the whole process, which is something
novel this thesis wants to deliver. The following chapters will present the results of
the simulation and optimisation of the process as well as the implementation of heat
integration.

59
 Chapter 3 – Mathematical models and model
validation

Abstract

This chapter presents the mathematical models of all the units involved in the
production of ethanol from lignocellulosic biomass. The kinetic models of
pretreatment, detoxification, simultaneous saccharification and co-fermentation
(SSCF) and pervaporation are validated against experimental data found in
literature. The models of evaporation and distillation are formulated from mass and
energy balances. gPROMS is used for the simulation of all the units in the process
and ChemCAD is used in the validation of the mathematical model for the distillation
column.

3.1 Introduction

Simulations and modelling are often used in chemical engineering as tools to provide
information regarding the operation and capacity of the plant, the energy required for
its functioning and the concentration of the desired product. It is a cheaper way to
predict the results of any process before engaging in any large-scale production.
However, there must be sensible experimental data regarding the performance of the
units in the process so the simulation can be compelling and reliable. This
information usually comes from experimentation that has been done for a specific
unit or mathematical formulations based on physical principles or both (Esteghlalian
et al., 1997; Morales-Rodriguez et al., 2011; Phisalaphong et al., 2006).

Empirical models are obtained from experimental data usually at laboratory scale
since the conditions at which the experiments are carried out can be easily controlled
in order to get low margins of error. This also permits the reproduction of a specific
experiment as many times as required to guarantee reliability in the final results. In
the production of ethanol from biomass there are many mathematical models

60
available in literature and each one is derived from specific experiments for specific
raw materials.

The ethanol production process from lignocellulosic biomass considers two main
sections: a pretreatment section and a dehydration section as illustrated in Figure 3.1
(Bai et al., 2008; Balat et al., 2008; Karlsson et al., 2014; Wei et al., 2014). The
pretreatment section comprises the pretreatment of the raw material using a chemical
agent, the detoxification of the acid hydrolysates and the simultaneous
saccharification and co-fermentation (SSCF) of the available substrate for the
production of ethanol. For the pretreatment section, the kinetic models are derived
from experimental data specifically obtained from different lignocellulosic materials.

Similarly, the kinetic models for detoxification, enzymatic hydrolysis and

fermentation are derived from experimental data fitted into mathematical formulae.
These models, however, are restricted to the operating conditions at which their
parameters were obtained, meaning that if the parameters of a model were acquired
in a temperature range between 10 and 20 ºC, any results above or below this range
cannot be trusted. The versatility of these models can be observed in their respective
works showing an accurate description of the phenomena involved.

The dehydration section consists of the separation units where ethanol is recovered
and purified. The aim of this chapter is to present reliable mathematical models that
can accurately describe the different units of the ethanol process for both the
distillation column and the pervaporation module. The distillation column is
modelled using dynamic mass and energy balances considering aspects such as non-
ideality of the mixture for a VLE approach. The set of algebraic-differential
equations is solved using gPROMS and validated with simulation results using
ChemCAD.

The mathematical models of the pervaporation systems presented in this work are
also based on experiments which serve to calculate the molar flux through the
membrane. This work presents two models: one for an organophilic membrane that
separates volatile compounds on the permeate side and one model for a hydrophilic
membrane that removes water to increase the concentration of ethanol on the
61
retentate side. This work will list the models that will be used in the simulation and
optimisation of the ethanol production process from corn stover. These models will
be implemented in the formulation of the mass and energy balances for each unit.

3.2 Process description

Figure 3.1 presents a configuration for the ethanol process using corn stover as raw
material. This flowsheet has been proposed by other authors and considers SSCF as
the method to produce ethanol, which represents savings in investment of equipment
and the utilisation of services (Cardona and Sánchez, 2007; Grisales et al., 2008).
Additionally, this method prevents the inhibition by substrate in the enzymatic
hydrolysis and the inhibition by product in the fermentation (Morales-Rodriguez et
al., 2012, 2011).

The separation section consists of a distillation system at atmospheric pressure in

which the concentration of ethanol in the distillate stream will be close the azeotropic
point and a set of pervaporation modules where ethanol is expected to be dehydrated
up to concentrations higher than 99 % (w/w). The design of the pervaporation section
presented in Figure 3.1 was proposed by Marriott and Sørensen, (2003).

3.3 Dilute acid pretreatment

The chemical characterisation of corn stover is shown in Table 3.1 (Esteghlalian et
al., 1997; Rašković et al., 2010). The pretreatment stage consists of a reactor in
which the raw material gets in contact with a chemical agent (e.g. acid, base or
water) at high temperatures and pressures to degrade the lignin and to convert the
hemicellulose into xylose and by-products (e.g. furfural, hydroxyl methyl furfural
and organic acids) (Abedinifar et al., 2009; Drapcho et al., 2008; Mosier et al., 2005).

The pretreatment method of choice in this work is dilute-acid pretreatment. The

application of this pretreatment method for lignocellulosic materials has shown
promising results in industry in terms of lignin degradation, high hemicellulose
conversion and xylose production (Demirbas, 2007; Drapcho et al., 2008;
Esteghlalian et al., 1997; Mosier et al., 2005; Sreenath et al., 2001; Sreenath and
Jeffries, 2000). The reactions that are considered in the work of Esteghlalian et al.,

62
(1997) and take place in the pretreatment of the lignocellulosic biomass are the
following:

𝑘𝑓 𝑘𝑥
𝐻𝑒𝑚𝑖𝑐𝑒𝑙𝑙𝑢𝑙𝑜𝑠𝑒 (𝑓𝑎𝑠𝑡) → 𝑋𝑦𝑙𝑜𝑠𝑒 → 𝐵𝑦 − 𝑝𝑟𝑜𝑑𝑢𝑐𝑡𝑠

𝑘𝑠 𝑘𝑥
𝐻𝑒𝑚𝑖𝑐𝑒𝑙𝑙𝑢𝑙𝑜𝑠𝑒 (𝑠𝑙𝑜𝑤) → 𝑋𝑦𝑙𝑜𝑠𝑒 → 𝐵𝑦 − 𝑝𝑟𝑜𝑑𝑢𝑐𝑡𝑠

According to Esteghlalian et al., (1997), the percentage of fast hemicellulose in corn

stover is 64.4 % and the rest is slow-reacting hemicellulose. The reason to choose
this kinetic model is because it considers the effect of the concentration of sulphuric
acid in the system, the reactivity of the raw material (which noticeably varies from
on material to another (see Section 1.5)) and the effect of temperature in the reactor.
This model assumes that the raw material consists of two fractions of hemicellulose:
fast and slow. Fast hemicellulose refers to the part of the material that converts
completely during the pretreatment and in a high reaction rate.

Slow hemicellulose refers to the fraction of the material whose reaction rate is very
low due to the competitive nature of its counterpart. Equations 3.1 to 3.4 are the
expressions of the reaction rates for fast hemicellulose, slow hemicellulose, xylose
and by-products.

63
Figure 3.1: Flowsheet of the ethanol production process from corn stover

64
Table 3.1: Chemical characterisation of corn stover (dry basis % w/w) reported by Bhandari et
al., (1984); Čuček et al., (2011); Esteghlalian et al., (1997) and Liu and Chen, (2016)
Composition
Component
(% w/w)
Cellulose 36.0
Fast xylan 12.8
Slow xylan 7.00
Galactan 1.30
Arabinan 2.80
Lignin 19.7
Ash 7.20
Other 13.2

Equations 3.5 to 3.7 are Arrhenius-type equations for the reactions rate constants of
the previous set of equations as functions of the operating temperature. Equation 3.8
is the actual concentration of sulphuric acid used in the pretreatment of corn stover
(Esteghlalian et al., 1997). To summarise:

𝑓𝑎𝑠𝑡 𝑓𝑎𝑠𝑡
𝑟ℎ𝑒𝑚𝑖 = −𝑘𝑓 𝐶ℎ𝑒𝑚𝑖 Eq. 3.1

𝑓𝑎𝑠𝑡 𝑠𝑙𝑜𝑤
𝑟ℎ𝑒𝑚𝑖 = −𝑘𝑠 𝐶ℎ𝑒𝑚𝑖 Eq. 3.2

𝑓𝑎𝑠𝑡 𝑠𝑙𝑜𝑤
𝑟𝑥𝑦 = 𝑘𝑓 𝐶ℎ𝑒𝑚𝑖 + 𝑘𝑠 𝐶ℎ𝑒𝑚𝑖 − 𝑘𝑥 𝐶𝑋𝑦 Eq. 3.3

𝑟𝑏𝑝 = 𝑘𝑥 𝐶𝑋𝑦 Eq. 3.4

𝐸𝑓𝑜
𝑘𝑓 = 𝐴𝑓𝑜 𝐶𝑎1.5 𝑒𝑥𝑝 (− ) Eq. 3.5
𝑅𝑇

𝐸𝑠𝑜
𝑘𝑠 = 𝐴𝑠𝑜 𝐶𝑎1.6 𝑒𝑥𝑝 (− ) Eq. 3.6
𝑅𝑇

65
 𝐸𝑥𝑜
𝑘𝑥 = 𝐴2𝑜 𝐶𝑎0.5 𝑒𝑥𝑝 (− ) Eq. 3.7
𝑅𝑇

𝐶𝑎 = 𝐶 − (0.1 𝑁𝑒𝑢𝑡𝑟𝑎𝑙𝑖𝑧𝑖𝑛𝑔 𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝑜𝑓 𝑡ℎ𝑒 𝑟𝑎𝑤 𝑚𝑎𝑡𝑒𝑟𝑖𝑎𝑙) Eq. 3.8

where:

Afo : Pre-exponential factor for fast hemicellulose (6.7x1016 min-1)

Aso : Pre-exponential factor for slow hemicellulose (6.9x1019 min-1)

Axo : Pre-exponential factor for xylose (3.7x1010 min-1)

𝐶 : Real concentration of sulphuric acid in the reactor (mass fraction)

𝐶𝑎 : Actual concentration of sulphuric acid (mass fraction)

𝐶𝑓𝑢𝑟 : Concentration of furfural produced during pretreatment (g/L)

𝑓𝑎𝑠𝑡
𝐶ℎ𝑒𝑚𝑖 : Concentration of fast hemicellulose (g/L)

𝑠𝑙𝑜𝑤
𝐶ℎ𝑒𝑚𝑖 : Concentration of slow hemicellulose (g/L)

𝐶𝑋𝑦 : Concentration of xylose (g/L)

Efo : Activation energy for fast hemicellulose conversion (129.8 kJ/mole)

Eso : Activation energy for slow hemicellulose conversion (167.6 kJ/mole)

Exo : Activation energy for xylose degradation (98.4 kJ/mole)

𝑘𝑓 : Reaction rate constant for the conversion of fast hemicellulose (min-1)

𝑘𝑠 : Reaction rate constant for the conversion of slow hemicellulose (min-1)

𝑘𝑥 : Reaction rate constant for the conversion of xylose (min-1)

𝑓𝑎𝑠𝑡
𝑟ℎ𝑒𝑚𝑖 : Reaction rate of fast hemicellulose (g L-1 min-1)

𝑠𝑙𝑜𝑤
𝑟ℎ𝑒𝑚𝑖 : Reaction rate of slow hemicellulose (g L-1 min-1)

66
𝑟𝑥𝑦 : Reaction rate of xylose (g L-1 min-1)

𝑟𝑏𝑝 : Reaction rate of by-products produced during pretreatment (g L-1 min-1)

Although the kinetic model for the pretreatment using dilute acid considers several
operating conditions involved in the reaction, it does not consider the formation of
other by-products and other reducing sugars.

3.4 Evaporation
The outlet stream from the pretreatment reactor is filtered and divided into two
streams: one liquid fraction (rich in xylose) and one solid fraction (rich in unreacted
cellulose). The liquid fraction is evaporated with the main purpose of increasing the
concentration of sugar and to remove some volatile components that can inhibit the
performance of the fermenting strains (Binod et al., 2010; Kiran and Jana, 2015;
Petersson et al., 2007; Wei et al., 2014, 2014). The mathematical model for the
evaporation is formulated from the mass and energy balances which are unfolded in
Appendix A.

3.5 Detoxification
The concentrated solution leaving the evaporation unit is treated with Ca(OH)2 in a
process known as detoxification. This process is carried out, at constant temperature
and pH, to reduce the concentration of furfural and other by-products, which have a
high inhibitory effect on the performance of the fermenting microorganisms, by
converting them into insoluble salts (Abedinifar et al., 2009; Bai et al., 2008;
Ballesteros et al., 2004; Purwadi et al., 2004). The kinetic model of this unit and the
mass balance are taken from Purwadi et al., (2004) and is based on a series of
reactions that are assumed to take place during the detoxification of the hydrolysates
and are represented as follows:

𝑍+𝐴↔ {𝑍𝐴}

{𝑍𝐴} → 𝑃+𝑍

67
where:

Z : Concentration of Ca2+ cation (g/L)

A : Concentration of furfural (g/L)

P : Concentration of insoluble salts (g/L)

{𝑍𝐴} : Concentration of Ca complex (g/L)

The reaction scheme above suggests the formation of a furan – Ca+2 complex that
rapidly disassociates into insoluble salts and Ca+2 ions. In other words, Ca(OH)2
serves as a base in a neutralization reaction in which the acid hydrolysate is
converted into salts. The mathematical model proposed by Purwadi et al., (2004) is
an empirical model which was formulated taking into account pH and operating
temperature. Unfortunately, this is the only model available in literature since current
research only focuses in finding new experimental techniques for the detoxification
of the lignocellulosic hydrolysates (see Section 2.3).

The limitations of this model in terms of temperature dependence will have a direct
effect on the optimisation of the overall process since it will restrain the evaluation of
different operating conditions. The equations of this model are presented as follows
and the mass and energy balances are described in more detail in Appendix A. See
Purwadi et al. (2004) for the values of the parameters of the model at pH values.

𝑟𝑓𝑢𝑟 = −𝑘1 𝐶𝑓𝑢𝑟 (𝐶𝑍,0 − 𝐶𝑍𝐴 ) + 𝑘2 𝐶𝑍𝐴 Eq. 3.9

𝑟𝑍𝐴 = −(𝑘2 + 𝑘3 ) 𝐶𝑍𝐴 + 𝑘1 𝐶𝑓𝑢𝑟 (𝐶𝑍,0 − 𝐶𝑍𝐴 ) Eq. 3.10

𝑟𝑃 = −𝑘3 𝐶𝑍𝐴 Eq. 3.11

The pH level is related to the initial amount of cation 𝐶𝑍,0 as shown in Eq. 3.12:

68
 𝑁
𝐶𝑍,𝑜 = 𝑘𝑍 10−(14−𝑝𝐻) + 𝐶𝑍,0 Eq. 3.12

where:

𝐶𝑓𝑢𝑟 : Concentration of furfural (g/L)

𝐶𝑃 : Concentration of insoluble salts (g/L)

𝐶𝑍𝐴 : Concentration of Ca complex (g/L)

𝑁
𝐶𝑍,0 : Concentration of Ca(OH)2 to neutralise hydrolysates from acid state (g/L)

𝑘𝑖 : Reaction rate constants (min-1)

𝑟𝑓𝑢𝑟 : Reaction rate of furfural (g L-1 min-1)

𝑟𝑃 : Reaction rate of insoluble salts (g L-1 min-1)

𝑟𝑍𝐴 : Reaction rate of Ca+2 complex (g L-1 min-1)

3.6 Simultaneous saccharification and co-fermentation (SSCF)

The stream leaving the detoxification stage is mixed with the solid fraction obtained
during the pretreatment which will be later used as substrate in the production of
ethanol. The simultaneous saccharification and co-fermentation (SSCF) stage
comprises two operations: Enzymatic hydrolysis of cellulose and co-fermentation of
both pentoses and hexoses. Both operations take place simultaneously in order to
reduce the size of the plant and to further the production of ethanol by reducing the
inhibitory effects over the activity of the enzymes and the metabolism of the
fermenting strains (Kadam et al., 2004; Kadar et al., 2004; Leksawasdi et al., 2001;
Morales-Rodriguez et al., 2011).

3.6.1 Enzymatic hydrolysis

The first operation is the enzymatic hydrolysis or saccharification where the
unreacted cellulose from the pretreatment stage is degraded into cellobiose and
glucose using cellulases (Bezerra and Dias, 2005; Kadam et al., 2004; Öhgren et al.,

69
2007b; Saha et al., 2005). The set of reactions for the enzymatic saccharification are
listed below.

𝑘1
2𝐶𝑒𝑙𝑙𝑢𝑙𝑜𝑠𝑒 + 𝐻2 𝑂 → 𝐶𝑒𝑙𝑙𝑜𝑏𝑖𝑜𝑠𝑒
𝑘2
𝐶𝑒𝑙𝑙𝑢𝑙𝑜𝑠𝑒 + 𝐻2 𝑂 → 𝐺𝑙𝑢𝑐𝑜𝑠𝑒
𝑘3
𝐶𝑒𝑙𝑙𝑜𝑏𝑖𝑜𝑠𝑒 + 𝐻2 𝑂 → 2𝐺𝑙𝑢𝑐𝑜𝑠𝑒
𝑘4
𝐻𝑒𝑚𝑖𝑐𝑒𝑙𝑙𝑢𝑙𝑜𝑠𝑒 + 𝐻2 𝑂 → 𝑋𝑦𝑙𝑜𝑠𝑒
𝑘5
𝑋𝑦𝑙𝑜𝑠𝑒 → 𝐹𝑢𝑟𝑓𝑢𝑟𝑎𝑙 +3𝐻2 𝑂

The kinetic model presented by Kadam et al., (2004) for the saccharification of
cellulose using cellulase includes a term for absorption/desorption reaction Kiad and a
term for the total concentration of enzymes as shown in Equations 3.13 and 3.14.

𝐸𝑖𝑚𝑎𝑥 𝐾𝑖𝑎𝑑 𝐸𝑖𝐹 𝐶𝑐𝑒𝑙𝑙

𝐸𝑖𝐵 = Eq. 3.13
1 + 𝐾𝑖𝑎𝑑 ∙ 𝐸𝑖𝐹

𝐸𝑖𝑇 = 𝐸𝑖𝐹 + 𝐸𝑖𝐵 Eq. 3.14

i =1 for cellulase; i =2 for β-glucosidase

The reaction rates for cellulose, glucose and cellobiose are shown next. These
models present the concentration of enzymes and a constant Kinh which represents the
inhibition effect of the production of glucose, cellobiose and xylose over the activity
of the enzymes. The reaction rates for each component are shown as follows:

𝑘1 𝐸1𝐵 𝑅𝑠 𝐶𝑐𝑒𝑙𝑙
𝑟𝑐𝑒𝑙𝑙 =
𝐶 𝐶 𝐶𝑋𝑦 Eq. 3.15
1 + 𝐾 𝐶𝑏 + 𝐾 𝐺𝑙 + 𝐾
1.𝑖𝑛ℎ.𝐶𝑏 1.𝑖𝑛ℎ.𝐶𝑏 1.𝑖𝑛ℎ.𝐶𝑏

𝑘2 (𝐸1𝐵 + 𝐸2𝐵 ) 𝑅𝑠 𝐶𝑐𝑒𝑙𝑙

𝑟𝐺𝑙 =
𝐶 𝐶 𝐶𝑋𝑦 Eq. 3.16
1 + 𝐾 𝐶𝑏 + 𝐾 𝐺𝑙 + 𝐾
2.𝑖𝑛ℎ.𝐶𝑏 2.𝑖𝑛ℎ.𝐶𝑏 2.𝑖𝑛ℎ.𝐶𝑏

70
 𝑘3 𝐸2𝐹 𝑅𝑠 𝐶𝐶𝑏
𝑟𝐶𝑏 =
𝐶 𝐶𝑋𝑦 Eq. 3.17
𝐾3𝑀 (1 + 𝐾 𝐺𝑙 + 𝐾 ) + 𝐶𝐶𝑏
3.𝑖𝑛ℎ.𝐺𝑙 3.𝑖𝑛ℎ.𝑋𝑦

A variation for hemicellulose conversion into xylose and further degradation into
furfural during the enzymatic hydrolysis has been proposed by Ballesteros et al.
(2004).

𝑟ℎ𝑒𝑚𝑖 = 𝑘4 𝐶ℎ𝑒𝑚𝑖 Eq. 3.18

𝑟𝑋𝑦 = 𝑘5 𝐶𝑋𝑦 Eq. 3.19

3.6.2 Co-fermentation
The second operation is the co-fermentation of reducing sugars (e.g. xylose, glucose,
cellobiose, etc.) into ethanol and other by-products (i.e. acetic acid, CO2, cells, etc.)
via anaerobic metabolism. The kinetic model presented by Leksawasdi et al., (2001)
for recombinant bacterium Zymomonas mobilis ZM4(pZB5) includes the inhibition
effect caused by the substrate, the main product and toxic by-products such as
furfural. This mathematical approach is based on the Monod cell growth model. The
mass and energy balances of the combined processes are summarized in Appendix A.
The kinetic model for co-fermentation using bacteria is the following:

𝐶𝐺𝑙 𝐶𝐸𝑡ℎ −𝐶𝐸𝑡ℎ,𝑖𝑥,𝐺𝑙 𝐾𝑖𝑥,𝐺𝑙

𝑟𝑋,𝐺𝑙 = 𝜇𝑚𝑎𝑥,𝐺𝑙 (𝐾 ) (1 − 𝐶 ) (𝐾 ) Eq. 3.20
𝑠𝑥,𝐺𝑙 +𝐶𝐺𝑙 𝐸𝑡ℎ,𝑚𝑥,𝐺𝑙 −𝐶𝐸𝑡ℎ,𝑖𝑥,𝐺𝑙 𝑖𝑥,𝐺𝑙 +𝐶𝐺𝑙

𝐶𝑋𝑦 𝐶𝐸𝑡ℎ −𝐶𝐸𝑡ℎ,𝑖𝑥,𝑋𝑦 𝐾𝑖𝑥,𝑋𝑦

𝑟𝑋,𝑋𝑦 = 𝜇𝑚𝑎𝑥,𝑋𝑦 (𝐾 ) (1 − 𝐶 ) (𝐾 ) Eq. 3.21
𝑠𝑥,𝑋𝑦 +𝐶𝑋𝑦 𝐸𝑡ℎ,𝑚𝑥,𝑋𝑦 −𝐶𝐸𝑡ℎ,𝑖𝑥,𝑋𝑦 𝑖𝑥,𝑋𝑦 +𝐶𝑋𝑦

𝑟𝑋 = [𝛼 𝑟𝑋,𝐺𝑙 + (1 − 𝛼) 𝑟𝑋,𝑋𝑦 ] 𝐶𝑋 Eq. 3.22

𝐶𝐺𝑙 𝐶𝐸𝑡ℎ −𝐶𝐸𝑡ℎ,𝑖𝑠,𝐺𝑙 𝐾𝑖𝑠,𝐺𝑙

𝑟𝐺𝑙 = 𝛼 𝑞𝑠,𝑚𝑎𝑥,𝐺𝑙 (𝐾 ) (1 − 𝐶 ) (𝐾 ) 𝐶𝑋 Eq. 3.23
𝑠𝑠,𝐺𝑙 +𝐶𝐺𝑙 𝐸𝑡ℎ,𝑚𝑠,𝐺𝑙 −𝐶𝐸𝑡ℎ,𝑖𝑠,𝐺𝑙 𝑖𝑠,𝐺𝑙 +𝐶𝐺𝑙

71
 𝐶𝐺𝑙 𝐶𝐸𝑡ℎ −𝐶𝐸𝑡ℎ,𝑖𝑝,𝐺𝑙 𝐾𝑖𝑝,𝐺𝑙
𝑟𝑃,𝐺𝑙 = 𝑞𝑚𝑎𝑥,𝐺𝑙 (𝐾 ) (1 − 𝐶 ) (𝐾 ) Eq. 3.24
𝑠𝑝,𝐺𝑙 +𝐶𝐺𝑙 𝐸𝑡ℎ,𝑚𝑝,𝐺𝑙 −𝐶𝐸𝑡ℎ,𝑖𝑝,𝐺𝑙 𝑖𝑝,𝐺𝑙 +𝐶𝐺𝑙

𝐶𝑋𝑦 𝐶𝐸𝑡ℎ −𝐶𝐸𝑡ℎ,𝑖𝑝,𝑋𝑦 𝐾𝑖𝑝,𝑋𝑦

𝑟𝑃,𝑋𝑦 = 𝑞𝑚𝑎𝑥,𝑋𝑦 (𝐾 ) (1 − 𝐶 ) (𝐾 ) Eq. 3.25
𝑠𝑝,𝑋𝑦 +𝐶𝑋𝑦 𝐸𝑡ℎ,𝑚𝑝,𝑋𝑦 −𝐶𝐸𝑡ℎ,𝑖𝑝𝑋𝑦 𝑖𝑝,𝑋𝑦 +𝐶𝑋𝑦

𝑟𝑃 = [𝛼 𝑟𝑃,𝐺𝑙 + (1 − 𝛼) 𝑟𝑃,𝑋𝑦 ] 𝐶𝑋 Eq. 3.26

where:

𝐶𝑐𝑒𝑙𝑙 : Concentration of cellulose (g/L)

𝐶𝐶𝑏 : Concentration of cellobiose (g/L)

𝐶𝐸𝑡ℎ : Concentration of ethanol (g/L)

𝐶𝐸𝑡ℎ,𝑖𝑝,𝐺𝑙 : Threshold inhibition concentration of ethanol from glucose (g/L)

𝐶𝐸𝑡ℎ,𝑖𝑠,𝐺𝑙 : Threshold conc. of ethanol from glucose that affects glucose uptake
(g/L)

𝐶𝐸𝑡ℎ,𝑖𝑥,𝐺𝑙 : Threshold conc. of ethanol from glucose that causes inhibition (g/L)

𝐶𝐸𝑡ℎ,𝑖𝑝,𝑋𝑦 : Threshold inhibition concentration of ethanol from xylose (g/L)

𝐶𝐸𝑡ℎ,𝑖𝑠,𝑋𝑦 : Threshold conc. of ethanol from xylose that affects xylose uptake
(g/L)

𝐶𝐸𝑡ℎ,𝑖𝑥,𝑋𝑦 : Threshold conc. of ethanol from xylose that causes inhibition (g/L)

𝐶𝐸𝑡ℎ,𝑚𝑝,𝐺𝑙 : Maximum inhibition concentration of ethanol from glucose (g/L)

𝐶𝐸𝑡ℎ,𝑚𝑝,𝑋𝑦 : Maximum inhibition concentration of ethanol from xylose (g/L)

𝐶𝐸𝑡ℎ,𝑚𝑠,𝐺𝑙 : Maximum conc. of ethanol from glucose that affects glucose uptake
(g/L)

72
 𝐶𝐸𝑡ℎ,𝑚𝑠,𝑋𝑦 : Maximum conc. of ethanol from glucose that affects xylose uptake
(g/L)

𝐶𝐸𝑡ℎ,𝑚𝑥,𝐺𝑙 : Maximum conc. of ethanol from glucose that causes inhibition (g/L)

𝐶𝐸𝑡ℎ,𝑚𝑥,𝑋𝑦 : Maximum conc. of ethanol from xylose that causes inhibition (g/L)

𝐶𝐺𝑙 : Concentration of glucose (g/L)

𝐶ℎ𝑒𝑚𝑖 : Concentration of hemicellulose (g/L)

𝐶𝑋 : Concentration of cells (g/L)

𝐶𝑋𝑦 : Concentration of xylose (g/L)

EiB : Bound concentration of enzymes on the substrate (g/L)

EiF : Concentration of free enzymes in solution (g/L)

EiT : Total enzyme concentration (g/L)

Eimax : Maximum mass of active enzyme/substrate (kg of protein/kg of

substrate)

Kiad : Dissociation const. for absorption/desorption reaction i (i= reactions

1, 2, 3 m3/kg of protein)

ri : Reaction rates for equation 16 – 28 (g L-1 min-1)

Rs : Substrate reactivity

Kj:inh:k : Inhibition const. for the reactions j = 1, 2, 3. k = cellobiose, glucose,

xylose (g/L)

Equations 3.20 to 3.26 are the reaction rates for the consumption of xylose and
glucose as well as the production of ethanol and cells during the co-fermentation
stage. These equations are governed by two important aspects. The first one is the
inhibition effect by both substrate and product which can be identified as Kis and Kip,

73
respectively. The second aspect is the maximum concentration of ethanol that can be
produced without inhibiting the strain.

This is a detailed model that covers several aspects of the fermentation of reducing
sugars and therefore will be used in the overall simulation of the ethanol production
process (Morales-Rodriguez et al., 2011). The kinetic models for both enzymatic
hydrolysis and co-fermentation are limited by the number of components that the
authors investigated in their respective works. This issue is addressed later in this
thesis in order to determine how these limitations affect the simulation and
optimisation of the overall process.

3.7 Separation
Once the SSCF operation stops, the content of the reactor is filtered in order to
remove cells and other solids. The liquid, consisting mainly of ethanol and water, is
sent to a separation section where ethanol is purified to biofuel standard
concentrations (>99% w/w.). The separation stage comprises two parts as described
next.

3.7.1 Distillation
The first part of the separation stage is the removal of ethanol from the fermentation
broth via distillation. A binary mixture of ethanol and water is recovered and purified
to its azeotropic point. Two distillation columns are often used in order to reach this
separation (Kiss and Olujic, 2014; Knapp and Doherty, 1992; Lutze and Gorak,
2013). The mass and energy balances and the assumptions used in this work are
summarised in Appendix A.

3.7.2 Membrane-based operation

The second stage is the dehydration of the binary azeotropic mixture via membrane
separation. This methodology is able to reach concentrations of ethanol of 99 % w/w
or even higher. The membrane separation stage comprises different modules in
parallel and/or in-series to increase the purity of ethanol with low energy
consumption (Koch et al., 2013; Lutze and Gorak, 2013; Roth et al., 2013;

74
Valentinyi et al., 2013). The model used in this work to simulate the pervaporation
module is the one introduced by Tsuyumoto et al., (1997). The mass and energy
balances for the fibre side and the shell side are shown in Appendix A. Eq. 3.27
presents the partial molar flux of ethanol as a function of the pressure gradient
between feed and permeate and also the mole composition in the feed stream.

𝐽𝐸𝑡ℎ𝑎𝑛𝑜𝑙 = 1.72𝑥10−10 𝑥𝐸𝑡ℎ𝑎𝑛𝑜𝑙 (𝑃𝑓𝑒𝑒𝑑 − 𝑃𝑝𝑒𝑟𝑚𝑒𝑎𝑡𝑒 ) Eq. 3.27

Eq. 3.28 is the partial molar flux for water permeating through the membrane. This is
a more complex formulation as it includes the activity between the vapour phase and
the liquid phase. This equation is a function of feed temperature, permeate pressure,
mole fraction of water in feed, mole fraction of water in permeate, the thickness of
the membrane. Tsuyumoto et al., (1997) and (1995) obtained the parameters of this
model for permeate pressures between 10 and 10,000 Pa, temperatures between 40
and 70 ºC and concentrations of ethanol in feed between 99 and 99.5 % (w/w).

11500
8.086𝑥106 𝑒𝑥𝑝 (− ) 𝑃𝑝𝑒𝑟𝑚𝑒𝑎𝑡𝑒
𝐽𝑊𝑎𝑡𝑒𝑟 = 𝑇 (𝛾𝑓𝑒𝑒𝑑 𝑥𝑤𝑎𝑡𝑒𝑟 − 𝑌𝑤𝑎𝑡𝑒𝑟 )
𝑠𝑎𝑡
𝛿𝑚 𝑃𝑊𝑎𝑡𝑒𝑟
3390
3.441𝑥10−3 𝑒𝑥𝑝 (− )
+ 𝑇 ((𝛾𝑓𝑒𝑒𝑑 ∙ 𝑥𝑤𝑎𝑡𝑒𝑟 )
2 Eq. 3.28
2𝛿𝑚
2
𝑃𝑝𝑒𝑟𝑚𝑒𝑎𝑡𝑒
−( 𝑠𝑎𝑡 𝑦𝑤𝑎𝑡𝑒𝑟 ) )
𝑃𝑊𝑎𝑡𝑒𝑟

3.8 Results
The mass and energy balances and other equations used in this work (and outlined
and Appendix A) are solved using gPROMS (Process System Enterprises, 2015). In
order to solve the models and validate their formulation, a revision of the algorithms
and solvers used in gPROMS should be introduced. gPROMS presents two
mathematical solvers for the simultaneous solution of differential and algebraic (DA)
systems: DASOLV and SRADAU. Both are very effective in the solution of DA
systems but they differ from each other in the method of integration. DASOLV uses
Backward Finite Differences (BFD) whereas SRADAU uses a 4th order Runge-Kutta
method. The stability and effectiveness of these models depends on discontinuities
75
and oscillations in the system. In this work, DASOLV is used since computational
times with this method are substantially shorter than with SRADAU (5 sec shorter)
(Process System Enterprises, 2015). Both solvers automatically adjust the integration
step so the criterion shown in Eq. 3.29 can be met. The difference in CPU times is

𝑛𝑑
2
1 ∈𝑖
√ ∑( ) Eq. 3.29
𝑛𝑑 𝑎 + 𝑟|𝑥𝑖 |
𝑖=1

where:

𝑛𝑑 : The number of differential variables in the problem

I : The solver's estimate for the local error in the ith differential variable
xi : The current value the ith differential variable
a : Absolute error tolerance
r : Relative error tolerance

Each model is validated against experimental results (in the case of the reactors and
the pervaporation modules) and ChemCAD (for the distillation columns).

3.9 Pretreatment stage

The pretreatment stage is validated against the experimental results obtained by
Esteghlalian et al., (1997) and this is shown in Figure 3.2. The results in this figure
initially suggest good concordance for the percentage of xylose remaining in the raw
material between the experimental data and the mathematical model. Figure 3.2 also
shows the effect of temperature in the pretreatment as the conversion of
hemicellulose. Higher temperatures not only increase the conversion of
hemicellulose but also reduce the reaction time as hemicellulose degrades more
quickly than as seen in the curves of 140 ºC and 180 ºC for a concentration of
sulphuric acid of 0.6 % (w/w). Conversely, the concentration of sulphuric acid in the
reaction also contributes to the fast degradation of hemicellulose into xylose. Figure
3.2 indicates a reduction of the amount of xylose remaining in the material as the

76
concentration of sulphuric acid doubles up. According to Esteghlalian et al., (1997),
yields of xylose above 80 % from corn stover can be reached at temperatures of 180
ºC and concentrations of sulphuric acid of 1.2 % (w/w).

3.10 Detoxification stage or Overliming

The overliming model proposed by Purwadi et al., (2004) has been compared with
experimental results at 30 oC for different pH levels. As seen in Figure 3.3, the
concentration of furfural in the hydrolysates decreases as pH increases, suggesting a
faster neutralisation of the acid hydrolysates. The experimental data obtained by
Purwadi et al. (2004), are close to the trend predicted by the mathematical model
which justifies the implementation of this model into the main process. However,
this model is restricted to the operating configurations at which the parameters were
obtained. The parameters for the detoxification model are available in the work
Purwadi and collaborators at an operating temperature of 30 oC.

3.11 Enzymatic hydrolysis stage

During the enzymatic hydrolysis of cellulose, cellulases cut the chain of the
biopolymer to produce monomeric units known as cellulose and cellobiose. For the
formulation of the model and therefore the estimation of its parameters, it is required
to measure the concentration of most of the species involved in the reaction.

In the case of the work presented by Kadam et al., (2004), the only components that
could be measured in laboratory were glucose and cellobiose. These sugars are
usually quantified in laboratory using instrumental analysis or colourimetry
(Najafpour and Shan, 2003; Saini et al., 2013; Xiao et al., 2004).

77
 100

(%) Xylose remaining in raw material

10

140 oC, 0.6 % wt acid

140 oC, 1.2 % wt acid
180 oC, 0.6 % wt acid
1
0 10 20 30 40 50 60
Time (min)

Figure 3.2: Xylose remaining in the solid fraction in the hydrolysis of hemicellulose. The left side
represents the results obtained with gPROMS. The right side is the experimental data presented
by Esteghlalian et al., (1997)

0.40 pH = 9
pH = 10
pH = 11
0.35
Concentration furfural (g/L)

pH = 12

0.30

0.25

0.20

0.15

0.10

0.05

0 10 20 30 40 50 60 70 80 90
Time (min)

Figure 3.3: Model validation for the detoxification of acid hydrolysates. The left side shows the
results using gPROMS. The right side shows the experimental data obtained by Purwadi et al.,
(2004)

78
 60 30

50 25

Concentration cellobiose (g/L)

Concentration glucose (g/L)
40 20
Glucose (left)
Cellibiose (right)
30 15

20 10

10 5

0 0
0 24 48 72 96 120 144 168
Time (hour)

Figure 3.4: Validation of the kinetic model for enzymatic saccharification. The left side presents
the results using gPROMS. The right side presents the experimental data presented by Kadam
et al., (2004)

60 6
Concentration glucose, xylose, ethanol (g/L)

50 5
Concentration cells (g/L)

Glucose
40 4
Xylose
Ethanol
30 Cells 3

20 2

10 1

0 0
0 5 10 15 20 25 30 35 40 45
Time (hour)

Figure 3.5: Validation of the fermentation model proposed by Leksawasdi et al., (2001). The left
side presents the results using gPROMS. The right side presents the experimental data

The experimental results in Figure 3.4 were obtained using 10% w/w corn stover
solids with no added background sugars at 45 °C and, as seen in Figure 3.4, the
production of glucose and the production and subsequent conversion of cellobiose in
the enzymatic hydrolysis take place in a period of 168 h. The kinetic model proposed
by Kadam et al., 2004 is chosen in this work because it considers inhibition by
xylose, a major sugar in hemicellulose-derived hydrolysates. Other models not only
neglect the effect of the concetration of xylose in the production of glucose but also
do not consider the benefits of supplementing the reaction medium with β-
glucosidase for higher yields of glucose.
79
 1.0
0.9
0.8

Liquid mass fraction

0.7
0.6 Ethanol gPROMS
Water gPROMS
0.5
Ethanol ChemCAD
0.4 Water ChemCAD
0.3
0.2
0.1
0.0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Number of trays
Figure 3.6: Validation profile for liquid weight fraction of binary mixture ethanol/water using
gPROMS and ChemCAD

Table 3.2: Estimated ethanol purity for an ethanol feed concentration of 94% (w/w)
Ethanol concentration (% w/w)
Feed flow
Marriott and Sorensen, (2003);
rate
Tsuyumoto et al., (1997) Marriott et al., (2001) This work
(kg/h)
(3D model)
97.20 97.18
44.9 97.30
(Experimental data) (Calculated)

3.12 Co-fermentation stage

Figure 3.5 presents the validation of the co-fermentation model proposed by
Leksawasdi et al., (2001). This model describes the fermentation of both xylose and
glucose into ethanol using a recombinant bacterium known as Zymomonas mobilis
ZM4 (pZB5). The results of the kinetic model show agreement with the experimental
data obtained by Leksawasdi et al., (2001). It is also clear that by using this
microorganism, a higher conversion of both xylose and glucose and lower production
of cells can be expected in comparison to conventional strains often used in industry.

80
 3.13 Distillation stage
The distillation column system has also been validated. The results of the dynamic
model for a distillation column are compared against ChemCAD when the column
has reached steady-state. For the simulation, a flow rate of 1.18 kg/sec (or 50
mol/sec) of a solution 20 % mole of ethanol is fed into a distillation column
consisting of 15 trays, one total condenser and one reboiler. The number of stages is
determined with the Fenske-Underwood-Gilliland (FUG) approach using
ChemCAD’s shortcut model. The activity model used in both simulations is NRTL
and the Equation of State is PSRK for the evaluation of thermodynamic properties.
As seen in Figure 3.6, the results using the MESH equations in gPROMS for the
distillation column follow a similar trend as the profiles obtained using ChemCAD.

3.14 Pervaporation stage

Table 3.2 shows the validation of the model for the pervaporation module against the
experimental results obtained by Tsuyumoto et al., (1997). It also shows the results
reported by Marriott and Sorensen which consider radial and axial distribution of
ethanol and water inside the membrane module as well as the operation time. The
concentration of ethanol in the retentate obtained in this work considers a plug-flow
approach inside the membrane module and is very similar to the results obtained with
the 3D approach (a difference of 0.12 %). This suggests that a more robust
formulation does not necessarily increase accuracy or even required and often incurs
in longer computational times due to the complexity that entails solving partial
differential equations. Although radial and axial distributions are important for a
thorough design of the module and the performance of the membrane, this work will
not consider this approach since the main objective is to integrate all the sources of
energy in the process for a more efficient and profitable production of ethanol.

3.15 Conclusions
This chapter presented the different models that describe the different units involved
in the production of ethanol from corn stover. The empirical models used for the
pretreatment, detoxification, enzymatic hydrolysis and fermentation were validated
against experimental data used for their own formulation which are available in the

81
respective publications. These comparisons were made in order to guarantee an
accurate depiction of the reactions that take place in the production of ethanol.
However, it is important to notice that these models are restricted to a range of
operating conditions and this can limit the opportunities to explore different designs
and arrangements of the process with higher efficiencies. The system of equations
that described the distillation process was developed from dynamic mass and energy
balances and the results of the simulation were compared to licensed software giving
similar trends in the profiles of the mass fractions along the column. The
mathematical formulation of the distillation column using gPROMS considered the
non-ideality of the mixture ethanol/water as well as the implementation of equations
of state to predict thermodynamic properties.

The model for the pervaporation using a hydrophilic membrane proposed by

Tsuyumoto et al., (1997) has been used in several works since it considers the effect
of the different operating conditions in the system as well as the concentration of
ethanol in the feed. Some of these works have considered a more complex
formulation of the distribution of the species inside the membrane module. However,
the results obtained in this work have shown that a more simplistic focus on the
simulation of the membrane was as effective as a more rigorous and detailed
approach. In other words, a plug-flow formulation gave similar results in the
concentration of ethanol to the 3D model. The following chapters will present the
incorporation of all these models into an overall simulation of the ethanol production
process leading to the heat integration thereof.

Chapter 4 presents the development of a mathematical model based on

experimentations of an organophilic membrane for the removal of ethanol from
aqueous solutions. This model will be used in the simulation of the complete process
in order to determine the viability in terms of separation rates, energy consumption
and total annualised cost.

82
 Chapter 4 – Ethanol recovery from aqueous
solutions using a PERVAPTM 4060 organophilic
membrane

Abstract

Pervaporation is one of the most investigated and most applied membrane-based

methods for ethanol dehydration in industry as it generally is less energy intensive
than distillation. This work presents the formulation of a model that can describe the
performance of a PERVAPTM 4060 organophilic membrane used for the separation
of ethanol from a fermentation broth in order to reduce the energy consumption of
the separation stages in the ethanol production process.

4.1 Introduction
The literature review presented in Chapter 2 showed a variety of works using
different techniques to recover ethanol from aqueous solutions. Most of these works
included membrane-assisted methods to overcome the azeotrope and reduce energy
consumption (Cho et al., 2010; Golias et al., 2002; Sánchez and Cardona, 2012;
Siqueira et al., 2008).

Alternative technologies for fluid separations include membrane-based operations

such as pervaporation, reverse osmosis and vapour permeation (Gaykawad et al.,
2013, 2012; Wei et al., 2014). Membrane-based operations can be implemented as
stand-alone processes or in combination with other units (i.e. hybrid separation
processes) to complete a separation task (Kreis and Górak, 2006). Currently,
membrane separations are seldom considered as stand-alone operations since they
have a low throughput and require larger membrane areas, resulting in high
investment and operating costs. If large streams or high purities are required,
membrane separations are best applied within hybrid processes which are membrane
83
systems coupled with thermal and/or extractive separation technologies (e.g.
distillation, absorption, esterification reactors, etc.) (Lutze and Gorak, 2013).

Some of the benefits of using hybrid processes (e.g. membrane-assisted distillation

processes) include: large flexibility (the system can deal with a wide range of
operating conditions, feed concentrations, capacity and purity requirements),
improvement in energy efficiency (the separation stages can achieve high separation
rates with low energy inputs) and the reduction of the size of some of the units in the
process (Koch et al., 2013; Lutze and Gorak, 2013; Valentinyi et al., 2013).

This work will considered a system consisting of a distillation column coupled with a
membrane-based operation to remove ethanol from the fermentation broth.
Pervaporation is one of the most potentially interesting types of membrane-assisted
separation methods used for solvent dehydration, organic solvents removal, breaking
azeotropes, etc. Pervaporation consists of two steps: (a) permeation of one or more of
the components of the feed through the membrane due to the gradient in the chemical
potential because of pressure and concentration differences between the permeate
and the retentate side and (b) evaporation into the vapour phase on the permeation
side of the membrane (Yakovlev et al., 2013).

Configurations for the dehydration of ethanol often include a distillation system

followed by a pervaporation network with hydrophilic membranes (Baker, 2012;
Roth et al., 2013). The distillate stream of the distillation system leaves the column
with a concentration close to the azeotrope and is passed through a hydrophilic
membrane network where water is removed in the permeate side and ethanol reaches
concentrations higher than 99 % (Marriott and Sorensen, 2003; Nagasawa et al.,
2016; Tsuyumoto et al., 1997).

This chapter proposes the inclusion of a pervaporation module into the separation
stage using a selective membrane to remove ethanol from the fermentation broth
through the permeate side. The pervaporation module will be located between the

84
fermentation stage and the distillation column and is expected to operate at the same
temperatures as the fermentation tank.

A polymeric membrane PERVAPTM 4060 will be considered in this work. This

particular membrane removes volatile organic components through the permeate side
instead of removing water as in the case of hydrophilic membranes. The PERVAP TM
4060 organophilic membrane has been studied for the removal of butanol,
benzaldehyde, ethanol, heptane, 1-octen-3-ol from aqueous solutions (Ben Soltane et
al., 2013; Claes et al., 2010; Martinez et al., 2013).

In order to assess the effectiveness of the implementation of the PERVAPTM 4060

organophilic membrane into the ethanol production a mathematical model of the
system ethanol/water, based on experimentation, is required. However, there are no
publications regarding the formulation of an empirical mathematical model of this
specific system at the operating conditions expected to be obtained from a
fermentation reactor. This chapter presents the formulation of an empirical model for
the pervaporation of ethanol through an organophilic membrane. The membrane used
TM
in this work is a PERVAP 4060 polymeric membrane provided by Sulzer
(Holtbruegge et al., 2013; Koch and Gorak, 2014). The operating conditions and the
experimental procedure are explained in more detail in Sections 4.2 and 4.3,
respectively. Sections 4.4 and 4.5 present the parameter estimation of the
mathematical model using the experimental data and the statistical analysis thereof.

4.2 Experimental materials and methods

In order to estimate the parameters of a model that describes the permeation of
ethanol, experimental data are required. This set of experiments is obtained in
collaboration with the Laboratory of Fluid Separations at Technische Universität
Dortmund, Germany. The mathematical model of the membrane has to be able to
determine the permeance and partial fluxes of ethanol and water through said
membrane when feed concentration, feed temperature and permeate pressure are
specified.

85
The model also needs to show good fit with the experimental data given that its
formulation is purely empirical. Additional experiments are carried out in order to
evaluate the influence of some common fermentation by-products over the
performance of membrane since these components may affect the performance of the
pervaporation module.

4.2.1 Materials
The feed concentration of ethanol for the set of experiments presented in Table 4.1 is
the range of concentrations that can be obtained after the filtration of the unreacted
solids from the fermentation stage (Baeyens et al., 2015; Fan et al., 2014; Gaykawad
et al., 2012). The range of feed temperatures includes the operating temperatures at
which the microorganisms and the enzymes can simultaneously perform without
compromising their activity and metabolism (Haelssig et al., 2012; O’Brien et al.,
2004).

The range of operation for permeate pressure is between 400 and 3500 Pa and as
Holtbruegge et al., 2013 have demonstrated, the optimal operating costs can often be
found at maximum pressure differences between the feed and the permeate side.
Table 4.1 summarises the operating conditions used in the estimation of the
parameters of the mathematical model. Figure 4.1 shows the experimental design
proposed for the development of the mathematical model. This experimental set-up
considers a number of experiments that will cover the different operating conditions
in order to guarantee a good fit of the parameters. The chemical system consists
mainly of ethanol and water which are obtained after the fermentation stage (Erdei et
al., 2013). However, during the fermentation of lignocellulosic hydrolysates, other
by-products such as organic acids, glycerol, CO2, can also be produced, although in
lower concentrations (Sánchez and Cardona, 2012). Table 4.2 presents some of the
most common by-products in the fermentation of lignocellulosic hydrolysates.

86
Figure 4.1: Experimental design for the evaluation of the parameters of the mathematical model
of the PERVAPTM 4060 membrane (C = feed concentration, T = feed temperature, P = permeate
pressure)

Table 4.1: Operating conditions used in the experiments

Ethanol Feed Permeate
Point feed conc. temperature pressure
(% w/w) (oC) (mbar)
1 5 (C1) 30 (T1) 4 (P1)
2 7 (C2) 40 (T2) 20 (P2)
3 13 (C3) 50 (T3) 35 (P3)

4.2.2 Membrane and module

Figure 4.2 shows the flowsheet for the pervaporation module used in the Laboratory
of Fluid Separations at TU Dortmund. The pervaporation plant (Figures 4.2 and 4.3)
consists of a membrane module (3) inside of a thermal oven (4) that controls the
temperature of the system and a jacketed tank (1) where the mixture is fed,
completely insulated in order to avoid heat losses (see Figures 4.4 to 4.6).

The plant also features a set of cooling traps (5) (see also Figure 4.7) in which the
condensate of the vapour that permeates through the membrane is collected for
further measurements. The system initially operates in a loop in order to guarantee
the steady-state conditions prior to the measurement of concentration in the samples.
Once these conditions have been reached, the sampling can commence.
87
Figure 4.2: Simplified flowsheet of the laboratory-scale set-up for pervaporation experiments.
(1) Feed tank, (2) Heat exchanger, (3) Membrane module, (4) Thermal oven, (5) Cooling traps,
(6) Vacuum pump, (7) Exhaust, (8) Heating cabinet and (9) Back-pressure regulator

88
 Table 4.2: List of components produced during the fermentation of lignocellulosic hydrolysates
Ethanol Acetic acid Glycerol Glucose Xylose
Raw material Pretreatment Microorganism Reference
(g/L) (g/L) (g/L) (g/L) (g/L)
Saccharomyces
Whole wheat Steam explosion 27 – 35 2.1 - 3 3.7- 4.3 0.3 - 0.7 15 - 17 (Erdei et al., 2013)
cerevisiae TMB3400
Steam explosion Saccharomyces c
Whole wheat 44 – 45 2.1 - 3 4.7- 4.8 0.9 - 1 14 - 17 (Erdei et al., 2013)
+ Amyloglucosidases cerevisiae TMB3400
Saccharomyces
Wheat straw Steam explosion 22 5.6 - 5.7 - 0-5 15 - 16 (A. Moreno et al., 2013)
cerevisiae F12
Saccharomyces
Paja brava Steam explosion 35 - 40 2.9 - 3.5 3.5- 4 <0.1 8 – 8.5 (Carrasco et al., 2011)
cerevisiae TMB3400
Hydrolysate Dilute acid Yeast CGMCC 2661 20 - 30 - - <0.5 <0.5 (Tian et al., 2009)
Sugar cane
Dilute acid E. coli SL100 20 - 30 - - <0.5 4.0 - 6 (Geddes et al., 2011)
Bagasse
Hydrolysate Dilute acid Mucor indicus 15 - 20 - 1.0 - 2 <0.5 2.0 - 4 (K Karimi et al., 2006)
Saccharomyces
Barley straw Concentrated acid 21 - 30 2.04 - <0.5 - (Gaykawad et al., 2012)
cerevisiae CEN.PK113-7D
Willow Saccharomyces
Mild alkaline 8.0 - 12 5.43 - <0.5 - (Gaykawad et al., 2012)
wood chips cerevisiae CEN.PK113-7D
Saccharomyces
Barley straw Mild alkaline 13 - 19 4.75 - <0.5 - (Gaykawad et al., 2012)
cerevisiae CEN.PK113-7D
Prepared
- - 30 - 40 - 1 5 5 (Chovau et al., 2011)
Mixture
Corn fibre Dilute acid E. coli strain KO 11 30 - 40 <0.5* - <0.5 5.0 - 10 (O’Brien et al., 2004)
SO2 impregnation
Saccharomyces
Spruce chips and exposed to a 51 - 52 9.6 - 9.8 * - 1.5 3.7 (Ishola et al., 2013)
cerevisiae (commercial)
pressure of 22 bar
*After detoxification and neutralisation

89
Figure 4.3: Pervaporation plant Figure 4.4: Membrane module

Figure 4.5: Thermal oven Figure 4.6: Jacketed tank

Figure 4.7: Cooling traps

90
 Table 4.3: Specifications for the optimal operation of the PERVAP TM 4060 membrane
(Sulzer Chemtech, 2014)
Item Value
Max concentration in feed
Organic components 90 % (w/w)
Organic acids <1 % (w/w)
Inorganic acids <0.1% (w/w)
Maximum temperature
Short operating times 85 oC
Long operating times 80 oC

The PERVAPTM 4060 membrane uses an active polyvinyl alcohol separating layer
(the idealised formula [CH2CH(OH)]n), differently cross-linked to adapt the
membrane separation performance to different operating conditions and components
in the feed stream (Ben Soltane et al., 2013; Claes et al., 2010; Sulzer Chemtech,
2014). The PERVAPTM 4060 membrane is mainly used to selectively separate
volatile organic components (VOC’s) such as benzene, benzaldehyde, ethanol,
isobutanol, methanol, among others showing high separation rates with smaller
membrane areas (Claes et al., 2012, 2010; Setlhaku et al., 2013; Yakovlev et al.,
2013). The membrane area used in these experiments is 124.6 cm2 which fits in the
membrane module provided by TU Dortmund. The specifications for an optimal
operation of this membrane are shown in Table 4.3.

4.2.3 Gas chromatography

A gas chromatograph (Shimadzu GC14A with software GCSolution) equipped with
a flame ionisation detector (FID) is used to analyse feed, retentate and permeate
samples. It has a column Restek RTX 5 which uses diphenyl dimethyl polysiloxane
as a stationary component for sampling using an internal standard component. The
specifications of the GC column are the following (RESTEK, 2016):

 Length = 30 mm
 Internal diameter = 0.25mm
 Film diameter =1µm
91
The mass fractions of organic compounds are obtained directly from the evaluation
of the chromatogram based on single-component calibration curves using acetonitrile
as the internal standard. Each sample is analysed three times and average values are
used for further estimation of the parameters taking into account the standard
deviation of each experiment (Koch and Gorak, 2014; Roth et al., 2013). The quality
of the GC analysis is validated during the experiments by the analysis of test
mixtures with known compositions. The FID detector used in the GC analysis is able
to analyse only the mass fraction of organic components such as ethanol and
acetonitrile in the samples. The mass fraction of water is calculated from the mass
fractions of ethanol and acetonitrile determined by the GC analysis using a total mass
balance. The operating conditions of the Shimadzu GC14A, used in this work, are
presented next:

 Helium as carrier gas with pressure of 65 kPa

 Split around 1:80 1µl injection-volume
 Detector: FID (with synth. Air and Hydrogen)

Temperature programme:

 Detector: 300 °C
 Injection-temperature: 275 °C
 Column: (only Ethanol) 80 °C hold for 3.5 min at constant temperature
 Column: (with Acetic Acid) 80 °C hold for 3.2 min heat up with 30 °C/min to
150 °C hold for 0.5 min

4.3 Experimental procedure

The experimental procedure implemented in this work is as follows:

1. Open the bypass valves and close the valves that feed the membrane module
(3) to circulate the mixture during the stabilisation of the process (see Figure
4.2).
2. Open the valves of the feed tank (1).

92
3. Turn on the control system of the thermal oven (3) and set the desired
temperature.
4. Switch on the vacuum pump and set the permeate pressure at which the
separation will take place.
5. Switch on the pump used for the feed (1) and circulation of the mixture at a
flow rate of 50 L/h. The feed flow rate has to be high since in this experiment
what is being determined is only the flux through the membrane and the
concentration in permeate. The concentration in retentate is expected to be
the same as in the feed.
6. Turn on the control system for the temperature and set a feed temperature.
7. Set the feed concentration of the mixture by adding ethanol or water to the
feed tank (GC analysis is used here in order to determine the feed
concentration).
8. When the temperature has reached the set point and is no longer changing,
the feed valve of the membrane module is opened and the bypass is closed. It
is recommendable to wait for 30 more minutes after steady-state has been
reached in order to guarantee reliable data.
9. Place the empty cooling traps (5) (previously weighed) in the containers with
liquid nitrogen and open the valve for the vacuum pump to collect the
sample.
10. Samples of the feed and retentate are taken at time 0 min. The sampling time
for feed, permeate and retentate is every 10 minutes or even longer depending
on the operating conditions and the mass of permeate collected, (some
operating conditions require longer times in order to collect sufficient
permeate for the GC analysis). The experiment stops once three samples of
permeate have been collected. Average time per experiment: 30 minutes.
11. The same procedure is carried out using different fermentation by-products at
different concentrations. In this work, glucose, xylose, glycerol and acetic
acid are considered. Although furfural is usually found in the ethanol
production process from lignocellulosic biomass, this work does not contain
any experiment with this component since its handling requires high safety
conditions due to its high toxicity.

93
The experimental procedure presented above consists of a series of steps which
allows regulating and fully operating the pervaporation plant in order to guarantee
reliable experimental data. The operating flow rate of the feed is set at 50 L/h in
order to secure the same concentration in both feed and retentate since the idea of
this experiment is to determine the flux and the concentration of ethanol through the
membrane not to evaluate the rates of separation in the retentate side. The
experimental procedure also seeks to minimise losses of ethanol in the sampling and
in the feed tank by constantly checking the concentration of the feed tank and the
other outlet streams as well as the stability in the measurement of both temperature
and permeate pressure.

4.4 Modelling
The experimental procedure previously outlined helped to obtain the different
experimental results that will be used in the estimation of the parameters of the
mathematical model. This section presents the methodology and tools implemented
in the estimation of the parameters of an empirical mathematical model used to
describe the pervaporation of aqueous solutions of ethanol under different operating
conditions. The approach for the parameter estimation in the model uses the weight
of permeates, the concentration of ethanol, the time of collection of samples and the
area of the membrane to determine the partial fluxes of the components through the
membrane. All calculations are carried out using gPROMS v. 4.1.0 (Process System
Enterprises, 2015).

4.4.1 Approach
The driving force for mass-transfer of one or more components between two phases
is given by their differences in the chemical potentials (DFi) (Holtbruegge et al.,
2013). For pervaporation, the driving force for mass transfer is commonly expressed
as the difference in fugacities between the feed and the permeate side of the
membrane as shown in Eq. 4.1 (Valentinyi et al., 2013):

𝑓𝑒𝑒𝑑
𝐷𝐹𝑖 = (𝑥𝑖 𝛾𝑖 𝑃𝑖𝑠𝑎𝑡 𝜑𝑖𝑠𝑎𝑡 ) − (𝑥𝑖𝑝𝑒𝑟𝑚 𝑃𝑝𝑒𝑟𝑚 𝜑𝑖𝑣𝑎𝑝 ) 𝑖 = 1,2, … 𝑁𝐶 Eq. 4.1

94
The partial flux of component i is evaluated from the experimental data as the
relation between the mass of permeate, the membrane area and the time of the
collection of the samples. The permeance of ethanol and water Qi (g 𝑚−2 𝑠 −1 𝑃𝑎−1 )
can be calculated from Eq. 4.2. The number of components NC is two (i.e. only
ethanol and water are considered).

𝐽𝑖 = 𝑄𝑖 𝐷𝐹𝑖 𝑖 = 1,2, … 𝑁𝐶 Eq. 4.2

4.4.2 Pure component and mixture property models

The thermodynamic properties of all components (including fermentation by-
products) are evaluated using Multiflash 4.1.43 (Infochem, 2013). NRTL is used for
the calculation of the activity coefficients of the components in the mixture. The
PSRK equation of state is used to evaluate the fugacity coefficients and Antoine’s
equation is used for the calculation of vapour pressures.

4.4.3 Models for permeance

This work introduces three equations for the permeance of the membrane in order to
conduct a comparative study between different formulations for the calculation of the
partial fluxes. This analysis will determine if the number of experiments and the
number of parameters are relevant factors for an accurate depiction of the
pervaporation of ethanol through the membrane and if the model is suitable to be
implemented in the main process.

Eq. 4.3 is taken from the work presented by Koch and Gorak, (2014). This equation
represents the variation of the permeance as a function of the feed temperature T with
an Arrhenius-type equation along with the influence of the feed concentration Wfeed.i.
However, the effect of the variation of the permeate pressure is not considered for
this particular formulation. According to the work of Koch and Gorak, (2014), the
approach presented in Eq. 4.3 has been found to accurately describe systems of
binary mixtures since it incorporates two operating conditions that have great
influence over the process.

95
 𝐸𝑖 1 1
𝑄𝑖 = 𝑄0,𝑖 𝐸𝑋𝑃 (− ( − )) 𝑊𝑓𝑒𝑒𝑑.𝑖 𝐴𝑖 𝑖 = 1,2, … 𝑁𝐶 Eq. 4.3
𝑅 𝑇 𝑇0

Another variation of Eq. 4.3 can be used to obtain the parameters of the
mathematical model taking into consideration the effect of the permeate pressure and
vapour pressure. This approach has been presented in the work of Holtbruegge et al.,
(2013).

𝐵𝑖
𝐸𝑖 1 1 𝑃𝑝𝑒𝑟𝑚
𝑄𝑖 = 𝑄0,𝑖 𝐸𝑋𝑃 (− ( − )) 𝑊𝑓𝑒𝑒𝑑.𝑖 𝐴𝑖 ( 𝑠𝑎𝑡 ) 𝑖 = 1,2, … 𝑁𝐶
𝑅 𝑇 𝑇0 𝑃𝑖 Eq. 4.4

Eq. 4.5 is a variation of the work presented by Holtbruegge et al., (2013). The
approach presented in Eq. 4.5 presents a different form of the term that includes the
permeate pressure and the vapour pressure and also features an additional parameter:

𝑠𝑎𝑡
𝐸𝑖 1 1 𝐴𝑖 (𝑃𝑖 − 𝑃𝑝𝑒𝑟𝑚 ) + 𝐶𝑖
𝑄𝑖 = 𝑄0,𝑖 𝐸𝑋𝑃 (− ( − )) 𝑊𝑓𝑒𝑒𝑑.𝑖 ( 𝑠𝑎𝑡 )𝑖
𝑅 𝑇 𝑇0 (𝑃𝑖 − 𝑃𝑝𝑒𝑟𝑚 ) + 𝐷𝑖 Eq. 4.5
= 1,2, … 𝑁𝐶

The mathematical model will have the feed temperature T, the permeate pressure
Pperm and the feed mass concentration of ethanol Wfeed.i as input variables to
calculate the permeance Qi of each component in the system (Equations 4.3 to 4.5).
The molar fraction of ethanol in the permeate side 𝑥𝑖𝑝𝑒𝑟𝑚 is determined as follows:

𝐽𝑖
𝑥𝑖𝑝𝑒𝑟𝑚 = 𝑁𝐶 𝑖 = 1,2, … 𝑁𝐶 Eq. 4.6
∑𝑖=1 𝐽𝑖

With the mole fractions of the permeate and feed sides and the temperature of the
feed, the driving forces DFi can be estimated using Eq. 4.1 and the partial fluxes with
Eq. 4.2. The parameters to be fitted from the experimental data are listed next:

96
  𝑄0,𝑖 : Pre-exponential permeance value for the Arrhenius-
type equation for the influence of feed temperature
in the performance of the separation.
(g 𝑚−2 𝑠 −1 𝑃𝑎−1 )
 𝐸𝑖 : Activation energy. (J mol-1 K-1)
 𝐴𝑖 , 𝐵𝑖 , 𝐶𝑖 , 𝐷𝑖 : Constants

4.5 Results
Table 4.4 presents the measured data and the calculated data from the set of
experiments previously outlined using Eq. 4.2. The results reported in Table 4.4
show that the operating conditions have a direct influence over the performance of
the membrane and therefore the partial fluxes of ethanol and water. When feed
concentration of ethanol increases (from 5% to 13% w/w ethanol) at constant feed
temperature (30 oC) and permeate pressure (400 Pa) the partial fluxes increase too
(59 % increase for ethanol and 27% for water).

Similarly, when the temperature increases from 30 oC to 50 oC at constant feed

concentration (5% w/w ethanol) and permeate pressure (400 Pa), the partial fluxes
increase 56 % and 61 % for ethanol and water, respectively. Conversely, when the
permeate pressure increases from 400 Pa to 3500 Pa, the partial fluxes tend to
decrease as reported by Marriott and Sorensen (2003), Roth et al., (2013) and Wei et
al., (2014). The decrease of the partial fluxes (see Table 4.4) at constant T (30 oC)
and Wfeed (5% w/w ethanol) is reported to be 71 % for ethanol 80 % for water. The
results of the experiments suggest that permeate pressure is an operating condition
that has a significant impact in the amount of material passing through the membrane
and should be taken into consideration for the formulation of the mathematical
model. This analysis also indicates that, within the range of operating conditions,
higher fluxes and higher separation rates of ethanol through the membrane can be
achieved.

Table 4.5 presents the results of the partial fluxes of ethanol and water at different
concentrations of impurities. These are the by-product concentrations expected to be

97
obtained after the fermentation stage. To investigate the influence of different
fermentation by-products, first a pure ethanol/water mixture is subjected to
pervaporation experiments, which will serve as a reference for the comparative
analysis.

The operating conditions for the pure mixture ethanol/water are: 40 oC, 2000 Pa and
a feed concentration of ethanol of 5 % (w/w). For components that exhibit low
vapour pressure, driving forces such as pressure gradient and concentration gradient
are very low. Hence, these components (i.e. glycerol, glucose, xylose) can be
considered as impermeable and should not be detected in the permeate stream
(Chovau et al., 2011). However, an increase of the total flux through the membrane
is observed when these components are introduced into the mixture.

Acetic acid is detected in the permeate stream and an increase in the flux of ethanol
of approximately 40% can be observed. However, and as previously mentioned,
higher concentrations of organic acids can compromise the membrane in its structure
and its durability in the long term (Sulzer Chemtech, 2014).

4.5.1 Model discrimination

This work compares three empirical models of the PERVAPTM 4060 membrane
taking into account different approaches in the formulation of the equation for
permeance of each component. The parameters of these models are adjusted from the
experimental data shown in Table 4.4. The parameter estimation model using gEST
and gPROMS considers a solver of Maximum-Likelihood Estimation (MLE) for
complex models using both dynamic and steady-state experimental data (Process
System Enterprises, 2015). gPROMS attempts to determine values for the uncertain
physical and variance model parameters that maximise the probability that the
mathematical model will predict the measurement values obtained from the
experiments.

Table 4.5 presents the results of the experiments with the membrane using by-
products often found in the fermentation of hydrolysates of lignocellulosic biomass.

98
The calculations featured in this work are carried out using Dell OptiPlex 9010
Intel® Core™ i7-3770 CPU @3.40 GHz and 16 GB RAM running Windows 7®
Enterprise (64-bit operating system). The results of the parameter estimation for
equations 4.3 to 4.5 are presented in Table 4.6.

99
 Table 4.4: Experimental results for the removal of ethanol from aqueous solutions using the PERVAPTM 4060 membrane
Measured data Calculated data
Ethanol Water Ethanol Water Permeate Mass Ethanol Water Permeance Permeance
Sample Temp 5
feed feed permeate permeate pressure permeate flux flux ethanol 10 water 105
(oC)
(w/w) (w/w) (w/w) (w/w) (mbar) (g) (𝐠 𝒎−𝟐 𝒔−𝟏 ) (𝐠 𝒎−𝟐 𝒔−𝟏 ) (𝐠 𝒎−𝟐 𝒔−𝟏 𝑷𝒂−𝟏 ) (𝐠 𝒎−𝟐 𝒔−𝟏 𝑷𝒂−𝟏 )
C1T1P1 0.06±0.003 0.94±0.003 0.25±0.01 0.75±0.01 30.53±0.03 4.35±0.15 2.59±0.07 0.058±0.002 0.173±0.008 5.63±0.28 4.42±0.20
-4 -4
C1T1P=10 0.06±1x10 0.94±1x10 0.26±0.01 0.74±0.01 30.60±0.10 9.70±0.01 2.10±0.01 0.055±0.002 0.151±0.002 5.61±0.20 4.37±0.08
-4 -4 -4
C1T1P2 0.05±0.001 0.95±0.001 0.27±0.01 0.73±0.01 30.34±0.05 19.8±0.13 1.6±1x10 0.04±4x10 0.11±4x10 5.20±0.35 4.15±0.03
-4 -4
C1T1P=28 0.06±2x10 0.94±2x10 0.29±0.01 0.71±0.01 30.42±0.01 28.3±0.22 1.69±0.04 0.033±0.001 0.080±0.003 4.99±0.17 4.39±0.02
C1T1P3 0.05±0.003 0.95±0.003 0.33±0.01 0.67±0.01 30.58±0.03 34.8±0.27 0.6±0.005 0.017±0.001 0.034±0.001 4.11±0.37 2.43±0.12
-4 -4
C2T1P1 0.07±6x10 0.93±6x10 0.3±0.004 0.7±0.004 30.63±0.02 4.33±0.27 2.94±0.11 0.078±0.002 0.184±0.008 5.85±0.19 4.71±0.20
-4 -4
C1T2P1 0.04±5x10 0.96±5x10 0.23±0.02 0.77±0.02 40.21±0.07 4.08±0.21 2.56±0.05 0.080±0.007 0.262±0.009 5.81±0.61 3.75±0.10
-4 -4
C1T3P1 0.04±3x10 0.96±3x10 0.23±0.02 0.77±0.01 49.71±0.06 4.63±0.23 4.29±0.03 0.131±0.004 0.442±0.001 5.25±0.19 3.83±0.46
C1T3P3 0.05±0.001 0.95±0.001 0.27±0.01 0.73±0.01 49.79±0.05 34.8±0.28 3.34±0.16 0.122±0.001 0.326±0.021 5.58±0.05 3.62±0.24
C3T1P1 0.13±0.002 0.87±0.002 0.38±0.01 0.62±0.01 30.37±0.16 4.09±0.21 4.24±0.08 0.142±0.001 0.236±0.006 6.62±0.15 6.15±0.15
C3T3P1 0.13±0.003 0.87±0.003 0.44±0.02 0.56±0.02 50.32±0.27 3.98±0.18 6.56±0.10 0.382±0.018 0.495±0.012 6.05±0.36 4.01±0.16
-4
C3T1P3 0.13±0.001 0.87±0.001 0.44±0.02 0.56±0.02 30.58±0.02 35.4±0.13 2.44±0.08 0.09±4x10 0.121±0.076 6.81±0.17 8.16±0.80
C3T3P3 0.14±0.004 0.86±0.004 0.4±0.001 0.6±0.001 50.11±0.26 35.24±0.44 7.83±0.47 0.290±0.017 0.408±0.026 4.90±0.10 4.53±0.28

C = concentration of ethanol in the feed stream, T = feed temperature, P = permeate pressure. 1, 2, and 3 represent the points shown in Figure 4.1. P = 10 and 28 mbar
represent intermediate points taken to improve the fit of the parameters of the model. The uncertainty of the results is obtained as the mean of the standard deviation of all
the experiments used for each point.

100
 Table 4.5: Partial fluxes of ethanol and water under the influence of different impurities at 40 oC and a permeate pressure of 20 mbar
Ethanol Water By-product Ethanol Water By-product Ethanol Water By-product Total
Concentration
By-product feed feed feed permeate permeate permeate flux flux flux flux
(g/L)
(% w/w) (% w/w) (% w/w) (% w/w) (% w/w) (% w/w) (𝐠 𝒎−𝟐 𝒔−𝟏 ) (𝐠 𝒎−𝟐 𝒔−𝟏 ) (𝐠 𝒎−𝟐 𝒔−𝟏 ) (𝐠 𝒎−𝟐 𝒔−𝟏 )
No by-product 0 0.050±0.001 0.950±0.001 - 0.26±0.011 0.74±0.011 - 0.076±0.002 0.217±0.007 - 0.293±0.009
Glucose 1 0.051±0.001 0.947±0.001 0.002±0.001 0.24±0.003 0.76±0.003 - 0.071±0.002 0.226±0.002 - 0.297±0.004
-4 -4 -4
Glucose 2 0.05±1x10 0.95±1x10 0.004±1x10 0.26±0.007 0.74±0.003 - 0.074±0.002 0.210±0.002 - 0.284±0.004
Xylose 5 0.047±0.001 0.948±0.001 0.005±0.001 0.25±0.002 0.75±0.002 - 0.076±0.005 0.235±0.015 - 0.311±0.020
Xylose 10 0.048±0.001 0.942±0.001 0.010±0.001 0.25±0.006 0.75±0.006 - 0.081±0.002 0.250±0.006 - 0.328±0.008
-4 -4 -4
Glycerol 3 0.057±3x10 0.94±3x10 0.003±3x10 0.28±0.013 0.72±0.013 - 0.092±0.001 0.242±0.013 - 0.332±0.014
-4 -4 -4
Glycerol 6 0.050±1x10 0.944±1x10 0.006±1x10 0.26±0.007 0.74±0.007 - 0.088±0.001 0.249±0.006 - 0.337±0.007
-4 -4 -4
Acetic acid 1 0.046±5x10 0.953±5x10 0.001±1x10 0.228±0.003 0.685±0.002 0.088±0.001 0.138±0.003 0.414±0.008 0.053±0.001 0.605±0.012
-4 -4
Acetic acid 2 0.051±1x10 0.947±1x10 0.002±0.001 0.269±0.011 0.648±0.012 0.083±0.002 0.121±0.006 0.291±0.003 0.037±0.001 0.499±0.010

101
 Table 4.6: Parameters of the mathematical model of the organophilic membrane
Ethanol
Parameter
Eq. 4.3 Eq. 4.4 Eq. 4.5
Qo,i (105) 7.40±1.72 6.98±1.650 7.67±0.78
Ei -1883±752 -4420±1520 -3640±825
Ai (102) 9.60±0.75 10.6±0.750 10.2±0.77
Bi - -0.045±0.03 -
Ci - - 8.57±2.31
Di - - -85.4±20.7
Water
Qo,i (105) 3.46±0.73 3.51±0.80 3.45±0.17
Ei -11399±4774 -8550±3240 -9640±1042
Ai -5.24±1.87 -5.26±1.85 -4.88±2.09
Bi - 0.037±0.01 -
Ci - - 26.6±8.74
Di - - -2.68±0.60

Table 4.6 present the results of the parameter estimation of the different approaches
for the evaluation of the permeance of each component. From this table, it can be
seen the similarities in the values of the pre-exponential term Qo,i in each equation
for both ethanol and water. For ethanol, for example, the values go from 6.98 x10-5 to
7.67x10-5 g m-2 s-1 Pa-1 (a difference of approximately 10%).

On the other hand, the activation energy terms (Ei) exhibit differences that can go up
to 57 % for ethanol and 25 % for water. The power term Ai, which is also found in all
the approaches, exhibits values that go from 0.096 to 0.106 for ethanol (approx. 9%)
and from -5.26 to -4.88 for water (approx. an increment of 7%). The results of the
parameter estimation indicate that all the parameters are susceptible to changes when
a different formulation is implemented or when additional operating conditions (i.e.
permeate pressure) are considered.

From the previous section it can be seen that fermentation by-products influence the
performance of the organophilic membrane. However, their contribution in the
mathematical model is not considered in this work since most of these components
are either degraded during the pretreatment or the overliming or present low
102
concentration in the feed stream of the separation section. Additional experiments are
conducted using different fermentation by-products in order to assess their effect
over the membrane performance. The influence of temperature over partial fluxes
using the parameters of equations 4.3 to 4.5 is shown in Figure 4.8.

As seen in Table 4.4, the partial fluxes increase with temperature as the result of the
increase of the driving force and the vapour pressures (Marriott and Sorensen, 2003;
Roth et al., 2013; Wei et al., 2014). It can also be seen that Eq. 4.5 has a better fit of
the experimental data as curves for ethanol and water are closer to the experimental
data also plotted in Figure 4.8. From this figure, it can also be concluded that a good
adjustment of the experimental data can be obtained when the number of parameters
in the model is higher.

An additional experiment, only for the binary mixture ethanol/water, was conducted
in order to validate the mathematical model using a point that was not included in the
parameter estimation. This experimental point consists of a feed concentration of 7 %
(w/w) of ethanol, a feed temperature T of 40 oC and a permeate pressure of 2000 Pa.
Using the additional set of experiments (C2T2P2), the mathematical model using Eq.
4.5 can be validated.

The results of the model verification presented in Table 4.7 show a good fit between
experimental data and predicted results reporting errors of ± 4.4 % in the
concentration of ethanol in the permeate side. Figure 4.9 shows the parity plots for
the mathematical model of the organophilic membrane and the experimental values.
The standard deviation between the experimental data and the predicted values of the
permeance for ethanol and water are 6.9x10-6 g m-2 s-1 Pa-1 and 1.32x10-5 g m-2 s-1
Pa-1, respectively. The closing of the mass and component balances for these
experiments can be quite difficult, because all measured variables are inevitably
corrupted by experimental errors.

Given that errors are inevitable during experimentation, a statistical analysis is

required in order to quantify said errors and determine the reliability of the

103
mathematical model and how it differs from the experimental data. Tables 4.8 to 4.9
show the error analysis of the estimation of the parameters in the mathematical
model for the membrane using Equations 4.3 to 4.5. The Lack-of-fit test, which is
commonly used for testing independence and goodness of fit between observed and
expected values, is considered in this work. This chapter will use the Pearson’s chi-
square test which is the most used method for the assessment of accuracy in the
estimation of parameters for empirical models. Equation 4.7 defines the difference
between the expected outcome frequencies and the observed ones:

2
𝑂
𝑛 𝑛 ( 𝑘⁄𝑝𝑖 − 𝑝𝑖 )
2
(𝑂𝑘 − 𝐸𝑘 )2 Eq. 4.7
𝜒 =∑ = 𝑁∑
𝐸𝑘 𝑝𝑖
𝑘=1 𝑘=1

104
 0.50
Ethanol (Eq. 4.3)
0.45 Water (Eq. 4.3) 
Ethanol (Eq. 4.4)
0.40 Water (Eq. 4.4)
Partial flux (g m-2 s-1) 0.35 Ethanol (Eq. 4.5)
Water (Eq. 4.5)
0.30



0.25
0.20



0.15

0.10



0.05 

304 306 308 310 312 314 316 318 320 322
Temperature (K)
Figure 4.8: Partial flux for ethanol and water using the different formulations proposed in
Equations 4.3 to 4.5

Table 4.7: Model validation using additional experimental data

Experimental Predicted
Operating
JEthanol JWater JEthanol JWater
Conditions 𝑾𝑬𝒕𝒉𝒂𝒏𝒐𝒍
𝑷𝒆𝒓𝒎𝒆𝒂𝒕𝒆 𝑾𝑬𝒕𝒉𝒂𝒏𝒐𝒍
𝑷𝒆𝒓𝒎𝒆𝒂𝒕𝒆
(𝒈 𝒎−𝟐 𝒔−𝟏 ) (𝒈 𝒎−𝟐 𝒔−𝟏 ) (𝒈 𝒎−𝟐 𝒔−𝟏 ) (𝒈 𝒎−𝟐 𝒔−𝟏 )
T = 40.4 oC
0.339±0.029 0.115±0.013 0.224±0.004 0.324 0.115 0.223
P= 20.05 mbar

Table 4.8: Lack-of-fit test for the parameter estimation of the membrane model
Weighted
Equation χ2 – value (95 %)
Residual (χ)
4.3 15.04 31.41
4.4 13.34 28.87
4.5 15.42 26.30

105
Table 4.8 shows the results of the statistical analysis of the experimental data using
the Pearson’s chi-square method described in Eq. 4.7. Here Ok is the observed
frequency for bin k, Ek is the expected (theoretical) frequency for bin k and n is the
number of bins used in the analysis. The Pearson’s chi-square test states that when
the χ2-value (at 95%) ≥ χ2 for a p-value of 0.95 (or 95 %), the mathematical model
can be concluded at a significance level of confidence of 95% to be a good
representation of the experimental process and the parameters are well-adjusted.
Conversely, χ2-value (at 95%) ≤ χ2 rejects the null hypothesis that the model is
representing the observed data.

The results of the lack-of-fit test show that all the models have a good fit against the
experimental data, which indicates that the number of experiments is enough to
predict the permeance for either component at the range of operating conditions.
Tables 4.9 to 4.11 show the error analysis of the mathematical model of the
membrane using equations 4.3 to 4.5.

The results of these tables show the errors and the standard deviations for the three
models. These results also support the aforementioned statement which made
reference to the good fit of all the models used in this work. However, and as seen in
Figures 4.8 and 4.9, the results for Eq. 4.5 are closer to the experimental data and
when those results are combined with the statistical analysis provided in Table 4.8,
one can conclude that this model is accurate enough to describe the permeation of
ethanol though the organophilic membrane within the operating conditions at which
these models are restricted.

106
 7x10-5 7x10-5 7x10-5

Predicted ethanol permeance

Predicted ethanol permeance

Predicted ethanol permeance
6x10-5 6x10-5 6x10-5

(g m-2 s-1 Pa-1)

(g m-2 s-1 Pa-1)

(g m-2 s-1 Pa-1)

5x10-5 5x10-5 5x10-5

4x10-5 4x10-5 4x10-5

4x10-5 5x10-5 6x10-5 7x10-5 4x10-5 5x10-5 6x10-5 7x10-5 4x10-5 5x10-5 6x10-5 7x10-5
Experimental ethanol permeance Experimental ethanol permeance Experimental ethanol permeance
(g m-2 s-1 Pa-1) (g m-2 s-1 Pa-1) (g m-2 s-1 Pa-1)

(a) (b) (c)

7x10-5 7x10-5 7x10-5

Predicted water permeance

Predicted water permeance

Predicted water permeance

6x10-5 6x10-5 6x10-5

(g m-2 s-1 Pa-1)

(g m-2 s-1 Pa-1)

(g m-2 s-1 Pa-1)

5x10-5 5x10-5 5x10-5

4x10-5 4x10-5 4x10-5

3x10-5 3x10-5 3x10-5

2x10-5 2x10-5 2x10-5

2x10-5 3x10-5 4x10-5 5x10-5 6x10-5 7x10-5 8x10-5 9x10-5 2x10-5 3x10-5 4x10-5 5x10-5 6x10-5 7x10-5 8x10-5 9x10-5 2x10-5 3x10-5 4x10-5 5x10-5 6x10-5 7x10-5 8x10-5 9x10-5
Experimental water permeance Experimental water permeance Experimental water permeance
(g m-2 s-1 Pa-1) (g m-2 s-1 Pa-1) (g m-2 s-1 Pa-1)

(d) (e) (f)

Figure 4.9: Parity plots with the fitted parameters for ethanol (top) and water (bottom) using: (a) Eq. 4.3, (b) Eq. 4.4, (c) Eq. 4.5, (d) Eq. 4.3, (e), Eq. 4.4,
(f) Eq. 4.5

107
 Table 4.9: Parameter estimation problem for the permeance of ethanol and water using Eq. 4.3
Ethanol Water
Permeance Permeance
ethanol Deviation water Deviation
(𝐠 𝒎−𝟐 𝒔−𝟏 𝑷𝒂−𝟏 ) (𝐠 𝒎−𝟐 𝒔−𝟏 𝑷𝒂−𝟏 )
Experimental Model Experimental Model
Absolute Percentage Absolute Percentage
Measurement Prediction 7
Weighted Measurement Prediction 7
Weighted
5 5
10 (%) 5 5
10 (%)
10 10 10 10
5.63 5.53 9.20 1.63 0.13 4.42 4.32 9.76 2.41 0.04
5.61 5.53 7.30 1.30 0.11 4.37 4.31 5.23 1.28 0.04
5.89 5.49 39.6 6.73 0.43 4.57 4.22 34.9 7.13 0.26
4.97 5.53 -5.61 -11.3 -0.96 4.35 4.31 3.97 -0.04 0.03
4.04 5.47 -143 -35.5 -2.23 2.40 4.17 -177 -78.1 -1.34
5.85 5.68 16.9 2.88 0.10 4.71 4.76 -4.81 -0.16 -0.04
5.81 5.25 55.8 9.61 0.89 3.73 3.45 27.2 5.25 0.21
5.25 5.18 6.64 1.26 0.34 3.83 3.10 72.7 15.8 0.55
5.63 5.19 43.5 7.73 0.87 3.66 3.11 54.2 11.3 0.41
6.62 6.00 61.9 9.35 0.71 6.11 6.61 -50.8 -5.1 -0.38
5.90 5.71 18.7 3.18 0.49 4.15 4.89 -74.1 -19.6 -0.56
6.81 6.00 80.7 11.8 0.96 8.16 6.64 151 19.1 1.15
4.83 5.76 -93.4 -19.3 -1.12 4.52 5.25 -73.2 -16.5 -5.55

108
 Table 4.10: Parameter estimation problem for the permeance of ethanol and water using Eq. 4.4
Ethanol Water
Permeance Permeance
ethanol Deviation water Deviation
(𝐠 𝒎−𝟐 𝒔−𝟏 𝑷𝒂−𝟏 ) (𝐠 𝒎−𝟐 𝒔−𝟏 𝑷𝒂−𝟏 )
Experimental Model Experimental Model
Absolute Percentage Absolute Percentage
Measurement Prediction 7
Weighted Measurement Prediction 7
Weighted
5 5
10 (%) 5 5
10 (%)
10 10 10 10
5.63 5.77 -13.7 -2.43 -0.19 4.42 4.11 30.4 6.88 0.23
5.61 5.56 4.91 0.88 0.07 4.37 4.24 13.4 3.06 0.10
5.89 5.33 55.5 9.42 0.80 4.57 4.26 31.2 6.82 0.24
4.97 5.29 -32.4 -6.52 -0.47 4.35 4.40 -4.92 -1.13 -0.04
4.04 5.18 -114 -28.3 -1.66 2.40 4.29 -189 -78.8 -1.45
5.85 5.93 -8.24 -1.41 -0.12 4.71 4.53 17.8 3.77 1.36
5.81 5.45 35.9 6.17 0.52 3.73 3.32 41.0 10.9 3.15
5.25 5.29 -4.16 -0.79 -0.06 3.83 3.05 78.3 20.4 0.60
5.63 4.83 79.2 14.1 1.15 3.66 3.31 35.4 9.67 0.27
6.62 6.37 25.1 3.79 0.36 6.11 6.27 -15.9 -2.61 -0.12
5.90 5.92 -2.29 -0.39 -0.03 4.15 4.83 -68.0 -16.4 -0.52
6.81 5.73 107 15.8 1.56 8.16 6.85 131 16.10 1.01
4.83 5.42 -59.4 -12.3 -0.86 4.52 5.57 -105 -23.3 -0.81

109
 Table 4.11: Parameter estimation problem for the permeance of ethanol and water using Eq. 4.5
Ethanol Water
Permeance Permeance
ethanol Deviation water Deviation
(𝐠 𝒎−𝟐 𝒔−𝟏 𝑷𝒂−𝟏 ) (𝐠 𝒎−𝟐 𝒔−𝟏 𝑷𝒂−𝟏 )
Experimental Model Experimental Model
Absolute Percentage Absolute Percentage
Measurement Prediction 8
Weighted Measurement Prediction 7
Weighted
10 (%) 10 (%)
105 105 105 105
5.63 5.62 5.18 0.09 0.01 4.42 4.31 10.7 2.41 0.08
5.61 5.63 -15.4 -0.27 -0.22 4.37 4.31 5.59 1.28 0.04
5.89 5.59 298 5.06 0.43 4.57 4.24 32.6 7.13 0.25
4.97 5.64 -666 -13.4 -0.96 4.35 4.35 -0.16 -0.04 -0.001
4.04 5.58 -1540 -38.0 -2.23 2.40 4.27 -187 -78.1 -1.44
5.85 5.78 72.8 1.24 0.10 4.71 4.72 -0.76 -0.16 -0.01
5.81 5.19 619 10.6 0.89 3.73 3.53 19.6 5.25 0.15
5.25 5.02 235 4.47 0.34 3.83 3.22 60.5 15.8 0.47
5.63 5.03 601 10.7 0.87 3.66 3.25 41.5 11.3 0.32
6.62 6.13 490 7.41 0.71 6.11 6.43 -31.5 -5.1 -0.24
5.90 5.56 342 5.79 0.49 4.15 4.96 -81.2 -19.6 -0.62
6.81 6.15 664 9.74 0.96 8.16 6.60 156 19.1 1.20
4.83 5.61 -7.76 -16.1 -1.12 4.52 5.27 -74.7 -16.5 -0.58

110
 4.6 Conclusions
This chapter presented an experimental design for the formulation of an empirical
model that could be used to describe the permeation of ethanol from aqueous
solutions using a PERVAPTM 4060 organophilic membrane. Several works have
assessed the benefits of using this particular membrane for different organic solvents
but none of them has explored its application for the ethanol production process from
lignocellulosic biomass or even formulated a mathematical model to predict the
partial fluxes. These experiments consisted of the variation of different operating
conditions during the pervaporation process in which the ethanol was expected to be
removed from the aqueous solutions on the permeate side of the membrane module.

The results of the experiments showed that the operating conditions had a significant
influence over the performance of membrane. The partial fluxes for ethanol and
water increased as the feed temperature (Tfeed) and the feed concentration of ethanol
(Wfeed) increased. Conversely, it was observed that the partial fluxes decreased along
with the increase of the permeate pressure (Pperm) (see Table 4.4). These trends were
consistent with the observations reported by several authors who had used the same
membrane in the separation of other organic solvents from aqueous solutions.

The results of the experiments were used to evaluate the parameters of an empirical
model using gPROMS. Three different mathematical models for the calculation of
the permeance of ethanol and water were considered (see Eq. 4.3 to 4.5). The first
approach (Eq. 4.3) considered the product of an Arrhenius-type term for the
influence of temperature and the feed concentration powered to a parameter (Ai). The
second approach (Eq. 4.4) added another term to the model presented in Eq. 4.3,
which considers a ratio between the vapour pressure and the permeate pressure
powered to a term (Bi).

The third approach (Eq. 4.5) considers a different term for the influence of the
permeate pressure in the pervaporation. The results showed that Eq. 4.4 had a good
fit for the experimental results. However, Eq. 4.5 had a better fit since the curves for
partial fluxes under the influence of feed temperature (see Figure 4.8) were closer to

111
the experimental data and the error percentages were smaller than in Eq. 4.3 and 4.4.
A statistical analysis was conducted in this chapter in order to quantify the errors
obtained during the collection of samples and the measurements. This analysis
showed that all the models presented small deviations from the experimental data
and low errors.

However, the errors obtained with Eq. 4.5 were less significant than the errors
obtained with the other two equations. Additional experiments were conducted using
common fermentation by-products such as glucose, xylose, glycerol and acetic acid.
These results were not included in the parameter estimation due to the limited
number of experiments. However, an increase in the total flux could be appreciated
when the concentration of the impurities was increased. Acetic acid was the only
component detected in the permeate stream. The presence of acetic acid in the
permeate side suggested that a more rigorous experimental design for a more detailed
mathematical model might be required. The implementation of the mathematical
model into the ethanol production process from lignocellulosic biomass and the
optimisation and energy evaluation will be presented in Chapter 5.

112
 Chapter 5 – Optimisation of the separation
section of the ethanol production process

Abstract

The ethanol production process from lignocellulosic biomass shows high levels of
energy demand (mainly in the separation section) making its implementation in
industry less profitable. This chapter focuses on the optimal design and operation of
the separation section of the ethanol production process from corn stover. In
particular, the chapter provides an assessment of the impact of the implementation of
the organophilic membrane before the distillation column.

5.1 Introduction
Ethanol production has been studied extensively for its potential to generate energy
from renewable sources. However, much is still needed to be done in terms of energy
efficiency to make this process profitable and competitive against traditional fuels
such as coal and oil. In this work, ethanol is produced from corn stover, a
lignocellulosic residue with high organic content (Bhandari et al., 1984; Liu and
Chen, 2016). The separation section of the ethanol production process from corn
stover shown in Chapter 3 consists of two distillation columns and a pervaporation
network (Cardona et al., 2010; Drapcho et al., 2008; Garcia et al., 2010; Morales-
Rodriguez et al., 2011). Different publications have concluded that one of the main
obstacles for the implementation of the ethanol production from lignocellulosic
biomass in industry is the energy consumption associated with the separation of this
alcohol from the fermentation medium (Baeyens et al., 2015; Karlsson et al., 2014;
Kiss and Suszwalak, 2012; Kravanja et al., 2013; Triana et al., 2015).

Alternative configurations to distillation techniques for ethanol separation such as

double effect distillation, pressure-swing distillation, heat-integrated distillation

113
columns, etc., seek to reduce the heat duty in the reboiler and increase the
concentration of ethanol in distillate as close as possible to the azeotrope (Kiss and
Olujic, 2014; Knapp and Doherty, 1992; Pohlmeier and Rix, 1996). In the case of
pressure-swing distillation and extractive distillation, the primary goal is to overcome
the azeotrope more than reduce energy consumption (Towler and Sinnott, 2013).
These configurations have shown to be very successful in the purification of ethanol
from aqueous solutions (Caballero, 2015; Kookos, 2003; Sudhoff et al., 2015).
However, their implementation may incur in additional operating costs such as
controllability of the column and operating conditions, solvent recovery,
compression, etc. (Kiss and Olujic, 2014). In this thesis, only simple distillation will
be considered since this work seeks to evaluate simpler technologies rather than
embarking in more elaborate configurations. The optimal arrangement of the
separation section (i.e. number of distillation columns, number of trays in the
distillation columns, reflux ratio, the number of pervaporation stages, permeate
pressure, etc.), which minimises the total annualised cost (TAC), will be obtained by
using rigorous optimisation methods.

This chapter will focus on the implementation of a practical methodology for the
optimisation of the entire separation section of the ethanol production process. This
methodology consists of using optimisation results of individual units (i.e. distillation
column or pervaporation modules) as initial guesses for the overall separation
sections since this will narrow the feasibility region in which the optimum solution
may be located and will help reduce computational times as the set of algebraic-
differential equations is highly non-linear. This chapter presents initially the
optimisation of the individual units and will lead to the solution and analysis of the
different configurations.

5.2 Design of distillation columns

The number of trays required in a distillation column to achieve a separation with a
specific concentration of product can be determined in several ways. Some of these
methods include shortcut methods, group methods, aggregated methods, rigorous

114
tray-by-tray optimisation methods, among others. This section will briefly introduce
some of these methods.

5.2.1 Shortcut methods

One of the most recognised methods for the design of distillation columns is the
method of Fenske-Underwood-Gilliland (FUG). The FUG method presents the
assumptions of constant molar overflow and constant relative volatilities in all the
trays along the distillation column. This method considers two extremely ideal
assumptions (Sorensen, 2014; Towler and Sinnott, 2013):

1. The distillation column operates at total reflux (i.e. no lateral or additional

feed is entering or exiting from the column), which allows the calculation of
the minimum number of trays for a given separation of two key components
2. The column operates at infinite number of trays, which allows calculating the
minimum reflux

The optimal solution of a single distillation column will lie between these two
extreme cases. Although this method can be applied for the separation of different
mixtures, it is not recommended for systems with highly non-ideal nature like those
that present azeotropes (Sorensen, 2014).

5.2.2 Group methods

Group methods are equation-based methods that can be easily implemented in
mathematical programming. Group methods (GM) use approximate calculations to
relate the outlet stream properties to the thermodynamic state of the inlet stream and
number of equilibrium trays (Caballero and Grossmann, 2014a; Dünnebier and
Pantelides, 1999). These approximation procedures are called group methods
because they provide only an overall treatment of the trays in the cascade without
considering detailed changes in the temperature and composition of individual trays.
However, they are easy to solve because they involve few variables and constraints.
They can be represented as a cascade of trays in counter-current operations like

115
absorption, stripping, distillation, leaching or extraction (Dünnebier and Pantelides,
1999; Kamath et al., 2010; Kiss and Olujic, 2014) .

5.2.3 Rigorous tray-by-tray optimisation models

As previously mentioned, the economic optimisation of a distillation column
involves the selection of the number of trays and feed location, as well as the
operating conditions, to minimise the total investment and operating costs.
Continuous decisions are related to the operational conditions and energy involved in
the separation, while discrete decisions are related to the design of the unit (i.e. total
number of trays, and the tray positions of each feed and product streams). A major
challenge is to perform the optimisation using tray-by-tray models that assume phase
equilibrium because of non-linearity of the thermodynamic models and the equations
of state used therein (Caballero & Grossmann 2014a ;Caballero & Grossmann
2014b; Gomez-Castro et al. 2011; Viswanathan & Grossmann 1993).

5.3 Optimisation of a distillation system

Section 2.7 presented a brief introduction to the definition of MINLP problems for
systems with both continuous and discrete variables. In this work, the system
consists of a distillation column divided in three sections as shown in Figure 5.1. The
first section is rectification. This part of the column comprises the upper trays of the
column, the condenser and reflux. The second section of the column is the feed
which consists of three fixed trays where the feed stream is fed into the column.
There is no restriction as to how many trays there should be in the feed section (i.e.
there could be higher than three).

However, a higher number of trays in the feed section could lead to unfeasible
results, especially regarding the feed location since it may not represent a physically
realistic solution. On the other hand, fewer trays in the feed section may result in
convergence issues during the solution of the optimisation problem, especially if a
concentration close to the azeotrope is specified. The third section is stripping where
the liquid passes through a partial reboiler to generate the vapour for the mass and
heat transfer within the upper stages (Caballero and Grossmann, 2014a).

116
Figure 5.1: Sections of the distillation column used in the rigorous tray-by-tray optimisation

(a) (b)
Figure 5.2: a) Rectification section with side reflux stream and b) stripping section with side
boilup stream

117
The multiple recycles in both the rectification and stripping sections will determine
the trays into which the reflux and the boilup streams are fed. The total number of
trays in the column is calculated as the summation of all the active trays in
rectification, stripping and feed. Figure 5.2 shows the configurations and variables
used in both rectification and stripping sections in finding the optimal number of
trays in the distillation column. The tray-by-tray method is a very useful method for
the evaluation of the optimal parameters of a distillation column given fixed
operating conditions for feed. However, the implementation of this methodology
entails longer CPU times than shortcut methods given the highly non-linear nature of
the system of equations (i.e. mass and energy balances, equation of state, activity
coefficient models, etc.).

The streams and the mole fractions presented in Figure 5.2 are described next:

 Lj : Liquid flow rate leaving tray j (mole/sec)

 xij : Liquid mole fraction of component i leaving tray j
 Tj : Outlet temperature tray j (K)
 Vj : Vapour flow rate leaving tray j (mole/sec)
 yij : Vapour mole fraction of component i leaving tray j
 αj : Binary variable used to determine the position of the reflux stream
 βj : Binary variable used to determine the position of the boilup stream

The objective function to be minimised is presented in Eq. 5.1. This function

comprises the capital investment which is the cost of the different units included in
the distillation system (e.g. cost of condenser, reboiler, trays and column shell)
within a payback period that may go between 3 and 10 years, depending on the
projections of production and earnings in order to recover the money invested in the
project (Peters et al., 2003).

𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝑖𝑛𝑣𝑒𝑠𝑡𝑚𝑒𝑛𝑡
𝑇𝐴𝐶 = ∑ 𝐴𝑛𝑛𝑢𝑎𝑙 𝑜𝑝𝑒𝑟𝑎𝑡𝑖𝑛𝑔 𝑐𝑜𝑠𝑡 + ∑ Eq. 5.1
𝑝𝑎𝑦𝑏𝑎𝑐𝑘 𝑝𝑒𝑟𝑖𝑜𝑑

118
However, some authors suggest that for separation units such as distillation a
payback period of four years is recommended (Caballero and Grossmann, 2014a;
Dünnebier and Pantelides, 1999; Koch et al., 2013; Luyben, 2006).

The objective function also features the annual operating cost which represents the
costs related to the utilisation of services such as steam, cooling water and electricity.
The objective function is the sum of all the different costs for operation and
equipment and the optimisation will find the minimum value of this function. The
solution of this problem also requires additional constraints to guarantee the viability
of the process in terms of purity and design. The constraints of the optimisation
problem considered in this work are shown in Equations 5.2 to 5.5 (Caballero and
Grossmann, 2014a; J Viswanathan and Grossmann, 1993; Viswanathan and
Grossmann, 1990). Figure 5.3 presents the two configurations considered in the
solution of the optimisation problem.

Eq. 5.2 presents the constraint related to the purity specifications at the top of the
distillation column in order to obtain a concentration of ethanol that meets the feed
requirements for the pervaporation system. Equations 5.3 and 5.4 ensure that the
reflux and the steam leaving the reboiler only enter into one tray.

(a) (b)

Figure 5.3: Configurations for the distillation system considered in this work: a) single
distillation column and b) double distillation system with mixed waste

119
 𝑆𝑝𝑒𝑐
𝑊𝐷,𝐸𝑡ℎ ≥ 𝑊𝐷,𝑒𝑡ℎ Eq. 5.2

Equations 5.3 and 5.4 refer to the constraints necessary to ensure that the recycles in
both sides of the column (e.g rectification and stripping) are fed into one single tray.

𝑁𝑡𝑟𝑎𝑦_𝑟𝑐

∑ 𝛼𝑗 = 1 𝑗 = 1, 2, … , 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑟𝑎𝑦𝑠 𝑖𝑛 𝑟𝑒𝑐𝑡𝑖𝑓𝑖𝑐𝑎𝑡𝑖𝑜𝑛 Eq. 5.3

𝑗=1

𝑁𝑡𝑟𝑎𝑦𝑠_𝑠𝑡

∑ 𝛽𝑘 = 1 𝑘 = 1, 2, … , 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑟𝑎𝑦𝑠 𝑖𝑛 𝑠𝑡𝑟𝑖𝑝𝑝𝑖𝑛𝑔 Eq. 5.4

𝑘=1

One the other hand, Equations 5.5 and 5.6 represent the number of active trays that
can be counted in both sections of the column and are used in the calculation of its
height.

𝑡𝑜𝑡𝑎𝑙
𝑁𝑅𝑒𝑐 = ∑ 𝛼𝑗 𝑅𝑐𝑡𝑗 𝑅𝑐𝑡𝑗 = {1, 2, 3 … 𝑁𝑡𝑟𝑎𝑠_𝑟𝑐 } Eq. 5.5
𝑗

𝑡𝑜𝑡𝑎𝑙
𝑁𝑆𝑡𝑟 = ∑ 𝛽𝑘 𝑆𝑡𝑟𝑘 𝑆𝑡𝑟𝑘 = {𝑁𝑡𝑟𝑎𝑦_𝑠𝑡 , 𝑁𝑡𝑟𝑎𝑦_𝑠𝑡−1 , … ,2, 1} Eq. 5.6
𝑘

The set of constraints that correlate the continuous variables of the system with the
discrete variables used to identify the recycle of both distillate and steam into the
column are defined as follows (Caballero and Grossmann, 2014a):

𝑅𝐹𝑇 = {𝑗 | 𝐶𝑎𝑛𝑑𝑖𝑑𝑎𝑡𝑒 𝑡𝑟𝑎𝑦 𝑜𝑓 𝑟𝑒𝑓𝑙𝑢𝑥 𝑟𝑒𝑡𝑢𝑟𝑛} Eq. 5.7

𝑅𝐵𝑇 = {𝑘 | 𝐶𝑎𝑛𝑑𝑖𝑑𝑎𝑡𝑒 𝑡𝑟𝑎𝑦 𝑜𝑓𝑟𝑒𝑏𝑜𝑖𝑙 𝑟𝑒𝑡𝑢𝑟𝑛} Eq. 5.8

120
Let LRx, LRb be the reflux and reboil flow rate returned to the column, respectively.
Let 𝛼𝑗 𝑗 ∈ 𝑅𝐹𝑇 ; 𝛽𝑘 𝑘 ∈ 𝑅𝐹𝑇 be binaries that take the value 1 if the reflux/reboil is
returned to tray j and k, respectively. Equations

𝑟𝑒𝑓𝑡𝑟 ≤ 𝐿𝑅𝑥 𝛼𝑗 𝑗 ∈ 𝑅𝐹𝑇 Eq. 5.9

𝑟𝑒𝑏𝑡𝑟 ≤ 𝐿𝑅𝑏 𝛽𝑘 𝑘 ∈ 𝑅𝐵𝑇 Eq. 5.10

𝑟𝑒𝑓𝑡𝑟 ≥ 0, 𝛼𝑗 ∈ {0,1} ∀𝑗 ∈ 𝑅𝐹𝑇 Eq. 5.11

𝑟𝑒𝑏𝑡𝑟 ≥ 0, 𝛽𝑘 ∈ {0,1} ∀𝑘 ∈ 𝑅𝐵𝑇 Eq. 5.12

Equations 5.13 and 5.14 correlate the vectors of recycles with the actual flow rate to
be fed into the column.

𝐿𝑅𝑥_𝑖𝑛 = ∑ 𝑟𝑒𝑓𝑡𝑟 Eq. 5.13

𝑗∈𝑅𝐹𝑇

𝐿𝑅𝑏_𝑖𝑛 = ∑ 𝑟𝑒𝑏𝑡𝑟 Eq. 5.14

𝑘∈𝑅𝐵𝑇

An additional constraint is introduced for the double-column system, which

represents the amount of ethanol lost during the separation as shown in Figure 5.3.b.

𝑚𝑖𝑛 𝑚𝑎𝑥
𝑤𝐿𝑜𝑠𝑠.𝐸𝑡ℎ ≤ 𝑤𝐿𝑜𝑠𝑠.𝐸𝑡ℎ ≤ 𝑤𝐿𝑜𝑠𝑠.𝐸𝑡ℎ Eq. 5.15

Several authors have presented the cost estimating formulae and values of the
parameter for the evaluation of the annual cost for the different parts of the
distillation column as well as the costs for utilities such as cooling water and steam
(Kiran and Jana, 2015). These equations are also presented in the works of Chung et
al. (2015), Kookos (2003), Olujić et al. (2006), Szitkai et al. (2002), Wei et al.
(2013), etc.

121
 𝑀&𝑆
𝐶𝑜𝑠𝑡 𝑐𝑜𝑙𝑢𝑚𝑛 𝑠ℎ𝑒𝑙𝑙 = ( ) 101.9 𝐷𝑐1.066 𝐿0.802
𝑐 (𝑐𝑖𝑛 + 𝑐𝑚 𝑐𝑝 )
280 Eq. 5.16
< 𝑈𝑆$ >

where Dc is the column diameter (ft), Lc the column height (ft), the M&S (the
Marshal & Swift index for 2015) is 1625.9 and the coefficients cin =2.18, cm =3.67
and co =1.2.

𝑀&𝑆
𝐶𝑜𝑠𝑡 𝑐𝑜𝑙𝑢𝑚𝑛 𝑡𝑟𝑎𝑦 = ( ) 4.7𝐷𝑐1.55 𝐿𝑐 (𝑐𝑠 + 𝑐𝑡 + 𝑐𝑚 ) < 𝑈𝑆$ > Eq. 5.17
280

where coefficients cs =1, ct =0 and cm =1.7

The cost function used for both the condenser and reboiler are presented in the works
of Luyben (2006).

𝐶𝑜𝑠𝑡 𝑐𝑜𝑛𝑑𝑒𝑛𝑠𝑒𝑟 = 7296 (𝐴𝑐𝑜𝑛𝑑 0.65 ) Eq. 5.18

𝐶𝑜𝑠𝑡 𝑟𝑒𝑏𝑜𝑖𝑙𝑒𝑟 = 7296 (𝐴𝑟𝑒𝑏𝑜𝑖𝑙𝑒𝑟 0.65 ) Eq. 5.19

The area for both condenser and reboiler can be determined as follows:

𝑄𝑡𝑟𝑎𝑛𝑠 = 𝑈 𝐴 𝐿𝑀𝑇𝐷 Eq. 5.20

Log Mean Temperature Difference (LMTD) is calculated for counter-current heat

exchanging processes. The average overall heat transfer coefficients (U) for the
condenser and reboiler are 852 and 568 𝑊 ⁄𝐾 𝑚2 , respectively, (Luyben, 2006).
These overall heat transfer coefficients are suggested based on the materials normally
used in condensers and reboilers and the types of fluids used therein. The costs of
services such as cooling water, low pressure steam (≤ 6 bar), high pressure steam
(≥10 bar) and electricity for an annual operating period of 8000 hours, are 0.16
US$/GJ, 7.78 US$/GJ, 9.98 US$/GJ and 16.8 US$/GJ, respectively (Caballero, 2015;
Wang et al., 2015).
122
5.3.1 Specifications of the distillation columns
This section presents the specifications used in the optimisation of the distillation
column system including: feed flow rate, feed concentration, concentration of
distillate, number of trays in both rectification and stripping, pressure and, in the case
of a double-column system, the losses of ethanol in the waste stream (see Figure
5.3.b).

The optimisation of a single distillation column and an arrangement of two columns

are carried out in order to determine which configuration has the minimum TAC (see
Eq. 5.1). The MINLP problems are solved using gPROMS v.4.1.0 on a Dell
OptiPlex 9010 Intel® Core™ i7-3770 CPU @3.40 GHz and 16 GB RAM running
Windows 7® Enterprise (64-bit operating system). The tool known as gOPT allows
solving optimisation problems using a dynamic approach. This tool includes several
solvers whose effectiveness are dependent on the nature of the system of equations
(i.e. the complexity in terms the number of variables, constraints, application, etc.).

The solver used in this work for the integration of the algebraic and ordinary
differential equations is called DASOLV and uses a Backward Finite Difference
(BFD) formulation. gPROMS uses an Outer Approximation (OA) algorithm for the
solution of the MINLP problem and the steps it follows are shown in Figure 5.4. This
algorithm is implemented in the optimisation of both the distillation systems as well
as the pervaporation network (Process System Enterprises, 2015). The input
variables are shown in Table 5.1.

123
Figure 5.4: Outer Approximation (OA) algorithm for the solution of the MINLP problem using
gPROMS (Process System Enterprises, 2015)

Table 5.1: Input variables used in the optimisation of the distillation systems
Feature Value
Feed flow rate (mole/sec) 50
Conc. ethanol feed (% w/w) 5
Pressure (kPa) 101.3
Feed temperature (oC) Saturated liquid at specified pressure

124
This work will consider the separation of 50 mole/sec (or 3344.63 kg/h) of a solution
5% (w/w) of ethanol. The concentration of ethanol in the feed stream has been set at
5 % (w/w) since this is the average concentration that is expected to be obtained
during the fermentation of lignocellulosic hydrolysates using the genetically
modified bacterium Zymomonas mobilis ZM4 (pZB5). Pressure inside the column has
been set at atmospheric conditions in order to avoid the implementation of additional
devices such as pumps since the feed stream comes out directly from the fermenter.
The feed temperature corresponds to the temperature of saturated liquid at the
operating pressure in the column in order to guarantee energy efficiency within the
system (Kunnakorn et al., 2013).

5.3.2 Results of the optimisation of the distillation system

This section presents the results of the optimisation of the different units included in
the separation stage of the ethanol production process from corn stover. These
calculations are carried out using the tool in gPROMS known as gOPT. As
previously mentioned, this chapter seeks to obtain the solution of each unit
individually and later use these results as initial estimates for the solution of the
overall problem. This methodology is a practical way of reducing the number of
calculations and therefore reducing computational times since the system of
equations featured in this problem possess a highly non-linear formulation. Appendix
A presents the equations used in the formulation of a distillation column. Table 5.2
presents the initial guesses used in the optimisation of the distillation systems as well
as the bounds of the optimisation variables considered in the solution of the MINLP
problem.

125
 Table 5.2: Initial guesses and bounds used in the optimisation of the distillation systems
Two columns
Single Lower Upper
Parameter Column Column
column bound bound
1 2
Number of trays in rectification 25 20 20 1 25

Number of trays in stripping 25 20 20 1 25

Reflux ratio 2 1 3 0.01 100

Mass fraction distillate (𝑊𝐷,𝐸𝑡ℎ ) 0.93 0.93* 0.93 1.0

Waste (𝒘𝑳𝒐𝒔𝒔.𝑬𝒕𝒉 ) % (W/W) N/A 7 5 10

* Refers to the concentration of ethanol in the distillate stream of the second column

Table 5.3: Results of the optimisation of the distillation systems

Two columns
Parameter Single column
Column 1 Column 2

Number of trays rectification 19 13 11

Number of trays stripping 12 15 1*

Total number of trays 34 31 15

Feed tray 21 15 13

Diameter column (m) 0.43 0.36 0.38

Height column (m) 20.7 18.9 9.15

Reflux ratio 3.81 1.29 2.89

Heat duty reboiler (kW) 224.6 217.9 160.6

Mass fraction distillate 0.93* 0.64 0.93*

Cost column shell (US$/yr. 104) 4.15 3.48 2.11

Cost trays (US$/yr. 103) 1.07 0.81 0.44

Cost reboiler (US$/yr. 104) 1.25 1.22 1.01

Cost condenser (US$/yr. 103) 7.03 7.10 5.71

Cost steam (US$/yr. 104) 5.51 5.35 3.94

Cost cooling water (US$/yr. 103) 1.06 1.08 0.77

CPU time (sec) 8912 20,852

Payback period 4 years

TAC (US$/yr. 105) 1.18 1.87

* Value on bound
126
Table 5.3 and Figures 5.5 and 5.6 show the results of the optimisation of the
distillation column systems shown in Figure 5.3. The results suggest that the
implementation of a single distillation column in the separation of ethanol from the
fermentation broth can be more economical than having a double-column system for
the same distillate concentration. This conclusion is also supported by the amount of
energy required in the reboiler for both systems with heat requirements in the
reboiler of 224.6 kW for a single column and 378.5 kW in total for the double-
column system (savings of up to 41 %). In terms of economics, the configuration
shown in Figure 5.6 reported a TAC of 1.87x105 US$/year which is 42 % more
expensive than the single distillation column.

In terms of the design, the number of trays in the single column is related to the
concentration in the distillate stream. In binary mixtures that present azeotropes, the
number of trays increase as the concentration in the distillate gets closer to the
azeotropic point. The heat duty at the reboiler also increases as the concentration of
ethanol increases, which directly correlates to the operating costs. The results of the
optimisation of the double-column system show that the energy requirement in the
first column accounts only for 97 % of the energy used in the reboiler of the single
column, which makes the installation of the second column less practical from the
point of view of the usage of utilities and controllability of additional units.

5.4 Optimisation of the pervaporation system using a PAN-B5 hydrophilic

membrane
This section presents the optimisation of the membrane network for the
pervaporation of the distillate stream from the distillation column chosen in the
previous section. Similar to the case of the distillation column systems, the
evaluation of the optimal arrangement of pervaporation modules will be based on the
minimum TAC using the algorithm presented in Figure 5.4. For the subsequent
analysis and optimisation of the pervaporation network, the values for feed flow rate,
concentration of ethanol, feed temperature and feed pressure, will be taken from the
values obtained in the distillate stream of the single distillation column configuration
shown in Table 5.1 and Figure 5.5.

127
 Figure 5.5: Streams and concentrations in a single distillation column using the results of the
optimisation presented in Table 5.3

Figure 5.6: Streams and concentrations in a double-column distillation system using the results
of the optimisation presented in Table 5.3
128
 Figure 5.7: Pervaporation network design using PAN-B5 hydrophilic membrane

Figure 5.7 presents the general configuration of the membrane network. The
methodology to find the optimal configuration consists of determining the optimal
number of in-parallel modules and operating conditions for one stage. Then, the
optimal conditions for two in-series stages is determined, then three stages, etc. The
configuration with the minimum TAC is the configuration to be used for the overall
optimisation along with the distillation column (Marriott and Sørensen, 2003;
Marriott et al., 2001).

In industrial pervaporation processes, due to the evaporation of the components in

the feed mixture and the permeation through the membrane into the permeate side,
the temperature in the retentate side decreases (Baker, 2012). Since the flux has a
high dependence on the feed temperature, operating temperatures within the system
should be as high as possible, and temperature drops should be limited to a
minimum. Hence, intermediate heating is required in the system to ensure that the
feed temperature for all the stages is the same. This requirement has to be balanced
against the additional costs for a larger number of sub-divisions of membrane
compartments and for the heat exchangers (Baeyens et al., 2015; Kunnakorn et al.,
2013; Sander and Soukup, 1988; Tusel and Brüschke, 1985). For the optimisation of

129
the pervaporation network, an objective function similar to Eq. 5.1 is required. This
equation needs to have costs related to the size of the membrane, the installation of
the modules, compression and heat exchangers:

𝑀&𝑆
𝐶𝑜𝑠𝑡 𝐶𝑜𝑚𝑝𝑟𝑒𝑠𝑠𝑜𝑟 = ( ) 517.9 𝐵𝐻𝑃0.82 3.11 < 𝑈𝑆$ > Eq. 5.21
280

BHP, which is also known as Brake horsepower, is the work done by the compressor
over the vapour and is defined as:

ℎ𝑝
𝐵𝐻𝑃 = Eq. 5.22
𝜂

Where hp is the work done by an isothermal compressor and η is the efficiency of the
compressor assumed to be 0.95 (Marriott and Sorensen, 2003).

𝐹𝑖𝑛 𝑃𝑖𝑛 𝑃𝑜𝑢𝑡

ℎ𝑝 = 𝑉
𝑙𝑛 ( ) Eq. 5.23
𝜌 𝜂 𝑃𝑖𝑛

The following constraints are required for the optimisation of the pervaporation
network. Eq. 5.16 is the temperature difference allowed after the permeation of the
mix through the membrane. This temperature difference constraint is only applicable
for two pervaporation stages or more. For a single stage, this constraint does not
apply since the flux through the membrane, in order to meet the desire specifications,
is too high and the temperature will plummet beyond a difference of 20 K due to the
evaporation heat.

𝑚𝑖𝑛 𝑚𝑎𝑥
∆𝑇𝑓𝑒𝑒𝑑,𝑟𝑒𝑡𝑒𝑛𝑡 ≤ ∆𝑇𝑓𝑒𝑒𝑑,𝑟𝑒𝑡𝑒𝑛𝑡 ≤ ∆𝑇𝑓𝑒𝑒𝑑,𝑟𝑒𝑡𝑒𝑛𝑡 Eq. 5.24

Eq. 5.17 is the constraint related to the purity of the product in the retentate stream
which needs to meet biofuel standards (Drapcho et al., 2008; Palmarola-Adrados et
al., 2005; Sánchez and Cardona, 2012):

130
 𝑚𝑖𝑛 𝑚𝑎𝑥
𝑊𝑟𝑒𝑡𝑒𝑛𝑡,𝑒𝑡ℎ ≤ 𝑊𝑟𝑒𝑡𝑒𝑛𝑡,𝑒𝑡ℎ ≤ 𝑊𝑟𝑒𝑡𝑒𝑛𝑡,𝑒𝑡ℎ Eq. 5.25

Eq. 5.18 is the permeate pressure range at which the PAN-B5 hydrophilic membrane
model is valid. This is presented in the experimental work developed by Tsuyumoto
et al., (1995), (1997):

𝑚𝑖𝑛 𝑚𝑎𝑥
𝑃𝑝𝑒𝑟𝑚 ≤ 𝑃𝑝𝑒𝑟𝑚 ≤ 𝑃𝑝𝑒𝑟𝑚 Eq. 5.26

The cost of the PAN-B5 hydrophilic membrane is 217.4 US$/m2, the cost of each
membrane module is 140.2 US$/m2 and their replacement takes place every four
years (Kiran and Jana, 2015).

5.4.1 Specifications of the pervaporation network using a hydrophilic

membrane
The flow rate obtained in the distillate of the distillation column shown in Figure 5.5
will be used as the feed flow rate for the pervaporation network presented in Figure
5.7. The distillate stream is initially cooled down to operating feed temperature of the
membrane and the passed through the pervaporation stages in order to remove water
from the mixture.

Table 5.4: Input variables used in the optimisation of the pervaporation system using the PAN-
B5 hydrophilic membrane
Feature Value
Feed flow rate (mole/sec) Distillate flow rate
Conc. ethanol feed (% w/w) 93
Pressure (kPa) 101.3
Feed temperature (oC) Saturated liquid at specified pressure

131
 Table 5.5: Initial guesses and bounds used in the optimisation of the PAN-B5 hydrophilic
membrane

Lower Upper
Parameter Stage 1 Stage 2 Stage 3 Stage 4 Stage 5
bound bound

Number of in-parallel modules 1 1 1 1 1 1 25

Permeate pressure (Pa) 4000 2000 400 400 400 10 10,000

Conc. retentate (𝑊𝑟𝑒𝑡𝑒𝑛𝑡,𝐸𝑡ℎ ) 0.993 0.993 0.993 0.993 0.993 0.993 0.998

∆𝑇𝑓𝑒𝑒𝑑,𝑟𝑒𝑡𝑒𝑛𝑡 20 20 20 20 20 10 20

Table 5.4 presents the input variables used in the optimisation of the pervaporation
system using the PAN-B5 hydrophilic membrane. Table 5.5 presents the initial
guesses and boundaries used in the solution of the MINLP problem.

5.4.2 Results of the optimisation of the pervaporation network using a

hydrophilic membrane
Figure 5.8 presents the results of the optimisation of the pervaporation network for
different in-series membrane stages. Figure 5.8 shows a decrease of the TAC as the
number of in-series stages increases and after three stages, the TAC increases for
four and five stages meaning that the optimal configuration for the pervaporation
network may have three stages. However, for this particular case study, the curves
for three and four stages present a narrow gap, and in some cases they overlap
making this comparison more difficult. Therefore, another criterion should be taken
into consideration in order to determine the optimal number of pervaporation stages
and their membrane area.

Figure 5.9 shows a trade-off between the TAC and energy consumption for the
arrangements of three and four in-series pervaporation stages in order to find an
optimal configuration that can be considered in the optimisation of the whole
separation section. For membrane areas between 100 and 150 m2, the energy
consumption is high for both arrangements and then decreases until 400 m2.

Table 5.6 presents the results of the optimisation of the different arrangements for the
pervaporation network in the dehydration of ethanol. It can be observed that the total
132
number of in-parallel modules decreases as the area increases. For membrane areas
larger than 250 m2, the number of stages remains constant. This suggests that
configurations with larger membrane areas are economically inviable as they only
represent an increment in the capital and operating costs with similar performance as
those configurations with smaller membrane areas. For instance, for four and five
stages, the number of models remains constant after a membrane area of 250 m2. The
only difference between these arrangements is the permeate pressure which increases
as the membrane area increases.

However, the partial fluxes and the separation rates of ethanol through the permeate
side in these configurations with larger areas do not differ from those with smaller
areas. Moreover, for larger areas in the configuration consisting of five stages, a
permeate pressure higher than 10,000 Pa is required in order to find an optimal (see
Eq. 5.16 for the boundaries of permeate pressure used in the membrane module).
These results are marked with an asterisk in Table 5.6.

3.5x105

3.0x105

2.5x105
TAC (US$/yr.)

2.0x105

1.5x105
1 stage
2 stages
1.0x105 3 stages
4 stages
5 stages
5.0x104
100 200 300 400 500
2
Membrane area (m )

Figure 5.8: Optimisation of the hydrophilic pervaporation system at different membrane areas
for the system consisting of a single distillation column shown in Figure 5.5

133
 2.0x105 9.8
9.6
1.8x105
9.4

Energy consumption (kW)

5
TAC (US$/yr.) 1.6x10 9.2

1.4x105 9.0
8.8
5
1.2x10 3 stages (left)
8.6
4 stages (left)
3 stages (right)
1.0x105 4 stages (right)
8.4
8.2
8.0x104
8.0
100 150 200 250 300 350 400 450 500
Membrane area (m2)
Figure 5.9: Comparison of total annualised cost vs. energy consumption for three and four in-
series pervaporation stages shown in Figure 5.8

For four stages, for instance, the energy consumption starts to increase along with the
TAC from 400 m2 onwards. This supports the conclusion previously stated that
having more than three stages with larger areas is inefficient not only in terms of
separations rates but also in terms of energy requirements. This energy assessment
also shows that for three stages, with areas between 150 and 300 m2, both the TAC
and the energy consumption are lower than for four stages. It can be concluded based
on this analysis that the optimal configuration for the dehydration of ethanol using a
hydrophilic membrane should consist of three stages with a membrane area of 250
m2 .

134
 Table 5.6: Optimal parameters for a minimum TAC in the PAN-B5 hydrophilic membrane
system
Membrane Area (m2)
Stage Parameters
100 150 200 250 300 350 400 500
One Stage
TAC (US$/yr.) 254,470 266,503 262,350 258,307 279,299 292,345 296,724 279,299
Number of modules 27 19 14 11 10 9 8 6
1 Permeate pressure
57 138 112 85 211 278 298 211
(Pa)
Two Stages
TAC (US$/yr.) 155,013 163,849 164,090 181,613 190,517 221,635 217,225 226,200
Number of modules 1 1 1 1 1 1 1 1
1 Permeate pressure
3324 5139 5972 6495 6880 7190 7456 7915
(Pa)
Number of modules 15 10 8 6 6 5 4 4
2 Permeate pressure
59 46 133 21 276 227 86 374
(Pa)
Three Stages
TAC (US$/yr.) 74,898 83,737 110,095 115,339 121,720 128,102 145,958 193,282
Number of modules 1 1 1 1 1 1 1 1
1 Permeate pressure
3324 5139 5972 6495 6880 7190 7456 7915
(Pa)
Number of modules 2 2 1 1 1 1 1 1
2 Permeate pressure
74 1134 480 243 1232 1880 2369 3122
(Pa)
Number of modules 5 3 3 2 2 2 3 3
3 Permeate pressure
74 71 251 243 415 538 1073 1215
(Pa)
Underlined values in bold and italic font represent the optimal solution of the pervaporation network

135
 Table 5.6: Optimal parameters for a minimum TAC in the PAN-B5 hydrophilic membrane
system (cont.)
Membrane Area (m2)
Stage Parameters
100 150 200 250 300 350 400 500
Four Stages
TAC (US$/yr.) 73,077 94,715 116,354 120,948 125,541 134,469 152,206 188,999
Number of modules 1 1 1 1 1 1 1 1
1 Permeate pressure
3324 5365 5972 8044 8059 8064 8060 8500
(Pa)
Number of modules 1 1 1 1 1 1 1 1
2 Permeate pressure
325 316 1097 2360 2670 2880 3032 6000
(Pa)
Number of modules 3 2 3 1 1 1 1 1
3 Permeate pressure
325 316 1097 437 754 991 1187 800
(Pa)
Number of modules 2 1 1 1 1 1 1 1
4 Permeate pressure
325 316 1097 437 754 991 1187 800
(Pa)
Five Stages
TAC (US$/yr.) 81,436 103,279 134,778 139,321 143,863 165,940 188,166 232,798
Number of modules 1 1 2 1 1 1 1 1
1 Permeate pressure
8079 8845 10,702 9735 10,102* 10,441* 10,770* 11,427*
(Pa)
Number of modules 1 1 1 1 1 1 1 1
2 Permeate pressure
3720 4848 7376 7726 8178 8562 8934 9063
(Pa)
Number of modules 1 1 1 1 1 1 1 1
3 Permeate pressure
494 800 1125 3602 3046 3085 3111 3242
(Pa)
Number of modules 3 3 2 1 1 1 1 1
4 Permeate pressure
494 800 814 973 762 1000 1196 1524
(Pa)
Number of modules 2 1 1 1 1 1 1 1
5 Permeate pressure
494 800 814 17 762 1000 1196 1524
(Pa)
* Indicates values of permeate pressure out of the boundaries of the mathematical model for the
membrane the PAN-B5 developed by Tsuyumoto et al., (1997)

136
 Figure 5.10: Results of the optimisation of the pervaporation network using PAN-B5
hydrophilic membrane and a membrane area of 250 m2 for the distillation system shown in
Figure 5.5

The results of the optimisation and simulation of the configuration illustrated in

Figure 5.10 show that a 98 % of the ethanol fed into the system is recovered in the
retentate stream from the last stage. In terms of operating conditions, the results show
a temperature drop for the first and the second stages of almost 20 K which is
necessary in order to remove as much water as possible from the mix.

This optimisation problem is mainly influenced by the concentration at the end of the
pervaporation plant. If concentrations of ethanol higher than 99.8 % (w/w) were
fixed, the number of in-parallel modules and the total membrane area would
increase. The operating costs of the plant would increase as well since lower
permeate pressures would be required in order to increase the concentration of
ethanol in the retentate, which will impact on the consumption of energy in the
compressor and the amount of steam to provide heat to the intermediate heat
exchangers.

137
 5.5 Optimisation of the pervaporation system using a PERVAPTM 4060
organophilic membrane
Chapter 4 presented the formulation of a mathematical model, based on
experimentation, of an organophilic membrane for the removal of ethanol from the
fermentation broth. This mathematical model (see Eq. 4.5) describes the permeation
of ethanol through the membrane into the permeate side for a specific range of
operating conditions (see Table 4.1).

As explained in Chapter 4, the reason to use the organophilic membrane is to reduce

the energy consumption in the distillation column as well as the size of some units in
the plant by reducing the feed flow rate into the column and increasing the
concentration of ethanol. The pervaporation network with the PERVAPTM 4060
organophilic membrane will be optimised in the same fashion as the PAN-B5
hydrophilic membrane shown in Figure 5.7.

Unlike the pervaporation network using the hydrophilic membrane, the removal of
ethanol using PERVAPTM 4060 will require an additional heat exchanger that will
heat the permeate stream to a temperature corresponding to the dew point of the
mixture at atmospheric conditions (i.e. 101.3 kPa). Once the mixture has reached the
desired temperature, the vapour is passed through a compressor to increase the
pressure from the vacuum pressure of the pervaporation module to atmospheric
conditions before entering the distillation column as saturated vapour to reduce heat
losses and improve energy efficiency. Figure 5.11 illustrates this configuration.

The feed conditions are 50 mole/sec (or 3344.63 kg/h) of a solution 5% (w/w) of
ethanol which is the concentration expected to be obtained at the end of the
fermentation stage at 50 oC. The cost of the PERVAPTM 4060 organophilic
membrane is 214.53 US$/m2 and its replacement takes place every three years
(Sulzer Chemtech, 2014). The cost of the each membrane module is 140.2 US$/m2
(Gaykawad et al., 2013; Kiran and Jana, 2015).

138
 Figure 5.11: Pervaporation network design using PERVAPTM 4060 organophilic membrane

Similar to the optimisation of the hydrophilic membrane, a set of constraints need to

be included in the formulation of the MINLP problem. Eq. 5.19 can be applied for
two or more pervaporation stages, unlike a single stage in which the evaporation heat
transferred from the feed side to the permeate side can cause a sharp decrease in the
retentate temperature lower than these limits. The following are the constraints for
the optimisation of the pervaporation network using an organophilic membrane:

𝑚𝑖𝑛 𝑚𝑎𝑥
∆𝑇𝑓𝑒𝑒𝑑,𝑟𝑒𝑡𝑒𝑛𝑡 ≤ ∆𝑇𝑓𝑒𝑒𝑑,𝑟𝑒𝑡𝑒𝑛𝑡 ≤ ∆𝑇𝑓𝑒𝑒𝑑,𝑟𝑒𝑡𝑒𝑛𝑡 Eq. 5.27

Eq. 5.20 is presents the desired concentration of ethanol in the retentate stream. In
other words, all the ethanol present in the feed stream is expected to permeate and be
fed to the distillation system.

𝑚𝑖𝑛 𝑚𝑎𝑥
𝑊𝑟𝑒𝑡𝑒𝑛𝑡,𝑒𝑡ℎ ≤ 𝑊𝑟𝑒𝑡𝑒𝑛𝑡,𝑒𝑡ℎ ≤ 𝑊𝑟𝑒𝑡𝑒𝑛𝑡,𝑒𝑡ℎ Eq. 5.28

139
Table 5.7: Input variables used in the optimisation of the pervaporation system using the PAN-
B5 hydrophilic membrane
Feature Value
Feed flow rate (mole/sec) 50
Conc. ethanol feed (% w/w) 5.0
Pressure (kPa) 101.3
Feed temperature (oC) 50

Table 5.8: Initial guesses and bounds used in the optimisation of the PAN-B5 hydrophilic
membrane

Lower Upper
Parameter Stage 1 Stage 2 Stage 3
bound bound

Number of in-parallel modules 1 1 1 1 25

Permeate pressure (Pa) 3500 2000 400 400 3500

Conc. retentate (𝑊𝑟𝑒𝑡𝑒𝑛𝑡,𝐸𝑡ℎ ) 0.001 0.001 0.001 0.0001 0.001

∆𝑇𝑓𝑒𝑒𝑑,𝑟𝑒𝑡𝑒𝑛𝑡 20 20 20 10 20

Eq. 5.21 is the range of permeate pressure used in the formulation of the
mathematical model of the membrane (see Chapter 4, sections 4.4 and 4.5).

𝑚𝑖𝑛 𝑚𝑎𝑥
𝑃𝑝𝑒𝑟𝑚 ≤ 𝑃𝑝𝑒𝑟𝑚 ≤ 𝑃𝑝𝑒𝑟𝑚 Eq. 5.29

5.5.1 Specifications for the pervaporation network using an organophilic

membrane
Following a similar approach as in the specifications for the hydrophilic membrane,
Table 5.7 presents the input variables used in the optimisation of the pervaporation
system using the organophilic membrane. The feed flow rate and the feed
concentration are set as in the distillation systems (see Table 5.3), which are expected
to be the outlet stream conditions from the fermentation stage. Table 5.8 presents the
initial guesses and the bounds used in the solution of the optimisation problem.

140
5.5.2 Results of the optimisation of the pervaporation network using an
organophilic membrane
Table 5.9 and Figure 5.12 present the results of the optimisation of the pervaporation
network using the organophilic membrane. The results show that the TAC increases
with the number of stages in the pervaporation network. Similar to the case of the
hydrophilic membrane, for larger areas, higher permeate pressures are required. For
one, two and three stages, and membrane areas of 200 m2, the optimal permeate
pressures are above 3500 Pa which rules out any implementation of these
configurations into the main process since the operating conditions are out of the
bounds in which the mathematical model of the membrane was formulated.

The TAC for all the cases decreases at 200 m2 which can be attributed to the
reduction in the energy consumption in the compressor that increases the pressure to
atmospheric conditions. Also notice that for all the cases, the permeate pressures are
higher than 1000 Pa and as seen in Section 4.5, the higher the permeate pressure the
lower the separation rates in the pervaporation sections.

The results of permeate pressure also suggest that for two and three stages and
membrane areas of 50 m2 and above, the flux rates in the first two stages are
expected to be lower (as shown in Table 4.4, the partial fluxes decrease as permeate
pressure increases) which basically means that the last stage would do all the work in
the pervaporation plant. This trend in the partial fluxes through the membrane
supports the implementation of just one stage with multiple in-parallel modules since
it is more practical from the point of view of controllability, size of the units,
intermediate heating and maintenance. Hence, for the optimisation of the overall
process, a membrane area of 20 m2 and one pervaporation stage will be considered.
Figure 5.12 shows the results of the optimisation of the pervaporation network using
the PERVAPTM organophilic membrane for the removal of ethanol from the
fermentation broth.

141
 Table 5.9: Optimal parameters for a minimum TAC in the PERVAP 4060TM organophilic
membrane system
Membrane area (m2)
Stage Parameters
20 50 80 100 150 180 200
One Stage
TAC (US$/yr.) 206,516 231,248 181,064 197,388 177,431 181,831 197,388
Number of modules 19 7 5 4 4 4 2
1
Permeate pressure (Pa) 3431 1988 6959* 4223* 8579* 9829* 4223*
Two Stages
TAC (US$/yr.) 231,584 236,389 223,821 250,206 229,312 225,983 226,133
Number of modules 3 1 2 1 1 1 1
1
Permeate pressure (Pa) 3500 3500 8329* 3500 7891* 9063* 10,000*
Number of modules 19 8 7 4 4 3 3
2 * * *
Permeate pressure (Pa) 3758 4453 8011* 4652 8526* 7713* 8509*
Three Stages
TAC (US$/yr.) 257,363 261,168 267,632 293,957 276,321 273,654 270,952
Number of modules 3 1 1 1 1 1 1
1
Permeate pressure (Pa) 3500 3500 8169* 3500 10,000* 12,109* 10,000*
Number of modules 3 1 1 1 1 1 1
2
Permeate pressure (Pa) 5828* 3500 8022* 3500 10,000* 11,740* 10,000*
Number of modules 18 7 7 4 4 2 3
3
Permeate pressure (Pa) 3164 2669 8028* 4975* 8493* 3078* 8455*
* Indicates permeate pressure out of the boundaries of the empirical model
Underlined values in bold and italic font represent the optimal solution of the pervaporation network

142
 Figure 5.12: Results of the optimisation of the pervaporation network using PERVAPTM 4060
organophilic membrane and a membrane area of 20 m2 and 19 in-parallel modules

Table 5.10: Input variables used in the optimisation of the distillation column linked to the
organophilic membrane system
Feature Value
Permeate flow rate leaving the
Feed flow rate (mole/sec)
compressor (see Figure 5.12)
Conc. of permeate leaving
Conc. ethanol feed (% w/w)
the pervaporation module
Pressure (kPa) 101.32
Saturated vapour at
Feed temperature (oC)
specified pressure

Table 5.10 presents the list of input variables used in the solution of the optimisation
problem of the distillation column using the permeate stream from the PERVAPTM
4060 organophilic membrane system. Table 5.11 presents the results optimisation of
the distillation column using the outlet stream of the pervaporation stage with the
organophilic membrane. These results show a noticeable reduction in the energy
consumption in comparison with the results shown in Table 5.3 and this is due to the
conditions at which the feed stream is fed (saturated vapour). The size of the trays
and the shell of the column are also smaller than the single column in Figure 5.5.
143
Table 5.11: Results of the optimisation of the distillation systems using the permeate stream of
the pervaporation system shown in Figure 5.15 as feed

Single
Parameter
column

Number of trays rectification 21

Number of trays stripping 1*

Total number of trays 25

Feed tray 23

Diameter column (m) 0.35

Height column (m) 15.3

Reflux ratio 2.08

Heat duty reboiler (kW) 144.1

Mass fraction distillate 0.93*

Cost column shell (US$/yr. 104) 2.59

Cost trays (US$/yr. 102) 5.67

Cost reboiler (US$/yr. 103) 9.37

Cost condenser (US$/yr. 103) 5.19

Cost steam (US$/yr. 104) 3.54

Cost cooling water (US$/yr. 102) 6.68

CPU time (sec) 29,182

Payback period 4 years

TAC (US$/yr. 104) 7.71

* Value on bound

144
 Figure 5.13: Results of the optimisation of the distillation column under the specifications
presented in Table 5.10

3.5x105

3.0x105

2.5x105
TAC (US$/yr.)

2.0x105

1.5x105
1 stage
1.0x105 2 stages
3 stages
4 stages
5 stages
5.0x104
100 200 300 400 500
Membrane area (m2)

Figure 5.14: Optimisation of the hydrophilic pervaporation system at different membrane areas
for the system consisting of a single distillation column shown in Figure 5.13

Figure 5.14 presents the results optimisation of the pervaporation section using the
distillate stream of the distillation column shown in Figure 5.13. A similar trend as in

145
Figure 5.8 can be observed in Figure 5.14. The TAC decreases for the first three
stages and starts to increase after four stages. However, in this case, an energy
analysis is not necessary since there is a clear distinction between the curve that
represents three stages and the one for four stages.

Similarly to the case shown in Figure 5.8, for five stages and membrane areas larger
than 350 m2, the permeate pressure is higher than the upper bound in Eq. 5.16 which
means that these configurations are not viable for implementation into the main
process since the mathematical model has not validity beyond 10,000 Pa.

A minimum TAC is also obtained for a configuration consisting of three stages and a
membrane area of 150 m2 (see Table 5.14). This configuration will serve as initial
estimate for the solution of the optimisation of the overall separation process. Figure
5.15 shows the simulation of the optimal configuration which consists of three in-
series stages of pervaporation and their respective operating conditions.

Table 5.12: Initial guesses and bounds used in the optimisation of the PAN-B5 hydrophilic
membrane in the configuration with the organophilic membrane

Lower Upper
Parameter Stage 1 Stage 2 Stage 3
bound bound

Number of in-parallel modules 1 1 1 1 25

Permeate pressure (Pa) 3500 2000 400 400 3500

Conc. retentate (𝑾𝒓𝒆𝒕𝒆𝒏𝒕,𝑬𝒕𝒉 ) 0.001 0.001 0.001 0.0001 0.001

∆𝑻𝒇𝒆𝒆𝒅,𝒓𝒆𝒕𝒆𝒏𝒕 20 20 20 10 20

146
 Table 5.13: Initial guesses and bounds used in the optimisation of the PAN-B5 hydrophilic
membrane

Lower Upper
Parameter Stage 1 Stage 2 Stage 3 Stage 4 Stage 5
bound bound

Number of in-parallel modules 1 1 1 1 1 1 25

Permeate pressure (Pa) 4000 2000 400 400 400 10 10,000

Conc. retentate (𝑾𝒓𝒆𝒕𝒆𝒏𝒕,𝑬𝒕𝒉 ) 0.993 0.993 0.993 0.993 0.993 0.993 0.998

∆𝑻𝒇𝒆𝒆𝒅,𝒓𝒆𝒕𝒆𝒏𝒕 20 20 20 20 20 10 20

Table 5.14: Optimal parameters for a minimum TAC in the pervaporation network using the
distillate stream of the single distillation column shown in Figure 5.13
Membrane Area (m2)
Stage Parameters
100 150 200 250 300 350 400 500
One Stage
TAC (US$/yr.) 264,961 270,127 264,961 280,459 280,459 264,961 264,961 280,459
Number of modules 28 15 14 12 10 8 7 6
1
Permeate pressure (Pa) 32 10 32 137 137 32 32 137
Two Stages
TAC (US$/yr.) 159,234 173,387 174,572 190,685 199,444 199,922 225,689 234,805
Number of modules 1 1 1 1 1 1 1 1
1
Permeate pressure (Pa) 2984 4958 5836 6379 6772 7085 7351 7804
Number of modules 15 11 8 7 6 5 5 4
2
Permeate pressure (Pa) 10 115 56 179 208 157 330 311
Three Stages
TAC (US$/yr.) 87,755 86,200 93,478 95,972 110,833 120,788 146,919 183,405
Number of modules 1 1 1 1 1 1 1 1
1
Permeate pressure (Pa) 2984 4958 5836 6379 6772 7085 7351 7804
Number of modules 1 2 1 1 1 1 1 1
2
Permeate pressure (Pa) 75 649 179 137 978 1672 2182 2948
Number of modules 8 3 3 2 2 2 2 2
3
Permeate pressure (Pa) 56 80 179 137 371 498 597 746
Underlined values in bold and italic font represent the optimal solution of the pervaporation network

147
 Table 5.14: Optimal parameters for a minimum TAC in the pervaporation network using the
distillate stream of the single distillation column shown in Figure 5.13 (cont.)
Membrane Area (m2)
Stage Parameters
100 150 200 250 300 350 400 500
Four Stages
TAC (US$/yr.) 84,926 92,058 99,983 100,545 117,462 134,980 152,678 188,385
Number of modules 1 1 1 1 1 1 1 1
1
Permeate pressure (Pa) 2995 8643 6686 8038 8056 8063 8063 8085
Number of modules 1 1 1 1 1 1 1 1
2
Permeate pressure (Pa) 511 1656 3113 2255 2589 2814 2976 3255
Number of modules 2 1 1 1 1 1 1 1
3
Permeate pressure (Pa) 10 214 448 334 668 913 1111 1411
Number of modules 3 3 2 1 1 1 1 1
4
Permeate pressure (Pa) 401 214 448 334 668 913 1111 1365
Five Stages
TAC (US$/yr.) 86,657 95,430 105,551 122,670 144,097 166,111 188,310 232,917
Number of modules 1 1 1 1 1 1 1 1
1
Permeate pressure (Pa) 5827 8059 9215 9647 9995 10,322* 10,636* 11,259
Number of modules 1 1 1 1 1 1 1 1
2
Permeate pressure (Pa) 4325 6008 7116 7658 8056 8426 8783 9114
Number of modules 1 1 1 1 1 1 1 1
3
Permeate pressure (Pa) 66 1597 2581 2974 3031 3073 3102 3200
Number of modules 3 2 1 1 1 1 1 1
4
Permeate pressure (Pa) 20 22 10 342 676 921 1120 1446
Number of modules 2 1 1 1 1 1 1 1
5
Permeate pressure (Pa) 20 22 37 342 676 921 1120 1446
* Indicates values of permeate pressure out of the boundaries of the mathematical model for the
membrane the PAN-B5 developed by Tsuyumoto et al., (1997)

148
 Figure 5.15: Results of the optimisation of the pervaporation network using PAN-B5
hydrophilic membrane and a membrane area of 150 m2 for the distillation system shown in
Figure 5.14

(a)
(b)

Figure 5.16: Superstructure of the separation section of the ethanol production process: a)
single distillation and pervaporation network with PAN-B5 hydrophilic and b) pervaporation
network with PERVAPTM organophilic membrane linked to a single distillation column and a
pervaporation network with PAN-B5 hydrophilic membrane

149
 Table 5.15: Input variables used in the optimisation of the separation section with and without
the organophilic membrane
Feature Value
Feed flow rate (mole/sec) 50
Conc. ethanol feed (% w/w) 5.0
Pressure (kPa) 101.3
Feed temperature without Saturated liquid at
o
organophilic membrane ( C) specified pressure
Feed temperature with
50
organophilic membrane (oC)

5.6 Optimisation of the complete separation section – Case study

Figure 5.16 presents the schematics of the two different cases of separation stages
considered in this chapter. As mentioned before, the results of the optimisation of the
individual units of the separation section (i.e. distillation, hydrophilic pervaporation
network and organophilic pervaporation network) are used in the solution of the
optimisation problem in the superstructures presented in Figure 5.16 as initial
estimates.

5.6.1 Specifications of the separation section

The specifications for the complete separation sections are the same as for the
individual units. Table 5.15 presents the input variables for the optimisation of the
complete separation section used in the solution of the MINLP problem. The initial
guesses for the superstructure without the organophilic membrane are taken from the
results presented in Tables 5.3 and 5.6. The initial guesses for the superstructure
featuring the organophilic membrane are taken from the results presented in Tables
5.9, 5.11 and 5.14. The lower and upper bounds and the specifications of purity for
the complete system are the same as used in the induvial units (see Tables 5.2, 5.5
and 5.8).

For the optimisation of the complete separation stage, an additional constraint needs
to be included in the formulation of the solution of the optimisation problem. That
constraint represents the amount of ethanol removed in the waste stream as shown in
Figure 5.16 and is presented in Eq. 5.22:
150
 𝑚𝑖𝑛 𝑚𝑎𝑥
𝑤𝐿𝑜𝑠𝑠.𝐸𝑡ℎ ≤ 𝑤𝐿𝑜𝑠𝑠.𝐸𝑡ℎ ≤ 𝑤𝐿𝑜𝑠𝑠.𝐸𝑡ℎ Eq. 5.30

This section presents the optimisation of the separation sections with and without the
organophilic membrane. As previously mentioned, the combined process will use the
optimisation results of the individual units as initial guess in order to narrow the
feasibility region in which the optimal solutions can be found.

5.6.2 Results of the separation sections

For the structure presented in Figure 5.17, the results of each unit present similarities
to those obtained individually. For instance, the total number of trays in the
distillation column is 34, which is the same solution obtained for the individual case
(see Figure 5.5). For the superstructure, the feed stream is located at tray 20 whereas
in the individual column it was located at tray 21. The number of in-parallel modules
is slightly different since in the individual solution (see Figure 5.10) the total number
of modules is 4 whereas the superstructure presents 5. These small variations indicate
that the superstructure can be considered as a completely new and different system of
equations with different behaviour and possibly different solutions. However, these
comparisons also suggest that using the individual solution as an initial guess for the
optimisation of the overall separation section does facilitate the solution of the
MINLP.

A similar comparison can be applied to the distillation column systems in Figure

5.13 and Figure 5.19. The feed tray in both cases is close to the bottom of the column
and the number of trays is 25 and 24 for the individual solution and the
superstructure, respectively. Other similarities can be observed in the permeate
pressures, which decrease as the concentration of ethanol in the retentate increases in
both the individual solution of the dehydration section and the superstructure. Table
5.16 presents a summary of the results of the optimisation for both superstructures,
including the TAC and the energy requirements.

151
Figure 5.17: Results of the optimisation of the separation section with a PAN-B5 hydrophilic membrane area of 250 m2

152
Figure 5.18: Results of the optimisation of the separation section including a PERVAP TM 4060 organophilic membrane with 19 in-parallel membrane
modules with an area of 20 m2 and a PAN-B membrane with a membrane area of 150 m2

153
 Table 5.16: Results of the optimisation of the superstructures of the separation stages of the
ethanol production process
Superstructure Superstructure
Parameters
1 2
Number of trays 34 24
Feed tray 20 22
Reflux ratio 3.80 2.08
2
Organophilic membrane area (m ) N/ A 20
2
Hydrophilic membrane area (m ) 250 150
Electricity (kW) 3.23 87.9
Heat duty reboiler (kW) 224.1 144.3
Total energy consumption (kW) 233.5 271.2
Ethanol losses in waste (%) 7.06 9.06
CPU time (sec) 54,720 106,672
TAC (US$/yr. 105) 2.56 3.68

Table 5.16 presents the results of the optimisation of both superstructures, providing
an evaluation of the of the overall energy consumption within the process as well as
the minimum TAC. For instance, the distillation column in Superstructure 2 presents
fewer trays and lower reflux ratio. The optimal design of this column shows that the
feed stream should be fed in stages close to the reboiler to guarantee higher mass and
heat transfer between phases in upper trays in order to increase the separation ethanol
with lower energy demands.

The number of in-parallel modules in Superstructure 2 increases as the result of a

lower flow rate from the distillate stream which directly translates into smaller total
membrane areas (900 m2 in Superstructure 2 against 1250 m2 in Superstructure 1)
and high separation rates.

Superstructure 2 also reports a reduction in the heat duty at the reboiler of the
distillation column. In the case of the energy demand, the reboiler in Superstructure 2
shows a reduction of 36 % in comparison to the results obtained in Superstructure 1.
This reduction can be attributed to the thermodynamic state, the feed flow rate and

154
the concentration of ethanol in the feed as well as the feed location in the distillation
column.

However the overall energy consumption of Superstructure 2 is 16 % higher than

Superstructure 1, which is mainly caused by the implementation of additional units
such as heat exchangers to heat the permeate stream until saturation conditions and
compressors to increase the pressure of the permeate stream from 3430 Pa to
atmospheric conditions (i.e. 101,325 Pa). These results have also influenced the
TAC, making Superstructure 2 more expensive than Superstructure 1 (an increment
of 44 %). Although several aspects like the lower energy consumption in the reboiler
and the smaller membrane areas in the dehydration section are appealing for some
units of the separation section with the organophilic membrane, the overall results
suggest that this membrane is not suitable for the separation of ethanol from the
fermentation broth.

5.7 Sensitivity analysis

The results presented in Figures 5.17 and 5.18 are obtained for a case study where
the concentration of ethanol in the feed stream is fixed at 5 % (w/w) and a distillate
concentration close to 93 % (w/w). However, the concentration of ethanol in the feed
is subject to variations due to changes in the chemical composition of the raw
material, the conversion of hemicellulose and cellulose during the pretreatment
stages, the operating conditions and performance of the fermenting strain during the
fermentation, among other aspects.

Therefore, this section will focus on the variations of the concentration of ethanol in
the feed and the distillate stream and their effect on the design of the overall
separation section with and without the organophilic membrane. Table 5.17 presents
the different case studies considered in the sensitivity analysis.

155
 Table 5.17: Case studies considered in the optimisation of the separation section
Characteristics
Base case 5 % (w/w ) feed and 93 % (w/w) distillate
Case 1 4 % (w/w ) feed and 93 % (w/w) distillate
Case 2 6 % (w/w ) feed and 93 % (w/w) distillate
Case 3 5 % (w/w ) feed and 90 % (w/w) distillate

Table 5.18 presents the results of the optimisation problem for three different cases
in which both the concentration of the feed stream and distillate are modified in order
to evaluate the optimal design of Superstructures 1 and 2. In Superstructure 1, the
number of distillation columns is one and the number of in-series pervaporation
stages is fixed at three with a membrane area of 250 m2. On the other hand,
Superstructure 2 presents a single pervaporation module for the organophilic
membrane with an area of 20 m2, a single distillation column and three in-series
pervaporation stages with an area of 150 m2. These specifications are taken from the
results presented in the previous sections.

The results in Table 5.18 suggest that a decrease in the concentration of ethanol in
the feed stream reduces the amount of ethanol in the distillate and therefore the
energy required to recover it at the top of the column. Conversely, the number of
trays in the distillation column increases alongside the reflux ratio in order to obtain
the concentration of ethanol specified in Table 5.17. The economic evaluation shows
a reduction in the TAC of 6 % and 7 % for Superstructure 1 and Superstructure 2,
respectively. These savings are the result of a reduction in the utilisation of steam
and electricity in the process as well as the cost of the membrane modules.

Case 2 presents an increment of the concentration of ethanol in the feed (form 5 % to

6 %). Unlike the results presented for Case1, in Case 2 the TAC and the total energy
consumption are higher than in the base case, which is mainly due to the cost of
membranes, steam and electricity. The number of in-parallel membrane modules in
both superstructures increase since the flow rate of the distillate stream increase with
the feed concentration, which can also be observed in Figures 5.19 to 5.21.

156
When the concentration of ethanol in the distillate stream changes from 93 % to 90
% (Case 3 in Table 5.17), the energy consumption increases 3% and 5% in
Superstructures 1 and 2, respectively. However, the most noticeable change in this
case is the size of distillation column and the pervaporation network. Since the
concentration of distillate is lower than in the base case, the number of trays is
expected to be lower. This results in lower costs for the distillation column,
especially in Superstructure 2 where savings of up to 28 % are reported. In the
pervaporation section, the number of in-parallel modules is higher than in Cases 1
and 2.

The partial flux of water through the membrane increases when the distillate
concentration of ethanol in the distillate decreases, causing a sharp decline in the
temperature of the retentate stream. Since the constraints used in the solution of the
optimisation problem allow a maximum temperature drop of 20 K, the first two
stages of pervaporation will attempt to increase the concentration in the retentate side
without letting the temperature drop below the maximum bound. This behaviour has
a clear effect on the last stage where basically more in-parallel modules and lower
permeate pressures are required in order to reach the desired concentration. In
Superstructure 1, there are 8 in-parallel membrane modules whereas Superstructure 2
presents 12.

The minimum TAC for each case can also be observed in Figure 5.22. The results
illustrated in Figure 5.22 showed that in all case studies, Superstructure 2 has the
highest TAC which is likely to be caused by the compression done over the permeate
stream before the distillation column and as seen in Table 5.18, the costs associated
with compression go from 1000 to 38,000 US$/yr.

157
 Table 5.18: Sensitivity analysis of the separation section
Without the organophilic membrane With the organophilic membrane
Feature
Base case Case 1 Case 2 Case 3 Base case Case 1 Case 2 Case 3
Number of trays 34 43 33 31 24 21 21 11
Feed location 20 30 27 22 22 17 18 6
Reflux ratio 3.80 4.71 3.51 3.55 2.08 2.20 1.95 2.08
Organophilic membrane
Number of modules stage 1 N/A N/A N/A N/A 19 19 19 19
Permeate pressure stage 1 (Pa) N/A N/A N/A N/A 3430.1 3375.7 3441.7 3430.1
Hydrophilic membrane
Number of modules stage 1 1 1 1 1 1 1 1 1
Permeate pressure stage 1 (Pa) 6374.7 8194.1 8181.0 10,000 4919.7 6028.9 4984.9 9892.2
Number of modules stage 2 1 1 1 1 2 1 2 1
Permeate pressure stage 2 (Pa) 179.8 2849.0 521.2 7162.4 468.3 392.7 987.1 5789.1
Number of modules stage 3 3 3 4 8 3 4 5 12
Permeate pressure stage 3 (Pa) 179.8 393.2 521.2 198.6 194.9 392.7 194.6 107.8
Ethanol losses in waste (% w/w) 7.06 7.55 7.03 9.00 9.06 9.06 8.04 10.0
Total energy consumption (kW) 233.5 218.70 261.0 241.0 271.2 235.6 307.3 285.6

158
 Table 5.18: Sensitivity analysis of the separation section (cont.)
Without the organophilic membrane With the organophilic membrane
Feature
Base case Case 1 Case 2 Case 3 Base case Case 1 Case 2 Case 3
Total cost of the distillation
0.62 0.69 0.66 0.59 0.40 0.34 0.42 0.29
system (US$/yr. 105)
Total cost of the organophilic
N/A N/A N/A N/A 0.34 0.34 0.34 0.34
membrane system (US$/yr. 105)
Total cost of the hydrophilic
1.28 1.12 1.34 2.23 0.81 0.80 1.07 1.84
membrane system (US$/yr. 105)
Cost of heat exchanger 1 (US$/yr. 105) 0.002 0.002 0.002 0.002 0.02 0.02 0.02 0.02
Cost of heat exchanger 2 (US$/yr. 105) 0.002 0.001 0.001 0.001 0.002 0.002 0.002 0.002
Cost of heat exchanger 3 (US$/yr. 105) 0.001 0.001 0.002 0.001 0.002 0.001 0.002 0.002
5
Cost of heat exchanger 4 (US$/yr. 10 ) N/A N/A N/A N/A 0.001 0.001 0.002 0.001
Total cost of compressors
0.08 0.06 0.08 0.12 1.32 1.23 1.41 1.37
(US$/yr. 105)
Cost of cooling water (US$/yr. 105) 0.01 0.01 0.01 0.01 0.01 0.006 0.008 0.007
Cost of steam (US$/yr. 105) 0.56 0.52 0.63 0.57 0.43 0.36 0.50 0.45
Cost of electricity (US$/yr. 105) 0.01 0.01 0.01 0.02 0.35 0.33 0.38 0.36
TAC (US$/yr. 105) 2.56 2.41 2.73 3.54 3.68 3.43 4.15 4.68

159
Figure 5.19: Case 1 - Results of the optimisation of the separation section of the ethanol production process with the PERVAPTM 4060 organophilic
membrane considering a feed concentration of ethanol of 4 % (w/w)

160
Figure 5.20: Case 2 - Results of the optimisation of the separation section of the ethanol production process with the PERVAPTM 4060 organophilic
membrane considering a feed concentration of ethanol of 6 % (w/w)

161
Figure 5.21: Case 3 - Results of the optimisation of the separation section of the ethanol production process with the PERVAPTM 4060 organophilic
membrane considering a concentration of ethanol in the distillate stream of 90 % (w/w)

162
 5x105
Without organophilic membrane
With organophilic membrane
TAC (US$/yr.) 4x105

3x105

2x105

1x105

0
e1

e2

e3
se
ca
as

as

as
se
C

C
Ba

Figure 5.22: Evaluation of the minimum TAC of the separation section with and without the
organophilic membrane considering different cases studies

5.8 Conclusions
This chapter presented the optimisation of two different configurations of the
separation section of the ethanol production process from lignocellulosic biomass
using a deterministic approach. The first configuration comprised a distillation
system linked to a pervaporation network with a hydrophilic membrane (PAN-B5).
The second configuration included the organophilic membrane (PERVAPTM 4060)
introduced in Chapter 4 in the overall process, located before the distillation column.
This pervaporation network is linked to a distillation column and a hydrophilic
membrane system.

The optimisation of both superstructures consisted in first optimising each unit

individually. Using specific feed conditions, the optimal design with the minimum
total annualised cost was determined. The first unit to be analysed was the distillation
column using a tray-by-tray approach presented by Caballero and Grossmann,
(2014a). With this approach, the total number of trays and the optimal feed location
that minimised the TAC could be determined. In fact, two arrangements for the

163
distillation section were optimised: a single distillation column and double-
distillation system. The results of this optimisation showed that a single distillation
column was more economically viable for the separation of a solution 5% (w/w) of
ethanol and a distillate concentration of 93 % (w/w). The single column was 42 %
cheaper than the double-column arrangement and reported energy savings of 41%.

The following step was to determine the design of the pervaporation network for
both membranes. The optimisation showed that for the PAN-B5 hydrophilic
membrane system, for both superstructures, a configuration of three in-series stages
was required. In the case of the PERVAPTM 4060 membrane, only one stage was
required in order to remove most of the ethanol from the feed. Once all the units
involved in the separation of ethanol from the fermentation broth were optimised, the
result were used in the solution of the superstructure as initial estimates as a way to
narrow down the feasibility region in which the solution might be found.

The results of these calculations showed that the implementation of the organophilic
membrane reduced the energy required in the reboiler of the distillation column in
Superstructure 2 by 36 % in comparison with distillation column in Superstructure 1.
However, the overall energy consumption in Superstructure 2 was 16 % higher than
in Superstructure 1. These results could be attributed to the use of a compressor unit
for the permeate stream leaving the organophilic membrane which directly
influenced the TAC in Superstructure 2. Table 5.16 showed that not only
Superstructure 2 had higher energy consumption than Superstructure 1, this
configuration was also more expensive and the implementation of additional units to
provide the necessary operating conditions only increased the investment costs.

A sensitivity analysis was also carried out in this chapter. The purpose of this
analysis was to determine the variations in the optimal design of both
superstructures, focusing mainly on the structure with the organophilic membrane
since this is the novelty of this work. The results showed that in all case studies, the
implementation of the organophilic membrane is more expensive and more energy
intensive than the configuration without said membrane. Chapter 6 will focus on the

164
heat integration of the entire plant as an additional step to reduce the usage of utilities
and distribute all the sources of energy within the plant in a more efficient fashion.

165
Chapter 6 – Heat integration across an ethanol
production process

Abstract

This chapter introduces a methodology for the implementation of heat storage

systems for the ethanol production process from lignocellulosic biomass. Initially,
this chapter presents the optimisation of the complete process without heat storage
so it can be taken as a reference case for other arrangements with heat storage. Two
arrangements with heat storage are considered in this work in which two tanks of
equal size are used to store the heat transfer fluid. The first tank will initially contain
the heat transfer fluid used to provide heat to the reactors and other units, and the
second tank will receive the outlet streams coming out of the jackets of the units. The
results of the optimisation of the process with heat storage show a decrease in the
Total Annualised Cost as well as the energy consumption throughout the plant.

6.1 Introduction
In the design of a chemical process, there are several design variables that need to be
considered such as size of the equipment, fouling, operating time, capital investment,
etc., and operating variables such as purity of the product, waste, temperature,
pressure, among others. One variable, of particular interest is energy consumption.
Energy consumption makes reference to the heat provided from external sources to
units in the process in order to function at the desire operating conditions. This heat
can be transferred from sources such as steam or electricity (Yee et al., 1990a,
1990b; Yee and Grossmann, 1990). On the other hand, chemical processes may also
require external cooling (e.g. cooling towers in power and nuclear plants) (Kiss and
Olujic, 2014; Wang et al., 2015). The costs related to the usage of these external
sources can have a significant impact on the profitability of a process. (Čuček et al.,
2011; Grisales et al., 2008; Kiran and Jana, 2015; Kravanja et al., 2013). Hence, heat
integration has emerged as a sophisticated way of reducing the energy demands by

166
distributing the different forms of heat (either supplied or released) that can be found
in the process in order to make the entire plant more energy efficient (Modarresi et
al., 2012; Rašković et al., 2010).

Batch operations are generally run on a smaller scale compared to continuous

operations and utility requirements are therefore considered lower than their
counterparts. Energy consumption in general is commonly estimated to be about 5%
- 10% of total costs (Vaklieva-Bancheva et al., 1996). However, some batch
industries have a much higher utility demands than others. For example, utility
requirements in the food industry, breweries, dairies, meat processing facilities,
biochemical plants and agrochemical facilities contribute largely to the total cost and
heat integration has proved successful in improving energy efficiency in these sectors
(Knopf et al., 1982; Rašković et al., 2010). In several chemical processes, for
instance pharmaceuticals and pigments, due to the existence of large amounts of used
solvents and volatile organic compounds to incinerate, there is an excess of available
heat which can be utilised for heat integration.

Heat integration for improvement and reduction in energy consumption in a plant,

may be achieved by one of the following paths: Direct Heat Integration (DHI) and
Indirect Heat Integration (IHI). These methods for heat integration and their
implementation depend on the type of production (e.g. continuous or batch) and the
different components involved in the process (Krummenacher, 2002). These methods
are discussed in more detail in the following section.

6.1.1 Direct heat integration (DHI)

In DIH, process streams can exchange heat with one another directly so the process
can only experience heat transfer once. This aspect suggests that the amount of heat
recovered in the process can be high due to a smaller temperature difference during
heat integration. However, because process streams cannot be mixed, when the
number of process streams involved in heat integration is large, a number of heat
transfer loops are required (Kiran and Jana, 2015; Kravanja et al., 2013). DHI
normally requires more heat exchange loops than in the indirect heat integration
approach which may result higher capital costs, although this is also dependent on

167
the area of the heat exchangers and the amount of heat that can be recovered (Wang
et al., 2015).

6.1.2 Indirect heat integration (IHI)

A second form of heat integration is Indirect Heat Integration (IHI). Energy may be
stored in several forms, including chemical energy (e.g. in biomass and
electrochemical batteries), mechanical energy (e.g. in pumped hydro, compressed air
and flywheels) and as heat (Dinçer and Rosen, 2010). For the IHI method, the heat
exchange between non-coexistent process streams becomes less limited so it is less
schedule-sensitive and it could provide a great deal of operating flexibility
(Krummenacher, 2002). IHI may sometimes be considered as an expensive form of
heat integration. Heat storage has only been considered as an alternative mode for
heat integration when opportunities of using DHI have been exhausted (Rodera and
Bagajewicz, 2001). When energy is stored in the form of heat, this is known as
thermal energy storage (TES) and such storage may further be in the form of either
Sensible TES or Latent TES as shown next:

Latent energy storage: It uses phase change materials (e.g. hydrated salts) to store
heat (generally by crystallization /melting). The temperature at which the phase
change occurs depends on the selected material; therefore, unlike sensible heat
storage, the operating temperature of Latent Heat Storage Units (HSU) has to be
selected among a discrete set of values (Schröder and Gawron, 1981). Latent heat
storage generally results in a significantly smaller volume when compared to sensible
heat storage, and is particularly recommended when large amounts of heat must be
stored with small temperature differences (Feczkó et al., 2016; Schröder and
Gawron, 1981; Sciacovelli et al., 2015; Zhang et al., 2016).

Sensible energy storage: Three main types of sensible heat storage systems may be
distinguished:

 Mixing heat storage: the storage of heat is achieved by an increase of the

overall mean temperature of the storage fluid in the tank. By construction, or
by operation, the heat storage fluid features an almost homogenous
temperature in the HSU (Krummenacher, 2002).
168
  Stratified heat storage: stratification results in a hot (or cold) fluid storage
tank when different regions in the tank are at different temperatures. The
density variations which result from those temperature differences will cause
the hotter fluid masses to rise to the top and the cooler fluid masses to fall to
the bottom of the storage tank, with a middle layer transition zone (the
thermocline), separating the hot upper zone from the cold lower zone
(Altuntop et al., 2005; Han et al., 2009).
 Fixed temperature / variable mass (FTVM) heat storage: FTVM heat storage
operates in the same way as stratified heat storage, but prevents the
temperature degradation problem by keeping the volumes presenting different
temperatures in separate tanks (HSU’s) (Sciacovelli et al., 2015; Zalba, 2003;
Zhang et al., 2016).

For any of these three types, given a heat storage capacity, the required volume is
inversely proportional to the temperature difference between the "charged state" and
"empty state".

6.1.3 Mixed direct-indirect heat integration (MDIHI)

This mode of heat integration combines the approaches of both direct and indirect
heat integration where process streams are initially stored in order to provide heat to
other streams at different times. The indirect mode often implies higher capital costs,
so in order to avoid these costs, process rescheduling is considered, which increases
the potential of direct heat exchanges between process streams. Similarly, heat
integration of multi-product and multi-purpose plants is often focused on the
scheduling of batch operations accounting for direct heat exchanges opportunities.

6.2 Heat transfer fluids

Heat integration using heat storage often requires a fluid which must be carefully
selected for a specific purpose. Table 6.1 presents a list of heat transfer fluids often
used in industry for heat storage. Air and water in the form of steam are the cheapest
and most accessible transfer fluids available. Their implementation usually requires
larger volumes of HSU’s which is very impractical from the point of view of costs
and also the use of additional equipment such as compressors and more heat
exchangers (Krummenacher, 2002; Zalba, 2003).
169
 Table 6.1: List of heat transfer fluids (Vignarooban et al., 2015)
Melting point Stability limit Heat capacity Average
Name o o -1 -1
( C) ( C) (kJ kg K ) cost ($/kg)
Air - - 1.03 (127 oC) 0
o
Water 0 - 1.90 (127 C) ~0
Thermal oils
Mineral oil -20 300 1.67 0.3
Synthetic oil -20 350 1.94 3.0
Silicone oil -20 400 1.09 5.0
Organic
Biphenyl/Diphenyl
Oxide 12 393 1.93 100
(Therminol VP-1)
Molten-salts
Solar salt 220 650 1.60 0.50
Hitec 142 535 1.56 0.93
Na-K-Li nitrates 130 600 1.10 1.10
Liquid metals
Na-K -12 785 0.87 (600 oC) 2.00
o
Na 98 883 1.25 (600 C) 2.00
Pb-Bi 125 1533 0.15 (600 oC) 13.0

Heat storage using water is limited by the operating conditions of the rest of the units
of the process. For instance, units with high operating temperatures would require a
transfer fluid with high heat capacities and, in the case of steam, pressurised tanks
would be required which reflects negatively on the costs of the process and the
controllability of the operating conditions. Furthermore, steam systems experience
large heat losses due to condensation (Lienhard, 2010; Zalba, 2003).

Thermal oils are widely used to carry thermal energy in process heating, metal
working and machine cooling applications. They are mainly used in high temperature
process applications where the optimum bulk fluid operating temperatures (between
150ºC and 400ºC) are safer and more efficient than steam, electrical, or direct fire
heating methods (Vignarooban et al., 2015; Zhang et al., 2016). Thermal oils allow
the use of low pressure heat transfer systems to achieve high temperatures which
would otherwise have necessitated high pressure steam systems (Mawire et al., 2014;
Veses et al., 2016). Steam systems are subject to statutory and regulatory
170
requirements due to the inherent risk from pressure and the increased cost of
installation and routine insurance inspection requirements (Mawire et al., 2014;
Zhang et al., 2016).

Molten salts refer to the type of ionic liquids that are currently used for heat transfer
and heat storage due to the elevated temperatures at which they can be operated. The
growing interest in energy applications of molten salts is justified by several of their
properties (Amusat et al., 2015; Serrano-López et al., 2013a). The advantages of
molten salts as heat transfer fluid and thermal storage systems promise a great
development during next decades. The cost for the required volume of heat
exchangers and pumps are highly reduced by the use of liquid salts instead of other
coolants due to their higher volumetric heat capacity without the need of pressurising
(Vignarooban et al., 2015; Zhang et al., 2016). One of the most prominent
applications for molten salts is solar thermal power which is a promising way of
providing renewable electricity (Feczkó et al., 2016; Serrano-López et al., 2013b).
Unlike other renewable energy technologies, solar thermal power plants have the
ability to store thermal energy (Zaversky et al., 2013).

Liquid Metals are a specific class of heat transfer fluid. Their basic advantage is a
high molecular thermal conductivity which, for identical flow parameters, enhances
heat transfer coefficients. Another characteristic of liquid metals is the low pressure
of their vapours, which allows their use in power engineering equipment at high
temperatures and low pressure, thus improving solution of mechanical strength
problems (Zeigarnik, 2011). One of the major applications for liquid metals has been
in nuclear industries which have been in development since the 1940’s. Nowadays,
liquid metals are currently being studied for use in solar thermal systems as energy
storage media with purposes of green technologies and sustainability (Vignarooban
et al., 2015).

Organic fluids are also used in the chemical industry for heat transfer and heat
storage applications in processes within oil and gas, plastic processing,
pharmaceuticals, solar energy, etc. (Eastman, 2016; Krummenacher, 2002; Schröder
and Gawron, 1981; Zalba, 2003). The most known material is Therminol which is a
mix of several organic compounds with high stability and relatively low prices
171
(Weiguo et al., 2016; Xu et al., 2016). Its implementation in the chemical industry
has been reported by several authors showing promising results in terms of lower
energy losses, lower fire risks and easier acquisition. Therefore, organic fluids will
be considered in the development of a heat transfer configuration for the ethanol
production process from corn stover.

Eastman Ltd has a variety of Therminol fluids for heat transfer puroposes but has
reported that the most successful in sales is Therminol 66 (Eastman, 2016; Weiguo et
al., 2016). Table 6.2 presents the properties of Therminol 66 used in the calculations
carried out in this chapter.

6.3 Simulation procedure

Heat integration can be used to reduce the consumption of utilities in processes such
as ethanol production from lignocellulosic biomass considered in this work. The
process has high heat demands in units such as the pretreatment reactor, the
evaporator and the distillation system. For this work, heat integration will consider
the implementation of heat storage units in which a heat transfer fluid will be used to
substitute steam in order to reduce operating costs.

Chapter 5 presented the optimisation of the separation section of the ethanol

production process based on a total annual cost function (TAC). Chapter 6 will
follow a similar approach for the optimisation of the separation section, except that
in this part of the work, the entire process considering heat storage will be optimised.

6.3.1 Configuration of the ethanol process without heat storage

The optimisation problem presented in Figure 6.1 is the production of ethanol from
corn stover, and is based on the configuration illustrated in Figure 2.1. The steps of
the production of ethanol from corn stover are the following:

172
 Table 6.2: List properties of Therminol 66 (Eastman, 2016)
Item Characteristics
Appearance Clear, pale yellow liquid
Composition Modified terphenyl
Boiling point 359°C (678°F)
Molecular weight 252
Autoignition temperature 374°C (705°F)
Cost (US$/kg) 6.5*
*Taken from Rajkumar et al., 2015

1. The initial load of raw material is processed in the pretreatment reactor

2. The outlet stream is filtered and the liquid fraction is concentrated in an
evaporator and detoxified with Calcium Hydroxide Ca(OH)2 in the
detoxification reactor
3. The solid fraction of the pretreatment and the detoxified stream are mixed
and sent to the Simultaneous Saccharification and Co-Fermentation (SSCF)
reactor where ethanol is produced using the cocktail of enzymes cellulase/β-
glucosidases and fermenting bacteria Zymomonas mobilis ZM4 (pZB5)
4. The solids of the SSCF process (i.e. unreacted cellulose and hemicellulose,
cells, insoluble solids, etc.) are filtered and the liquid is pumped into the
distillation column
5. The distillate stream from the distillation column is cooled down to 70 oC
(maximum operating temperature at which the partial flux of water is the
highest) before the hydrophilic membrane network
6. Ethanol is dehydrated to concentrations higher than 99 % (w/w) in the
pervaporation network

This chapter introduces the implementation of heat integration into the ethanol
process and the optimisation of the design and operation of the plant by minimising
costs. In other words, the solution of the optimisation problem needs to show the
optimal number of trays in the distillation column, the optimal number of in-parallel
pervaporation modules, the optimal size of each reactor, the operating conditions of
each unit, etc. The configuration shown in Figure 6.1 will be used as reference for
the comparative analysis against the configurations featuring heat integration. The

173
optimisation variables for the process without heat integration (see Figure 6.1) are
the following:

1. Pretreatment reactor volume (𝑚3 )

2. Detoxification reactor volume (𝑚3 )
3. SSCF reactor volume (𝑚3 )
4. Inlet stream of inoculum of cells (kg/s)
5. Concentration of cells in inoculum (% (w/w))
6. Reflux ratio
7. Number of trays in the stripping and rectification sections of the distillation
column
8. Number of in-parallel modules in each pervaporation stage
9. Permeate pressure (Pa)

This optimisation methodology assumes that the number of in-series pervaporation

stages is fixed as well as the number of distillation columns. These assumptions
come from the evaluation of the separation section presented in Chapter 5, which
showed that two distillation columns are more energy intensive and more expensive
than just one column for the same distillate mole fraction.

The optimisation of the entire process considers steady-state conditions in the

operation of the pretreatment stages, assuming that all flow rates have average
values. This facilitates the calculations of the optimal design variables and operating
conditions which is the main scope of this work. This methodology is also
implemented in the optimisation of the configurations with heat storage tanks.

174
Figure 6.1: Flowsheet of the ethanol production process from corn stover without heat integration

175
The number of in-series pervaporation modules is assumed to be three since the
results presented in Chapter 5 suggested that this arrangement presents the lowest
total annual cost with a low energy consumption in the intermediate heaters as well
as the low required area of each membrane module. The solution of an optimisation
problem requires a set of constraints that will narrow down the convergence region to
find an optimum under specifications of production rate and/or purity of ethanol.
Equations 6.1 to 6.10 show the constraints used in the optimisation of the ethanol
process without heat storage.

The range of conversion of hemicellulose during pretreatment is set as shown in Eq.

6.1. The reason for this choice is explained in the work of Esteghlalian et al., (1997)
which states that higher conversions of hemicellulose could not only produce higher
concentrations of xylose, but also higher concentrations of by-products which
influence the performance of the detoxification stage. Therefore, the minimum
conversion in the system should be 75 % (see Table 6.3):

𝑠𝑝𝑒𝑐
𝑋ℎ𝑒𝑚𝑖𝑐𝑒𝑙𝑙𝑢𝑙𝑜𝑠𝑒 ≥ 𝑋ℎ𝑒𝑚𝑖𝑐𝑒𝑙𝑙𝑢𝑙𝑜𝑠𝑒 Eq. 6.1

The constraint for conversion of furfural in the detoxification stage is shown in Eq.
6.2. With this constraint, a higher degradation of by-products in the acid hydrolysates
is expected. Given the parameters and the operating conditions obtained by Purwadi
et al., (2004) the minimum conversion of by-products is the following:

𝑠𝑝𝑒𝑐
𝑋𝑏𝑦−𝑝𝑟𝑜𝑑𝑢𝑐𝑡𝑠 ≥ 𝑋𝑏𝑦−𝑝𝑟𝑜𝑑𝑢𝑐𝑡𝑠 Eq. 6.2

The constraint for the production of ethanol during SSCF is shown in Eq. 6.3. This
constraint seeks to maintain the concentration of other by-products such as
cellobiose, xylose and cells to the lowest possible (Morales-Rodriguez et al., 2012,
2011):

𝑚𝑖𝑛 𝑚𝑎𝑥
𝑊𝑒𝑡ℎ𝑎𝑛𝑜𝑙 ≤ 𝑊𝑒𝑡ℎ𝑎𝑛𝑜𝑙 ≤ 𝑊𝑒𝑡ℎ𝑎𝑛𝑜𝑙 Eq. 6.3

176
Eq. 6.4 presents the constraint related to the maximum concentration achievable
during the distillation process. This constraint is chosen in order to secure a feed
composition in the pervaporation network applicable for the selected membranes
(Tsuyumoto et al., 1997).

𝑆𝑝𝑒𝑐
𝑊𝐷,𝑒𝑡ℎ ≥ 𝑊𝐷,𝑒𝑡ℎ Eq. 6.4

Due to the evaporation of the volatile components in the feed which diffuse through
the membrane, temperature drops in the retentate side. Feed temperature needs to be
kept constant in all stages of the pervaporation network in order to maintain high
rates of flux through the membrane. As seen in Figure 6.1, there are intermediate
heat exchangers to maintain feed temperature at 70 oC. Equations 6.5 and 6.6 present
the range of temperature difference between feed and retentate for the optimisation
of the process (Baker, 2012; Marriott and Sorensen, 2003; Sander and Janssen,
1991):

𝑚𝑖𝑛 𝑚𝑎𝑥
∆𝑇𝑝𝑒𝑟𝑣𝑎𝑝_1 ≤ ∆𝑇𝑝𝑒𝑟𝑣𝑎𝑝_1 ≤ ∆𝑇𝑝𝑒𝑟𝑣𝑎𝑝_1 Eq. 6.5

𝑚𝑖𝑛 𝑚𝑎𝑥
∆𝑇𝑝𝑒𝑟𝑣𝑎𝑝_2 ≤ ∆𝑇𝑝𝑒𝑟𝑣𝑎𝑝_2 ≤ ∆𝑇𝑝𝑒𝑟𝑣𝑎𝑝_2 Eq. 6.6

In order to meet the standards of purity required for ethanol for energy generation
purposes (Binod et al., 2010; Cardona and Sánchez, 2007; Drapcho et al., 2008), the
concentration obtained in the retentate side of the last pervaporation stage is set as
shown in Eq. 6.7 (see Table 6.3):

𝑚𝑖𝑛 𝑚𝑎𝑥
𝑊𝑟𝑒𝑡𝑒𝑛𝑡,𝑒𝑡ℎ ≤ 𝑊𝑟𝑒𝑡𝑒𝑛𝑡,𝑒𝑡ℎ ≤ 𝑊𝑟𝑒𝑡𝑒𝑛𝑡,𝑒𝑡ℎ Eq. 6.7

The streams leaving the distillation column and the retentate of the third
pervaporation stage are mixed. The constraint presented in Eq. 6.8 seeks to keep
minimum losses of ethanol in this stream.

𝑚𝑖𝑛 𝑚𝑎𝑥
𝑤𝐿𝑜𝑠𝑠.𝐸𝑡ℎ ≤ 𝑤𝐿𝑜𝑠𝑠.𝐸𝑡ℎ ≤ 𝑤𝐿𝑜𝑠𝑠.𝐸𝑡ℎ Eq. 6.8

177
For the optimisation of the distillation column, additional constraints are required for
the evaluation of the total number of trays and feed location (see Figure 5.1). These
constraints were discussed in more detail in Chapter 5 and are based on the tray-by-
tray approach proposed by Caballero and Grossmann, (2014).

𝑁𝑡𝑟𝑎𝑦_𝑟𝑐

∑ 𝛼𝑗 = 1 𝑗 = 1, 2, … , 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑟𝑎𝑦𝑠 𝑖𝑛 𝑟𝑒𝑐𝑡𝑖𝑓𝑖𝑐𝑎𝑡𝑖𝑜𝑛 Eq. 6.9

𝑗=1

𝑁𝑡𝑟𝑎𝑦𝑠_𝑠𝑡

∑ 𝛽𝑘 = 1 𝑘 = 1, 2, … , 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑟𝑎𝑦𝑠 𝑖𝑛 𝑠𝑡𝑟𝑖𝑝𝑝𝑖𝑛𝑔 Eq. 6.10

𝑘=1

Equations 6.11 and 6.12 represent the number of active trays that can be counted in
both sections of the column and are used in the calculation of its height, as shown in
section 5.3.1.

𝑡𝑜𝑡𝑎𝑙
𝑁𝑅𝑒𝑐 = ∑ 𝛼𝑗 𝑅𝑐𝑡𝑗 𝑅𝑐𝑡𝑗 = {1, 2, 3 … 𝑁𝑡𝑟𝑎𝑠_𝑟𝑐 } Eq. 6.11
𝑗

𝑡𝑜𝑡𝑎𝑙
𝑁𝑆𝑡𝑟 = ∑ 𝛽𝑘 𝑆𝑡𝑟𝑘 𝑆𝑡𝑟𝑗 = {𝑁𝑡𝑟𝑎𝑦_𝑠𝑡 , 𝑁𝑡𝑟𝑎𝑦_𝑠𝑡−1 , … ,2, 1} Eq. 6.12
𝑘

As presented in Section 5.3, the set of constraints that correlate the continuous
variables of the system with the discrete variables used to identify the recycle of both
distillate and steam into the column are defined as follows (Caballero and
Grossmann, 2014a):

𝑅𝐹𝑇 = {𝑗 | 𝐶𝑎𝑛𝑑𝑖𝑑𝑎𝑡𝑒 𝑡𝑟𝑎𝑦 𝑜𝑓 𝑟𝑒𝑓𝑙𝑢𝑥 𝑟𝑒𝑡𝑢𝑟𝑛} Eq. 6.13

𝑅𝐵𝑇 = {𝑘 | 𝐶𝑎𝑛𝑑𝑖𝑑𝑎𝑡𝑒 𝑡𝑟𝑎𝑦 𝑜𝑓𝑟𝑒𝑏𝑜𝑖𝑙 𝑟𝑒𝑡𝑢𝑟𝑛} Eq. 6.14

Let LRx, LRb be the reflux and reboil flow rate returned to the column, respectively.
Let 𝛼𝑗 𝑗 ∈ 𝑅𝐹𝑇 ; 𝛽𝑘 𝑘 ∈ 𝑅𝐹𝑇 be binaries that take the value 1 if the reflux/reboil is
returned to tray j and k, respectively. Equations
178
𝑟𝑒𝑓𝑡𝑟 ≤ 𝐿𝑅𝑥 𝛼𝑗 𝑗 ∈ 𝑅𝐹𝑇 Eq. 6.15

𝑟𝑒𝑏𝑡𝑟 ≤ 𝐿𝑅𝑏 𝛽𝑘 𝑘 ∈ 𝑅𝐵𝑇 Eq. 6.16

𝑟𝑒𝑓𝑡𝑟 ≥ 0, 𝛼𝑗 ∈ {0,1} ∀𝑗 ∈ 𝑅𝐹𝑇 Eq. 6.17

𝑟𝑒𝑏𝑡𝑟 ≥ 0, 𝛽𝑘 ∈ {0,1} ∀𝑘 ∈ 𝑅𝐵𝑇 Eq. 6.18

Equations 6.19 and 6.20 correlate the vectors of recycles with the actual flow rate to
be fed into the column as previously presented in Chapter 5.

𝐿𝑅𝑥_𝑖𝑛 = ∑ 𝑟𝑒𝑓𝑡𝑟 Eq. 6.19

𝑗∈𝑅𝐹𝑇

𝐿𝑅𝑏_𝑖𝑛 = ∑ 𝑟𝑒𝑏𝑡𝑟 Eq. 6.20

𝑘∈𝑅𝐵𝑇

The objective function shown in Eq. 6.21 is the same used in Chapter 5. It presents
the total annual cost or the total investment on the production of ethanol from
lignocellulosic biomass and it covers the cost of all the equipment and the usage of
utilities and services.

𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝑖𝑛𝑣𝑒𝑠𝑡𝑚𝑒𝑛𝑡
𝑇𝐴𝐶 = ∑ 𝐴𝑛𝑛𝑢𝑎𝑙 𝑜𝑝𝑒𝑟𝑎𝑡𝑖𝑛𝑔 𝑐𝑜𝑠𝑡 + ∑ Eq. 6.21
𝑝𝑎𝑦𝑏𝑎𝑐𝑘 𝑝𝑒𝑟𝑖𝑜𝑑

6.3.2 First superstructure of the ethanol production process with heat storage
The heat integration of the ethanol process requires two heat storage tanks of equal
size (HS1 and HS2). Heat storage tank 1 (HS1) initially contains the heat transfer
fluid and distributes it to the reactors of the process. Heat storage tank 2 (HS2),
collects the heat transfer fluid coming out of the jackets of the reactors and sends it to
the separation section of the process. In Figure 6.2, the red lines denote the hot
streams of Therminol 66 used in the transfer of heat to the units of the process. The

179
blue lines represent cooling water used in the detoxification reactor and the
condenser of the distillation column.

Figure 6.2 presents the first configuration of the process using heat storage. This
configuration is very similar to the one shown in Figure 6.1 but with a few
differences. The first difference is the utilisation of a pump prior to the distillation
system to increase the pressure inside the column. The second difference is the
implementation of a heat exchanger (HEx 5) to use the heat from the stream leaving
the reboiler to heat up the stream of Therminol 66 leaving HS2. The third difference
is another heat exchanger (HEx 4) in which high-pressure steam is used to provide
energy to the stream of Therminol 66 in order to increase its temperature to the same
temperature as the outlet streams of HS1. The steps of the configuration shown in
Figure 6.2 are summarised below:

1. The HS1 tank filled with Therminol 66 provides heat to the pretreatment,
evaporation, SSCF stages through their jackets
2. The outlet of the jackets is collected in HS2 and sent to the other units in the
process
3. The outlet of HS2 is passed through heat exchanger (HEx 5) where the
bottom stream of the distillation column is used to heat the Therminol stream
4. The stream of Therminol heated in heat exchanger (HEx 5) is used in Heat
Exchangers 2 and 3
5. The stream of Therminol is passed through Heat Exchanger 4 to increase the
temperature of the heat transfer fluid before sending it back to HS1

The additional optimisation variables when considering heat storage are:

1. Distillation column pressure (Pa)

2. Outlet flowrates leaving HS1 and passed through the jackets of the different
reactors (kg/s)
3. HS1 outlet temperature
4. Temperature difference between outlet streams in both shell and tube in Heat
Exchanger 5 (Superstructure 1)

180
The assumptions for the simulation and optimisation of the ethanol production with
and without heat storage process are the following:

1. Perfect mixing in all the reactors as well as in each tray of the distillation
column and heat storage tanks (see Appendix A for mass and energy
balances)
2. The HS tanks are adiabatic as in all the piping in the process. This assumption
is supported by the reports presented by Zaversky et al., (2013) and Zhang et
al., (2016), where insulation prevents heat losses higher than 0.5 %
3. There is no mixing between process streams and the transfer fluid Therminol
66
4. Solids or other soluble components in the feed of the distillation column are
not considered
5. The strain Zymomonas mobilis used in these calculations is assumed to be
thermotolerant and its capability is not diminished by the operating
temperatures used in this case study (Baeyens et al., 2015; Hasunuma and
Kondo, 2012)

181
Figure 6.2: Superstructure 1 of the ethanol production process from corn stover with heat storage (using the bottom stream from the distillation column
to heat the stream of Therminol 66)

182
 Table 6.3: Initial guesses used in the optimisation of the process with and without heat
integration
Lower Upper
Variable Initial guess
bound bound
Conversion xylose 0.75 0.75 1
Temperature pretreatment (oC) 114 114 120
Volume pretreatment reactor (m3) 1 0.1 100
Conversion by-products 0.70 0.70 1
Volume detoxification reactor (m3) 1 0.1 100
Conc. ethanol fermentation (% w/w) 4 4 5
Intel flow rate cells (kg/sec) 1 0.01 100
Volume fermenter (m3) 1 0.1 100
Reflux ratio 3 1 100
Pressure column (kPa) 101.3 101.3 2026
Mass fraction ethanol in distillate 0.93 0.93 1
Number of membrane modules 1 1 25
Mass fraction ethanol in retentate 0.995 0.993 0.998
Permeate pressure (kPa) 400 0.01 10
Outlet temperature HS1 (oC) 114 114 350

Table 6.4: Operating conditions for the reactors and evaporator

Unit Operating conditions
Pretreatment reactor 0.6 % (w/w) of acid in a solution 1:10 liquid/solid ratio
Evaporator 60 % of liquid evaporated
Detoxification reactor 30 oC and pH of 12
SSCF reactor 50 oC

6.3.3 Second superstructure of the ethanol production process with heat

storage
The second superstructure with heat storage (see Figure 6.3) presents a similar
configuration as in the previous section. The main difference lies in the usage of the
Therminol 66 stream leaving HS2. Instead of using a heat exchanger to provide heat

183
from the stream leaving the bottom of the distillation column, the stream coming out
of HS2 is heated by using the energy released from the condenser.

6.4 Results
The ethanol production system without heat integration (Figure 6.1) is going to be
used as the base case for the comparative analysis with two systems with heat
integration. Table 6.3 presents the initial guesses used in the solution of the
optimisation problem and Table 6.4 shows the operating conditions considered for
the different configurations for the pretreatment, detoxification, evaporation and
SSCF stages.

6.4.1 Specifications
The specifications for the different models in each unit and the initial guesses use in
the optimisation problem are listed in Tables 6.3 and 6.4 and are also described as
follows:

1. For the pretreatment stage, the operating temperature and acid concentration
were selected based on the findings by Esteghlalian et al., (1997) which state
that, although higher temperatures and acid concentrations reduces the
reaction times and increase the conversion of hemicellulose, they also
increase the rate of degradation of xylose into by-products.

2. As stated in Section 3.8, the kinetic model for the detoxification of the acid
hydrolysates is very limited in terms of the operating conditions. The
parameters available in literature correspond to a temperature of 30 oC
(Purwadi et al., 2004). However, the mathematical model was obtained as a
function of the pH of the solution and, as seen in Section 3.8, the highest
levels of conversion of furfural can be achieved at a pH of 12.

3. Similarly, the operating conditions of the SSCF stage are limited since only
the enzymatic hydrolysis model is a function of temperature. The results
presented by Kadam et al., (2004) show a high production of glucose at high
temperatures. Morales-Rodriguez et al., (2011, 2012) suggest that the
operating temperature for the SSCF should also be high only under the
184
 assumption that the fermenting microorganisms can reproduce themselves at
those conditions. Current research (see Chapter 2) has shown that
thermotolerant strains are a viable option for the production of ethanol from
lignocellulosic biomass.

4. The percentage of evaporation has been suggested by different companies

that produce ethanol around the world to increase the concentration of sugars
and to ensure a reduction in the concentration of other volatile compounds in
the fermentation medium (BP, 2016b; Incauca, 2016).

6.4.2 Optimisation results – pretreatment stages

Table 6.5 presents the results of the optimisation of all the superstructures for the
pretreatment stages. It can be seen from Table 6.5 that the configuration and design
of the reactors and evaporator in all the configurations are very similar. This means
that this part of the process is mainly governed by the mass balances rather than the
energy. The largest reactor in the process is for SSCF which indicates that in order to
obtain the necessary concentration of ethanol, the residence time during fermentation
has to be long. Table 6.6 shows the results of the streams and concentrations of the
optimal design of the pretreatment stages of ethanol production process.

The conversions in the pretreatment, detoxification and SSC stages are 76%, 90%
and 83%, respectively. It can also be observed from Table 6.6 that components such
as cellobiose, furfural and cells have low concentrations in their respective outlet
streams. This indicates that, as expected, the operating conditions favoured the
selectivity of the desired products for each stage.

Table 6.5: Optimal results of the pretreatment stages for all the configurations
Process without Superstructure Superstructure
Variable
heat storage 1 2
Pretreatment reactor (m3) 3.21 3.20 3.21
Detoxification reactor (m3) 1.19 1.20 1.19
SSCF reactor (m3) 19.0 19.0 19.0
Inlet cells stream (kg/s) 0.17 0.16 0.16

185
Figure 6.3: Superstructure 2 of the ethanol production process from corn stover with heat storage using the heat released in the condenser to heat the
stream of Therminol 66

186
Figure 6.4: Results of the simulation and optimisation of the pretreatment stages of the ethanol production process

187
 Table 6.6: Results of the optimisation of the pretreatment stages and composition of the process streams (% w/w)
Stream Hemicellulose Xylose Furfural Cellulose Glucose Cellobiose Ethanol Cells Others* Water
(1) 1.98 - - 3.60 - - - - 4.96 89.46
(2) 0.48 1.39 0.11 3.60 - - - - 4.96 89.46
(3) - 1.59 0.12 - - - - - 1.10 97.19
(4) 3.80 - - 28.6 - - - - 31.9 35.61
(5) - 3.99 0.21 - - - - - 1.17 94.63
(6) - - - - - - - - 1.09** 98.91
(7) - 3.98 0.02 - - - - 9.37 94.63
(8) - - - - - - - - 3.76 96.24
(9) - 4.19 0.02 - - - - - 1.24 94.55
(10) 1.04 3.04 0.02 7.86 - - - - 9.66 78.38
(11) - - - - - - - 25.0 - 75.00
(12) 0.41 - 0.02 0.75 0.19 0.68 4.5 1.87 11.9 79.69

* Refers to lignin, ashes other by-products in the process

** Volatile components such as Sulphuric Acid found in the vapour stream in the evaporator
Values below 0.001 have been ignored

188
 Figure 6.5: Results of the optimisation of the separation stages without heat storage

Figure 6.6: Results of the simulation and optimisation of the separation stages with heat storage
(Superstructure 1)

189
Figure 6.7: Results of the simulation and optimisation of the separation stages with heat storage
(Superstructure 2)

6.4.3 Optimisation results – separation section

The results of the separation section are presented in Tables 6.7 to 6.9. These tables
include the concentration of ethanol at the top and bottom of the column, number of
trays in the distillation column, the feed tray location, the reflux ratio, the outlet
pressure of the pump, the permeate pressures in the pervaporation network, the
temperature of the retentate streams, and the final concentration of ethanol. The
membrane area is chosen based on the results presented in Chapter 5 where 250 m2
were used for a flow rate of 50 mole/sec. For all the configurations in this chapter the
flow rate to the separation system is almost twice the capacity of the configurations
in Chapter 5. Therefore, a membrane area of 500 m2 will be considered for the
pervaporation stages.

As presented in Chapter 5, the optimisation of the distillation column consists of

three sections: rectification, feed and stripping. Rectification and stripping have
initially 25 trays each as initial estimates. The feed section consists of three fixed
trays being where the middle tray is where the feed stream is fed. The pervaporation
network consists of three stages which were found to be less expensive according to
the results presented in Section 5.4.2.

190
 Table 6.7: Results of the optimisation of the separation stages without heat storage
(membrane area of 500 m²)
Distillation Pervaporation Pervaporation Pervaporation
Variable
column 1 2 3

Number of trays 40 - - -

Feed tray from top 32 - - -

Pressure column (kPa) 101.3 - - -

Heat duty reboiler (kW) 521.9 - - -

Reflux ratio 3.51 - - -

Ethanol concentration in distillate 0.93* - - -

Ethanol concentration bottoms 0.003 - - -

Number of modules - 1* 1* 4

Permeate pressure (Pa) - 6767 545.1 381.5

Temperature in retentate (ºC) - 52.41 51.59 55.52

Ethanol mass fraction in retentate - 0.95 0.98 0.995

Ethanol mass fraction in permeate - 0.11 0.12 0.44

* Values on lower or upper bounds

Table 6.8: Results of the optimisation of the separation stages in Superstructure 1

(membrane area of 500 m²)
Distillation Pervaporation Pervaporation Pervaporation
Variable
column 1 2 3

Number of trays 36 - - -

Feed tray from top 26 - - -

Pressure column (kPa) 935.9 - - -

Heat duty reboiler (kW) 579.7 - - -

Reflux ratio 4.68 - - -

Ethanol concentration in distillate 0.93* - - -

Ethanol concentration bottoms 0.003 - - -

Number of modules - 1* 1* 3

Permeate pressure (Pa) - 6313 134 134

Temperature in retentate (ºC) - 52.41 51.59 57.86

Ethanol mass fraction in retentate - 0.96 0.98 0.995

Ethanol mass fraction in permeate - 0.11 0.12 0.44

* Values on lower or upper bounds

191
 Table 6.9: Results of the optimisation of the separation stages in Superstructure 2
(membrane area of 500 m²)
Distillation Pervaporation Pervaporation Pervaporation
Variable
column 1 2 3

Number of trays 38 - - -

Feed tray from top 13 - - -

Pressure column (kPa) 872.3 - - -

Heat duty reboiler (kW) 566.9 - - -

Reflux ratio 4.43 - - -

Ethanol concentration in distillate 0.93* - - -

Ethanol concentration bottoms 0.003 - - -

Number of modules - 1* 1* 3

Permeate pressure (Pa) - 6701 356 12

Temperature in retentate (ºC) - 49.98 49.97 56.75

Ethanol mass fraction in retentate - 0.95 0.98 0.995

Ethanol mass fraction in permeate - 0.09 0.11 0.41

* Values on lower or upper bounds

Variables marked with an asterisk represent values on lower or upper bounds. In

Table 6.7, the optimisation for the column without heat integration shows that the
rectification section has 30 stages and the stripping section has 7 stages, giving a
total of trays of 40 with the feed stream being located in tray number 32.

For the distillation column in superstructure 1 with heat integration (see Table 6.8),
the number of trays is 36 with the feed location on tray 26. The design of this
distillation column includes a pump before the distillation column instead of a heat
exchanger. The state in which the feed enters the column is subcooled liquid whereas
in the original configuration (see Figure 6.1), the feed enters as saturated liquid.

As seen in Table 6.8, Superstructure 2 with heat integration has 38 trays and the feed
location is on tray 13. The design and operation of the distillation column is highly
influenced by the pressure in the column, the thermodynamic state and concentration
in which the feed stream enters the column.

192
 Table 6.10: Energy requirements for the different configurations of the ethanol process (kW)
With heat storage
Without
Unit Superstructure Superstructure
heat storage
1 2
Pretreatment 2386* 2386 2386
Evaporator 5806* 5806 5806
Detoxification -609.5 -609.6 -609.6
SSCF 368.5* 368.5 368.5
Condenser -489.9 -489.2 -481.7
Reboiler 521.9* 579.7* 566.9*
Heat exchanger 1 393.3* -38.2 -37.1
Heat exchanger 2 -3.5 6.9 7.9
Heat exchanger 3 7.7* 6.8 7.5
Heat exchanger 4 7.4* 7995* 8094*
Heat exchanger 5 - -578.8 -
Total energy demand 9491 8574 8660
* Heat sources that are considered in the calculation of the total energy demand

The pervaporation system for the configuration without heat integration (see Table
6.7) shows a total of 6 in-parallel module for a final concentration of ethanol of 99.5
% (w/w). The number of in-parallel modules in the pervaporation stages increases as
the permeate pressure decreases from 6767 Pa to 381 Pa which agrees with what
was presented in Chapter 4 (i.e. the lower the permeate pressure the higher the
separation rate through the membrane).

Superstructures 1 (see Table 6.8) and 2 (see Table 6.9) with heat integration present
similar behaviour compared with the base case without heat integration. One key
difference between the base case and the two superstructures with heat integration is
the number of in-parallel modules (only three modules in the last stage) which
directly correlates to the permeate pressure in each stage. The concentration of
ethanol in retentate for all the configurations is 99.5% (w/w) as specified.

193
 Table 6.11: Economic analysis for equipment used in the ethanol production process 1x103
(US$/year)
With heat storage
Without heat
Unit Superstructure Superstructure
storage
1 2
Pretreatment 30.83 30.83 30.83
Evaporator 106.8 106.8 106.8
Detoxification 21.19 21.19 21.19
SSCF 89.79 89.79 89.79
Condenser 12.16 12.15 12.08
Reboiler 21.64 23.17 22.83
Column shell 73.89 79.89 79.92
Column trays 2.41 2.73 2.70
Pervaporation modules 268.2 223.5 223.5
Heat exchanger 1 2.49 0.87 0.82
Heat exchanger 2 0.20 0.21 0.23
Heat exchanger 3 0.13 0.21 0.23
Heat exchanger 4 0.13 16.79 13.64
Heat exchanger 5 - 7.81 -
Compressor 16.32 17.46 25.63
Pump - 2.12 2.12
Heat storage tanks - 6.01 6.01
Total cost 646.2 641.5 638.3

Table 6.12: Economic analysis for utilities and services used in the ethanol production process
1x103 (US$/year)
Without heat Superstructure Superstructure
Item
storage 1 2
Cost of steam 2700 2439 2464
Cost of cooling water 5.08 2.81 2.81
Cost of Therminol 66 - 14.21 14.71
Cost of electricity 3.4 5.16 7.54
Total cost 2708.5 2461.2 2489.1

194
 Table 6.13: Area and cost for heat exchangers in all configurations
With heat storage
Without heat
Unit Superstructure Superstructure
storage
1 2
Cost 1x103 (US$/yr.)
Heat exchanger 1 2.49 0.87 0.82
Heat exchanger 2 0.20 0.21 0.23
Heat exchanger 3 0.13 0.21 0.23
Heat exchanger 4 0.13 16.79 13.64
Heat exchanger 5 - 7.81 -
Heat transfer area (m2)
Heat exchanger 1 3.35 0.66 0.59
Heat exchanger 2 0.07 0.07 0.08
Heat exchanger 3 0.04 0.07 0.08
Heat exchanger 4 0.04 63.03 45.75
Heat exchanger 5 - 19.42 -

6.4.4 Optimisation results – energy demand and economic analysis

Additional calculations are required to assess the overall performance of the plant in
terms of energy and cost. The results of energy demand in the process are shown in
Table 6.10. The energy consumption is indicated by positive values, whilst energy
released from a unit is represented by negative values. The total energy demand is
calculated as the amount of heat transferred from steam into the system and is
distinguished with an asterisk in Table 6.10. Therefore, the base case configuration
(i.e. without heat storage) takes into account the amount of energy in the
pretreatment reactor, evaporator, SSCF reactor, reboiler and intermediate heat
exchangers, whereas superstructures 1 and 2 only consider the contribution from the
reboilers and the heat exchanger where Therminol 66 needs to be re-heated.

A reduction in the energy requirements of 10 % and 8.7 % from the base case
configuration can be observed for superstructures 1 and 2 due to the implementation
of the heat storage tanks and the heat integration in the process. Figures 6.8 and 6.9
present the temperature of the heat transfer fluid and the flow rates across the

195
superstructures. Superstructure 2 operates at lower temperatures than Superstructure
1 and therefore requires higher flow rates (a total of 67.97 kg/sec against 39.36
kg/sec) in order to provide the necessary heat to the other units.

An economic analysis is presented in Tables 6.11 and 6.12 which will be helpful in
determining the most appropriate configuration. Table 6.11 presents the economic
evaluation of the equipment in the process, Table 6.12 shows an economic
assessment of the utilities and services and Table 6.13 presents the information of the
areas and costs of the heat exchangers. The total annual costs (TAC) for the base
case configuration, Superstructure 1 and Superstructure 2 are 3355, 3103, 3127
k$/year, respectively.

Compared to the original configuration, superstructure 1 reported a decrease in TAC

of 6.9 %, whereas Superstructure 2 had a reduction of 6.2 %. The costs of equipment
for the two superstructures are higher than the base case process since more units
such as heat exchangers and heat storage tanks are included. In the case of heat
exchangers, larger areas are required to heat up the stream of Therminol 66 that is
recycled back to HS1 which has a direct impact in the overall costs of equipment. On
the other hand, a reduction in the usage of utilities in both superstructures led to a
reduction of 9 % in the costs of utilities and services.

6.5 Conclusions
This work presented a comparative study between three configurations of the process
with and without heat integration using robust optimisation methods and dynamic
mass and energy balances. The novelty of this work lies on the methodology to solve
the system of equations and the optimisation of the process which had been
introduced in Chapter 5 with applications on bioprocesses. The overall goal of this
thesis was to prove that it was possible to reduce the energy consumption within the
process and therefore the investment costs in order to make the production of ethanol
for lignocellulosic biomass a more appealing alternative for energy generation from
renewable sources.

The optimisation of the process without heat integration showed that for higher
conversions, larger volumes were required. The results of operating conditions of the
196
units along with the corresponding constraints allowed obtaining low concentrations
of furfural, cellobiose and cells throughout the production of ethanol. The
optimisation also showed the arrangement of the separation section and the operating
conditions necessary to achieve the desired concentration of ethanol. For
Superstructure 1, the pressure inside the distillation column was higher than in
Superstructure 2, which directly influenced the heat duty in the reboiler and the
energy transferred to the stream of Therminol 66 leaving the reactor.

The implementation of the heat storage units showed a reduction of 6.9 % in the total
annual cost for Superstructure 1 which can be attributed to the usage of the heat
transfer fluid to substitute high pressure steam; and a reduction of TAC of 6.2 % was
obtained in Superstructure 2. The difference between these two configurations lies on
the recovery of heat in the stream of Therminol 66 coming out of HS2. For
Superstructure 1, the total flow rate of Therminol was lower than its counterpart and
the amount of heat transferred in Heat Exchanger 5 was higher.

Superstructure 1 is the optimal configuration for the ethanol production process since
it reduced the energy consumption in the pretreatment reactors and reduced the costs
related to equipment and usage of utilities in the process. Chapter 7 will summarise
the achievements and main conclusions of this thesis and will present proposals for
the future work for this project.

197
Figure 6.8: Temperatures of heat transfer fluid Therminol 66 across Superstructure 1

198
Figure 6.9: Temperatures of heat transfer fluid Therminol 66 across Superstructure 2

199
 Chapter 7 – Conclusions and future work

Abstract

This chapter summarises the work that has been presented in this thesis. Section 7.1
presents the contributions of this research to the modelling of the ethanol production
process, the formulation of the mathematical model of the organophilic membrane,
the optimisation of the separation stage and the heat integration of the process.
Section 7.2 presents the possible directions for future research and the application of
this work into different processes.

7.1 Conclusions
This worked has focused on developing a complete dynamic model of the production
of ethanol from lignocellulosic biomass and on the heat integration of this process
using heat transfer fluids to reduce the usage of external sources of heat such as
steam with the aims of finding more profitable and more efficient process
configurations. This thesis covered several aspects which are summarised as follows:

A review of the current status of fuel consumption around the world was presented in
Chapter 1. In this review, it could be seen that the consumption of crude oil and coal
has been increasing in the last decades which signifies a problem for society giving
that population and energy consumption go hand in hand. This chapter also presented
the global production of biofuels and how they represent a sustainable alternative for
energy generation from renewable sources. The chapter continued with the
production of ethanol and the different routes from which it can be produced (e.g.
hydration of ethylene and fermentation). In this work, the author opted for the
production of ethanol from lignocellulosic biomass, more specifically, from corn
stover. This raw material can be found in many countries and possess a high organic
content and has the potential to be used in the production of high-added-value
products of industrial interest.

200
In Chapter 2, a literature survey of the current state of the art on ethanol production
was introduced. Some works included the implementation of new pretreatment
methods, new strains, and new configurations for the separation stages. However,
most of these works only focused on individual units and operating conditions rather
than consider complete dynamic modelling and heat integration of the ethanol
production process which is the main objective of this thesis. All the work available
related to the optimisation of the process was only applicable to the design of the
separation stages and individual reaction units.

This thesis presented complete mathematical models for all units of the process that
were used in the calculations and subsequent simulations. In order to guarantee the
reliability and consistency of the simulations, each model was validated against at
least another source. In the cases of pretreatment, detoxification, enzymatic
hydrolysis, fermentation and pervaporation, these stages were validated against
experimental data. The distillation column was validated against simulations run in
ChemCAD which is a commercially available simulation tool.

This work presented the formulation of a mathematical model for a membrane used
in the separation of ethanol from the fermentation broth. This work, which was
developed in collaboration with TU Dortmund in Germany, consisted of a series of
experiments using a membrane called PERVAPTM 4060 provided by Sulzer. The
parameters for three different mathematical models were estimated using gPROMS
in order to evaluate the accuracy of different approaches. The results of these
calculations generated a reliable model which was able to accurately describe the
permeation of ethanol through the membrane which was later used in the
optimisation of the complete separation section. It is important to establish that the
implementation of membrane modules prior to the distillation system and their
impact on the design and energy consumption of the process have not been
investigated in the separation of ethanol from lignocellulosic hydrolysates.

A methodology for the optimisation of hybrid processes for the separation of ethanol
was also developed in this thesis. This methodology took individual parts of the
separation section (i.e. the pervaporation modules or the distillation column) and
optimised their design and operating conditions to later use their results as initial
201
guesses to optimise the entire hybrid process. This methodology is very practical
since it seeks for feasible regions in which the combined process can find a sensible
solution that satisfies the specifications of purity and energy consumption. Two main
configurations were compared: The first one consisted of a single distillation column
linked to a network of hydrophilic membranes. The second one included the
organophilic membrane before the distillation column. The results showed that
having the organophilic membrane could indeed reduce the heat duty in the reboiler
and the size of the distillation column and the pervaporation network. However, the
overall energy consumption and the Total Annualised Cost (TAC) did not support
the implementation of this membrane due to the requirement of additional units such
as compressors and heat exchangers. A sensitivity analysis was also included in this
work. The purpose of this analysis was to determine the variations in the optimal
design of both superstructures focusing mainly on the structure with the organophilic
membrane since this is the novelty of this work. The results showed that in all case
studies, the implementation of the organophilic membrane was more expensive and
more energy intensive than the configuration without said membrane.

Finally, the last section of this work focused on the optimisation and heat integration
of the ethanol production process from corn stove and compared three
configurations: The first configuration was a standard arrangement of the process as
presented in Chapter 3. This configuration served as the base case for the results of
the optimisation with heat integration.

The second configuration presented the introduction of heat storage units containing
Therminol 66 to provide heat to the main reactors and to recover the energy from the
stream at the bottom of the column. The third configuration also implemented heat
storage units but it used the heat released by the condenser to recover heat with the
transfer fluid.

The results of these optimisations showed that using heat transfer fluids reduced the
energy consumption in the process by 10% and the Total Annualised Cost (TAC) by
7%. Although these numbers seemed relatively small, they do have an impact,
especially in processes with high production rates. This thesis showed that biofuels
can be a competitive source of energy in the mainstream energy market and the
202
usage of lignocellulosic biomass can still be an appealing alternative to reduce the
consumption of oil and coal and to mitigate the harmful effects of global warming.
This production route is still far from being successfully implemented in industry for
sustainable energy generation and there are still issues that need to be addressed in
the future and they are suggested in the following section.

7.2 Future work

This section focuses on the areas that this thesis did not cover and could potentially
be interesting to address in future research. One of the most important aspects of the
thesis was the optimisation of the separation stages for a binary mixture
ethanol/water. However, the separation section of the process presented in this work
did not consider other components in the purification of ethanol. This assumption
was adopted because the concentration of most reducing sugars and soluble acids is
very low in the outlet streams of the fermentation stage. However, these compounds
could have a significant influence over the performance of some of the units in the
separation section, especially in the pervaporation modules. Compounds such as
acetic acid and other organic acids tend to reduce the lifetime of a membrane as seen
in Chapter 4 for the PERVAP 4060TM organophilic membrane. This could represent
an increase in the heat duty in the reboiler which also translates into more expensive
operating conditions and design of the process. For the organophilic membrane
model, it would be beneficial to include the effect of other by-products from the
pretreatment. As seen in Chapter 4, by-products from pretreatment and fermentation
influence the total flux through the membrane. The model would require more
experimental data in order to estimate the parameters in the model to describe the
separation between feed and permeate sides.

Another aspect that should be considered in the simulation of the separation stages is
the presence of solids. Soluble and insoluble solids produced during the fermentation
stage can also influence the design of the distillation columns as well as the heat duty
in the reboiler. This aspect not only affects the design of the unit but also affects the
maintenance and control of the distillation columns which are directly correlated to
the components found in the feed stream.

203
 Figure 7.1: Flowsheet of possible scenarios for the optimisation of the separation section

Chapter 5 presented the optimisation of the hybrid processes introducing a

methodology in which sub-optimal solutions, obtained from individual units, were
used as initial guesses for the solution of the MINLP problem of the overall
separation section. This methodology was found to be very practical and showed
reliable results in the design of the separation section of the process. An alternative
way to optimise this part of the process would be the introduction of binary variables
to determine which unit is included and active within the process. For instance, the
solution of the MINLP problem should provide the number of distillation columns
and the number of in-series pervaporation stages as outcomes of the optimisation of
the separation section. Figure 7.1 illustrates this proposal:

In Figure 7.1, the dashed blue lines represent the possible alternatives that could be
part of the optimal configuration. Scenario “I” represents the selection of the number
of distillation columns that could possibly be included in the separation of ethanol.
Scenarios “II” to “IV” indicate the number of pervaporation stages that could be
required during the dehydration of ethanol. In the dehydration section only one
scenario can take place. For instance, when Scenario “II” is considered, Scenarios
“III” and “IV” should not be included in the process.

This approach would reduce the number of calculations since it would have all the
units combined to optimise all the separation section at once. Although this

204
methodology seems to be more practical, it is important to remember that the system
of equations that describe the process is highly non-linear and the optimisation
thereof could take longer computational times.

The heat integration process presented in Chapter 6 introduced the optimisation of

the entire process using heat storage units. This methodology was implemented with
aims to reduce the usage of utilities which have a significant effect in the
minimisation of the total annualised costs. Something that should be considered for
future work is the generation of energy from the unreacted lignocellulosic solids in
the process. Several authors have presented works related to the generation of energy
from the unreacted material in order to make the process more sustainable and more
feasible for industrial applications. This approach will attempt to solve two issues in
the production of ethanol: one is the reduction in the usage of utilities and the other
one is the reduction and management of solids with high organic content coming out
of the process. This will ensure a more efficient way to produce ethanol from any
lignocellulosic material without risking the environment and natural sources.

Finally, this work could be expanded for the production of other metabolites of
industrial applications such as butanol, biopolymers, biodiesel and organic acids
from renewable materials. There is a demand for greener and more efficient
technologies that can solve the problem of agroindustrial pollution as well as the
management of solids or other residues that could potentially become into valuable
chemicals. The work developed in this thesis follows a series of logical steps, in
which every unit required is evaluated leading to a reliable simulation of a complete
process with applicability into any metabolite from lignocellulosic biomass. This will
ensure a reduction of energy consumption, costs of raw materials and an increase in
the sustainability and profitability of the process.

205
List of publications and events

1. D Lorenzo, C Perez-Galvan, CF Triana, A Santos, A Romero, IDL Bogle,

2016, Modelling of a reactive distillation in the production process of high
purity Cyclohexanone to produce Caprolactam. Proceedings of ESCAPE
2016. Portorož, Slovenia

2. CF Triana, ES Fraga & E Sorensen, 2015, Evaluation of the energy

consumption of the production of bioethanol from lignocellulosic biomass.
Proceedings of ESCAPE 2015 and PSE 2015. Copenhagen, Denmark

3. CF Triana, P Lutze, ES Fraga, E Sorensen, 2014, Implementation of an

organophilic membrane system to remove ethanol from lignocellulosic
biomass hydrolysates. AIChE Annual Meeting. Atlanta, GA, USA

4. CF Triana, ES Fraga & E Sorensen, 2014, Energy efficient separation

alternatives for the production of ethanol from lignocellulosic biomass. 10th
International Conference on Distillation & Absorption 2014. Friedrichshafen,
Germany

5. 125th BASF International Summer School. Ludwigshafen, Germany

6. CF Triana, ES Fraga & E Sorensen, 2013, Optimal energy consumption in the

production of bioethanol from lignocellulosic biomass. AIChE Annual
Meeting. San Francisco, CA, USA

206
 REFERENCES

Abedinifar, S., Karimi, K., Khanahmadi, M., Taherzadeh, M.J., 2009. Ethanol
production by Mucor indicus and Rhizopus oryzae from rice straw by separate
hydrolysis and fermentation. Biomass and Bioenergy 33, 828–833.

Agbogbo, F.K., Coward-Kelly, G., Torry-Smith, M., Wenger, K.S., 2006.

Fermentation of glucose/xylose mixtures using Pichia stipitis. Process Biochem.
41, 2333–2336.

Aguilar, R., Ramirez, J.A., Garrote, G., Vazquez, M., 2002. Kinetic study of the acid
hydrolysis of sugar cane bagasse. J. Food Eng. 55, 309–318.

Altuntop, N., Arslan, M., Ozceyhan, V., Kanoglu, M., 2005. Effect of obstacles on
thermal stratification in hot water storage tanks. Appl. Therm. Eng. 25, 2285–
2298.

Amusat, O., Shearing, P., Fraga, E.S., 2015. System Design of Renewable Energy
Generation and Storage Alternatives for Large Scale Continuous Processes, in:
Computer Aided Chemical Engineering. pp. 2279–2284.

AN, W., YU, F., DONG, F., HU, Y., 2008. Simulated Annealing Approach to the
Optimal Synthesis of Distillation Column with Intermediate Heat Exchangers.
Chinese J. Chem. Eng. 16, 30–35.

Avci, A., Saha, B.C., Dien, B.S., Kennedy, G.J., Cotta, M.A., 2013. Response
surface optimization of corn stover pretreatment using dilute phosphoric acid
for enzymatic hydrolysis and ethanol production. Bioresour. Technol. 130, 603–
12.

Baeyens, J., Kang, Q., Appels, L., Dewil, R., Lv, Y., Tan, T., 2015. Challenges and
opportunities in improving the production of bio-ethanol. Prog. Energy
Combust. Sci. 47, 60–88.

Bai, F.W., Anderson, W.A., Moo-Young, M., 2008. Ethanol fermentation

technologies from sugar and starch feedstocks. Biotechnol. Adv. 26, 89–105.

Baker, R., 2012. Pervaporation, in: Membrane Technology and Applications. Wiley,
pp. 355–392.

207
Balat, M., Balat, H., Öz, C., 2008. Progress in bioethanol processing. Prog. Energy
Combust. Sci. 34, 551–573.

Ballesteros, M., Oliva, J.M., Negro, M.J., Manzanares, P., Ballesteros, I., 2004.
Ethanol from lignocellulosic materials by a simultaneous saccharification and
fermentation process (SFS) with Kluyveromyces marxianus CECT 10875.
Process Biochem. 39, 1843–1848.

Balusu, R., Paduru, R.R., Kuravi, S.K., Seenayya, G., Reddy, G., 2005. Optimization
of critical medium components using response surface methodology for ethanol
production from cellulosic biomass by Clostridium thermocellum SS19. Process
Biochem. 40, 3025–3030.

Baños, R., Manzano-Agugliaro, F., Montoya, F.G., Gil, C., Alcayde, A., Gómez, J.,
2011. Optimization methods applied to renewable and sustainable energy: A
review. Renew. Sustain. Energy Rev. 15, 1753–1766.

Bansal, P., Hall, M., Realff, M.J., Lee, J.H., Bommarius, A.S., 2009. Modeling
cellulase kinetics on lignocellulosic substrates. Biotechnol. Adv. 27, 833–848.

Beck, M., Johnson, R., Baker, C., 1990. Comparison of three commercial cellulases
for production of glucose from acid-treated hardwood. Appl. Biochem.
Biotechnol. 24–25, 407–414.

Behera, S., Mohanty, R.C., Ray, R.C., 2010. Comparative study of bio-ethanol
production from mahula (Madhuca latifolia L.) flowers by Saccharomyces
cerevisiae and Zymomonas mobilis. Appl. Energy 87, 2352–2355.

Ben Soltane, H., Roizard, D., Favre, E., 2013. Effect of pressure on the swelling and
fluxes of dense PDMS membranes in nanofiltration: An experimental study. J.
Memb. Sci. 435, 110–119.

Berrin, J.-G., Herpoel-Gimbert, I., Ferreira, N.L., Margeot, A., Heiss-Blanquet, S.,
2014. Biotechnology and Biology of Trichoderma, Biotechnology and Biology
of Trichoderma. Elsevier.

Bezerra, R., Dias, A., 2005. Enzymatic kinetic of cellulose hydrolysis. Appl.
Biochem. Biotechnol. 126, 49–59.

Bhandari, N., Macdonald, D.G., Bakhshi, N.N., 1984. Kinetic studies of corn stover

208
 saccharification using sulphuric acid. Biotechnol. Bioeng. 26, 320–327.

Biegler, L.T., Grossmann, I.E., Westerberg, A.W., 1999. Systematic methods of

chemical process design. Prentice-Hall.

Binod, P., Sindhu, R., Singhania, R.R., Vikram, S., Devi, L., Nagalakshmi, S.,
Kurien, N., Sukumaran, R.K., Pandey, A., 2010. Bioethanol production from
rice straw: An overview. Bioresour. Technol. 101, 4767–4774.

Blum, C., Rol, A., Alba, E., 2005. Parallel metaheuristics: a new class of algorithms.
Wiley.

BP, 2016a. BP statistical review of world energy June 2016.

BP, 2016b. Etanol da cana de acucar - BP Brazil [WWW Document].

BP, 2012. BP statistical review of world energy June 2012.

Bryson, A.E., Ho, Y.-C., 1975. Applied optimal control: optimization, estimation and
control, 1st Editio. ed. CRC Press.

Caballero, J.A., 2015. Logic hybrid simulation-optimization algorithm for distillation

design. Comput. Chem. Eng. 72, 284–299.

Caballero, J.A., Grossmann, I.E., 2014a. Chapter 11 - Optimization of Distillation

Processes, in: Sorensen, A.G. (Ed.), Distillation. Academic Press, Boston, pp.
437–496.

Caballero, J.A., Grossmann, I.E., 2014b. Optimal synthesis of thermally coupled

distillation sequences using a novel {MILP} approach. Comput. Chem. Eng. 61,
118–135.

Cara, C., Moya, M., I, B., José, M., 2007. Influence of solid loading on enzymatic
hydrolysis of steam exploded or liquid hot water pretreated olive tree biomass.
Process Biochem. 42, 1003–1009.

Cardona, C.A., Quintero, J.A., Paz, I.C., 2010. Production of bioethanol from
sugarcane bagasse: Status and perspectives. Bioresour. Technol. 101, 4754–
4766.

Cardona, C.A., Sánchez, Ó.J., 2007. Fuel ethanol production: Process design trends
and integration opportunities. Bioresour. Technol. 98, 2415–2457.

209
Carrasco, C., Baudel, H., Penarrieta, M., Solano, C., Tejeda, L., Roslander, C.,
Galbe, M., Liden, G., 2011. Steam pretreatment and fermentation of the straw
material Paja Brava using simultaneous saccharification and co-fermentation. J.
Biosci. Bioeng. 111, 167–174.

Chandel, A.K., Kapoor, R.K., Singh, A., Kuhad, R.C., 2007. Detoxification of
sugarcane bagasse hydrolysate improves ethanol production by Candida
shehatae NCIM 3501. Bioresour. Technol. 98, 1947–1950.

Chemstations, 2015. ChemCAD.

Chen, H.-G., Zhang, Y.-H.P., 2015. New biorefineries and sustainable agriculture:
Increased food, biofuels, and ecosystem security. Renew. Sustain. Energy Rev.
47, 117–132.

Cheng, J.-R., Liu, X.-M., Chen, Z.-Y., 2015. Methane Production from Rice Straw
Hydrolysate Treated with Dilute Acid by Anaerobic Granular Sludge. Appl.
Biochem. Biotechnol. 178, 9–20.

Cho, D.H., Shin, S.J., Bae, Y., Park, C., Kim, Y.H., 2010. Enhanced ethanol
production from deacetylated yellow poplar acid hydrolysate by Pichia stipitis.
Bioresour. Technol. 101, 4947–4951.

Chovau, S., Gaykawad, S., Straathof, A., der Bruggen, B., 2011. Influence of
fermentation by-products on the purification of ethanol from water using
pervaporation. Bioresour. Technol. 102, 1669–1674.

Chung, Y.-H., Peng, T.-H., Lee, H.-Y., Chen, C.-L., Chien, I.-L., 2015. Design and
Control of Reactive Distillation System for Esterification of Levulinic Acid and
n-Butanol. Ind. Eng. Chem. Res. 54, 3341–3354.

Claes, S., Vandezande, P., Mullens, S., De Sitter, K., Peeters, R., Van Bael, M.K.,
2012. Preparation and benchmarking of thin film supported PTMSP-silica
pervaporation membranes. J. Memb. Sci. 389, 265–271.

Claes, S., Vandezande, P., Mullens, S., Leysen, R., De Sitter, K., Andersson, A.,
Maurer, F.H.J., Van den Rul, H., Peeters, R., Van Bael, M.K., 2010. High flux
composite PTMSP-silica nanohybrid membranes for the pervaporation of
ethanol/water mixtures. J. Memb. Sci. 351, 160–167.

210
Čuček, L., Martín, M., Grossmann, I.E., Kravanja, Z., 2011. Energy, water and
process technologies integration for the simultaneous production of ethanol and
food from the entire corn plant. Comput. Chem. Eng. 35, 1547–1557.

Davis, L., Jeon, Y.J., Svenson, C., Rogers, P., Pearce, J., Peiris, P., 2005. Evaluation
of wheat stillage for ethanol production by recombinant Zymomonas mobilis.
Biomass and Bioenergy 29, 49–59.

Demirbas, A., 2007. Progress and recent trends in biofuels. Prog. Energy Combust.
Sci. 33, 1–18.

Diaz, A.B., Moretti, M.M. de S., Bezerra-Bussoli, C., Carreira Nunes, C. da C.,
Blandino, A., da Silva, R., Gomes, E., 2015. Evaluation of microwave-assisted
pretreatment of lignocellulosic biomass immersed in alkaline glycerol for
fermentable sugars production. Bioresour. Technol. 185, 316–23.

Dien, B.S., Hespell, R.B., Wyckoff, H.A., Bothast, R.J., 1998. Fermentation of
hexose and pentose sugars using a novel ethanologenic Escherichia coli strain.
Enzyme Microb. Technol. 23, 366–371.

Dieterle, F., Kieser, B., Gauglitz, G., 2003. Genetic algorithms and neural networks
for the quantitative analysis of ternary mixtures using surface plasmon
resonance. Chemom. Intell. Lab. Syst. 65, 67–81.

Dinçer, İ., Rosen, M.A., 2010. Energy Storage Systems, in: Thermal Energy Storage.
John Wiley & Sons, Ltd, pp. 51–82.

Divne, C., Stahlberg, J., Tuula, T.T., TA., J., 1998. High-resolution crystal structures
reveal how a cellulose chain is bound in the 50 A long tunnel of
cellobiohydrolase I from Trichoderma reesei. J. Mol. Biol. 275, 309–325.

Drapcho, C.M., Nhuan, N.P., Walker, T.H., 2008. Ethanol Production, Biofuels
Engineering Process Technology, The McGraw-Hill Companies. McGraw-Hill.

Dünnebier, G., Pantelides, C.C., 1999. Optimal Design of Thermally Coupled

Distillation Columns. Ind. Eng. Chem. Res. 38, 162–176.

Eastman, 2016. Therminol.

Elgharbawy, A.A., Alam, M.Z., Moniruzzaman, M., Goto, M., 2016. Ionic liquid
pretreatment as emerging approaches for enhanced enzymatic hydrolysis of

211
 lignocellulosic biomass. Biochem. Eng. J. 109, 252–267.

Engel, P., Krull, S., Seiferheld, B., Spiess, A.C., 2012. Rational approach to optimize
cellulase mixtures for hydrolysis of regenerated cellulose containing residual
ionic liquid. Bioresour. Technol. 115, 27–34.

Erdei, B., Galbe, M., Zacchi, G., 2013. Simultaneous saccharification and co-
fermentation of whole wheat in integrated ethanol production. Biomass and
Bioenergy 56, 506–514.

Escobar, J.C., Lora, E.S., Venturini, O.J., Yáñez, E.E., Castillo, E.F., Almazan, O.,
2009. Biofuels: Environment, technology and food security. Renew. Sustain.
Energy Rev. 13, 1275–1287.

Esteghlalian, A., Hashimoto, A.G., Fenske, J.J., Penner, M.H., 1997. Modeling and
optimization of the dilute-sulfuric-acid pretreatment of corn stover, poplar and
switchgrass. Bioresour. Technol. 59, 129–136.

Fan, S., Xiao, Z., Zhang, Y., Tang, X., Chen, C., Li, W., Deng, Q., Yao, P., 2014.
Enhanced ethanol fermentation in a pervaporation membrane bioreactor with
the convenient permeate vapor recovery. Bioresour. Technol. 155, 229–234.

Feczkó, T., Trif, L., Horák, D., 2016. Latent heat storage by silica-coated polymer
beads containing organic phase change materials. Sol. Energy 132, 405–414.

Felix, E., Tilley, D.R., 2009. Integrated energy, environmental and financial analysis
of ethanol production from cellulosic switchgrass. Energy 34, 410–436.

Fornell, R., Berntsson, T., 2012. Process integration study of a kraft pulp mill
converted to an ethanol production plant – Part A: Potential for heat integration
of thermal separation units. Appl. Therm. Eng. 35, 81–90.

Freer, S.N., Detroy, R.W., 1983. Characterization of cellobiose fermentations to

ethanol by yeasts. Biotechnol. Bioeng. 25, 541–557.

Freer, S.N., Detroy, R.W., 1982. Direct fermentation of cellodextrins to ethanol by

Candida wickerhamii and C. Lusitaniae. Biotechnol. Lett. 4, 453–458.

Garcia, J.F., Cuevas, M., Bravo, V., Sanchez, S., 2010. Ethanol production from
olive prunings by autohydrolysis and fermentation with Candida tropicalis.
Renew. Energy 35, 1602–1608.

212
García, V., Pongrácz, E., Phillips, P.S., Keiski, R.L., 2013. From waste treatment to
resource efficiency in the chemical industry: recovery of organic solvents from
waters containing electrolytes by pervaporation. J. Clean. Prod. 39, 146–153.

Gaykawad, S.S., van der Wielen, L.A.M., Straathof, A.J.J., 2012. Effects of yeast-
originating polymeric compounds on ethanol pervaporation. Bioresour.
Technol. 116, 9–14.

Gaykawad, S.S., Zha, Y., Punt, P.J., van Groenestijn, J.W., van der Wielen, L.A.M.,
Straathof, A.J.J., 2013. Pervaporation of ethanol from lignocellulosic
fermentation broth. Bioresour. Technol. 129, 469–476.

Geddes, C., Mullinnix, M., Nieves, I., Peterson, J., Hoffman, H., York, S., Yomano,
L., Miller, E., Shanmugam, K., Ingram, L., 2011. Simplified process for ethanol
production from sugarcane bagasse using hydrolysate-resistant Escherichia coli
strain MM160. Bioresour. Technol. 102, 2702–2711.

Gendreau, M., Potvin, J.-Y., 2010. Handbook of metaheuristics, Internatio. ed.

Springer.

Gil, I.D., Gómez, J.M., Rodríguez, G., 2012. Control of an extractive distillation
process to dehydrate ethanol using glycerol as entrainer. Comput. Chem. Eng.
39, 129–142.

Golias, H., Dumsday, G.J., Stanley, G.A., Pamment, N.B., 2002. Evaluation of a
recombinant Klebsiella oxytoca strain for ethanol production from cellulose by
simultaneous saccharification and fermentation: comparison with native
cellobiose-utilising yeast strains and performance in co-culture with
thermotolerant yeas. J. Biotechnol. 96, 155–168.

Gomez-Castro, F.I., Rodriguez-Angeles, M.A., Segovia-Hernandez, J.G., Gutierrez-

Antonio, C., Briones-Ramirez, A., 2011. Optimal Designs of Multiple Dividing
Wall Columns. Chem. Eng. Technol. 34, 2051–2058.

Gorak, A., Sorensen, E., 2014. Principles of binary distillation, in: Distillation:
Fundamentals and Principles. Academic Press, pp. 145–186.

Gorji-Bandpy, M., Yahyazadeh-Jelodar, H., Khalili, M., 2011. Optimization of heat

exchanger network. Appl. Therm. Eng. 31, 779–784.

213
Grisales, R., Cardona, C.A., Gutierrez, L.F., Sanchez, O.J., 2008. Heat integration of
fermentation and recovery steps for fuel ethanol production from lignocellulosic
biomass, in: 2nd Mercosur Congress on Chemical Engineering and 4th
Mercosur Congress on Process Systems Engineering. Costa Verde, Rio De
Janeiro.

Haelssig, J.B., Tremblay, A.Y., Thibault, J., 2012. A new hybrid membrane
separation process for enhanced ethanol recovery: Process description and
numerical studies. Chem. Eng. Sci. 68, 492–505.

Hahn-Hagerdal, B., Galbe, M., Gorwa-Grauslund, M.F., Liden, G., Zacchi, G., 2006.
Bio-ethanol: the fuel of tomorrow from the residues of today. Trends
Biotechnol. 24, 549–556.

Han, Y.M., Wang, R.Z., Dai, Y.J., 2009. Thermal stratification within the water tank.
Renew. Sustain. Energy Rev. 13, 1014–1026.

Hasunuma, T., Kondo, A., 2012. Consolidated bioprocessing and simultaneous

saccharification and fermentation of lignocellulose to ethanol with
thermotolerant yeast strains. Process Biochem. 47, 1287–1294.

Hoekman, S.K., 2009. Biofuels in the U.S. – Challenges and Opportunities. Renew.
Energy 34, 14–22.

Holtbruegge, J., Wierschem, M., Steinruecken, S., Voss, D., Parhomenko, L., Lutze,
P., 2013. Experimental investigation, modeling and scale-up of hydrophilic
vapor permeation membranes: Separation of azeotropic dimethyl
carbonate/methanol mixtures. Sep. Purif. Technol. 118, 862–878.

Holtzapple, M., Cognata, M., Shu, Y., Hendrickson, C., 1990. Inhibition of
Trichoderma reesei cellulase by sugars and solvents. Biotechnol. Bioeng. 36,
275–287.

Huang, C.-F., Lin, T.-H., Guo, G.-L., Hwang, W.-S., 2009. Enhanced ethanol
production by fermentation of rice straw hydrolysate without detoxification
using a newly adapted strain of Pichia stipitis. Bioresour. Technol. 100, 3914–
3920.

Huang, H.-J., Ramaswamy, S., Al-Dajani, W., Tschirner, U., Cairncross, R.A., 2009.
Effect of biomass species and plant size on cellulosic ethanol: A comparative
214
 process and economic analysis. Biomass and Bioenergy 33, 234–246.

Huang, Z., Guan, H., lee Tan, W., Qiao, X.-Y., Kulprathipanja, S., 2006.
Pervaporation study of aqueous ethanol solution through zeolite-incorporated
multilayer poly(vinyl alcohol) membranes: Effect of zeolites. J. Memb. Sci.
276, 260–271.

Incauca, 2016. Incauca S.A. Alcohol carburante [WWW Document].

Infochem, 2013. Multiflash.

Ishola, M., Jahandideh, A., Haidarian, B., Brandberg, T., Taherzadeh, M., 2013.
Simultaneous saccharification, filtration and fermentation (SSFF): A novel
method for bioethanol production from lignocellulosic biomass. Bioresour.
Technol. 133, 68–73.

Jackson de Moraes Rocha, G., Martin, C., Soares, I.B., Souto Maior, A.M., Baudel,
H.M., Moraes de Abreu, C.A., 2011. Dilute mixed-acid pretreatment of
sugarcane bagasse for ethanol production. Biomass and Bioenergy 35, 663–670.

Jamai, L., Ettayebi, K., Yamani, J.E., Ettayebi, M., 2007. Production of ethanol from
starch by free and immobilized Candida tropicalis in the presence of amylase.
Bioresour. Technol. 98, 2765–2770.

Jeon, Y.J., Svenson, C.J., Joachimsthal, E.L., Rogers, P.L., 2002. Kinetic analysis of
ethanol production by an acetate-resistant strain of recombinant Zymomonas
mobilis. Biotechnol. Lett. 24, 819–824.

Jervis, E.J., Haynes, C.A., Kilburn, D.G., 1997. Surface Diffusion of Cellulases and
Their Isolated Binding Domains on Cellulose. J. Biol. Chem. 272, 24016–
24023.

Ji, X.J., Huang, H., Du, J., Zhu, J.G., Ren, L.J., Li, S., Nie, Z.K., 2009. Development
of an industrial medium for economical 2,3-butanediol production through co-
fermentation of glucose and xylose by Klebsiella oxytoca. Bioresour. Technol.
100, 5214–5218.

Jin, M., Gunawan, C., Balan, V., Lau, M.W., Dale, B.E., 2012. Simultaneous
saccharification and co-fermentation (SSCF) of AFEX(TM) pretreated corn
stover for ethanol production using commercial enzymes and Saccharomyces

215
 cerevisiae 424A(LNH-ST). Bioresour. Technol. 110, 587–94.

Jin, M., Lau, M.W., Balan, V., Dale, B.E., 2010. Two-step SSCF to convert AFEX-
treated switchgrass to ethanol using commercial enzymes and Saccharomyces
cerevisiae 424A(LNH-ST). Bioresour. Technol. 101, 8171–8.

Jung, Y.H., Kim, I.J., Kim, H.K., Kim, K.H., 2013. Dilute acid pretreatment of
lignocellulose for whole slurry ethanol fermentation. Bioresour. Technol. 132,
109–14.

Kadam, K.L., Forrest, L.H., Jacobson, W.A., 2000. Rice straw as a lignocellulosic
resource: collection, processing, transportation, and environmental aspects.
Biomass and Bioenergy 18, 369–389.

Kadam, K.L., Rydholm, E.C., McMillan, J.D., 2004. Development and Validation of
a Kinetic Model for Enzymatic Saccharification of Lignocellulosic Biomass.
Biotechnol. Prog. 20, 698–705.

Kadar, Z., Szengyel, Z., Reczey, K., 2004. Simultaneous saccharification and
fermentation (SSF) of industrial wastes for the production of ethanol. Ind. Crops
Prod. 20, 103–110.

Kamath, R.S., Grossmann, I.E., Biegler, L.T., 2010. Aggregate models based on
improved group methods for simulation and optimization of distillation systems.
Comput. Chem. Eng. 34, 1312–1319.

Karimi, K., Emtiazi, G., Taherzadeh, M.J., 2006. Production of ethanol and mycelial
biomass from rice straw hemicellulose hydrolyzate by Mucor indicus. Process
Biochem. 41, 653–658.

Karimi, K., Emtiazi, G., Taherzadeh, M.J., 2006. Ethanol production from dilute-acid
pretreated rice straw by simultaneous saccharification and fermentation with
Mucor indicus, Rhizopus oryzae, and Saccharomyces cerevisiae. Enzyme
Microb. Technol. 40, 138–144.

Karlsson, H., Barjesson, P., Hansson, P., Ahlgren, S., 2014. Ethanol production in
biorefineries using lignocellulosic feedstock-GHG performance, energy balance
and implications of life cycle calculation methodology. J. Clean. Prod. 83, 420–
427.

216
Kim, Y., Hendrickson, R., Mosier, N.S., Ladisch, M.R., Bals, B., Balan, V., Dale,
B.E., 2008. Enzyme hydrolysis and ethanol fermentation of liquid hot water and
{AFEX} pretreated distillers’ grains at high-solids loadings. Bioresour.
Technol. 99, 5206–5215.

Kiran, B., Jana, A.K., 2015. A hybrid heat integration scheme for bioethanol
separation through pressure-swing distillation route. Sep. Purif. Technol. 142,
307–315.

Kiss, A.A., Ignat, R.M., 2012. Innovative single step bioethanol dehydration in an
extractive dividing-wall column. Sep. Purif. Technol. 98, 290–297.

Kiss, A.A., Olujic, Z., 2014. A review on process intensification in internally heat-
integrated distillation columns. Chem. Eng. Process. Process Intensif. 86, 125–
144.

Kiss, A.A., Suszwalak, D.J.-. P.C., 2012. Enhanced bioethanol dehydration by

extractive and azeotropic distillation in dividing-wall columns. Sep. Purif.
Technol. 86, 70–78.

Klasson, K.T., Dien, B.S., Hector, R.E., 2013. Simultaneous detoxification,

saccharification, and ethanol fermentation of weak-acid hydrolyzates. Ind.
Crops Prod. 49, 292–298.

Knapp, J.P., Doherty, M.F., 1992. A new pressure-swing-distillation process for

separating homogeneous azeotropic mixtures. Ind. Eng. Chem. Res. 31, 346–
357.

Knopf, F.C., Okos, M.R., Reklaitis, G. V., 1982. Optimal design of

batch/semicontinuous processes. Ind. Eng. Chem. Process Des. Dev. 21, 79–86.

Koch, K., Gorak, A., 2014. Pervaporation of binary and ternary mixtures of acetone,
isopropyl alcohol and water using polymeric membranes: Experimental
characterisation and modelling. Chem. Eng. Sci. 115, 95–114.

Koch, K., Sudhoff, D., Kreiß, S., Górak, A., Kreis, P., 2013. Optimisation-based
design method for membrane-assisted separation processes. Chem. Eng.
Process. Process Intensif. 67, 2–15.

Koizumi, T., 2015. Biofuels and food security. Renew. Sustain. Energy Rev. 52,

217
 829–841.

Koltuniewicz, A., 2010. 4.05 - Integrated Membrane Operations in Various Industrial

Sectors, in: Drioli, E., Giorno, L. (Eds.), Comprehensive Membrane Science
and Engineering. Elsevier, Oxford, pp. 109–164.

Kookos, I.K., 2003. Optimal Design of Membrane/Distillation Column Hybrid

Processes. Ind. Eng. Chem. Res. 42, 1731–1738.

Kravanja, P., Modarresi, A., Friedl, A., 2013. Heat integration of biochemical
ethanol production from straw – A case study. Appl. Energy 102, 32–43.

Kreis, P., Górak, A., 2006. Process Analysis of Hybrid Separation Processes. Chem.
Eng. Res. Des. 84, 595–600.

Krishnan, M., Ho, N., Tsao, G., 1999. Fermentation kinetics of ethanol production
from glucose and xylose by recombinant Saccharomyces 1400(pLNH33). Appl.
Biochem. Biotechnol. 78, 373–388.

Krummenacher, P., 2002. Contribution to the heat integration of batch processes

(with or without heat storage). École polytechnique fédérale de Lausanne.

Kumar, P., Barrett, D.M., Delwiche, M.J., Stroeve, P., 2009. Methods for
Pretreatment of Lignocellulosic Biomass for Efficient Hydrolysis and Biofuel
Production. Ind. Eng. Chem. Res. 48, 3713–3729.

Kumar, R., Wyman, C.E., 2008. An improved method to directly estimate cellulase
adsorption on biomass solids. Enzyme Microb. Technol. 42, 426–433.

Kunnakorn, D., Rirksomboon, T., Siemanond, K., Aungkavattana, P.,

Kuanchertchoo, N., Chuntanalerg, P., Hemra, K., Kulprathipanja, S., James,
R.B., Wongkasemjit, S., 2013. Techno-economic comparison of energy usage
between azeotropic distillation and hybrid system for water–ethanol separation.
Renew. Energy 51, 310–316.

Larran, A., Jozami, E., Vicario, L., Feldman, S.R., Podestá, F.E., Permingeat, H.R.,
2015. Evaluation of biological pretreatments to increase the efficiency of the
saccharification process using Spartina argentinensis as a biomass resource.
Bioresour. Technol. 194, 320–5.

Le, Q.-K., Halvorsen, I.J., Pajalic, O., Skogestad, S., 2015. Dividing wall columns

218
 for heterogeneous azeotropic distillation. Chem. Eng. Res. Des. 99, 111–119.

Lee, H.J., Cho, E.J., Kim, Y.G., Choi, I.S., Bae, H.J., 2012. Pervaporative separation
of bioethanol using a polydimethylsiloxane/polyetherimide composite hollow-
fiber membrane. Bioresour. Technol. 109, 110–115.

Leksawasdi, N., Joachimsthal, E., Rogers, P., 2001. Mathematical modelling of

ethanol production from glucose/xylose mixtures by recombinant Zymomonas
mobilis. Biotechnol. Lett. 23, 1087–1093.

Lienhard, J.H., 2010. Heat Transfer. J. Heat Transfer, McGraw-Hill Higher

Education 82, 198.

Liguori, R., Ventorino, V., Pepe, O., Faraco, V., 2015. Bioreactors for lignocellulose
conversion into fermentable sugars for production of high added value products.
Appl. Microbiol. Biotechnol. 100, 597–611.

Liu, L., Du, J., El-Halwagi, M.M., Ponce-Ortega, J.M., Yao, P., 2013. A systematic
approach for synthesizing combined mass and heat exchange networks.
Comput. Chem. Eng. 53, 1–13.

Liu, Z.-H., Chen, H.-Z., 2016. Simultaneous saccharification and co-fermentation for
improving the xylose utilization of steam exploded corn stover at high solid
loading. Bioresour. Technol. 201, 15–26.

Lutze, P., Gorak, A., 2013. Reactive and membrane-assisted distillation: Recent
developments and perspective. Chem. Eng. Res. Des. 91, 1978–1997.

Luyben, W.L., 2006. Distillation Economic Optimization, in: Distillation Design and
Control Using AspenTM Simulation. John Wiley & Sons, Inc., pp. 85–97.

Mabee, W.E., McFarlane, P.N., Saddler, J.N., 2011. Biomass availability for
lignocellulosic ethanol production. Biomass and Bioenergy 35, 4519–4529.

Marriott, J., Sorensen, E., 2003. A general approach to modelling membrane

modules. Chem. Eng. Sci. 58, 4975–4990.

Marriott, J., Sørensen, E., 2003. The optimal design of membrane systems. Chem.
Eng. Sci. 58, 4991–5004.

Marriott, J.I., Sørensen, E., Bogle, I.D.L., 2001. Detailed mathematical modelling of
membrane modules. Comput. Chem. Eng. 25, 693–700.
219
Martinez, R., Sanz, M.T., Beltran, S., 2013. Concentration by pervaporation of
brown crab volatile compounds from dilute model solutions: Evaluation of
PDMS membrane. J. Memb. Sci. 428, 371–379.

Mawire, A., Phori, A., Taole, S., 2014. Performance comparison of thermal energy
storage oils for solar cookers during charging. Appl. Therm. Eng. 73, 1323–
1331.

Mielenz, J.R., Bardsley, J.S., Wyman, C.E., 2009. Fermentation of soybean hulls to
ethanol while preserving protein value. Bioresour. Technol. 100, 3532–3539.

Modarresi, A., Kravanja, P., Friedl, A., 2012. Pinch and exergy analysis of
lignocellulosic ethanol, biomethane, heat and power production from straw.
Appl. Therm. Eng. 43, 20–28.

Mohagheghi, A., Ruth, M., Shell, D.J., 2006. Conditioning hemicellulose

hydrolysates for fermentation: Effects of overliming pH on sugar and ethanol
yields. Process Biochem. 41, 1806–1811.

Morales-Rodriguez, R., Meyer, A.S., Gernaey, K. V., Sin, G., 2012. A framework for
model-based optimization of bioprocesses under uncertainty: Lignocellulosic
ethanol production case. Comput. Chem. Eng. 42, 115–129.

Morales-Rodriguez, R., Meyer, A.S., Gernaey, K. V, Sin, G., 2011. Dynamic model-
based evaluation of process configurations for integrated operation of hydrolysis
and co-fermentation for bioethanol production from lignocellulose. Bioresour.
Technol. 102, 1174–1184.

Moreno, A., Tomas-Pejo, E., Ibarra, D., Ballesteros, M., Olsson, L., 2013. Fed-batch
SSCF using steam-exploded wheat straw at high dry matter consistencies and a
xylose-fermenting Saccharomyces cerevisiae strain: effect of laccase
supplementation. Biotechnol. Biofuels 6, 1–10.

Moreno, A.D., Tomás-Pejó, E., Ibarra, D., Ballesteros, M., Olsson, L., 2013. In situ
laccase treatment enhances the fermentability of steam-exploded wheat straw in
SSCF processes at high dry matter consistencies. Bioresour. Technol. 143, 337–
43.

Mosier, N., Wyman, C., Dale, B., Elander, R., Lee, Y.Y., Holtzapple, M., Ladisch,
M., 2005. Features of promising technologies for pretreatment of lignocellulosic
220
 biomass. Bioresour. Technol. 96, 673–686.

Mulakala, C., Reilly, P.J., 2005. Hypocrea jecorina (Trichoderma reesei) Cel7A as a
molecular machine: A docking study. Proteins Struct. Funct. Bioinforma. 60,
598–605.

Nagasawa, H., Matsuda, N., Kanezashi, M., Yoshioka, T., Tsuru, T., 2016.
Pervaporation and vapor permeation characteristics of BTESE-derived
organosilica membranes and their long-term stability in a high-water-content
IPA/water mixture. J. Memb. Sci. 498, 336–344.

Najafpour, G.D., Shan, C.P., 2003. Enzymatic hydrolysis of molasses. Bioresour.

Technol. 86, 91–94.

Niemisto, J., Kujawski, W., Keiski, R.L., 2013. Pervaporation performance of

composite poly(dimethyl siloxane) membrane for butanol recovery from model
solutions. J. Memb. Sci. 434, 55–64.

Nigam, J.N., 2001. Ethanol production from wheat straw hemicellulose hydrolysate
by Pichia stipitis. J. Biotechnol. 87, 17–27.

Niu, H., Shah, N., Kontoravdi, C., 2016. Modelling of amorphous cellulose
depolymerisation by cellulases, parametric studies and optimisation. Biochem.
Eng. J. 105, 455–472.

Noureddini, H., Byun, J., 2010. Dilute-acid pretreatment of distillers’ grains and corn
fiber. Bioresour. Technol. 101, 1060–1067.

O’Brien, D.J., Senske, G.E., Kurantz, M.J., Craig, J.C., 2004. Ethanol recovery from
corn fiber hydrolysate fermentations by pervaporation. Bioresour. Technol. 92,
15–19.

Öhgren, K., Bura, R., Lesnicki, G., Saddler, J., Zacchi, G., 2007a. A comparison
between simultaneous saccharification and fermentation and separate hydrolysis
and fermentation using steam-pretreated corn stover. Process Biochem. 42,
834–839.

Öhgren, K., Bura, R., Lesnicki, G., Saddler, J., Zacchi, G., 2007b. A comparison
between simultaneous saccharification and fermentation and separate hydrolysis
and fermentation using steam-pretreated corn stover. Process Biochem. 42,

221
 834–839.

Okoli, C.O., Adams, T.A., 2015. Design of dividing wall columns for butanol
recovery in a thermochemical biomass to butanol process. Chem. Eng. Process.
Process Intensif. 95, 302–316.

Olofsson, K., Rudolf, A., Liden, G., 2008. Designing simultaneous saccharification
and fermentation for improved xylose conversion by a recombinant strain of
Saccharomyces cerevisiae. J. Biotechnol. 134, 112–120.

Olujić, Ž., Sun, L., de Rijke, A., Jansens, P.J., 2006. Conceptual design of an
internally heat integrated propylene-propane splitter. Energy 31, 3083–3096.

Palmarola-Adrados, B., Choteborska, P., Galbe, M., Zacchi, G., 2005. Ethanol
production from non-starch carbohydrates of wheat bran. Bioresour. Technol.
96, 843–850.

Palmqvist, E., Hahn-Hagerdal, B., 2000a. Fermentation of lignocellulosic

hydrolysates. II: inhibitors and mechanisms of inhibition. Bioresour. Technol.
74, 25–33.

Palmqvist, E., Hahn-Hagerdal, B., 2000b. Fermentation of lignocellulosic

hydrolysates. I: inhibition and detoxification. Bioresour. Technol. 74, 17–24.

Patle, S., Lal, B., 2008. Investigation of the potential of agro-industrial material as
low cost substrate for ethanol production by using Candida tropicalis and
Zymomonas mobilis. Biomass and Bioenergy 32, 596–602.

Peters, M.S., Timmerhaus, K.D., West, R.E., 2003. Profitability, alternative

investments, and replacements, in: Plant Design and Economics for Chemical
Engineers. Mc Graw Hill, pp. 324–330.

Petersson, A., Thomsen, M.H., Hauggaard-Nielsen, H., Thomsen, A.B., 2007.

Potential bioethanol and biogas production using lignocellulosic biomass from
winter rye, oilseed rape and faba bean. Biomass and Bioenergy 31, 812–819.

Petrobras, 2016. Ethanol plants [WWW Document].

Phisalaphong, M., Srirattana, N., Tanthapanichakoon, W., 2006. Mathematical

modeling to investigate temperature effect on kinetic parameters of ethanol
fermentation. Biochem. Eng. J. 28, 36–43.

222
Pohlmeier, J., Rix, A., 1996. Interactive plant and control design of a double-effect
distillation column. Comput. Chem. Eng. 20, 395–400.

Prasad, S., Singh, A., Joshi, H.C., 2007. Ethanol as an alternative fuel from
agricultural, industrial and urban residues. Resour. Conserv. Recycl. 50, 1–39.

Process System Enterprises, 2015. gPROMS.

Purwadi, R., Niklasson, C., Taherzadeh, M.J., 2004. Kinetic study of detoxification
of dilute-acid hydrolyzates by Ca(OH)2. J. Biotechnol. 114, 187–198.

Rajkumar, P., Shankar, R., Srinivas, T., 2015. Economic analysis of solar collector
with different thermic fluids of LiBr-water vapour absorption refrigeration
system. Int. J. Appl. Eng. Res. 10, 1673–1675.

Raman, S., Mohr, A., Helliwell, R., Ribeiro, B., Shortall, O., Smith, R., Millar, K.,
2015. Integrating social and value dimensions into sustainability assessment of
lignocellulosic biofuels. Biomass and Bioenergy.

Ranjan, R., Thust, S., Gounaris, C.E., Woo, M., Floudas, C.A., Keitz, M. V,
Valentas, K.J., Wei, J., Tsapatsis, M., 2009. Adsorption of fermentation
inhibitors from lignocellulosic biomass hydrolyzates for improved ethanol yield
and value-added product recovery. Microporous Mesoporous Mater. 122, 143–
148.

Rašković, P., Anastasovski, A., Markovska, L., Meško, V., 2010. Process integration
in bioprocess indystry: waste heat recovery in yeast and ethyl alcohol plant.
Energy 35, 704–717.

Rasmussen, M.L., Shrestha, P., Khanal, S.K., III, A.L.P., van Leeuwen, J. (Hans),
2010. Sequential saccharification of corn fiber and ethanol production by the
brown rot fungus Gloeophyllum trabeum. Bioresour. Technol. 101, 3526–3533.

Rattanachomsri, U., Tanapongpipat, S., Eurwilaichitr, L., Champreda, V., 2009.

Simultaneous non-thermal saccharification of cassava pulp by multi-enzyme
activity and ethanol fermentation by Candida tropicalis. J. Biosci. Bioeng. 107,
488–493.

Reijnders, L., 2006. Conditions for the sustainability of biomass based fuel use.
Energy Policy 34, 863–876.

223
RESTEK, 2016. GC columns: Installation guide.

Rivera, E.C., Costa, A.C., Atala, D.I.P., Maugeri, F., Maciel, M.R.W., Filho, R.M.,
2006. Evaluation of optimization techniques for parameter estimation:
Application to ethanol fermentation considering the effect of temperature.
Process Biochem. 41, 1682–1687.

Roberts, J.D., Caseiro, M.C., 1977. Basic princuples of organic chemistry, Second.
ed. W. A. Benjamin, Inc.

Rocha, G.J.M., Gonçalves, A.R., Nakanishi, S.C., Nascimento, V.M., Silva, V.F.N.,
2015. Pilot scale steam explosion and diluted sulfuric acid pretreatments:
Comparative study aiming the sugarcane bagasse saccharification. Ind. Crops
Prod. 74, 810–816.

Rodera, H., Bagajewicz, M.J., 2001. Multipurpose Heat-Exchanger Networks for

Heat Integration Across Plants. Ind. Eng. Chem. Res. 40, 5585–5603.

Romaní, A., Pereira, F., Johansson, B., Domingues, L., 2015. Metabolic engineering
of Saccharomyces cerevisiae ethanol strains PE-2 and CAT-1 for efficient
lignocellulosic fermentation. Bioresour. Technol. 179, 150–8.

Romero, I., Sanchez, S., Moya, M., Castro, E., Ruiz, E., Bravo, V., 2007.
Fermentation of olive tree pruning acid-hydrolysates by Pachysolen
tannophilus. Biochem. Eng. J. 36, 108–115.

Roth, T., Kreis, P., Gorak, A., 2013. Process analysis and optimisation of hybrid
processes for the dehydration of ethanol. Chem. Eng. Res. Des. 91, 1171–1185.

Saha, B.C., Cotta, M.A., 2008. Lime pretreatment, enzymatic saccharification and
fermentation of rice hulls to ethanol. Biomass and Bioenergy 32, 971–977.

Saha, B.C., Cotta, M.A., 2007. Enzymatic saccharification and fermentation of

alkaline peroxide pretreated rice hulls to ethanol. Enzyme Microb. Technol. 41,
528–532.

Saha, B.C., Iten, L.B., Cotta, M.A., Wu, Y. V, 2005. Dilute acid pretreatment,
enzymatic saccharification and fermentation of wheat straw to ethanol. Process
Biochem. 40, 3693–3700.

Saini, J.K., Anurag, R.K., Arya, A., Kumbhar, B.K., Tewari, L., 2013. Optimization

224
 of saccharification of sweet sorghum bagasse using response surface
methodology. Ind. Crops Prod. 44, 211–219.

Sanchez, O.J., Cardona, C.A., 2005. Produccion biotecnologica de alcohol

carburante I: obtencion a partir de diferentes materias primas. Interciencia 30,
671–678.

Sánchez, Ó.J., Cardona, C.A., 2012. Conceptual design of cost-effective and

environmentally-friendly configurations for fuel ethanol production from
sugarcane by knowledge-based process synthesis. Bioresour. Technol. 104,
305–314.

Sanchez, S., Bravo, V., Moya, A.J., Castro, E., Camacho, F., 2004. Influence of
temperature on the fermentation of d-xylose by Pachysolen tannophilus to
produce ethanol and xylitol. Process Biochem. 39, 673–679.

Sander, U., Janssen, H., 1991. Industrial application of vapour permeation. J. Memb.
Sci. 61, 113–129.

Sander, U., Soukup, P., 1988. Design and operation of a pervaporation plant for
ethanol dehydration. J. Memb. Sci. 36, 463–475.

Sasaki, K., Tsuge, Y., Sasaki, D., Teramura, H., Inokuma, K., Hasunuma, T., Ogino,
C., Kondo, A., 2015. Mechanical milling and membrane separation for
increased ethanol production during simultaneous saccharification and co-
fermentation of rice straw by xylose-fermenting Saccharomyces cerevisiae.
Bioresour. Technol. 185, 263–8.

Sassner, P., Galbe, M., Zacchi, G., 2006. Bioethanol production based on
simultaneous saccharification and fermentation of steam-pretreated Salix at high
dry-matter content. Enzyme Microb. Technol. 39, 756–762.

Sassner, P., Mårtensson, C.-G., Galbe, M., Zacchi, G., 2008. Steam pretreatment of
H2SO4-impregnated Salix for the production of bioethanol. Bioresour. Technol.
99, 137–145.

Saxena, R.C., Adhikari, D.K., Goyal, H.B., 2009. Biomass-based energy fuel through
biochemical routes: A review. Renew. Sustain. Energy Rev. 13, 167–178.

Schröder, J., Gawron, K., 1981. Latent heat storage. Int. J. Energy Res. 5, 103–109.

225
Sciacovelli, A., Gagliardi, F., Verda, V., 2015. Maximization of performance of a
PCM latent heat storage system with innovative fins. Appl. Energy 137, 707–
715.

Serrano-López, R., Fradera, J., Cuesta-López, S., 2013a. Molten salts database for
energy applications. Chem. Eng. Process. Process Intensif. 73, 87–102.

Serrano-López, R., Fradera, J., Cuesta-López, S., 2013b. Molten salts database for
energy applications. Chem. Eng. Process. Process Intensif. 73, 87–102.

Setlhaku, M., Heitmann, S., Gorak, A., Wichmann, R., 2013. Investigation of gas
stripping and pervaporation for improved feasibility of two-stage butanol
production process. Bioresour. Technol. 136, 102–108.

Shuai, L., Yang, Q., Zhu, J.Y., Lu, F.C., Weimer, P.J., Ralph, J., Pan, X.J., 2010.
Comparative study of SPORL and dilute-acid pretreatments of spruce for
cellulosic ethanol production. Bioresour. Technol. 101, 3106–14.

Siqueira, P.F., Karp, S.G., Carvalho, J.C., Sturm, W., Rodriguez-Leon, J.A.,
Tholozan, J.L., Singhania, R.R., Pandey, A., Soccol, C.R., 2008. Production of
bio-ethanol from soybean molasses by Saccharomyces cerevisiae at laboratory,
pilot and industrial scales. Bioresour. Technol. 99, 8156–8163.

Sreenath, H.K., Jeffries, T.W., 2000. Production of ethanol from wood hydrolyzate
by yeasts. Bioresour. Technol. 72, 253–260.

Sreenath, H.K., Koegel, R.G., Moldes, A.B., Jeffries, T.W., Straub, R.J., 2001.
Ethanol production from alfalfa fiber fractions by saccharification and
fermentation. Process Biochem. 36, 1199–1204.

Streitwieser, A., Heathcock, C.H., 1976. Introduction to organic chemistry.

Macmillan.

Stuart, P.R., El-Halwagi, M.M., 2012a. Integrated Biorefineries: Design, Analysis,

and Optimization. CRC Press.

Stuart, P.R., El-Halwagi, M.M., 2012b. Integrated Biorefineries: Design, Analysis,

and Optimization.

Su, R., Ma, Y., Qi, W., Zhang, M., Wang, F., Du, R., Yang, J., Zhang, M., He, Z.,
2012. Ethanol Production from High-Solid SSCF of Alkaline-Pretreated

226
 Corncob Using Recombinant Zymomonas mobilis CP4. BioEnergy Res. 6, 292–
299.

Su, Y., Zhang, P., Su, Y., 2015. An overview of biofuels policies and
industrialization in the major biofuel producing countries. Renew. Sustain.
Energy Rev. 50, 991–1003.

Sudhoff, D., Leimbrink, M., Schleinitz, M., Górak, A., Lutze, P., 2015. Modelling,
design and flexibility analysis of rotating packed beds for distillation. Chem.
Eng. Res. Des. 94, 72–89.

Sukumaran, R.K., Singhania, R.R., Mathew, G.M., Pandey, A., 2009. Cellulase
production using biomass feed stock and its application in lignocellulose
saccharification for bio-ethanol production. Renew. Energy 34, 421–424.

Sulzer Chemtech, 2014. Membrane Technology [WWW Document].

Swain, M.R., Krishnan, C., 2015. Improved conversion of rice straw to ethanol and
xylitol by combination of moderate temperature ammonia pretreatment and
sequential fermentation using Candida tropicalis. Ind. Crops Prod. 77, 1039–
1046.

Szitkai, Z., Lelkes, Z., Rev, E., Fonyo, Z., 2002. Optimization of hybrid ethanol
dehydration systems. Chem. Eng. Process. Process Intensif. 41, 631–646.

Tang, Y.Q., Koike, Y., Liu, K., An, M.Z., Morimura, S., Wu, X.L., Kida, K., 2008.
Ethanol production from kitchen waste using the flocculating yeast
Saccharomyces cerevisiae strain KF-7. Biomass and Bioenergy 32, 1037–1045.

Teh, Y.S., Rangaiah, G.P., 2003. Tabu search for global optimization of continuous
functions with application to phase equilibrium calculations. Comput. Chem.
Eng. 27, 1665–1679.

Telli-Okur, M., Eken-Saracoglu, N., 2008. Fermentation of sunflower seed hull

hydrolysate to ethanol by Pichia stipitis. Bioresour. Technol. 99, 2162–2169.

Tian, S., Zhou, G., Yan, F., Yu, Y., Yang, X., 2009. Yeast strains for ethanol
production from lignocellulosic hydrolysates during in situ detoxification.
Biotechnol. Adv. 27, 656–660.

Toth, A.J., Mizsey, P., 2015. Methanol removal from aqueous mixture with

227
 organophilic pervaporation: experiments and modelling. Chem. Eng. Res. Des.

Towler, G., Sinnott, R., 2013. Chapter 17 - Separation Columns (Distillation,

Absorption, and Extraction), in: Towler, G., Sinnott, R. (Eds.), Chemical
Engineering Design (Second Edition). Butterworth-Heinemann, Boston, pp.
807–935.

Triana, C.F., Fraga, E.S., Sorensen, E., 2015a. Energy assessment of different
configurations for the ethanol production process from lignocellulosic biomass,
in: Computer Aided Chemical Engineering. pp. 2285–2290.

Triana, C.F., Fraga, E.S., Sorensen, E., 2015b. 12th International Symposium on
Process Systems Engineering and 25th European Symposium on Computer
Aided Process Engineering, Computer Aided Chemical Engineering, Computer
Aided Chemical Engineering. Elsevier.

Triana, C.F., Quintero, J.A., Agudelo, R.A., Cardona, C.A., Higuita, J.C., 2011.
Analysis of coffee cut-stems (CCS) as raw material for fuel ethanol production.
Energy 36, 4182–4190.

Tsuyumoto, M., Akita, K., Teramoto, A., 1995. Pervaporative transport of aqueous
ethanol: Dependence of permeation rates on ethanol concentration and permeate
side pressures. Desalination 103, 211–222.

Tsuyumoto, M., Teramoto, A., Meares, P., 1997. Dehydration of ethanol on a pilot-
plant scale, using a new type of hollow-fiber membrane. J. Memb. Sci. 133, 83–
94.

Tusel, G.F., Brüschke, H.E.A., 1985. Use of pervaporation systems in the chemical
industry. Desalination 53, 327–338.

U.S. Department of Energy, 2015. International Energy Statistics [WWW

Document].

Vaklieva-Bancheva, N., Ivanov, B.B., Shah, N., Pantelides, C.C., 1996. Heat
exchanger network design for multipurpose batch plants. Comput. Chem. Eng.
20, 989–1001.

Valentinyi, N., Csefalvay, E., Mizsey, P., 2013. Modelling of pervaporation:

Parameter estimation and model development. Chem. Eng. Res. Des. 91, 174–

228
 183.

Vanderbei, R.J., 2010. Linear programming: foundations and extensions, Third Edit.
ed. Springer.

Vane, L.M., Alvarez, F.R., Rosenblum, L., Govindaswamy, S., 2013. Efficient
Ethanol Recovery from Yeast Fermentation Broth with Integrated Distillation–
Membrane Process. Ind. Eng. Chem. Res. 52, 1033–1041.

Veses, A., Aznar, M., Callén, M.S., Murillo, R., García, T., 2016. An integrated
process for the production of lignocellulosic biomass pyrolysis oils using
calcined limestone as a heat carrier with catalytic properties. Fuel 181, 430–437.

Vignarooban, K., Xu, X., Arvay, A., Hsu, K., Kannan, A.M., 2015. Heat transfer
fluids for concentrating solar power systems – A review. Appl. Energy 146,
383–396.

Viswanathan, J., Grossmann, I.E., 1993. Optimal feed locations and number of trays
for distillation columns with multiple feeds. Ind. Eng. Chem. Res. 32, 2942–
2949.

Viswanathan, J., Grossmann, I.E., 1993. An alternate {MINLP} model for finding
the number of trays required for a specified separation objective. Comput.
Chem. Eng. 17, 949–955.

Viswanathan, J., Grossmann, I.E., 1990. A combined penalty function and outer-
approximation method for {MINLP} optimization. Comput. Chem. Eng. 14,
769–782.

Wang, Y., Chang, C., Feng, X., 2015. A systematic framework for multi-plants Heat
Integration combining Direct and Indirect Heat Integration methods. Energy 90,
56–67.

Wei, H.-M., Wang, F., Zhang, J.-L., Liao, B., Zhao, N., Xiao, F., Wei, W., Sun, Y.-
H., 2013. Design and Control of Dimethyl Carbonate–Methanol Separation via
Pressure-Swing Distillation. Ind. Eng. Chem. Res. 52, 11463–11478.

Wei, P., Cheng, L.H., Zhang, L., Xu, X.H., Chen, H.I., Gao, C.J., 2014. A review of
membrane technology for bioethanol production. Renew. Sustain. Energy Rev.
30, 388–400.

229
Weiguo, X., Depeng, R., Qing, Y., Guodong, L., Huilin, L., Shuai, W., 2016.
Simulations and experiments of laminar heat transfer for Therminol heat
transfer fluids in a rifled tube. Appl. Therm. Eng. 102, 861–872.

Xiao, Z., Zhang, X., Gregg, D., Saddler, J., 2004. Effects of sugar inhibition on
cellulases and glucosidase during enzymatic hydrolysis of softwood substrates.
Appl. Biochem. Biotechnol. 115, 1115–1126.

Xie, R., Tu, M., Carvin, J., Wu, Y., 2015. Detoxification of biomass hydrolysates
with nucleophilic amino acids enhances alcoholic fermentation. Bioresour.
Technol. 186, 106–13.

Xu, J., Cheng, J.J., Sharma-Shivappa, R.R., Burns, J.C., 2010. Lime pretreatment of
switchgrass at mild temperatures for ethanol production. Bioresour. Technol.
101, 2900–2903.

Xu, W., Wang, S., Lu, H., Wang, Q., Zhang, Q., Lu, H., 2016. Thermo-hydraulic
performance of liquid phase heat transfer fluid (Therminol) in a ribbed tube.
Exp. Therm. Fluid Sci. 72, 149–160.

Yakovlev, A. V, Shalygin, M.G., Matson, S.M., Khotimskiy, V.S., Teplyakov, V. V,

2013. Separation of diluted butanol-water solutions via vapor phase by
organophilic membranes based on high permeable polyacetylenes. J. Memb.
Sci. 434, 99–105.

Yee, T.F., Grossmann, I.E., 1990. Simultaneous optimization models for heat
integration—II. Heat exchanger network synthesis. Comput. Chem. Eng. 14,
1165–1184.

Yee, T.F., Grossmann, I.E., Kravanja, Z., 1990a. Simultaneous optimization models
for heat integration—I. Area and energy targeting and modeling of multi-stream
exchangers. Comput. Chem. Eng. 14, 1151–1164.

Yee, T.F., Grossmann, I.E., Kravanja, Z., 1990b. Simultaneous optimization models
for heat integration—III. Process and heat exchanger network optimization.
Comput. Chem. Eng. 14, 1185–1200.

Zalba, B., 2003. Review on thermal energy storage with phase change: materials,
heat transfer analysis and applications. Appl. Therm. Eng. 23, 251–283.

230
Zaversky, F., García-Barberena, J., Sánchez, M., Astrain, D., 2013. Transient molten
salt two-tank thermal storage modeling for CSP performance simulations. Sol.
Energy 93, 294–311.

Zeigarnik, Y., 2011. Liquid-Metal Heat Transfer. Thermopedia.

Zhang, J., Xiongjun, S., Lynd, L.R., 2009. Simultaneous saccharification and co-
fermentation of paper sludge to ethanol by Saccharomyces cerevisiae RWB222.
Part II: Investigation of discrepancies between predicted and observed
performance at high solids concentration. Biotechnol. Bioeng. 104, 932–938.

Zhang, P., Ma, F., Xiao, X., 2016. Thermal energy storage and retrieval
characteristics of a molten-salt latent heat thermal energy storage system. Appl.
Energy 173, 255–271.

Zhang, X., Wenjuan, Q., Michael, G.P., Saddler, J.N., 2009. High consistency
enzymatic hydrolysis of hardwood substrates. Bioresour. Technol. 100, 5890–
5897.

Zhao, L., Zhang, X., Tan, T., 2008. Influence of various glucose/xylose mixtures on
ethanol production by Pachysolen tannophilus. Biomass and Bioenergy 32,
1156–1161.

Zhu, J.-Q., Qin, L., Li, W.-C., Zhang, J., Bao, J., Huang, Y.-D., Li, B.-Z., Yuan, Y.-
J., 2015. Simultaneous saccharification and co-fermentation of dry diluted acid
pretreated corn stover at high dry matter loading: Overcoming the inhibitors by
non-tolerant yeast. Bioresour. Technol. 198, 39–46.

Zhu, J.Y., Pan, X.J., 2010. Woody biomass pretreatment for cellulosic ethanol
production: Technology and energy consumption evaluation. Bioresour.
Technol. 101, 4992–5002.

231
NOMENCLATURE

𝐴𝑚𝑒𝑚 Membrane area m2

Afo Pre-exponential factor for fast hemicellulose (6.7x1016) min-1
Aso Pre-exponential factor for slow hemicellulose (6.9x1019) min-1
Axo Pre-exponential factor for xylose (3.7x1010) min-1
𝐶 Real concentration of sulphuric acid in the reactor -
𝐶𝑎 Actual concentration of sulphuric acid -
𝐶𝐶𝑏 Concentration of cellobiose g/L
𝐶𝑐𝑒𝑙𝑙 Concentration of cellulose g/L
𝐶𝐸𝑡ℎ Concentration of ethanol g/L
𝐶𝐸𝑡ℎ,𝑖𝑝,𝐺𝑙 Threshold inhibition conc. of ethanol from glucose g/L
𝐶𝐸𝑡ℎ,𝑖𝑠,𝐺𝑙 Threshold conc. of ethanol from glucose that affects glucose uptake g/L
𝐶𝐸𝑡ℎ,𝑖𝑥,𝐺𝑙 Threshold conc. of ethanol from xylose that causes inhibition g/L
𝐶𝐸𝑡ℎ,𝑖𝑝,𝑋𝑦 Threshold inhibition conc. of ethanol from xylose g/L
𝐶𝐸𝑡ℎ,𝑖𝑠,𝑋𝑦 Threshold conc. of ethanol from xylose that affects xylose uptake g/L
𝐶𝐸𝑡ℎ,𝑖𝑥,𝑋𝑦 Threshold conc. of ethanol from xylose that causes inhibition g/L
𝐶𝐸𝑡ℎ,𝑚𝑝,𝐺𝑙 Maximum inhibition conc. of ethanol from glucose g/L
𝐶𝐸𝑡ℎ,𝑚𝑝,𝑋𝑦 Maximum inhibition conc. of ethanol from xylose g/L
𝐶𝐸𝑡ℎ,𝑚𝑠,𝐺𝑙 Maximum conc. of ethanol from glucose that affects glucose uptake g/L
𝐶𝐸𝑡ℎ,𝑚𝑠,𝑋𝑦 Maximum conc. of ethanol from glucose that affects xylose uptake g/L
𝐶𝐸𝑡ℎ,𝑚𝑥,𝐺𝑙 Maximum conc. of ethanol from glucose that causes inhibition g/L
𝐶𝐸𝑡ℎ,𝑚𝑥,𝑋𝑦 Maximum conc. of ethanol from xylose that causes inhibition g/L
𝐶𝑓𝑢𝑟 Concentration of furfural produced during pretreatment g/L
𝐶𝐺𝑙 Concentration of glucose g/L
𝐶ℎ𝑒𝑚𝑖 Total Concentration of hemicellulose g/L
𝑓𝑎𝑠𝑡
𝐶ℎ𝑒𝑚𝑖 Concentration of fast hemicellulose g/L
𝑠𝑙𝑜𝑤
𝐶ℎ𝑒𝑚𝑖 Concentration of slow hemicellulose g/L
𝐶𝑃 Concentration of insoluble salts g/L
𝐶𝑋 Concentration of cells g/L
𝐶𝑋𝑦 Concentration of xylose g/L
𝐶𝑍𝐴 Concentration of Ca complex g/L
𝑁
𝐶𝑍,0 Concentration of Ca(OH)2 to neutralise hydrolysates g/L
DFi Gradient of activity through the membrane Pa
𝐸𝑖 Activation energy in organophilic membrane J mol-1 K-1
Efo Activation energy for fast hemicellulose conversion kJ/mol
EiB Bound concentration of enzymes on the substrate g/Kg

232
EiF Concentration of free enzymes in solution g/Kg
kg protein/
Eimax Maximum mass of active enzyme/substrate
kg subst
EiT Total enzyme concentration g/Kg
Eso Activation energy for slow hemicellulose conversion J/mol
Exo Activation energy for xylose degradation J/mol
F Feed flow rate kg/s
𝐻𝑙𝑖𝑞 Liquid enthalpy J/mol
𝑣𝑎𝑝
𝐻 Vapour enthalpy J/mole
𝐻𝐽 Enthalpy of permeate J/mole
𝑙𝑖𝑞
ℎ Liquid enthalpy J/sec
ℎ𝑣𝑎𝑝 Vapour enthalpy J/sec
𝐽𝑖 Partial flux of component i kg m-2 s-1
Kiad Dissociation constant for absorption/desorption reaction i m3/kg protein
𝐾𝑖𝑗 Relative volatility -
Inhibition constants for the reactions
Kj:inh:k g/L
j = 1, 2, 3; k = cellobiose, glucose, xylose
𝑘𝑖 Reaction rate constants min-1
𝐿 Liquid flow rate kg/s
𝑀𝑖 Liquid holdup of component i Kg
𝑀𝑡 Total liquid holdup Kg
M&S Marshal and Swift Index -
P Pressure Pa
𝑃𝑓𝑒𝑒𝑑 Feed pressure in pervaporation Pa
𝑃𝑝𝑒𝑟𝑚 Permeate pressure in pervaporation Pa
𝑃𝑖𝑠𝑎𝑡 Vapour pressure Pa
𝑄𝐶 Heat released in the condenser J/kg
𝑄𝐵 Heat duty in the reboiler J/kg
𝑄𝑖 Permeance of the membrane mol m-2 s-1Pa-1
𝑄0,𝑖 Pre-exponential factor for the evaluation of permeance in the membrane mol m-2 s-1Pa-1
R Reflux ratio -
𝑅𝐵 Boilup ratio -
𝑟𝑏𝑝 Reaction rate of by-products produced during pretreatment g L-1 min-1
𝑟𝑓𝑢𝑟 Reaction rate of furfural g L-1 min-1
𝑟𝑃 Reaction rate of insoluble salts g L-1 min-1
𝑟𝑍𝐴 Reaction rate of Ca+2 complex g L-1 min-1
ri Reaction rates for equation 16 – 28 g L-1 min-1
RS Substrate reactivity -
𝑟𝑥𝑦 Reaction rate of xylose g L-1 min-1

233
𝑢 Specific internal energy J/kg
U Internal energy J
V Vapour flow rate kg/s
Wfeed.i Feed mass fraction in pervaporation -
Xij Liquid mass fraction of component i in tray j -

Xifeed Feed mole fraction in pervaporation -

Yij Vapour mass fraction of component i in tray j -

Greek symbols

Α Weighting factor for glucose consumption -

αj Binary variable used in the optimisation of the rectification section of tray j -
βj Binary variable used in the optimisation of the stripping section of tray j -
γi Activity coefficient of component i in the mixture -
δm Membrane thickness m
η Efficiency of the compressor assumed to be 0.95 -
𝜑𝑖𝑠𝑎𝑡 Fugacity coefficient of vapour at saturated conditions -
𝑣𝑎𝑝
𝜑𝑖 Fugacity coefficient of vapour -

234
Appendix A – Kinetic models, mass and energy
balances for the production of ethanol from
corn stover

Abstract
This section presents all the different kinetic models, mass and energy balances used
in the simulation of the overall process. The kinetic models have been taken from
literature. The modelling of the distillation column presents a Vapour-Liquid
Equilibrium approach for the formulation of the mass and energy balances.

A.1. Pretreatment
The model proposed by Esteghlalian et al. (1997) presents a set of reaction rate
expressions taking into account the formation of xylose from both fast and slow
hemicellulose, and the production of by-products in an in-series reaction.

𝑓𝑎𝑠𝑡 𝑓𝑎𝑠𝑡
𝑟ℎ𝑒𝑚𝑖 = −𝑘𝑓 𝐶ℎ𝑒𝑚𝑖 Eq. A.1

𝑓𝑎𝑠𝑡 𝑠𝑙𝑜𝑤
𝑟ℎ𝑒𝑚𝑖 = −𝑘𝑠 𝐶ℎ𝑒𝑚𝑖 Eq. A.2

𝑓𝑎𝑠𝑡 𝑠𝑙𝑜𝑤
𝑟𝑥𝑦 = 𝑘𝑓 𝐶ℎ𝑒𝑚𝑖 + 𝑘𝑠 𝐶ℎ𝑒𝑚𝑖 − 𝑘𝑥 𝐶𝑋𝑦 Eq. A.3

𝑟𝑏𝑝 = 𝑘𝑥 𝐶𝑋𝑦 Eq. A.4

The initial conditions to solve this set of differential equations are given as the initial
concentrations of hemicellulose (fast and slow), cellulose, lignin and other
components found in the raw material (g/L). The initial concentration of the products
(xylose, furfural and others) is assumed to be zero. The reaction rate constants (kf , ks
and kx min-1) are defined by an Arrhenius-type temperature dependence:

235
 𝐸𝑓𝑜
𝑘𝑓 = 𝐴𝑓𝑜 𝐶𝑎1.5 𝑒𝑥𝑝 (− 𝑅 𝑇 ) Eq. A.5

𝐸
𝑘𝑠 = 𝐴𝑠𝑜 𝐶𝑎1.6 𝑒𝑥𝑝 (− 𝑅𝑠𝑜𝑇) Eq. A.6

𝐸
𝑘𝑥 = 𝐴2𝑜 𝐶𝑎0.5 𝑒𝑥𝑝 (− 𝑅𝑥𝑜𝑇 ) Eq. A.7

Eq. A.8 determines the real concentration of sulphuric acid in the system taking into
account the neutralising ability of the raw material:

𝐶𝑎 = 𝐶 − (0.1 𝑁𝑒𝑢𝑡𝑟𝑎𝑙𝑖𝑧𝑖𝑛𝑔 𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝑜𝑓 𝑡ℎ𝑒 𝑟𝑎𝑤 𝑚𝑎𝑡𝑒𝑟𝑖𝑎𝑙) Eq. A.8

The mass balance for each component and the energy balance are used to describe
the progress of the reaction in terms of operating conditions for the streams involved
and the conversion of the reactants. Equations A.9 to A.12 show the mass balance for
the components and Eq. A.13 presents the energy balance in the reactor. Figure A.1
presents the schematic of a jacketed reactor on which the mass and energy balances
for pretreatment, detoxification and SSCF are based:

Figure A.1: Scheme of a jacketed reactor used in the formulation of the mass and energy
balances

236
𝑑𝑀ℎ𝑒𝑚𝑖 𝑖𝑛 𝑜𝑢𝑡 𝑓𝑎𝑠𝑡 𝑠𝑙𝑜𝑤
= 𝐹𝑖𝑛 𝑤ℎ𝑒𝑚𝑖 − 𝐹𝑜𝑢𝑡 𝑤ℎ𝑒𝑚𝑖 − (𝑟ℎ𝑒𝑚𝑖 + 𝑟ℎ𝑒𝑚𝑖 )𝑉 Eq. A.9
𝑑𝑡

𝑓𝑎𝑠𝑡 𝑠𝑙𝑜𝑤
𝑀ℎ𝑒𝑚𝑖 = 𝑀ℎ𝑒𝑚𝑖 + 𝑀ℎ𝑒𝑚𝑖 Eq. A.10

𝑑𝑀𝑥𝑦 𝑖𝑛 𝑜𝑢𝑡
= 𝐹𝑖𝑛 𝑤𝑥𝑦 − 𝐹𝑜𝑢𝑡 𝑤𝑥𝑦 + 𝑟𝑥𝑦 𝑉 Eq. A.11
𝑑𝑡

𝑑𝑀𝑏𝑝 𝑖𝑛 𝑜𝑢𝑡
= 𝐹𝑖𝑛 𝑤𝑏𝑝 − 𝐹𝑜𝑢𝑡 𝑤𝑏𝑝 + 𝑟𝑏𝑝 𝑉 Eq. A.12
𝑑𝑡

The energy balance for the pretreatment reactor presents the heat of reaction which is
define as the summation of the heat of reactions for equations A.9, A.11 and A.12.

𝑑𝑈
= 𝐿𝑖𝑛 ℎ𝑖𝑛 − 𝐿𝑜𝑢𝑡 ℎ𝑜𝑢𝑡 + ((𝑟ℎ𝑒𝑚𝑖𝑐 + 𝑟𝑥𝑦 + 𝑟𝑏𝑝 ) (−∆𝐻 𝑅 ) 𝑉 )
𝑑𝑡 Eq. A.13
+ 𝑄𝑝𝑟𝑒𝑡

Eq. A.14 presents the heat transferred from the jacket to the reactor taking into
account overall heat transfer coefficients:

𝑇𝑗𝑎𝑐𝑘𝑒𝑡,𝑖𝑛 + 𝑇𝑗𝑎𝑐𝑘𝑒𝑡,𝑜𝑢𝑡
𝑄𝑝𝑟𝑒𝑡 = 𝑈𝐶 𝐴𝑡𝑟𝑎𝑛𝑠 ( − 𝑇𝑜𝑢𝑡 ) Eq. A.14
2

A.2. Detoxification
The kinetic models that describe the process of detoxification of the lignocellulosic
hydrolysates are the following (Purwadi et al., 2004):

𝑟𝑓𝑢𝑟 = −𝑘1 𝐶𝑓𝑢𝑟 (𝐶𝑍,0 − 𝐶𝑍𝐴 ) + 𝑘2 𝐶𝑍𝐴 Eq. A.15

𝑟𝑍𝐴 = −(𝑘2 + 𝑘3 ) 𝐶𝑍𝐴 + 𝑘1 𝐶𝑓𝑢𝑟 (𝐶𝑍,0 − 𝐶𝑍𝐴 ) Eq. A.16

𝑟𝑃 = −𝑘3 𝐶𝑍𝐴 Eq. A.17

237
With this model, the initial amount of cation 𝐶𝑍,0 is related to the pH level, according
to the Eq. A.13:

𝑁
𝐶𝑍,𝑜 = 𝑘𝑍 ∙ 10−(14−𝑝𝐻) + 𝐶𝑍,0 Eq. A.18

𝑁
Where 𝐶𝑍,0 is Ca(OH)2 needed to neutralise the acidic hydrolysates, and kZ is a
constant which serves as a function of solubility of Ca(OH)2, affinity of the reactants
to this ion and other unknown variables. See Purwadi et al., (2004) for the values of
the parameters of the model.

𝑑𝑀𝑓𝑢𝑟 𝑖𝑛 𝑜𝑢𝑡
= 𝐹𝑖𝑛 𝑤ℎ𝑒𝑚𝑖 − 𝐹𝑜𝑢𝑡 𝑤𝑓𝑢𝑟 + 𝑟𝑓𝑢𝑟 𝑉 Eq. A.19
𝑑𝑡

𝑑𝑀𝑍𝐴 𝑖𝑛 𝑜𝑢𝑡
= 𝐹𝑖𝑛 𝑤𝑍𝐴 − 𝐹𝑜𝑢𝑡 𝑤𝑍𝐴 + 𝑟𝑍𝐴 𝑉 Eq. A.20
𝑑𝑡

𝑑𝑀𝑃
= 𝐹𝑖𝑛 𝑤𝑃𝑖𝑛 − 𝐹𝑜𝑢𝑡 𝑤𝑃𝑜𝑢𝑡 + 𝑟𝑃 𝑉 Eq. A.21
𝑑𝑡

The energy balance for the detoxification process is presented in Eq. A.21. The heat
energy is assumed to be the heat of a neutralization reaction between a strong acid
and a strong base:

𝑑𝑈
= 𝐹𝑖𝑛 ℎ𝑖𝑛 − 𝐹𝑜𝑢𝑡 ℎ𝑜𝑢𝑡 + ((𝑟𝑓𝑢𝑟 + 𝑟𝑍𝐴 + 𝑟𝑃 ) (−∆𝐻 𝑅 ) 𝑉 ) − 𝑄𝐷𝑒𝑡𝑜𝑥 Eq. A.22
𝑑𝑡

Eq. A.23 presents the heat transferred from the jacket to the reactor taking into
account overall heat transfer coefficients using cooling water as utility stream:

𝑇𝑗𝑎𝑐𝑘𝑒𝑡,𝑖𝑛 + 𝑇𝑗𝑎𝑐𝑘𝑒𝑡,𝑜𝑢𝑡
𝑄𝐷𝑒𝑡𝑜𝑥 = 𝑈𝐶 𝐴𝑡𝑟𝑎𝑛𝑠 (𝑇𝑜𝑢𝑡 − ) Eq. A.23
2

A.3. Evaporation
The evaporation stage is used in the concentration of xylose in the hydrolysate
obtained from the pretreatment reaction. The approach considered in this work

238
consists of a dynamic mass balance, an energy balance and a VLE formulation for
the concentration of the components in the vapour phase.

𝑑𝑀𝑖
= 𝐿𝑖𝑛 𝑥𝑖𝑖𝑛 − 𝐿𝑜𝑢𝑡 𝑥𝑖𝑜𝑢𝑡 − 𝑉𝑜𝑢𝑡 𝑦𝑖𝑜𝑢𝑡 Eq. A.24
𝑑𝑡

𝑑𝑈 𝑙𝑖𝑞 𝑙𝑖𝑞 𝑣𝑎𝑝

= 𝐿𝑖𝑛 ℎ𝑖𝑛 − 𝐿𝑜𝑢𝑡 ℎ𝑜𝑢𝑡 − 𝑉𝑜𝑢𝑡 ℎ𝑜𝑢𝑡 + 𝑄𝐸𝑣𝑎𝑝 Eq. A.25
𝑑𝑡

𝑣𝑎𝑝 𝑄𝐸𝑣𝑎𝑝
ℎ𝑜𝑢𝑡 = Eq. A.26
∆𝐻 𝑣𝑎𝑝

𝑦𝑖𝑜𝑢𝑡 = 𝑥𝑖𝑜𝑢𝑡 𝐾𝑖 Eq. A.27

i =1, 2 … Number of components

Eq. A.28 presents the heat transferred from the jacket to the reactor taking into
account overall heat transfer coefficients considering high-pressure steam as the
utility of choice:

𝑇𝑗𝑎𝑐𝑘𝑒𝑡,𝑖𝑛 + 𝑇𝑗𝑎𝑐𝑘𝑒𝑡,𝑜𝑢𝑡
𝑄𝐸𝑣𝑎𝑝 = 𝑈𝐶 𝐴𝑡𝑟𝑎𝑛𝑠 ( − 𝑇𝑜𝑢𝑡 ) Eq. A.28
2

A.4. Simultaneous saccharification and co-fermentation (SSCF)

The parameters used in this work for the simulation of both enzymatic hydrolysis and
co-fermentation are presented in the works of Kadam et al., (2004) and Leksawasdi
et al., (2001).

A.4.1. Saccharification
The kinetic model for the enzymatic hydrolysis proposed by Kadam et al., (2004) is
the following:

𝐸𝑖𝑚𝑎𝑥 𝐾𝑖𝑎𝑑 𝐸𝑖𝐹 𝐶𝑐𝑒𝑙𝑙

𝐸𝑖𝐵 = Eq. A.29
1 + 𝐾𝑖𝑎𝑑 ∙ 𝐸𝑖𝐹

239
𝐸𝑖𝑇 = 𝐸𝑖𝐹 + 𝐸𝑖𝐵 Eq. A.30

i =1 for cellulase; i =2 for β-glucosidase

𝑘1 𝐸1𝐵 𝑅𝑠 𝐶𝑐𝑒𝑙𝑙
𝑟1 =
𝐶 𝐶 𝐶𝑋𝑦 Eq. A.31
1 + 𝐾 𝐶𝑏 + 𝐾 𝐺𝑙 + 𝐾
1.𝑖𝑛ℎ.𝐶𝑏 1.𝑖𝑛ℎ.𝐶𝑏 1.𝑖𝑛ℎ.𝐶𝑏

𝑘2 (𝐸1𝐵 + 𝐸2𝐵 ) 𝑅𝑠 𝐶𝑐𝑒𝑙𝑙

𝑟2 =
𝐶 𝐶 𝐶𝑋𝑦 Eq. A.32
1 + 𝐾 𝐶𝑏 + 𝐾 𝐺𝑙 + 𝐾
2.𝑖𝑛ℎ.𝐶𝑏 2.𝑖𝑛ℎ.𝐶𝑏 2.𝑖𝑛ℎ.𝐶𝑏

𝑘3 𝐸2𝐹 𝑅𝑠 𝐶𝐶𝑏
𝑟3 =
𝐶 𝐶𝑋𝑦 Eq. A.33
𝐾3𝑀 (1 + 𝐾 𝐺𝑙 + 𝐾 ) + 𝐶𝐶𝑏
3.𝑖𝑛ℎ.𝐺𝑙 3.𝑖𝑛ℎ.𝑋𝑦

A variation for hemicellulose conversion into xylose and further degradation into
furfural during the enzymatic hydrolysis has been proposed by Ballesteros et al.,
(2004).

𝑟4 = 𝑘4 𝐶ℎ𝑒𝑚𝑖 Eq. A.34

𝑟5 = 𝑘5 𝐶𝑋𝑦 Eq. A.35

A.4.2 Simultaneous saccharification and co-fermentation

Equations A.36 to A.42 show the growth rate of the strain from glucose and xylose
and the reaction rate for ethanol production from both sugars.

𝐶𝐺𝑙 𝐶𝐸𝑡ℎ −𝐶𝐸𝑡ℎ,𝑖𝑥,𝐺𝑙 𝐾𝑖𝑥,𝐺𝑙

𝑟𝑋,𝐺𝑙 = 𝜇𝑚𝑎𝑥,𝐺𝑙 (𝐾 ) (1 − 𝐶 ) (𝐾 ) Eq. A.36
𝑠𝑥,𝐺𝑙 +𝐶𝐺𝑙 𝐸𝑡ℎ,𝑚𝑥,𝐺𝑙 −𝐶𝐸𝑡ℎ,𝑖𝑥,𝐺𝑙 𝑖𝑥,𝐺𝑙 +𝐶𝐺𝑙

𝐶𝑋𝑦 𝐶𝐸𝑡ℎ −𝐶𝐸𝑡ℎ,𝑖𝑥,𝑋𝑦 𝐾𝑖𝑥,𝑋𝑦

𝑟𝑋,𝑋𝑦 = 𝜇𝑚𝑎𝑥,𝑋𝑦 (𝐾 ) (1 − 𝐶 ) (𝐾 ) Eq. A.37
𝑠𝑥,𝑋𝑦 +𝐶𝑋𝑦 𝐸𝑡ℎ,𝑚𝑥,𝑋𝑦 −𝐶𝐸𝑡ℎ,𝑖𝑥,𝑋𝑦 𝑖𝑥,𝑋𝑦 +𝐶𝑋𝑦

𝑟𝑋 = [𝛼 𝑟𝑋,𝐺𝑙 + (1 − 𝛼) 𝑟𝑋,𝑋𝑦 ] 𝐶𝑋 Eq. A.38

240
 Figure A.2: Scheme of a jacketed reactor used in the formulation of the mass and energy
balances

𝐶𝐺𝑙 𝐶𝐸𝑡ℎ −𝐶𝐸𝑡ℎ,𝑖𝑠,𝐺𝑙 𝐾𝑖𝑠,𝐺𝑙

𝑟𝐺𝑙 = 𝛼 𝑞𝑠,𝑚𝑎𝑥,𝐺𝑙 (𝐾 ) (1 − 𝐶 ) (𝐾 ) 𝐶𝑋 Eq. A.39
𝑠𝑠,𝐺𝑙 +𝐶𝐺𝑙 𝐸𝑡ℎ,𝑚𝑠,𝐺𝑙 −𝐶𝐸𝑡ℎ,𝑖𝑠,𝐺𝑙 𝑖𝑠,𝐺𝑙 +𝐶𝐺𝑙

𝑟𝑋𝑦 =
𝐶𝑋𝑦 𝐶𝐸𝑡ℎ −𝐶𝐸𝑡ℎ,𝑖𝑠,𝑋𝑦 𝐾𝑖𝑠,𝑋𝑦 Eq. A.40
(1 − 𝛼) 𝑞𝑠,𝑚𝑎𝑥,𝑋𝑦 ( ) (1 − 𝐶 ) (𝐾 ) 𝐶𝑋
𝐾𝑠𝑠,𝑋𝑦 +𝐶𝑋𝑦 𝐸𝑡ℎ,𝑚𝑠,𝑋𝑦 −𝐶𝐸𝑡ℎ,𝑖𝑠,𝑋𝑦 𝑖𝑠,𝑋𝑦 +𝐶𝑋𝑦

𝐶𝐺𝑙 𝐶𝐸𝑡ℎ −𝐶𝐸𝑡ℎ,𝑖𝑝,𝐺𝑙 𝐾𝑖𝑝,𝐺𝑙

𝑟𝑃,𝐺𝑙 = 𝑞𝑚𝑎𝑥,𝐺𝑙 ( ) (1 − )( ) Eq. A.41
𝐾𝑠𝑝,𝐺𝑙 +𝐶𝐺𝑙 𝐶𝐸𝑡ℎ,𝑚𝑝,𝐺𝑙 −𝐶𝐸𝑡ℎ,𝑖𝑝,𝐺𝑙 𝐾𝑖𝑝,𝐺𝑙 +𝐶𝐺𝑙

𝐶𝑋𝑦 𝐶𝐸𝑡ℎ −𝐶𝐸𝑡ℎ,𝑖𝑝,𝑋𝑦 𝐾𝑖𝑝,𝑋𝑦

𝑟𝑃,𝑋𝑦 = 𝑞𝑚𝑎𝑥,𝑋𝑦 ( ) (1 − )( ) Eq. A.42
𝐾𝑠𝑝,𝑋𝑦 +𝐶𝑋𝑦 𝐶𝐸𝑡ℎ,𝑚𝑝,𝑋𝑦 −𝐶𝐸𝑡ℎ,𝑖𝑝𝑋𝑦 𝐾𝑖𝑝,𝑋𝑦 +𝐶𝑋𝑦

𝑟𝑃 = [𝛼 𝑟𝑃,𝐺𝑙 + (1 − 𝛼) 𝑟𝑃,𝑋𝑦 ] 𝐶𝑋 Eq. A.43

Figure A.2 presents the schematic of a jacketed reactor used in the SSCF process.
Notice the addition of a second inlet stream in which the inoculum of the fermenting
microorganism is fed into the reaction medium. The mass balances are the
combination of both the enzymatic saccharification and the co-fermentation for each
component as shown below:

241
Hemicellulose:

𝑑𝑀ℎ𝑒𝑚𝑖 𝑖𝑛,1 𝑖𝑛,2 𝑜𝑢𝑡

= 𝐹𝑖𝑛,1 𝑤ℎ𝑒𝑚𝑖 + 𝐹𝑖𝑛,2 𝑤ℎ𝑒𝑚𝑖 − 𝐹𝑜𝑢𝑡 𝑤ℎ𝑒𝑚𝑖 + 𝑟5 𝑉 Eq. A.44
𝑑𝑡

Xylose:

𝑑𝑀𝑋𝑦 𝑖𝑛,1 𝑖𝑛,2 𝑜𝑢𝑡

= 𝐹𝑖𝑛,1 𝑤𝑋𝑦 + 𝐹𝑖𝑛,2 𝑤𝑋𝑦 − 𝐹𝑜𝑢𝑡 𝑤𝑥𝑦
𝑑𝑡 Eq. A.45
+ ((1.136 𝑟4 ) − 𝑟5 − 𝑟𝑋𝑦 ) 𝑉

Furfural:

𝑑𝑀𝑓𝑢𝑟 𝑖𝑛,1 𝑖𝑛,2 𝑜𝑢𝑡

= 𝐹𝑖𝑛,1 𝑤𝑓𝑢𝑟 + 𝐹𝑖𝑛,2 𝑤𝑋𝑦 − 𝐹𝑜𝑢𝑡 𝑤𝑋𝑦 + 𝑟5 𝑉 Eq. A.46
𝑑𝑡

Cellulose:

𝑑𝑀𝑐𝑒𝑙 𝑖𝑛,1 𝑖𝑛,2 𝑜𝑢𝑡

= 𝐹𝑖𝑛,1 𝑤𝑐𝑒𝑙 + 𝐹𝑖𝑛,2 𝑤𝑐𝑒𝑙 − 𝐹𝑜𝑢𝑡 𝑤𝑐𝑒𝑙 − (𝑟1 + 𝑟2 ) 𝑉 Eq. A.47
𝑑𝑡

Cellobiose:

𝑑𝑀𝐶𝑏 𝑖𝑛,1 𝑖𝑛,2 𝑜𝑢𝑡

= 𝐹𝑖𝑛,1 𝑤𝐶𝑏 + 𝐹𝑖𝑛,2 𝑤𝐶𝑏 − 𝐹𝑜𝑢𝑡 𝑤ℎ𝑒𝑚𝑖 + 𝑟5 𝑉 Eq. A.48
𝑑𝑡

Glucose:

𝑑𝑀𝐺𝑙 𝑖𝑛,1 𝑖𝑛,2 𝑜𝑢𝑡

= 𝐹𝑖𝑛,1 𝑤𝐺𝑙 + 𝐹𝑖𝑛,2 𝑤𝐺𝑙 − 𝐹𝑜𝑢𝑡 𝑤𝐺𝑙
𝑑𝑡 Eq. A.49
+ (1.111 𝑟2 + 1.053 𝑟3 − 𝑟𝐺𝑙 ) 𝑉

242
Ethanol:

𝑑𝑀𝐸𝑡ℎ 𝑖𝑛,1 𝑖𝑛,2 𝑜𝑢𝑡

= 𝐹𝑖𝑛,1 𝑤𝐸𝑡ℎ + 𝐹𝑖𝑛,2 𝑤𝐸𝑡ℎ − 𝐹𝑜𝑢𝑡 𝑤𝐸𝑡ℎ + 𝑟𝑃 𝑉 Eq. A.50
𝑑𝑡

Cells:

𝑑𝑀𝑋
= 𝐹𝑖𝑛,1 𝑤𝑋𝑖𝑛,1 + 𝐹𝑖𝑛,2 𝑤𝑋𝑖𝑛,2 − 𝐹𝑜𝑢𝑡 𝑤𝑋𝑜𝑢𝑡 + 𝑟𝑋 𝑉 Eq. A.51
𝑑𝑡

Water:

𝑑𝑀𝑊 𝑖𝑛,1
= 𝐹𝑖𝑛,1 𝑤𝑊 + 𝐹𝑖𝑛,2 𝑤𝑋𝑖𝑛,2 − 𝐹𝑜𝑢𝑡 𝑤𝑋𝑜𝑢𝑡
𝑑𝑡 Eq. A.52
+ (0.055 𝑟1 + 0.11 𝑟2 + 1.052 𝑟3 + 0.333 𝑟5 ) 𝑉

The energy balance is similar to the other reactors in the process. However, the heat
of reaction will be taken from the following stoichiometric reaction since the kinetic
model does not specify the reactions and the stoichiometry in its formulation:

𝐶6 𝐻12 𝑂6 → 2 𝐶2 𝐻6 𝑂 + 2 𝐶𝑂2

3 𝐶5 𝐻10 𝑂5 → 5 𝐶2 𝐻6 𝑂 + 5 𝐶𝑂2

𝑑𝑈
= 𝐹𝑖𝑛,1 ℎ𝑖𝑛,1 + 𝐹𝑖𝑛,2 ℎ𝑖𝑛,2 − 𝐿𝑜𝑢𝑡 ℎ𝑜𝑢𝑡 + (∑ 𝑟𝑗 (−∆𝐻 𝑅 )) 𝑉
𝑑𝑡 Eq. A.53
+ 𝑄𝑆𝑆𝐶𝐹

The equation for the heat transferred from the jacket to the bulk consists of a
difference between the average temperature in the jacket and the temperature in the
bulk of the reactor.

𝑇𝑗𝑎𝑐𝑘𝑒𝑡,𝑖𝑛 + 𝑇𝑗𝑎𝑐𝑘𝑒𝑡,𝑜𝑢𝑡
𝑄𝑆𝑆𝐶𝐹 = 𝑈𝐶 𝐴𝑡𝑟𝑎𝑛𝑠 ( − 𝑇𝑜𝑢𝑡 ) Eq. A.54
2

243
 A.5. Distillation column
The distillation column comprises three main sub-units: the tray section, the top and
the bottom of the column. The top is also composed by three units: the total
condenser in which the vapour coming out from the tray section is completely
condensed. The drum is used for the collection of the condensate. The splitter is the
unit that serves as a recycle of the distillate. Finally, the reboiler that serves as vapour
supplier to increase the mass and energy transfer within the tray section. This unit is
a partial reboiler since one liquid stream and one vapour stream are produced. This is
also known as a Kettle reboiler.

To solve the set of differential and algebraic equations that describe a tray section,
condenser and reboiler in a distillation column, the following assumptions were
taken into account since most of them have a negligible contribution in the solution
of the system and, as shown in Section 3.7.1, the results were validated against
ChemCAD (Chemstations, 2015).

1. Perfect mixing on each tray

2. Vapour holdup is neglected
3. Relative volatility is obtained with a γ – ϕ formulation
4. No chemical reaction
5. No pressure drop

A.5.1 Tray
For a single tray of the column, the following equations were formulated, according
to the scheme shown in Figure A.3:

Mass balance:

𝑑𝑀𝑖𝑗
= 𝐿𝑖𝑛,𝑗 𝑥𝑖𝑛,𝑖𝑗 + 𝑉𝑖𝑛,𝑗 𝑦𝑖𝑛,𝑖𝑗 + 𝐹𝑗 𝑧𝑖𝑗 − 𝐿𝑜𝑢𝑡,𝑗 𝑥𝑜𝑢𝑡,𝑖𝑗 − 𝑉𝑜𝑢𝑡,𝑗 𝑦𝑜𝑢𝑡,𝑖𝑗 Eq. A.55
𝑑𝑡

𝑀𝑖𝑗 = 𝑋𝑜𝑢𝑡𝑖𝑗 𝑀𝑙𝑖𝑞 Eq. A.56

244
 Figure A.3: Schematic of a distillation column tray

𝑁𝑐

∑ 𝑋𝑜𝑢𝑡,𝑖𝑗 = 1 Eq. A.57

𝑖=1

𝑖 = 1, 2, … 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡𝑠 𝑗 = 1, 2, … 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑟𝑎𝑦𝑠

Energy balance:

𝑑𝑈𝑗 𝑙𝑖𝑞 𝑣𝑎𝑝 𝑙𝑖𝑞 𝑙𝑖𝑞 𝑣𝑎𝑝

= 𝐿𝑖𝑛,𝑗 ℎ𝑖𝑛,𝑗 + 𝑉𝑖𝑛,𝑗 ℎ𝑖𝑛,𝑗 + 𝐹𝑗 ℎ𝑓𝑗 − 𝐿𝑜𝑢𝑡,𝑗 ℎ𝑜𝑢𝑡,𝑗 − 𝑉𝑜𝑢𝑡,𝑗 ℎ𝑜𝑢𝑡,𝑗 Eq. A.58
𝑑𝑡

𝑙𝑖𝑞
𝑈𝑗 = 𝑀𝑙𝑖𝑞 ℎ𝑜𝑢𝑡,𝑗 Eq. A.59

𝑗 = 1, 2, … 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑟𝑎𝑦𝑠

Vapour-Liquid Equilibrium:

𝑦𝑜𝑢𝑡,𝑖𝑗 = 𝑘𝑖𝑗 𝑥𝑜𝑢𝑡,𝑖𝑗 Eq. A.60

𝑖 = 1, 2, … 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡𝑠 𝑗 = 1, 2, … 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑟𝑎𝑦𝑠

245
 Figure A.4: Flowsheet of the top of the distillation column

Molar overflow:

𝐿𝑜𝑢𝑡,𝑗 = 𝐿𝑖𝑛,𝑗 + 𝐹𝑗 Eq. A.61

𝑉𝑜𝑢𝑡,𝑗 = 𝑉𝑖𝑛,𝑗 Eq. A.62

𝑖 = 1, 2, … 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡𝑠 𝑗 = 1, 2, … 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑟𝑎𝑦𝑠

A.5.2 Condenser
The condenser section is divided in three parts: Heat exchanger, drum and splitter.
Figure A.4 describes the total condenser, drum and splitter used in this calculation.
For the condenser, the vapour stream coming out of the first stage of the distillation
column is totally condensed, giving as a result the following set of equations:

𝐿𝐶 = 𝑉𝐶 Eq. A.63

𝐿𝐶 ℎ𝐶𝑙𝑖𝑞 + 𝑄𝐶 = 𝑉𝐶 ℎ𝐶𝑣𝑎𝑝 Eq. A.64

𝑥𝑖,𝐶 = 𝑦𝑖,𝐶 Eq. A.65

𝑖 = 1, 2, … 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡𝑠

246
 Figure A.5: Schematic of a Kettle reboiler

A.5.3 Drum
In the case of the drum, by using the same mass balance like in the tray section, and
implementing the same assumptions, the following equation can be obtained:

𝑑𝑀𝑖𝐷
= 𝐿𝐶 𝑥𝑖,𝐶 − 𝐿𝐷 𝑥𝑖𝐷 Eq. A.66
𝑑𝑡

𝑀𝑖𝐷 = 𝑥𝑖𝐷 𝑀𝑙𝑖𝑞,𝐷 Eq. A.67

𝑑𝑈𝐷
= 𝐿𝐶 ℎ𝐶𝑙𝑖𝑞 − 𝐿𝐷 ℎ𝐷𝑙𝑖𝑞 Eq. A.68
𝑑𝑡

𝑙𝑖𝑞
𝑈𝐷 = 𝑀𝑙𝑖𝑞,𝐷 ℎ𝐷 Eq. A.69

For the splitter and according to Figure A.2, the only equation considered is the
reflux ratio defined by:

𝐿0
𝑅= Eq. A.70
𝐷

247
 A.5.4 Reboiler
The reboiler has the same mathematical treatment as a single tray, except that this
does not include an inlet vapour stream. The following equations are based on Figure
A.5:

Mass balance:

𝑑𝑀𝑖𝐵
= 𝐿𝑁𝑇 𝑥𝑁𝑇𝑗 − 𝐿𝐵 𝑥𝑖𝐵 − 𝑉𝐵 𝑦𝑖𝐵 Eq. A.71
𝑑𝑡

𝑀𝑖𝐵 = 𝑥𝑖𝐵 𝑀𝑙𝑖𝑞 Eq. A.72

𝑁𝐶

∑ 𝑥𝑖𝐵 = 1 Eq. A.73

𝑖=1

𝑖 = 1, 2, … 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡𝑠

Energy balance:

𝑑𝑈𝐵 𝑙𝑖𝑞
= 𝐿𝑁𝑇 ℎ𝑁𝑇 − 𝐿𝐵 ℎ𝐵𝑙𝑖𝑞 − 𝑉𝐵 ℎ𝐵𝑣𝑎𝑝 + 𝑄𝑅 Eq. A.74
𝑑𝑡

𝑈𝑗 = 𝑀𝑙𝑖𝑞 ℎ𝑗𝑙𝑖𝑞 Eq. A.75

Vapour-Liquid Equilibrium:

𝑦𝑖𝐵 = 𝐾𝑖𝐵 𝑥𝑖𝐵 Eq. A.76

𝑖 = 1, 2, … 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡𝑠

Molar overflow:

𝐿𝑁𝑇 = 𝐿𝐵 + 𝑉𝐵 Eq. A.77

248
 A.5.5 Tray section
Additional equations are required in the simulation and modelling of distillation
systems with multiple trays. These equations allow connectivity between two trays
inside the column and the different compositions and temperatures therein.

The liquid stream:

𝐿𝑖𝑛,𝑗 = 𝐿𝑜𝑢𝑡,𝑗−1 Eq. A.78

𝑥𝑖𝑛,𝑖𝑗 = 𝑥𝑜𝑢𝑡,𝑖𝑗−1 Eq. A.79

𝑇𝑙𝑖𝑞,𝑖𝑛 = 𝑇𝑜𝑢𝑡,𝑗−1 Eq. A.80

The vapour stream:

𝑉𝑖𝑛,𝑗 = 𝑉𝑜𝑢𝑡,𝑗+1 Eq. A.81

𝑦𝑖𝑛,𝑖𝑗 = 𝑦𝑜𝑢𝑡,𝑖𝑗+1 Eq. A.82

𝑇𝑣𝑎𝑝,𝑖𝑛 = 𝑇𝑜𝑢𝑡,𝑗+1 Eq. A.83

The following equations show the connections of the streams being recycled into the
column and the other units in the distillation system (i.e. condenser and reboiler). Eq.
A.84 to A.86 connect the outlet stream of the drum with reflux stream into the
column.

𝐿𝐷 = 𝐿0 + 𝐷 Eq. A.84

𝑥𝑖0 = 𝑥𝑖𝐷 Eq. A.85

𝑇0 = 𝑇𝐷 Eq. A.86

Equations A.87 to A.89 connect the outlet stream of the splitter with the inlet liquid
stream of the first tray of the column.

𝐿0 = 𝐿𝑖𝑛,1 Eq. A.87

249
𝑥𝑖0 = 𝑥𝑖𝑛,𝑖1 Eq. A.88

𝑇0 = 𝑇𝑙𝑖𝑞,𝑖𝑛,1 Eq. A.89

Similarly, equations A.90 to A.92 connect the vapour outlet stream from the first tray
with the inlet vapour stream of the condenser.

𝑉𝑜𝑢𝑡,1 = 𝑉𝐶 Eq. A.90

𝑦𝑜𝑢𝑡,𝑖1 = 𝑦𝑖𝐶 Eq. A.91

𝑇𝑜𝑢𝑡,1 = 𝑇𝑐 Eq. A.92

Other equations are required for the connection between the bottom of the column
and the reboiler. These equations are the following:

𝐿𝑜𝑢𝑡,𝑁𝑇 = 𝐿𝑁𝑇 Eq. A.93

𝑥𝑜𝑢𝑡,𝑖𝑁𝑇 = 𝑥𝑖𝑁𝑇 Eq. A.94

𝑇𝑜𝑢𝑡,𝑁𝑇 = 𝑇𝑁𝑇 Eq. A.95

𝑉𝑖𝑛,𝑁𝑇 = 𝑉𝐵 Eq. A.96

𝑦𝑖𝑛,𝑖𝑁𝑇 = 𝑦𝑖𝐵 Eq. A.97

𝑇𝑖𝑛,𝑁𝑇 = 𝑇𝐵 Eq. A.98

250
Figure A.6: Schematic of the membrane module (Marriott and Sorensen, 2003; Marriott et al.,
2001; Tsuyumoto et al., 1997)

𝑉𝑗+1
= 𝑅𝐵 Eq. A.99
𝐿𝐵

A.6. Membrane system

In this work, a membrane model is considered for the ethanol dehydration. The
product from a simple distillation column reaches the azeotropic point, which at
atmospheric pressure is around 94 % (w/w). The selective membrane removes the
water by the principle of concentration gradient. In this case study, the model is
described as a fibre-side fed co-current model. The module is composed by three
sub-units: The fibre side, in which the inlet stream is fed and the shell side, which
uses an inert gas or vacuum to remove the water. Finally, the membrane model is
represented by the principle of Fick's law of mass transfer due to the concentration
gradient. The scheme of the membrane system is shown in Figure A.6.

A.6.1. Fibre side

The fibre side is where the feed stream is passed through and separated into retentate
and permeate. The mass balance for each component includes the inlet and outlet
streams and the flux through the membrane. In similar way, the energy balance
depicts the energy flows in the system with the particularity of considering the heat
of vaporisation of the mix which is responsible for the decrease of the feed
temperature.

251
𝑑𝑀𝑖
= 𝐹 𝑍𝑖 − 𝐹𝑟𝑒𝑡𝑒𝑛 𝑥𝑖,𝑟𝑒𝑡𝑒𝑛 − 𝐴𝑚𝑒𝑚 𝐽𝑖 Eq. A.100
𝑑𝑡

𝑀𝑖 = 𝑥𝑖,𝑟𝑒𝑡𝑒𝑛 𝑀𝑡 Eq. A.101

𝑑𝑈 𝑙𝑖𝑞
= 𝐹 ℎ𝑓𝑒𝑒𝑑 − 𝐹𝑟𝑒𝑡𝑒𝑛 ℎ𝑟𝑒𝑡𝑒𝑛 − 𝐴𝑚𝑒𝑚 𝐽 𝐻 𝐽 Eq. A.102
𝑑𝑡

𝑙𝑖𝑞
𝑈 = 𝑢𝑟𝑒𝑡𝑒𝑛 𝑀𝑡 Eq. A.103

A.6.2 Shell side

The balances applied to this part of the system are similar to those described in
equations A.79 to A.82, except that the terms 𝐴𝑚𝑒𝑚 𝐽𝑖 and 𝐴𝑚𝑒𝑚 𝐽 𝐻 𝐽 are positive
because vapour is permeating through the membrane into the permeate stream.

𝑑𝑀𝑖
= 𝐹 𝑍𝑖 − 𝐹𝑝𝑒𝑟𝑚 𝑥𝑖,𝑝𝑒𝑟𝑚 + 𝐴𝑚𝑒𝑚 𝐽𝑖 Eq. A.104
𝑑𝑡

𝑀𝑖 = 𝑥𝑖,𝑝𝑒𝑟𝑚 𝑀𝑡 Eq. A.105

𝑑𝑈 𝑣𝑎𝑝
= −𝐹𝑝𝑒𝑟𝑚 ℎ𝑝𝑒𝑟𝑚 + 𝐴𝑚𝑒𝑚 𝐽 𝐻 𝐽 Eq. A.106
𝑑𝑡

𝑣𝑎𝑝
𝑈 = 𝑢𝑟𝑒𝑡𝑒𝑛 𝑀𝑡 Eq. A.107

In this work, the mathematical model that describes the dehydration of ethanol in the
pervaporation system was introduced by Tsuyumoto et al., (1997). This model
considers the permeation of water and ethanol through the membrane taking into
account operating conditions such as feed temperature, feed concentration and
permeate pressure.

𝐽𝐸𝑡ℎ𝑎𝑛𝑜𝑙 = 1.72𝑥10−10 𝑥𝐸𝑡ℎ𝑎𝑛𝑜𝑙 (𝑃𝑓𝑒𝑒𝑑 − 𝑃𝑝𝑒𝑟𝑚𝑒𝑎𝑡𝑒 ) Eq. A.108

252
 11500
8.086𝑥106 𝐸𝑋𝑃 (− ) 𝑃𝑝𝑒𝑟𝑚𝑒𝑎𝑡𝑒
𝐽𝑊𝑎𝑡𝑒𝑟 = 𝑇 (𝛾𝑓𝑒𝑒𝑑 𝑥𝑤𝑎𝑡𝑒𝑟 − 𝑦𝑤𝑎𝑡𝑒𝑟 )
𝑠𝑎𝑡
𝛿𝑚 𝑃𝑊𝑎𝑡𝑒𝑟
3390
3.441𝑥10−3 𝐸𝑋𝑃 (− )
+ 𝑇 ((𝛾𝑓𝑒𝑒𝑑 𝑥𝑤𝑎𝑡𝑒𝑟 )
2 Eq. A.109
2𝛿𝑚
2
𝑃𝑝𝑒𝑟𝑚𝑒𝑎𝑡𝑒
−( 𝑠𝑎𝑡 𝑦𝑤𝑎𝑡𝑒𝑟 ) )
𝑃𝑊𝑎𝑡𝑒𝑟

where:

𝐹 : Feed flow rate (kg/sec)

𝐻𝐽 : Enthalpy of permeate (J/kg)

ℎ : Enthalpy (J/kg)

𝐽𝑖 : Flux of component i (kg. m-2. s-1)

ℎ : Enthalpy (J/kg)

𝐾𝑖𝑗 : Relative volatility

𝐿 : Liquid flow rate (kg/sec)

𝑀𝑖 : Liquid holdup of component i (kg)

𝑀𝑡 : Total liquid holdup (kg)

𝑃𝑓𝑒𝑒𝑑 : Feed pressure (Pa)

𝑃𝒑𝒆𝒓𝒎 : Permeate pressure (Pa)

𝑄𝑪 : Heat released in the condenser (J/kg)

𝑄𝑩 : Heat duty in the reboiler (J/kg)

𝑅 : Reflux ratio

𝑅𝑩 : Boilup ratio

𝑢 : Specific internal energy (J/kg)

𝑈 : Internal energy (J)

253
𝑉 : Vapour flow rate (kg/sec)

𝑋𝑖𝑗 : Liquid mass fraction of component i in tray j

xi : Liquid mole fraction of component i

𝑌𝑖𝑗 : Vapour mass fraction of component i in tray j

yi : Vapour mole fraction of component i

𝑍𝑖𝑗 : Feed mass fraction of component i in tray j

254