Bioresource Technology 102 (2011) 1174–1184

Contents lists available at ScienceDirect

Bioresource Technology

journal homepage: www.elsevier .com/locate /bior tech

Dynamic model-based evaluation of process configurations for integratedoperation of hydrolysis and co-fermentation for bioethanol productionfrom lignocellulose

Ricardo Morales-Rodriguez a, Anne S. Meyer b, Krist V. Gernaey c, Gürkan Sin a,⇑a CAPEC, Dept. of Chemical and Biochemical Engineering, Technical University of Denmark, DK-2800 Lyngby, Denmarkb Center of Bioprocess Engineering, Dept. of Chemical and Biochemical Engineering, Technical University of Denmark, DK-2800 Lyngby, Denmarkc Center for Process Engineering and Technology, Dept. of Chemical and Biochemical Engineering, Technical University of Denmark, DK-2800 Lyngby, Denmark

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

Article history:Received 21 June 2010Received in revised form 3 September 2010Accepted 9 September 2010Available online 17 September 2010

Keywords:BioethanolProcess configurationHydrolysis and co-fermentationDynamic modelsSSCF

0960-8524/$ - see front matter � 2010 Elsevier Ltd. Adoi:10.1016/j.biortech.2010.09.045

⇑ Corresponding author. Tel.: +45 4525 2806; fax: +E-mail address: gsi@kt.dtu.dk (G. Sin).

In this study a number of different process flowsheets were generated and their feasibility evaluatedusing simulations of dynamic models. A dynamic modeling framework was used for the assessment ofoperational scenarios such as, fed-batch, continuous and continuous with recycle configurations. Eachconfiguration was evaluated against the following benchmark criteria, yield (kg ethanol/kg dry-biomass),final product concentration and number of unit operations required in the different process configura-tions. The results show that simultaneous saccharification and co-fermentation (SSCF) operating in con-tinuous mode with a recycle of the SSCF reactor effluent, results in the best productivity of bioethanolamong the proposed process configurations, with a yield of 0.18 kg ethanol/kg dry-biomass.

� 2010 Elsevier Ltd. All rights reserved.

1. Introduction

Biofuels can potentially contribute to alleviate the current cli-mate change and energy resource challenges, which today’s societyis facing. However, turning biofuels production at industrial scaleinto a success story is only possible by solving a number of chal-lenges. This includes securing a sustainable feedstock supply aswell as optimizing the techno-economic feasibility of cellulosicbiomass conversion technologies by defining optimal process con-figurations (Gírio et al., 2010; Huber et al., 2006; Regalbuto, 2009).

Thus far the transfer of these conversion technologies fromproof-of-concept to industrial scale has been mainly done on anempirical basis that is typically inefficient and costly in terms oftime and resource consumption (Aden et al., 2002; Gnansounou,2010; Larsen et al., 2008). Although various flowsheet configura-tions have been reviewed and evaluated in the literature basedon steady state models (Alvarado-Morales et al., 2009; Cardonaand Sánchez, 2007; Lynd et al., 2008), quantitative modeling toolsfor the dynamic simulation and evaluation of different processflowsheet options have until now not been used for cellulosicethanol production processes.

ll rights reserved.

45 4593 2906.

The objective of this work was to develop a Dynamic Lignocel-lulosic Bioethanol (DLB1.0) modeling platform allowing the quan-titative simulation and comparison of different processconfigurations for 2nd generation (2G) bioethanol plants, therebyproviding a basis for evaluation of the most promising processflowsheet. The study has taken a conventional process configura-tion (Margeot et al., 2009) as a base case using the dimensionsand process conditions proposed by Aden et al. (2002). Dynamicmodels for each unit process operation, including pre-treatment,enzymatic hydrolysis and co-fermentation, have been imple-mented in one software platform (Matlab Simulink), and con-nected to obtain the plantwide dynamic model. This dynamicmodel has subsequently been used to simulate and evaluate differ-ent process configurations for 2G bioethanol production on the ba-sis of several benchmark criteria, notably the ethanol yield per unitbiomass.

2. Methods

2.1. DLB1.0 mathematical models: pre-treatment, hydrolysis, co-fermentation and simultaneous saccharification and co-fermentation(SSCF)

The model-based simulation framework involved two mainparts: (1) the collection, analysis and identification of the most

Nomenclature

ADB accumulated dry-biomassC continuos operationC_RECY continuous-recycle operationCin;i concentration of compound i in the feedstream of the

unit (g/kg), (g/L)CA arabinose concentration (g/kg)CAn arabinan concentration (g/kg)CASL acid-soluble lignin concentration (g/kg)CLn lignin concentration (g/kg)CEiT

total enzyme concentration (g/kg)CE1B bound concentration of CHB and EG (g/kg)CE2B bound concentration of b-glucosidase (g/kg)CE1F free enzyme concentration of CHB and EG in solution (g/kg)CE2F free enzyme concentration of b-glucosidase in solution

(g/kg)CEtG ethanol concentration from glucose fermentation (g/kg)CEtXy ethanol concentration from xylose fermentation (g/kg)CEt ethanol concentration (g/kg)CG glucose concentration (g/kg)CG2 cellobiose concentration (g/kg)CXy xylose concentration (g/kg)CXn xylan concentration (g/kg)CGn glucan (cellulose) concentration (g/kg)COC other compounds concentration (g/kg)CXG cell dry weight in glucose fermentation (g/L)CXXy cell dry weight in xylose fermentation (g/L)DLB1.0 Dynamic Lignocellulosic Bioethanol model version 1.0Ea activation energy = �5540 cal/molE1 max maximum mass of enzyme 1 that can be adsorbed onto

a unit mass of substrate, 0.06 g protein/g substrateE2 max maximum mass of enzyme 2 that can be adsorbed onto

a unit mass of substrate, 0.01 g protein/g substrateEtmax;G ethanol concentration above which cells do not grow in

glucose fermentation, 95:40 for Et 6 95:4 g=L, 129:90 for 95:4 < Et 6 129:9 g=L

Etmax;Xy ethanol concentration above which cells do not grow inxylose fermentation = 59.040 g/L

Et0max;G ethanol concentration above which cells do not produceethanol in glucose fermentation, 103 for Et 6 103 g=L,136:40 for 103 < Et 6 136:4 g=L

Et0max;Xy ethanol concentration above which cells do not produceethanol in xylose fermentation = 60.20 g/L

FB fed-batch operationK1ad dissociation constant for enzyme 1 = 0.4 g protein/

g substrateK2ad dissociation constant for enzyme 2 = 0.1 g protein/

g substratekEH

1;G reaction rate constant for glucose 1 in the enzymatichydrolysis = 7.18 g/mg h

kEH2;G reaction rate constant for glucose 1 in the enzymatic

hydrolysis = 285.5 h�1

kEHG2 reaction rate constant for cellobiose in the enzymatic

hydrolysis = 22.3 g/mg hKEH

1IEt inhibition constant for ethanol 1 in the SSCFunit = 0.15 g/kg

KEH1IG inhibition constant for glucose 1 = 0.1 g/kg

KEH2IG inhibition constant for glucose 2 = 0.04 g/kg

KEH3IG inhibition constant for glucose 3 = 3.9 g/kg

KEH1IG2 inhibition constant for cellobiose 1 = 0.015 g/kg

KEH2IG2 inhibition constant for cellobiose 2 = 132.0 g/kg

KEH1IXy inhibition constant for xylose 1 = 0.1 g/kg

KEH2IXy inhibition constant for xylose 2 = 0.2 g/kg

KEH3IXy inhibition constant for xylose 3 = 201.0 g/kg

KM substrate (cellobiose) saturation constant = 24.3 g/kgKCF

1G monod constant, for growth on glucose = 0.565 g/L

KCF2Xy monod constant, for growth on xylose = 3.4 g/L

K 0CF5IG inhibition constant, for product formation from glu-

cose = 4890.0 g/LK 0CF

6IXy inhibition constant, for product formation from xylo-se = 81.30 g/L

K 0CF5G monod constant, for product formation from glu-

cose = 1.342 g/LK 0CF

6Xy monod constant, for product formation from xylo-se = 3.4 g/L

KCF1XGIG inhibition constant, for growth on glucose = 283.7 g/L

KCF2IXy inhibition constant, for growth on xylose = 18.1 g/L

mG maintenance coefficient in glucose fermentation =0.097 h�1

mXy maintenance coefficient in xylose fermentation =0.067 h�1

Qin feed flow flow rate (kg/h)Qout outlet flow rate (kg/h)ri;j reaction rate of compound i for the different unit oper-

ations, g=kg h = g=L h, qmixture ¼ 1 kg=LrX;TOT total reaction rate for the biomass (g/L h)rEt;TOT total reaction rate for ethanol (g/L h)R universal gas constant, 1.9872 cal=mol KRGn substrate reactivityT temperature, KREt=dry-biomass ethanol/dry-biomass ratioURM unreacted raw materialV reaction volume (kg-slurry)xi mass fraction of glucose or xylose in the glucose and xy-

lose mixture = CG=ðCG þ CXyÞ and CXy=ðCG þ CXyÞ, respec-tively.

YEtG=G product yield constant (g-ethanol/g-glucose) = 0.470 g/gYEtXy=Xy product yield constant (g-ethanol/g-xylose) = 0.400 g/gYXG=G cell yield constant from glucose (g-cells/g-sub-

strate) = 0.115 g/gYXXy=Xy cell yield constant from xylose (g-cells/g-sub-

strate) = 0.162 g/g

Greek lettersa constant relating substrate reactivity with degree of

hydrolysis, 1bG constants in product inhibition model in glucose fer-

mentation1:29 for Et 6 95:4 g=L;0:25 for 95:4 < Et 6 129:9 g=L

bXy constant in the product inhibition model in xylose fer-mentation = 1.036 g/L

cG maximum specific rate of glucose formation1:42 for Et 6 95:4 g=L

cXy maximum specific rate of xylose formation = 0.608 g/Llmax;G maximum specific growth rate in glucose fermenta-

tion = 0.662 h�1

lmax;Xy maximum specific growth rate in xylose fermenta-tion = 0.190 h�1

mmax;G maximum specific rate of glucose formation = 2.005 h�1

mmax;Xy maximum specific rate of xylose formation = 0.250 h�1

R. Morales-Rodriguez et al. / Bioresource Technology 102 (2011) 1174–1184 1175

1176 R. Morales-Rodriguez et al. / Bioresource Technology 102 (2011) 1174–1184

promising mathematical models for pre-treatment, enzymatichydrolysis and co-fermentation, and, (2) the design, simulationand comparison of different integrated operational scenarios suchas, fed-batch, continuous and continuous-recycle (Sin et al., 2010).The chosen configurations employ separate hydrolysis and fermen-tation (SHF), where – as the name implies – the enzymatichydrolysis as well as the fermentation has been performed in dif-ferent unit operations. In addition, a model for the simultaneoussaccharification and co-fermentation (SSCF) process was devel-oped and configurations including SSCF reactors were simulatedand compared with the results from the base case.

2.1.1. Pre-treatmentThe mathematical model for the pre-treatment process (see

Supplementary data (S.1)), which is the operation aiming at break-ing down the structure of the lignocellulosic matrix of the raw bio-mass, has been taken from Lavarack et al. (2002).

2.1.2. Enzymatic hydrolysisAmong many competing models, the mathematical model

developed by Kadam et al. (2004) (see Table 1) has been chosento describe the enzymatic hydrolysis of lignocellulosic biomass,as it has already been extensively analyzed statistically (Sinet al., 2010), verified experimentally (Hodge et al., 2009) and hasalso been the subject of a thorough model validation (Morales-Rodriguez et al., 2010). The model of Kadam et al. (2004) quantifiesthe enzyme catalyzed decomposition of the cellulose content in thebiomass, where cellulose decomposes to cellobiose (Eq. (1)) andglucose (Eq. (2)) by the action of the enzymes endo-b-1,4-glucan-ases + exoglucanases (cellobiohydrolases), cellobiose is hydrolyzedto glucose (Eq. (3)) by b-glucosidases, and where the model also ac-counts for the enzyme adsorption (Eq. (4)), the levels of free andbound enzyme (Eq. (5)), the substrate reactivity (Eq. (6)), and theeffect of the temperature on the saccharification modeled by theArrhenius equation (Eq. (7)). The mathematical model also takesinto account the potential inhibition by cellobiose, glucose, andxylose on the different rate constants (Kadam et al., 2004). It isimportant to note that xylose is a product of the pretreatment sec-tion, which is transported in the slurry to the hydrolysis reactor(see supplementary data (S.1)).

2.1.3. Co-fermentationThe mathematical model employed in this work for the co-fer-

mentation process has been proposed by Krishnan et al. (1999)(see Table 2) and considers the simultaneous conversion of xyloseand glucose to ethanol. Based on new research findings that

Table 1Kinetic expressions of the enzymatic hydrolysis model (Kadam et al., 2004).

Cellulose to cellobiose (g/kg h)r1;EH ¼

kEHG2 CE1B

RGn CGn

1þCG2

KEH1IG2

þ CGKEH

1IG

þCXy

KEH1IXy

Eq. (1)

Cellulose to glucose (g/kg h) r2;EH ¼kEH

1;GðCE1BþCE2B

ÞRGn CGn

1þCG2

KEH2IG2

þ CGKEH

2IG

þCXy

KEH2IXy

Eq. (2)

Cellobiose to glucose (g/kg h)r3;EH ¼

kEH2;G CE2F

CG2

KM ð1þCG

KEH3IG

þCXy

KEH3IXy

ÞþCG2

Eq. (3)

Enzyme adsorption (g/kg h) CEiB¼ Ei max Kiad CEiF

CGn

1þKiad CEiF

Eq. (4)

Enzyme (g/kg) CEiT¼ CEiF

þ CEiBEq. (5)

Substrate reactivity RGn ¼ aCGn=C0Gn

Eq. (6)

Temp. dependence kEHirðT2Þ ¼ kEH

irðT1Þe�Ea=Rð1=T1�1=T2Þ;

30 �C 6 T 6 55 �C

Eq. (7)

Cellulose kinetic (g/kg h) rGn;EH ¼ �r1;EH � r2;EH Eq. (8)Cellobiose kinetic (g/kg h) rG2 ;EH ¼ 1:056r1;EH � r3;EH Eq. (9)Glucose kinetic (g/kg h) rG;EH ¼ 1:111r2;EH þ 1:053r3;EH Eq. (10)Water kinetic (g/kg h) rW ;EH ¼ �0:055r1;EH�

0:111r2;EH � 1:05263r3;EH

Eq. (11)

indicate that conversion of 5-carbon sugars has become possiblevia novel recombinant microorganisms, this work considers bio-ethanol production from the conversion of both xylose and glucoseto ethanol as a reliable future scenario (Bettiga et al., 2009). The co-fermentation model is based on the recombinant yeast Saccharo-myces cerevisiae strain 1400 (pLNH33). The use of this yeast asthe fermenting organism differs from the NREL model (Adenet al., 2002) that uses Zymomonas mobilis for the glucose and xylosefermentation in a series of different anaerobic batch fermentors.

The mathematical model for co-fermentation involves the reac-tion rates for: (1) cell growth on glucose (Eq. (12)) and xylose (Eq.(13)); (2) the total yeast cell mass production as the average prod-uct of the cell growth on glucose and xylose using the respectivemass fraction of these compounds present in the mixture (Eq.(14)); (3) consumption of glucose (Eq. (15)) and xylose (Eq. (16));(4) formation of ethanol from glucose (Eq. (17)) and xylose (Eq.(18)); and (5) overall formation of ethanol (Eq. (19)). The model ac-counts for substrates and product inhibition as well as the effect ofthe inoculum size that is employed for the cultivation.

For the sake of simplicity a detoxification step was not includedafter the pre-treatment; neither was the formation and removal ofgypsum – as considered in the NREL model (Aden et al., 2002) –included in the dynamic model.

2.1.4. Simultaneous saccharification and co-fermentation (SSCF)model

The development of the reaction kinetics for the simultaneoussaccharification and co-fermentation model has been carried outby combining the enzymatic hydrolysis and co-fermentation mod-els described above (Morales-Rodriguez et al., in press). It is knownthat the presence of ethanol during enzymatic hydrolysis can in-duce a certain level of inhibition on cellulose degradation aspointed out by Bezerra and Dias (2005). However, for ethanol con-centrations less than 4 M (equivalent to 18.42% wt/v ethanol) nosignificant inhibition of ethanol on the enzymatic activities in thesaccharification has been found (Ooshima et al., 1985). Moreover,Philippidis et al. (1993) have shown that the value of the inhibitioncoefficient of ethanol on cellulose (conversion to cellobiose) isapproximately 10 times higher than the inhibition coefficient ofcellobiose on cellulose conversion. Therefore, in this study, the va-lue of the inhibition constant of cellobiose on cellulose (KEH

1IG2) pro-posed by Kadam et al. (2004) for the enzymatic hydrolysis ofcellulose to cellobiose has been multiplied by a factor 10 in orderto obtain the inhibition constant for ethanol. The potential inhibi-tion of ethanol on the enzymatic hydrolysis of cellobiose to glucosewas neglected as it is assumed insignificant (Bezerra and Dias,

Table 2Kinetic expressions of the co-fermentation model (Krishnan et al., 1999).

BiomassGlucose

(g/L h)r1;CF ¼

dCXGdt ¼

lmax;GCG

KCF1GþCGþ

C2G

KCF1XG IG

1� CEtGEtmax;G

� �bG� �

Eq. (12)

BiomassXylose

(g/L h)r2;CF ¼

dCXXy

dt ¼lmax;Xy CXy

KCF2XyþCXyþ

C2Xy

KCF2IXy

1�CEtXy

Etmax;Xy

� �bXy� �

Eq. (13)

Biomass kinetic(g/L h)

rX;TOT ¼ xGr1;CF þ xXyr2;CF Eq. (14)

Glucose (g/L h) �r3;CF ¼ 1YEtG=G

dCEtGdt ¼

1YXG =G

dCXGdt þmGCXG

Eq. (15)

Xylose (g/L h) �r4;CF ¼ 1YEtXy=Xy

dCEtXy

dt ¼1

YXXy =Xy

dCXXy

dt þmXyCXXy

Eq. (16)

EthanolGlucose

(g/L h)r5;CF ¼ 1

CXG

dCEtGdt ¼

mmax;G CG

K 0CF5GþCGþ

C2G

K0CF5IG

1� CEtGEt0max;G

� �cG� � Eq. (17)

EthanolXylose

(g/L h) r6;CF ¼ 1CXXy

dCEtXy

dt ¼mmax;Xy CXy

K 0CF6XyþCXyþ

C2Xy

K0CF6IXy

1�CEtXy

Et0max;Xy

� �cXy� � Eq. (18)

Ethanol kinetic(g/L h)

rEt;TOT ¼ r5;CF þ r6;CF Eq. (19)

R. Morales-Rodriguez et al. / Bioresource Technology 102 (2011) 1174–1184 1177

2005; Philippidis et al., 1993). Thus, only the mathematical expres-sion for cellulose decomposition to cellobiose (Eq. (1)) has beenmodified by adding an additional ethanol inhibition factorCEt/K1IEt as shown in Eq. (20).

r1;EH ¼kEH

G2CE1B RGnCGn

1þ CG2KEH

1IG2þ CG

KEH1IGþ CXy

KEH1IXyþ CEt

KEH1IEt

ð20Þ

The units of model components are different in the separate hydro-lysis (g/kg) and co-fermentation models (g/L) (see above). To re-solve this issue, it is assumed in the mathematical model for theSSCF process that the density of the mixture is equal to 1 kg/L, suchthat the models can be integrated.

2.2. Mass balances

This study has employed fed-batch and continuous operationsfor hydrolysis, co-fermentation and SSCF units, and the genericmass balance for the different compounds present in the unit oper-ations involved in the process is represented as follows:

VdCi

dt¼ QinCin;i � Q outCi � ri;jV � Ci

dVdt

ð21Þ

where, Qin and Qout are the feed flow rate and outlet flow rate,respectively. Cin;i is the concentration of compound i in the feed,ri;j is the rate of generation(+)/degradation(–) of compound i inthe reaction for the j unit operation, V is the reactor volume, andCi is the actual concentration of compound i in the reactor, whichwill be equal to the concentration in the outlet during the drawingperiod of the reactor. As far as reaction volume is concerned, this isdescribed as follows:

dVdt¼ Qin � Qout ð22Þ

For fed-batch operation, Qin is different from zero during the load-ing period and Qout is different from zero for the drawing time. Withrespect to the continuous operation, Qin and Qout are assumed to beidentical during the operation of the enzymatic hydrolysis, co-fermentation and SSCF units, i.e. constant tank volume is assumed.

The mathematical model for the solid–liquid separation as-sumes an ideal separation in steady state. More details about themodel can be found in the Supplementary data (S.2).

3. Results and discussion

3.1. Process technology configurations: upstream

A number of process configurations have previously been pro-posed for bioethanol production (Aden et al., 2002; Cardona andSánchez, 2007; Larsen et al., 2008; Lynd et al., 2008; Margeotet al., 2009; Shell et al., 2004). Recently, Dutta et al. (2010) pro-posed the reconfiguration of the flowsheet presented by Adenet al. (2002) comparing the capabilities of different microorgan-isms for sugar fermentation as well as including a techno-eco-nomic study based on data from bench-scale experiments.

3.1.1. Base case: conventional bioethanol process flowsheetIn this study, a conventional process configuration (Margeot

et al., 2009) has been used as a base case (see Fig. 1a). This processconfiguration consists of four main sections: pre-treatment, enzy-matic hydrolysis, fermentation and downstream processes. First ofall, the feedstock is treated in the pre-treatment section (using di-luted acid pre-treatment), and the product from this operation ispassed to the enzymatic hydrolysis unit to perform the conversionof cellulose biomass to glucose. Afterwards, the effluent leaving theenzymatic hydrolysis unit passes through the solid–liquid separa-

tor where a percentage of solids is sent to the power generationsection (not shown in Fig. 1a), while the liquor stream is sent tothe fermentation to ferment the sugars into ethanol. The outputstream from the fermentation unit is then transferred to the down-stream operations to separate the most valuable products(ethanol) and recover those compounds that can be reused in theupstream sections – especially water.

3.1.2. SHF: Operational scenariosThe SHF process design for the hydrolysis and fermentation

processes is investigated using nine different configurationswhich refer to various combinations of fed-batch (FB), continu-ous (C) and continuous-recycle (C_RECY) operations (Figs. 1and 2).

3.1.2.1. Fed-batch enzymatic hydrolysis with fed-batch, continuousand continuous-recycle operation in the co-fermentation reac-tors. This section describes the three process configurations whichemploy fed-batch operation in the enzymatic hydrolysis units fol-lowed by fed-batch, continuous or continuous-recycle operationsin the co-fermentation units (Fig. 1a–c, respectively). Fed-batchoperation (see Fig. 1a) in the co-fermentation reactor is similar tothe base case described in Section 3.1.1. When the co-fermentationreactors are working in the continuous mode, (Fig. 1b) the liquorleaving the solid–liquid separator is distributed by a stream sepa-rator (‘‘splitter” in Fig. 1) to the reactors where the conversion ofsugars into ethanol is accomplished. The continuous-recycle con-figuration (Fig. 1c) operates similarly to the continuous operation(Fig. 1b), but the addition of two more unit operations per fermen-tor, one mixer and one settler tank, is necessary. The settler tankseparates the solids from the liquids in the effluent of the fermen-tor by gravity settling and recycles the solids back to the mixer unitwhich also contains the yeast. This recycling ensures that a highconcentration of solids is maintained in the co-fermentationreactors.

3.1.2.2. Continuous enzymatic hydrolysis with fed-batch, continuousand continuous-recycle operation in the co-fermentation reac-tors. Continuous operation in the enzymatic hydrolysis sectioncombined with the co-fermentation step working either in fed-batch, continuous or continuous-recycle mode is also investi-gated (see Fig. 1d–f, respectively). With enzymatic hydrolysisin batch mode as a reference (Fig. 1a–c), continuous operationrequires less unit operations to handle the biomass flow ratefrom the pre-treatment section in order to fulfill the necessaryresidence time in the hydrolysis reactors. Co-fermentation reac-tors are operated in the same manner as described in Sec-tion 3.1.2.1. However, some differences are found in otherparts of the process flowsheet. For example, in the co-fermenta-tion reactors operating in fed-batch mode (Fig. 1d), the liquorgenerated by the solid–liquid separator is fed to the co-fermen-tation reactors until reaching their maximum capacity while theremaining amount is stored in the buffer tank. For continuous(Fig. 1e) and continuous-recycle operation (Fig. 1f) of the fer-mentors, the effluent from the solid–liquid separator is conveyeddirectly to the co-fermentation section, thus avoiding the use ofthe separator unit (‘‘splitter”) whether it is compared with theconfigurations illustrated in Fig. 1b and 1c.

3.1.2.3. Continuous-recycle enzymatic hydrolysis with fed-batch,continuous and continuous-recycle operation in the co-fermentationreactors. Another process configuration in the enzymatic hydrolysissection is based on the recycle of the insoluble solids stream fromthe solid–liquid separator (Fig. 2a–c). This recycle stream is thenmixed with the effluent generated in the pre-treatment section be-fore entering the hydrolysis reactor. After solid–liquid separation,

Fig. 1. SHF process configurations for: (a) FB-FB, (b) FB-C, (c) FB-C_RECY, (d) C-FB, (e) C-C, and (f) C-C_RECY.

1178 R. Morales-Rodriguez et al. / Bioresource Technology 102 (2011) 1174–1184

the configurations for fed-batch (Fig. 2a), continuous (Fig. 2b) andcontinuous-recycle (Fig. 2c) in the co-fermentation section, are work-ing in a similar manner as described in Section 3.1.2.2.

3.1.3. SSCF operational scenariosThree different configurations are proposed for the integration

of enzymatic hydrolysis and co-fermentation in the same unit

Fig. 2. SHF process configurations for: (a) C_RECY-FB, (b) C_RECY-C, (c) C_RECY-C_RECY. SSCF process configurations for: (d) FB, (e) C, and (f) C_RECY.

R. Morales-Rodriguez et al. / Bioresource Technology 102 (2011) 1174–1184 1179

operation (Fig. 2d–f). During fed-batch operation of the SSCF unit(illustrated in Fig. 2d), this unit is fed continuously with the efflu-ent that is leaving the pre-treatment section. The effluent from theSSCF reactor passes through a solid–liquid separator where the li-quor is sent onwards to the downstream processing section, and

the solids are collected for subsequent power generation (notshown).

Continuous operation is also investigated for the SSCF unitoperation (see Fig. 2e) where the output stream from the pre-treat-ment process is split into three parts to feed the three parallel SSCF

1180 R. Morales-Rodriguez et al. / Bioresource Technology 102 (2011) 1174–1184

units. Afterwards, a solid–liquid separation is carried out wherethe resulting liquid stream is sent to the downstream processingsection to perform the recovery and purification of the ethanol.On the other hand, the solid streams are sent to the power gener-ation section for combustion.

Another configuration includes the recycle of the insoluble sol-ids stream to the SSCF unit (see Fig. 2f). This stream is mixed withthe liquid stream from the pre-treatment unit. This action aims toproduce the highest possible yield of ethanol per amount of pro-cessed raw biomass material, thereby reducing the waste of rawmaterials in the biofuel production plant.

3.1.4. Scheduling for operational scenariosWhen fed-batch processes are used (Figs. 1a–d, 2a and 2d), it is

assumed that parallel fed-batch reactors are operated following abatch scheduling scheme consisting of a sequence of differentoperational phases – for example fill, react, draw, idle – that are re-peated over time.

The schedule for fed-batch operation (see Fig. 3a) describing theoperation of the hydrolysis and co-fermentation units (used in theconfiguration in Fig. 1a) can be understood as follows: for reactornumber one a cycle of operation lasts 60 h. It starts with the load-ing, an operation that takes 12 h, and is followed by 36 h of reac-tion time. Finally it ends with 12 h of drawing/emptying thereactor contents. Upon the completion of the first cycle, the nextcycle starts again by repeating the same schedule. The first fermen-tation reactor therefore starts after 48 h then following the 12, 36and 12 h scheduling (Fig. 3a).

Regarding the co-fermentation unit, the loading period is as-sumed to start simultaneously with the drawing of the contentsfrom the hydrolysis unit, thus assuming that an ideal solid–liquidseparation operating in steady state is present between hydrolysisand co-fermentation. Some configurations (Figs. 1d and 2a) justemploy the scheduling strategy (shown in Fig. 3b) for the co-fer-mentation unit since the hydrolysis is operating in continuousmode. It is important to remark that a buffer tank is needed afterthe hydrolysis units (operating in continuous) to buffer the contin-uous flow before it is fed to the fed-batch operated fermentors.Similarly, when the SSCF configuration is operated in fed-batch

Fig. 3. (a) Scheduling for FB operation in the hydrolysis and fermentation sections (Fig. 1a(b) Scheduling for FB operation in the fermentation section for the configuration shown

mode (Fig. 2d) it also uses the scheduling strategy outlined inFig. 3b, since in this configuration the effluent from the pre-treat-ment section is directly fed to the SSCF units.

3.2. Process characteristics data and simulation platform used forsimulations

Process characteristics and information regarding the dimen-sion of the units have been taken from Aden et al. (2002) (seeTable 3). The DLB1.0 simulation of the proposed process flowsheetfor bioethanol production has been solved using the MatLab/Simu-link (R2008b) platform (The Mathworks, Natick, MA). The MatLabcode is available upon request from the authors.

3.3. Results: technological evaluation

3.3.1. Benchmark criteria for comparison of the DLB1.0 simulation ofthe configurations

The comparison of the performance of the different processflowsheets has been performed by using as evaluation criteria:the maximum ethanol/dry-biomass ratio, the minimum fractionof unreacted raw material and the maximum final ethanolconcentration.

3.3.1.1. Mathematical expressions for the ethanol/dry-biomass ratioand unreacted raw material fraction. The ethanol/dry-biomass ratiohas been calculated on the basis of the total amount of ethanol thatis transferred to the downstream processing section as follows:

REt=dry-biomass ¼Total Mass Et

Total Mass Dry Biomassð23Þ

The fraction of unreacted raw material (URM) has been calcu-lated using the accumulated dry-biomass (ADB) in the process plusthe dry-biomass separated in the solid–liquid separator unit versusthe total amount of dry-biomass fed in the operating time (Eq.(24)):

URM ¼ ADBþ Solid stream from S—L separatorTotal Mass Dry Biomass

ð24Þ

), FB operation in the hydrolysis unit in the configuration illustrated in Fig. 1b and c.in Fig. 1d and 2a.

Table 3Main process characteristic and conditions for the simulation of the different process configurations.

Feature Value Pretreatment

Feedstocka Corn stover Reactivea Diluted sulfuric acidDry feedstock capacitya 98,000 kg/h Concentration % (wt/v)a 1.1Operating time considered for the

evaluation348 h Residence timea 2 min

% Dry basis composition Temperature 443 KGlucana 37.4 Solid in the reactor % (wt/v)a 30Xylana 21.1 Feed flow rate 327696.66 kg/hArabinana 2.9 Treated mass in the reactor 10923.22 kgLignina 18Asha 5.2Other compounds 15.4Enzymatic hydrolysis Co-fermentationTemperaturea 338 K Temperature 303 KInitial solid concentration% (wt/v) 20 Inoculum levelb 10%Size of the vesselsa 3596 m3 (each) Size of the vesselsa 3596 m3 (each)Number of vesselsa 5 Number of vesselsa 5Enzyme Cellulases (EG, CBH and BDG) Organism Saccharomyces cerevisiae strain1400

(pLNH33)SSCF Solid–liquid separatorTemperature 308 K Solid separator efficiency 90%Inoculum levelb 10% Percent of water content in the insoluble

solid stream50%

Size of the vesselsa 3596 m3 (each) Solid settlerNumber of vesselsa 5 Solid separator efficiency 90%Enzymes Cellulases (EG, CBH and BDG) Percent of separated effluent 50%Organism Saccharomyces cerevisiae strain1400

(pLNH33)

a NREL report (Aden et al., 2002).b Fed-batch configuration.

R. Morales-Rodriguez et al. / Bioresource Technology 102 (2011) 1174–1184 1181

3.3.2. Comparison of the DLB1.0 simulations of the configurationsbased on: ratio ethanol/dry-biomass, unreacted raw material fractionand ethanol concentration

Among the different DLB1.0 simulations of the configurations,the maximum ethanol yield obtained is found for the SSCF processconfiguration operated with continuous feed with recycling of thesolids (see Fig. 2f). This outcome can be explained to a large extentby the positive effect of the recycle, which improves the processefficiency in two ways: (i) by recycling the unused raw material(that is cellulose) the amount of raw material wastage is decreased– this is illustrated in Table 4 – where the amount of unreacted rawmaterial for the SSCF-C_RECY is 0 and (ii) by recycling the yeast,the concentration of microorganisms maintained in the reactor isincreased significantly to 9.75% (wt/v) in comparison to2.13%(wt/v) in the SSCF-C configuration.

The second best yield was found for a SHF type process whereboth the hydrolysis and the co-fermentation units are operated con-tinuously with recycle (C_RECY-C_RECY) (Fig. 2c). This result dem-onstrates also that continuous operation with a recycle has themajor positive impact among the process flowsheet configurations.

The third best yield was found for a SHF type process configura-tion where hydrolysis is operated continuously with recycle whilethe co-fermentation is just in continuous mode (C_RECY-C)(Fig. 2b). The 0.0245 kg ethanol/kg dry-biomass decrease in theethanol yield is attributable to the lack of recycle in the co-fermen-tation reactor. Compared to the scenario with the best perfor-mance, there is 5.5% of the glucose and 71.2% of the xyloseunfermented (with respect to the feed) in the effluent of thefermentor.

The SSCF fed-batch (SSCF-FB) (Fig. 2d) operation is rankedfourth, even though a certain fraction of unreacted raw materialis presented (0.13). This configuration presents the highest ethanolconcentration (5.8% wt/v) in the final amount of product.

Fed-batch operation for the conversion of cellulose to glucose andcontinuous operations with a recycle stream in the co-fermentationof glucose and xylose to ethanol (FB-C_RECY, Fig. 1c), is the option

ranking 5th among the tested configurations (0.13 kg ethanol/kgdry-biomass). Although SSCF-FB and FB-C_RECY give the sameethanol/dry-biomass yield, the final ethanol concentration in theFB-C_RECY configuration is slightly lower (5.6% wt/v).

The remaining process flowsheet configurations were found toperform poorly with yields below 0.11 kg ethanol/kg dry biomass.

Fed-batch operation inherently involves some transient accu-mulation of reaction products in the process, which requires adynamic model to enable proper analysis of the process perfor-mance (see Supplementary data (S.3)). Moreover the implicationof accumulation of reaction products in the process is that it re-duces the total amount of treated biomass during the operationperiod (see Table 4).

Although the final ethanol concentration obtained from theSSCF-FB (Fig. 2d) was slightly higher (5.8% wt/v) than that obtainedwith the SSCF-C_RECY (5.5% wt/v) (Fig. 2f), the ratio of ethanol todry biomass was markedly higher in the SSCF-C_RECY configura-tion (0.18 kg/kg), which makes SSCF-C_RECY a better alternative(Table 4). The ethanol concentration from FB-C_RECY (Fig. 1c)was also relatively high (5.6% wt/v) but the unreacted raw materialwas also high (0.14) making it less effective compared to the SSCF-C_RECY configuration. Something similar is observed in the finalethanol concentration (5.5% Wt/v) for the FB-FB (Fig. 1a) configura-tion, which also has slightly lower ethanol/dry-biomass ratio(0.10 kg/kg).

A mass balance summary for the stream that is sent to thedownstream processes for the diverse proposed process configura-tions is included in the Supplementary data (S.4).

3.3.3. Reduction of the number of unit operations for the FB–FBconfiguration

Specifically for fed-batch operation, an additional simulationstudy has been performed for both enzymatic hydrolysis and co-fermentation units with the aim of evaluating whether it wouldbe possible to reduce the amount of unit operations (from fivereactors in parallel to three reactors in parallel) without compro-

Table 4Summary of the simulation results for the proposed configurations: ethanol/dry-biomass ratio, unreacted raw material fraction, and ethanol concentration.

Ethanol/dry-biomass ratio(kg/kg)

Unreactedraw materialfraction

Ethanolconcentration% (wt/v)

FB-FB 0.10 0.25 5.5FB-C 0.10 0.14 4.7FB-C_RECY 0.13 0.14 5.6C-FB 0.09 0.11 4.4C-C 0.11 0.00 4.3C-C_RECY 0.11 0.00 4.4C_RECY-FB 0.10 0.37 4.9C_RECY-C 0.14 0.00 4.2C_RECY-C_RECY 0.16 0.00 4.9SSCF-FB 0.13 0.14 5.8SSCF-C 0.12 0.00 4.6SSCF-C_RECY 0.18 0.00 5.5

Table 5Results for the FB–FB configuration using five and three operation units.

Characteristic FB(5)–FB(5) FB(3)–FB(3)

Ethanol/dry-biomass ratio 0.10 0.11Accumulation fraction in the process 0.25 0.12Loading (h) 12 12Reaction period (h) 36 12Drawing (h) 12 12

1182 R. Morales-Rodriguez et al. / Bioresource Technology 102 (2011) 1174–1184

mising the ethanol production. This implicitly involves reductionof the reaction time to 12 h rather than 36 h.

A comparison between both operational scenarios is shown inTable 5. The results show lower values of unreacted raw materialin the process with three parallel reactors, indicating more conver-sion of the raw material in the operating time, which is directly re-lated to the accumulation fraction of raw material in bothconfigurations. Moreover the ethanol/dry-biomass ratio is basicallythe same for both configurations, meaning that the shorterreaction period is sufficient to obtain a high conversion of thelignocellulosic biomass to ethanol. This indicates that the five reac-tors based NREL design is too conservative. In other words, thesafety factor in the NREL design is too high, since using only threereactors would have been sufficient on the basis of the dynamicmodel simulations. However, it is also important to highlight thedifferences in the process configuration in the NREL report and thiscase study. For example, the NREL report has employed simulta-neous saccharification and fermentation (SSF) with an additionalsection to metabolize xylose into ethanol, while this case studyemploys separate enzymatic hydrolysis and co-fermentation. Apossible explanation for our result could be related to the slowerxylose fermentation or inhibition generated by ethanol or otherinhibitors present in the SSF unit operation in the NREL case.

3.3.4. Towards further process optimization: analysis of reactionvolume versus ethanol/dry-biomass ratio

An analysis of the ethanol/dry-biomass ratio with respect to thetotal reaction volume (directly related with the number of reactorunits) has been performed, in order to investigate whether it ispossible to decrease the number of unit operations maintainingan acceptably high value of dry-biomass conversion.

Based on the results shown in Table 4, the scenario with thehighest value of the ethanol/dry-biomass ratio (SSCF-C_RECY,Fig. 2f), was chosen to perform a sensitivity analysis of the etha-nol/dry-biomass ratio with respect to the total reaction volume.To this end, ten scenarios with different (smaller) reaction volumeswere generated and simulated to calculate the resulting ethanol

yield. The reaction volume of the base case configuration(3932,360 kg slurry) is divided into three reactor units which pro-vided an ethanol/dry-biomass ratio of 0.18. Decreasing the reactionvolume by 38% and then running the reaction in two reactors(2411,423 kg slurry) the ethanol/dry-biomass ratio decreases just2.3% compared with the SSCF-C_RECY scenario (Table 4). Usingon the other hand just one reaction unit, the results show that anethanol/dry-biomass ratio of 0.1544 is obtained for a reaction vol-ume of 1033,467 kg slurry. In summary the sensitivity analysis re-sults show that there is room for further process optimization: (i)either the required number of reactors can be decreased withoutsignificantly compromising the ethanol yield (2 versus 3), (ii) orthe given three reactor-based design can also be used to processmore feedstock (loading could be increased) without compromis-ing the overall ethanol yield. It needs to be mentioned explicitlyhowever that the optimization results are based on the assumptionof a certain enzyme loading (40 g enzyme/kg of substrate) andkinetics for the enzymatic hydrolysis part and the kinetics andyield for the co-fermentation part.

A similar sensitivity analysis was performed for other configu-rations shown in Fig. 4 (C-C, C-C_RECY, C_RECY-C and C_RECY-C_RECY), yielding comparable results as mentioned above. Theseresults show that when continuous-recycle (C_RECY) is employedfor the operation of hydrolysis and/or co-fermentation units, theproductivity is maintained despite the reduction in the reactionvolume (up to a certain point). Similarly the ethanol/dry-biomassratio is also not affected significantly by the reduction of the reac-tion volume in the continuous process configurations. These con-siderations demonstrate the positive impacts of continuousoperation and recycle options on improving the processperformance.

3.4. General discussion

This study integrated, argued and demonstrated the use of amatlab based modeling platform, DLB1.0, for comparative evalua-tion of the operation and synergy among fed-batch, continuousand continuous with recycle units. The DLB1.0 modeling platformalso allows the implementation and evaluation of plantwide con-trol scenarios for the optimized operation. Moreover it enablesstudying transient and dynamic response of the plants to differentdisturbances (e.g. varying feedstock composition or varying flowlevels).

One of the other contributions in this study is the introductionof one mathematical model for SSCF unit operation, which is builthaving a combination of the enzymatic hydrolysis and co-fermen-tation models. The proposed model takes into account the inhibi-tion of ethanol on enzymatic conversion of cellulose, whichrequires further experimental validation. The DLB1.0 simulationsidentified five configurations, four of them involving a recycle flowas being the most feasible for maximizing the ethanol yields on aset amount of lignocellulose raw material (Table 4). Recycling ofmaterial results in maximizing the enzyme reaction time and fer-mentation time within a certain processing period. In turn, this al-lows for full reaction of the raw material in the reactors and thusachieves higher yields for both the enzyme catalysis reaction andthe fermentation.

The advantages of technology evaluation based on the DLB1.0simulations can be seen from three different perspectives: (a) max-imize the yield of the raw material, (b) the optimized operationpoint of view: get more out of the existing plant capacity/equip-ment, and (c) the process design point of view: use a minimumcapacity/equipment to achieve better process performance (designtarget), which has a direct impact on the capital cost of the bioeth-anol plant.

Fig. 4. Sensitivity analysis of the ethanol/dry-biomass ratio with respect to the reaction volume for: (a) C-C, (b) C-C_RECY, (c) C_RECY-C, and (d) C_RECY-C_RECYconfigurations. H: enzymatic hydrolysis unit, F: co-fermentation unit. Note that the scaling of the Z-axis is different for the different figures.

R. Morales-Rodriguez et al. / Bioresource Technology 102 (2011) 1174–1184 1183

One assumption in this study is that there is no inhibition ofbyproducts (such as, furfural and hydroxy methyl furfural) result-ing from the pre-treatment operations to the enzymatic hydrolysisand co-fermentation units implying a detoxification unit beforethese units operation. This study also assumes no significantchanges on the metabolic capacity as a result of the possible accu-mulation of inhibitors where some recycles are found in the config-urations. These are deemed acceptable assumptions as recentmolecular and protein engineering research efforts have demon-strated promising results with respect to improving the inhibitiontolerance of enzymes used for the hydrolysis as well as the co-fer-mentation organisms used for fermenting sugars (Almeida andHahn-Härgerdal, 2009; Bettiga et al., 2009; Karhumaa et al., 2007).

Last but not least, dynamic modeling has been demonstrated tobe a promising tool in evaluating different process configurationsin view of supporting process design and operation activities. Thepresent work provides a basis that may be expanded to includedownstream processing energy/heat integration as well as solids/lignin combustion for power co-generation. In addition, cost anal-ysis and optimization of the process operation and design will alsobe investigated in order to maximize the yield of bioethanol andreduce the energy consumption in the process.

4. Conclusion

A number of scenarios have been proposed, analyzed and com-pared for finding the most feasible process technology for inte-grated operation of various lignocellulosic bioethanol process

configurations using a dynamic modeling framework. The resultsshowed that five of those configurations produced the highest eth-anol yields per amount of dry-biomass: SSCF-C_RECY, C_RECY-C_RECY, C_RECY-C and SSCF-FB (0.18, 0.16, 0.14 and 0.13 kg/kg,respectively four of them involving a recycle flow). Sensitivityanalysis of the reaction volume with respect to process yield forethanol has shown the possibility of reducing the number of equip-ments without compromising the bioethanol production yield.

Acknowledgements

The authors acknowledge the Mexican National Council forScience and Technology (CONACyT, Project # 118903) and theDanish Research Council for Technology and Production Sciences(Project # 274-07-0339) for the financial support on the develop-ment of this project.

Appendix A. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.biortech.2010.09.045.

References

Aden, A., Ruth, M., Ibsen, K., Jechura, J., Neeves, K., Sheehan, J., Wallace, B.,Montague, L., Slayton, A., Lukas, J., 2002. Lignocellulosic Biomass to EthanolProcess Design and Economics Utilizing Co-Current Dilute Acid Prehydrolysisand Enzymatic Hydrolysis for Corn Stover. National Renewable EnergyLaboratory Technical Report. NREL/TP-510-32438.

1184 R. Morales-Rodriguez et al. / Bioresource Technology 102 (2011) 1174–1184

Almeida, J.R.M., Hahn-Härgerdal, B., 2009. Developing Saccharomyces cerevisiaestrains for second generation bioethanol: improving xylose fermentation andinhibitor tolerance. Int. Sugar J. 111, 172–180.

Alvarado-Morales, M., Terra, J., Gernaey, K.V., Woodley, J.M., Gani, R., 2009.Biorefining: Computer aided tools for sustainable design and analysis ofbioethanol production. Chem. Eng. Res. Des. 87, 1171–1183.

Bettiga, M., Bengtsson, O., Hahn-Härgerdal, B., Gorwa-Grauslund, M.F., 2009.Arabinose and xylose fermentation by recombinant Saccharomyces cerevisiaeexpressing a fungal pentose utilization pathway. Microb. Cell Fact. 8, 40.doi:10.1186/1475-2859-8-40 .

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

Cardona, C.A., Sánchez, O.J., 2007. Fuel ethanol production: process design trendsand integration opportunities. Bioresour. Technol. 98, 2457–2471.

Dutta, A., Dowe, N., Ibsen, K.N., Schell, D.J., Aden, A., 2010. An economic comparisonof different fermentation configurations to convert corn stover to ethanol usingZ. mobilis and Saccharomyces. Biotechnol. Prog. 26, 64–72.

Gírio, F.M., Fonseca, C., Carvalheiro, F., Duarte, L.C., Marques, S., Bogel-Lukasik, R.,2010. Hemicelluloses for fuel ethanol: a review. Bioresour. Technol. 101, 4775–4800.

Gnansounou, E., 2010. Production and use of lignocellulosic bioethanol in europe:current situation and perspectives. Bioresour. Technol. 101, 4842–4850.

Hodge, D.B., Karim, M.N., Schell, D.J., McMillan, J.D., 2009. Model-based fed-batchfor high-solids enzymatic cellulose hydrolysis. Appl. Biochem. Biotechnol. 152,88–107.

Huber, G.W., Iborra, S., Corma, A., 2006. Synthesis of transportation fuel frombiomass: chemistry, catalyst, and engineering. Chem. Rev. 106, 4044–4098.

Kadam, K.L., Rydholm, E.C., McMillan, J.D., 2004. Development and validation of akinetic model for enzymatic saccharification of lignocellulosic biomass.Biotechnol. Prog. 20, 698–705.

Karhumaa, K., Sanchez, R.G., Hahn-Hägerdal, H., Gorwa-Grauslund, M.F., 2007.Comparison of the xylose reductase-xylitol dehydrogenase and the xyloseisomerase pathways for xylose fermentation by recombinant Saccharomycescerevisiae. Microb. Cell Fact. 6, 5. doi:10.1186/1475-2859-6-5.

Krishnan, M.S., Ho, N.W.Y., Tsao, G.T., 1999. Fermentation kinetics of ethanolproduction from glucose and xylose by recombinant Saccharomyces1400(pLNH33). Appl. Biochem. Biotechnol. 77–79, 373–388.

Larsen, J., Petersen, M.Ø., Thirup, L., Li, H.W., Iversen, F.K., 2008. The IBUS process-lignocellosic bioethanol close to a commercial reality. Chem. Eng. Technol. 31,765–772.

Lavarack, B.P., Griffin, G.J., Rodman, D., 2002. The acid hydrolysis of sugarcanebaggase hemicellulose to produce, xylose, arabinose, glucose and otherproducts. Biomass Bioenergy 23, 367–380.

Lynd, L.E., Laser, M.S., Bransby, D., Dale, B.E., Davison, B., Hamilton, R., Himmel, M.,Keller, M., McMillan, J.D., Sheehan, J., Wyman, C.E., 2008. How biotech cantransform biofuels. Nat. Technol. 26, 169–172.

Margeot, A., Hahn-Hagerdal, B., Edlund, M., Slade, R., Monot, F., 2009. Newimprovements for lignocellulosic ethanol. Curr. Opin. Biotechnol. 20, 371–380.

Morales-Rodriguez, R., Capron, M., Huusom, J.K., Sin, G., 2010. Controlled fed-batchoperation for improving cellulose hydrolysis in 2G bioethanol production.Comput.-Aided Chem. Eng. 28, 1497–1502.

Morales-Rodriguez, R., Gernaey, K.V., Meyer, A.S., Sin, G., in press. Development of amathematical model describing hydrolysis and co-fermentation of C6 and C5sugars. Chin. J. Chem. Eng.

Ooshima, H., Ishitani, Y., Harano, Y., 1985. Simultaneous saccharification andfermentation of cellulose: effect of ethanol on enzymatic saccharification ofcellulose. Biotechnol. Bioeng. 27, 389–397.

Philippidis, G.P., Smith, T.K., Wyman, C.E., 1993. Study of the enzymatic hydrolysisof cellulose for production of fuel ethanol by the simultaneous saccharificationand fermentation process. Biotechnol. Bioeng. 41, 846–853.

Regalbuto, J.R., 2009. Cellulosic biofuel-got gasoline? Science 325, 822–824.Shell, D.J., Riley, C.J., Dowe, N., Farmer, J., Ibsen, K.N., Ruth, M.F., Toon, S.T., Lumpkin,

R.E., 2004. A bioethanol process development unit: initial operating experiencesand results with a corn fiber feedstock. Bioresour. Technol. 91, 179–188.

Sin, G., Meyer, A.S., Gernaey, K.V., 2010. Assessing reliability of cellulose hydrolysismodels to support biofuel process design – identifiability and uncertaintyanalysis. Comput. Chem. Eng. 34, 1385–1392.