Bioresource Technology 87 (2003) 113–124

Mathematical modeling of non-ideal mixing continuous flow

reactors for anaerobic digestion of cattle manure
a,b
A. Keshtkar , B. Meyssami b,*, G. Abolhamd b, H. Ghaforian a,
M. Khalagi Asadi a
a
Center of Renewable Energies for Research and Application, Atomic Energy Organization of Iran, Tehran, Iran
b
Faculty of Engineering, Department of Chemical Engineering, Tehran University, P.O. Box 11365-4563, Tehran, Iran
Received 6 August 2001; received in revised form 14 January 2002; accepted 23 April 2002

Abstract
Most conventional digesters used for animal wastewater treatment include continuously stirred-tank reactors. While imperfect
mixing patterns are more common than ideal ones in real reactors, anaerobic digestion models often assume complete mixing
conditions. Therefore, their applicability appears to be limited. In this study, a mathematical model for anaerobic digestion of cattle
manure was developed to describe the dynamic behavior of non-ideal mixing continuous flow reactors. The microbial kinetic model
includes an enzymatic hydrolysis step and four microbial growth steps, together with the effects of substrate inhibition, pH and
thermodynamic considerations. The biokinetic expressions were linked to a simple two-region liquid mixing model, which con-
sidered the reactor volume in two separate sections, the flow-through and the retention regions. Deviations from an ideal completely
mixed regime were represented by changing the relative volume of the flow-through region (a) and the ratio of the internal exchange
flow rate to the feed flow rate (b). The effects of the hydraulic retention time, the composition of feed, the initial conditions of the
reactor and the degree of mixing on process performance can be evaluated by the dynamic model. The simulation results under
different conditions showed that deviations from the ideal mixing regime decreased the methane yield and resulted in a reduced
performance of the anaerobic reactors. The evaluation of the impact of the characteristic mixing parameters (a) and (b) on the
anaerobic digestion of cattle manure showed that both liquid mixing parameters had significant effects on reactor performance.

2002 Published by Elsevier Science Ltd.

Keywords: Anaerobic digestion; Mathematical modeling; Continuous-flow reactor; Imperfect mixing; Cattle manure

1. Introduction
Anaerobic digestion is a biodegradation process,
which uses a consortium of natural bacteria to convert a
large portion of the organic solids in the wastewater into
biogas. The biogas is mainly a mixture of methane and
carbon dioxide, and if captured, is a gas fuel used for
heat and/or power generation. Most conventional di-
gesters used for animal wastewater treatment are either
continuously stirred-tank reactors or plug-flow reactors.
Most of the previous research on animal wastewater
treatment was constructed on these two types of the
anaerobic digesters (Varel et al., 1977; Hills and Me-
hlschan, 1984; Angelidaki and Ahring, 1993; Hansen
et al., 1999). In these two types of digesters, the HRT
*
Corresponding author. Fax: +98-21-6498982.
E-mail address: bmeysam@ut.ac.ir (B. Meyssami).
equals the SRT and the active biomass is removed from
the digester in the effluent on a daily basis. The HRT
needs to be long enough to ensure a sufficient SRT in the
digester so that a viable bacterial population necessary
to complete the anaerobic digestion process is main-
tained in the reactor. Typical HRTs of conventional
mesophilic (35 C) digesters for treating animal wastes
are usually controlled at 10–20 days, depending on the
solids content of the wastes. The long retention time
required for animal manure digestion may be attributed
not only to the presence of complex organic compounds,
but also to high concentrations of the ammonia nitro-
gen that affect the anaerobic decomposition process
(Zeeman et al., 1985).
The HRT is one of the most important design para-
meters affecting the economics of digesters. For a given
volume of wastewater, a shorter HRT translates into a
smaller digester and therefore more favorable econom-
ics. Digester developers have taken various approaches

0960-8524/03/$ - see front matter
2002 Published by Elsevier Science Ltd.

PII: S 0 9 6 0 - 8 5 2 4 ( 0 2 ) 0 0 1 0 4 - 9
114 A. Keshtkar et al. / Bioresource Technology 87 (2003) 113–124

Nomenclature

a mixing parameter R gas constant, atm l/mol K

b mixing parameter t time, d
C liquid concentration, g/l T temperature, K
[CO2 ] free CO2 in liquid concentration, mol/l Vg gas volume of reactor, l
f individual bacterial fraction in initial total Vl liquid volume of reactor, l
biomass VFA volatile fatty acids
fpr mass conversion factor of propionate to ac- X microorganisms concentration, g/l
etate ¼ 0.8108 ye yield factor used in Eq. (10)
fbut mass conversion factor of butyrate to acetate a flow-through region
¼ 0.6818 b retention region
Ft biogas transfer rate, mol/d h HRT, d
F(pH) pH function l specific growth rate, d1
H Henry’s constant, atm l/mol lmax maximum specific growth rate, d1
HRT hydraulic retention time
SRT sludge retention time Subscripts
k hydrolysis rate constant, d1 ac acetate
K0 non-inhibited hydrolysis rate constant, d1 am ammonia
Ka dissociation constant A acidogenic bacteria
kd bacterial decay rate constant, d1 AB butyric degrading acetogenic bacteria
Ki inhibition constant, g/l AP propionate degrading acetogenic bacteria
Ks Monod saturation constant, g/l but butyrate
m feed constant used in Eq. (1) c carbon dioxide
n feed constant used in Eq. (1) e exchange between zones
N gas transfer rate, g/d f feed
[NH3 ] free NH3 in liquid concentration, mol/l i component i
P pressure, atm is insoluble substrate
pKh constant used in Eq. (16) m methane
pKl constant used in Eq. (16) M methanogenic bacteria
Q volumetric flow rate, d1 pr propionate
rd bacterial decay rate, g/l d s soluble substrate
rh hydrolysis reaction rate, g/l d t total
rs substrate consumption rate, g/l d w water
rx bacterial growth rate, g/l d

in the past to reduce HRT, e.g., efficient mixing of the
reactor. Good mixing promotes the efficient transfer of
substrates and heat to the microorganisms, maintains
uniformity in other environmental factors and assures
effective use of the entire reactor volume by preventing
stratification and formation of dead spots and prevents
pockets of the VFA from forming. Scum formation can
also be greatly reduced or even eliminated by suitable
agitation. It is recognized that inhomogeneities in the
medium can have a profound influence, especially on
production of metabolites (Nielsen and Villadesen, 1992).
While imperfect mixing patterns are more common
than ideal ones in real reactors, anaerobic digestion
models often assume complete mixing conditions.
Therefore, their applicability appears to be limited.
However, by using an appropriate configuration that
corresponds to the hydraulic characteristics of the real
reactor flow pattern, it is possible to simulate non-ideal
reactor performance. For example, by using the proper
liquid mixing models based on the hydrodynamic con-
figuration of the reactor studied, Reinhold et al. (1996)
calculated and predicted the mixing behavior in a biogas
tower reactor to be scaled-up. One of the classical
mixing models is the two-region model which, despite its
simplicity, is used in chemical engineering for the de-
scription of retention time distribution in real reactors
(Levenspiel, 1972). It has proved to be a useful tool for
the theoretical study of the effects of inhomogeneity in
chemical and biological systems. For example, Bello-
Mendoza and Sharratt (1998) used this mixing model
for the effect of imperfect mixing on performance of
anaerobic sewage sludge digestion.
The objective of this study was to develop a mathe-
matical model, which combines the two-region mixing
model with a proper structured kinetic model, for the
simulation of anaerobic cattle manure digestion in non-
 A. Keshtkar et al. / Bioresource Technology 87 (2003) 113–124 115

ideal continuous flow reactors. By computer simulation, In Reaction (1), ye is the enzymatic efficiency or yield
the effect of mixing can be predicted on the performance factor and the subscript in represents the non-biode-
of the process, and the relation between mixing and the gradable inert organic material. The coefficients ye , n,
kinetic parameters can also be described. and m, together with the ratio of the soluble to the in-
soluble substrate depend on the type of manure. The
processes of hydrolysis and biomass decay are described
2. Mathematical model
by the first order reactions shown below:
2.1. Kinetic model rh ¼ kCis ð6Þ
rd ¼ k d X ð7Þ
The stoichiometry of anaerobic digestion of cattle
manure has been described by Hill (1982) and developed where k is the hydrolysis rate constant and kd is the
by Angelidaki et al. (1993). The kinetic model distin- decay rate constant. The hydrolytic reaction rate is as-
guishes five different processes: hydrolysis of particulate sumed to be inhibited by the presence of VFA. The in-
substrate by extracellular enzymes, consumption of hibition function chosen is non-competitive:
soluble substrates by acidogenic bacteria, consumption ki;VFA
k ¼ k0 P ð8Þ
of VFA and formation of acetate by propionate and VFA þ ki;VFA
butyrate degrading acetogenic bacteria, and finally X
consumption of acetate and generation of methane by VFA ¼ Cac þ fpr Cpr þ fbut Cbut ð9Þ
methanogenic bacteria. The model includes VFA inhi- Consumption of soluble substrates and volatile acids as
bition of the hydrolysis step, acetate inhibition of well as growth of anaerobic microorganisms, are as-
the acetogenic steps, free ammonia inhibition of the sumed to obey Monod-type kinetics, as follows:
methanogenic step, and pH inhibition of all biological
steps. In the model, the primary substrates in the ma- rs ¼ Ys=x lX ð10Þ
nure are represented as soluble (s) and insoluble (is) rx ¼ lX ð11Þ
carbohydrate units, with the basic formula (C6 H10 O5 )s
All the yield coefficients (Ys=x ), expressed as gram per
and (C6 H10 O5  nNH3 )is respectively. Cell mass is repre-
gram of bacteria synthesized, can be directly calculated
sented by the empirical formula. Model statements are
from the stoichiometric relationships. Specific growth
as follows:
rates (l) with non-competitive inhibition effects and pH
ðC6 H10 O5  nNH3 Þis ! ye ðC6 H10 O5 Þs modulation for the biological steps can be written as:
þ ð1  ye ÞðC6 H10 O5  mNH3 Þis Cs
lA ¼ lmax A ð12Þ
þ ðn  ð1  ye ÞmÞNH3 ð1Þ Kss þ Cs
Cpr Ki pr
ðC6 H10 O5 Þs þ 0:1115 NH3 ! 0:1115 C5 H7 NO2 lAP ¼ lmax AP FAP ðpHÞ ð13Þ
Kspr þ Cpr Ki pr þ Cac
þ 0:744 CH3 COOH
Cbut Ki but
þ 0:5 CH3 CH2 COOH lAB ¼ lmax AB FAB ðpHÞ ð14Þ
Ks but þ Cbut Ki but þ Cac
þ 0:4409 CH3 ðCH2 Þ2 COOH Cac Ki am
lM ¼ lmax M FM ðpHÞ ð15Þ
þ 0:6909 CO2 þ 0:0254 H2 O Ksac þ Cac Ki am þ Cam
ð2Þ where F ðpHÞ is the pH modulation function and is de-
scribed by a Michaelis pH function normalized to give a
CH3 CH2 COOH þ 0:06198 NH3 þ 0:314 H2 O value of 1.0 as the center value (Angelidaki et al., 1993):
! 0:06198 C5 H7 NO2 þ 0:9345 CH3 COOH 1 þ 2  100:5ðpKl pKh Þ
F ðpHÞ ¼ ð16Þ
þ 0:6604 CH4 þ 0:1607 CO2 ð3Þ 1 þ 10ðpHpKh Þ þ 10ðpKl pHÞ
CH3 ðCH2 Þ2 COOH þ 0:0653 NH3 þ 0:5543 CO2 The parameters pKl and pKh , denote the lower and the
upper pH drop off values, respectively, where growth
þ 0:5543 H2 O ! 0:0653 C5 H7 NO2
rates are approximately 50% of the uninhibited rate. In
þ 1:8909 CH3 COOH þ 0:4452 CH4 ð4Þ general, the coefficients pKl and pKh are different for
various microbial groups.
CH3 COOH þ 0:022 NH3 ! 0:022 C5 H7 NO2
þ 0:945 CH4 2.2. Two-region liquid mixing model
þ 0:945 CO2 In the two-region mixing model, it is assumed that the
þ 0:066 H2 O ð5Þ reactor volume is split into two sections: the flow-through
116 A. Keshtkar et al. / Bioresource Technology 87 (2003) 113–124
Fig. 1. Two-region mixing model.
(a) region and the retention (b) region. Both regions are
assumed to be perfectly mixed but the transfer of mate-
rials between the zones is limited. The retention region
has the features of the behavior shown by a stag-
nant zone. Different levels of mixing are accomplished by
adjusting the relative volume of the flow-through region
(a) and the ratio of the exchange flow rate between re-
gions to the feed flow rate (b). A conceptual represen-
tation of the two-region mixing model is illustrated in
Fig. 1.
2.3. Model development
The generic model consists of a set of ordinary dif-
ferential equations, which represent mass balances using
different variables (see Appendix A). Differential vari-
ables include total concentrations of substrates, different
intermediate products and bacterial groups. The model
is based on the following assumptions and consider-
ations:
1. The gas phase and the two liquid phases, a and b,
were each assumed to be uniform.
2. All reactions are effectively rate controlled, i.e., the
effects of diffusional limitations in the biomass aggre-
gates are constant and incorporated into the kinetic
terms.
3. The non-competitive type inhibition was considered
in all microbial steps as described in the previous sec-
tion.
4. First order reaction rates were applied for the bacte-
rial decay and the enzymatic hydrolytic steps.
5. Decay rate constants of the different bacterial groups
were assumed to be 5% of their maximum growth
rate (Angelidaki et al., 1993).
6. Mass transfer to the gas phase only occurs in liquid
phase a.
7. The influent and effluent streams are located in the
flow-through region.
8. The b liquid phase exchanges materials only with the
a liquid phase.
9. The system pressure and retention volume are con-
stant.
10. Energetic effects are not considered with temperature
perfectly controlled.
11. At the operational temperature and pressure, biogas
is considered to be an ideal gas.
12. The biogas consists of methane, CO2 and water.
13. The water vapor in the biogas stream is at the satu-
ration state.
14. The CO2 present in the a liquid phase is at thermody-
namic equilibrium with the CO2 in the gas phase and
it obeys Henry’s law.
15. The concentration of methane in the a liquid phase
is assumed to be negligible, i.e., it is immediately
transferred to the gas phase because of its low solu-
bility.
16. In the ionic charge balance (Eq. (A.32)), the algebraic
sum of the concentrations of other ionic compounds
in the process, [A Cþ ], are assumed to be constant
during the anaerobic digestion process and are calcu-
lated from the initial pH of the system for both a and
b liquid phases.
With these assumptions and considerations the resulting
material balances the gas phase and a and b liquid
phases and the ionic charge balance equations for the
two liquid phases are summarized and presented in
Appendix A.
The ionic charge balance equations should be itera-
tively solved for the pH calculation since the concen-
trations of the ionic compounds, in turn, are functions
of the pH according to Eqs. (A.25)–(A.31). Of course,
we need to use an additional iterative procedure for the
pH calculation of the a liquid phase, because according
to Eq. (A.24), the total CO2 in the a liquid phase is a
function of the pH of this phase and the partial pressure
of the CO2 in the gas phase. Therefore, the following
steps should be done for the calculation of pH and
different component concentrations of the a liquid
phase:
• Guess a pH for the time step of t (the pH of the pre-
vious time step is a good initial guess for the pH of
the next time step).
• Solve the set of ordinary differential equations for the
a liquid phase together with Eq. (A.24).
• Solve the ionic charge balance (Eq. (A.32)) by a
trial and error procedure for the calculation of
pH.
• Compare the calculated pH with the guessed pH.
• If jpHcalc  pHguess j > 103 , consider the calculated
pH as the new guess for pH and go to step 2 to repeat
the above steps. Otherwise, save the final results for
time step of t.
 A. Keshtkar et al. / Bioresource Technology 87 (2003) 113–124 117

2.4. Computer simulations Table 2

Physico-chemical parameters at 35 C (Dean, 1992)
For the dynamic prediction of anaerobic digestion of Parameter Value
cattle manure in non-ideal continuous flow reactors at Kw (M) 2:065  1014
different operating conditions and also for evaluating Ka1 (M) 4:909  107
Ka2 (M) 5:623  1011
effects of the characteristic mixing parameters (a and b)
Ka3 (M) 1:730  105
on reactor performance, the model described in this Ka4 (M) 1:445  105
paper was programmed in a generalized form in For- Ka5 (M) 1:445  105
tran, where a variable number of steps, feed composi- Ka6 (M) 1:567  109
tion, initial conditions of the a and b liquid and gas Hc (atm l/mol) 37.67a
a
phases, and the operating conditions could be specified Archer (1983).
through an input file. The simulations were performed
by numeric first order integration of the relevant equa-
to be 49% and 51%, respectively, whereas the ratio of the
tions with a fixed time step by a computer program
flow rate of the internal exchange to the feed flow rate
based on Euler’s method. The program created an out-
(b) was found to be equal to 0.333.
put data file in a format suitable for graphic processing.
Kinetic model parameters were taken directly from
the literature and are given in Table 1. Physico-chemical
model parameters at 35 C are given in Table 2. Values 3. Results and discussion
of the mixing parameters were selected on the basis of
information found in the literature. Tracer studies con- The manure composition used in the model simula-
ducted in full-scale anaerobic digesters have revealed tions is given in Table 3 and was based on the cattle
well-mixed portions of digester volumes ranging widely manure used in the experiments of Angelidaki and
from 23% to 88%, while the balances have been in Ahring (1993). The simulations were performed with the
stagnant volumes (Montieth and Stephenson, 1981). following conditions: 35 C, 15 days HRT, atmospheric
There is less evidence regarding the average interchange gas pressure, and a gas to liquid volume ratio of 0.1. On
rates between mixed and stagnant volumes in anaerobic the basis of the two-region mixing model described
digesters. Smith et al. (1993) applied the same tracer above, the liquid composition of the stream flowing out
technique used by Montieth and Stephenson (1981) to of the reactor has the same composition as the flow-
evaluate liquid mixing characteristics of a pilot scale through zone content. By definition, for a relative vol-
contact process anaerobic digester. In one of these ume in the flow-through region (a) close to unity and,
studies, they found the mixed and the stagnant volumes for any value of ‘a’, with an interchange rate of the
material between regions to feed flow rate ratio (b) ap-
proaching infinity, the dynamic model produces results
Table 1 closely approaching those of a completely mixed reac-
Kinetic parameters used in the model (Angelidaki et al., 1993) tor. Otherwise, for any ‘a’ with ‘b’ close to zero (i.e. no
Parameter Value interchange of material between regions) the system
consists of a reactor with a completely dead zone of
Kss (g/l) 0.5
Ks pr (g/l) 0.259 volume (1  a)Vl . For values of the mixing parameters
Ks but (g/l) 0.176 other than those mentioned above, the mathematical
Ksac (g/l) 0.12
Ki VFA (g/l) 0.33
Ki pr (g/l) 0.96 Table 3
Ki but (g/l) 0.72 Characteristics of the feed
Ki am (g/l) 0.26
K0 (d1 ) 1.0 Characteristic Value
lmax A (d1 ) 5.0 Insoluble substrate 30.4 (g/l)
lmax AP (d1 ) 0.54 Soluble substrate 5.4 (g/l)
lmax AB (d1 ) 0.68 Total acetate 4.5 (g/l)
lmax M (d1 ) 0.6 Total propionate 2.3 (g/l)
ye 0.55 Total butyrate 0.2 (g/l)
n 0.454 Total ammonia 3.0357 (g NH3 /l)
m 0.34 Total carbon dioxide 0.0 (g/l)
pKh AP 8.5 Total microbial biomass 0.2 (g/l)
pKl AP 6.0 Fraction of acidogens 0.65
pKh AB 8.5 Fraction of propionate acetogens 0.025
pKl AB 6.0 Fraction of butyrate acetogens 0.025
pKh M 8.5 Fraction of methanogens 0.3
pK1 M 6.0 pH 8.0
118 A. Keshtkar et al. / Bioresource Technology 87 (2003) 113–124

model simulates the performance of an imperfectly
mixed digester.
Simulations of the anaerobic digestion process ap-
plied to cattle manure with different degrees of mixing
consisting of ða; bÞ equal to ð0:9; 10Þ, ð0:6; 0:5Þ and
ð0:2; 0:2Þ are shown in Figs. 2–5. The three degrees of
mixing considered were chosen to simulate reactor be-
havior approaching a completely mixed reactor, an im-
perfectly mixed reactor and an incompletely mixed
reactor, respectively. The dynamic results of the insol-
uble substrate, total acetate, total propionate and total
ammonia concentrations are illustrated in these figures
for both regions, respectively. Significant differences
between the concentration patterns shown by these
systems arose due to the different degrees of mixing
considered. As can be seen from the figures, the medium
concentrations in both zones for the well-mixed reactor
are the same throughout the duration of the simulation,
because the materials quickly distribute from the flow-
through region to the retention region due to the good
degree of mixing applied. In contrast, for the poorly
mixed groups the dynamic simulation results of the
medium concentrations show non-homogeneous distri-
butions of components in the reactor and less volume
available for the active digestion due to the limited
interchange between zones. The resulting homoge-
neous and non-homogeneous medium concentrations
throughout the volume of the reactor due to the high
and low interchange rates used show the ability of the
two-region model to simulate anaerobic reactors with
either ideal or non-ideal mixing. It also shows the effect
of mixing parameters on the residence time distribu-
tion pattern as well as the distribution of components
in the reactor. Therefore, mixing influences rates in the
anaerobic digestion process. This influence on reac-
tion rates is a result of the number of non-linear rate

Fig. 2. Dynamic simulation of anaerobic digestion of cattle manure in a continuous flow reactor under HRT ¼ 15 days and different degrees of
mixing for prediction of the insoluble substrate concentration in flow-through and retention regions.

Fig. 3. Dynamic simulation of anaerobic digestion of cattle manure in a continuous flow reactor under HRT ¼ 15 days and different degrees of
mixing for prediction of the total acetate concentration in the flow-through and the retention regions.
 A. Keshtkar et al. / Bioresource Technology 87 (2003) 113–124 119

Fig. 4. Dynamic simulation of the anaerobic digestion of cattle manure in a continuous flow reactor under HRT ¼ 15 days and different degrees of
mixing for prediction of total propionate concentration in the flow-through and the retention regions.

Fig. 5. Dynamic simulation of the anaerobic digestion of cattle manure in a continuous flow reactor under HRT ¼ 15 days and different degrees of
mixing for prediction of total ammonia concentration in the flow-through and the retention regions.

expressions and the substrate-dependent Monod rela-
tionship by which anaerobic digestion is represented.
Dynamic simulation of the methane yield and the pH
of the liquid stream outflow of the reactor are shown in
Figs. 6 and 7 for the three different degrees of mixing
defined above. As the model results show, the methane
yield of the process depends on the variations of the pH
shown in Figs. 6 and 7 for the start-up period. As shown
in Fig. 6, the methane yield decrease with the degree of
mixing where this rate of decrease depends on how far
the conditions are from ideality. The methane yield in
the reactor with (a ¼ 0:6 and b ¼ 0:5) and in the reactor
with (a ¼ 0:2 and b ¼ 0:2) were respectively 3.2% and
85% lower than that of the reactor with (a ¼ 0:9 and
b ¼ 10). Conversely as can be seen from Figs. 2–4, the
concentrations of the insoluble substrate, acetate and
propionate in the effluent of the reactor with (a ¼ 0:6
and b ¼ 0:5) and the reactor with (a ¼ 0:2 and b ¼ 0:2)
were respectively (1.7; 3.3 times), (1.6; 137.6 times) and
(1.7; 53.4 times) higher than those of the reactor with
(a ¼ 0:9 and b ¼ 10). According to these results, it seems
likely that there is a threshold level of deviation from
ideal mixing where reactor performance declines sub-
stantially, however, ideal mixing may not be required to
have nearly ideal performance with regard to methane
yield and VFA concentrations in the effluent.
The effect of HRT on the methane yield under dif-
ferent mixing conditions was also evaluated. The steady-
state results are shown in Fig. 8. Methane yield showed
an increase with retention time and degree of mixing.
For HRT ¼ 15 days the methane yield for the poorer
mixed reactor (a ¼ 0:4 and b ¼ 0:5) was 8% lower than
that attained by the better mixed reactor (a ¼ 0:4 and
b ¼ 2). As shown in Fig. 8, extending the retention time
could improve the methane yield of imperfectly mixed
reactors. This is valid for reactors where the hydraulic
120 A. Keshtkar et al. / Bioresource Technology 87 (2003) 113–124
Fig. 6. Dynamic simulation of anaerobic digestion of cattle manure in
a continuous flow reactor under HRT ¼ 15 days and different degrees
of mixing for prediction of methane yield.
Fig. 7. Dynamic simulation of anaerobic digestion of cattle manure in
a continuous flow reactor under HRT ¼ 15 days and different degrees
of mixing for pH prediction of outlet stream from reactor.
characteristic time is greater than the mixing charac-
teristic time.
The methane yield for anaerobic digestion of cattle
manure as a function of the relative volume of the flow-
through region (a) is shown in Fig. 9 for different values
of the mixing parameter b. As expected, the steady-state
methane yield of the reactor increased with mixing
parameter a. It is obvious from the physical consider-
ations that increasing the volume of the flow-through
Fig. 8. Effect of HRT on the methane yield of anaerobic digestion of
cattle manure.
Fig. 9. Effect of the relative volume of the flow-through region (a) on
the methane yield of anaerobic digestion of cattle manure.
region results in more volume available for immedi-
ate anaerobic digestion activity and, therefore, higher
methane yield. It was also observed that decreasing the
mixing parameter a increased the effect of the mixing
parameter b on methane yield. On the other hand, the
model also showed that by increasing the mixing pa-
rameter a, the effect of mixing parameter b on methane
yield would decrease so that all the curves converge to a
fixed value. The steady-state methane yield as a function
of mixing parameter b is shown in Fig. 10 for different
values of the relative volume of the flow-through region.
As expected, with decreasing b values, the steady-state
 A. Keshtkar et al. / Bioresource Technology 87 (2003) 113–124 121

Fig. 10. Effect of ratio of the internal exchange flow rate to the feed
flow rate (b) on the methane yield of anaerobic digestion of cattle Fig. 11. Model prediction versus experimental data (Dugba and
manure. Zhang, 1999) of methane yield––organic loading rate for selecting the
most appropriate set of mixing parameters.

methane yield of the reactor would decrease. How-

ever, in the range of values evaluated, mixing parame-
ter b showed a lower impact on the performance of the
steady-state anaerobic digestion than mixing para-
meter a.

3.1. Potential applicability of the two-region model of

liquid mixing

The validity of the kinetic model described was

evaluated by Angelidaki et al. (1993) for thermophilic
conditions and Keshtkar et al. (2001) for mesophilic
conditions. Although, the simulation results obtained
are qualitatively in agreement with what could be ex-
pected theoretically and with the published experimental
observations (Perot et al., 1988; Lin and Pearce, 1991),
the proposed dynamic model requires experimental
verification in order to assess its applicability. In Figs.
11 and 12, model simulation results were compared to Fig. 12. Comparison between experimental data (Dugba and Zhang,
experimental data given by Dugba and Zhang (1999). 1999) and prediction of methane yield as a function of organic loading
They conducted a series of experiments for anaerobic rate at a temperature of 35 C.
digestion of dairy manure in a two-stage anaerobic se-
quencing batch reactor system. The volume and the
Table 4
operating parameters of their experiments are shown in Operating parameters of the reactor
Table 4.
Operational parameters Values
To evaluate the applicability of the model, prelimi-
Total volume 15 l
nary simulations were compared to sequencing batch
Temperature 35 C
experimental runs measuring methane yield at various pH Controlled at 6.7–7.3
organic loading rate for an HRT of 3 days to determine Mixing of reactor 1 min every hour
the most appropriate set of mixing model parameters. In VS loading rate for HRT ¼ 3 days 2, 3, 4, 6, 8 g VS/l/day
Fig. 11, the best fit curve for the experimental data is VS loading rate for HRT ¼ 6 days 2, 3, 4 g VS/l/day
shown. The estimated a and HRT=b mixing parameters
of the reactor are equal to 0.3 and 4.0, respectively. then predicted for different organic loading rates using
Steady-state methane yields for an HRT of 6 days were the mixing parameters estimated. Predicted values are
122 A. Keshtkar et al. / Bioresource Technology 87 (2003) 113–124

compared with experimental data in Fig. 12. As seen, Appendix A. Material balances
since the appropriate adjustments to this model were
made for a sequencing batch reactor, good agreement A.1. Liquid phase
was obtained between the predicted values and the ex-
perimental data. Microbial biomass, Xi , i ¼ A, AP, AB, M
For future application of the model, methods to dXia Xi;f  Xia Xib  Xia
measure the characteristic mixing parameters are nee- ¼ þ þ ðlai  bi ÞXia ðA:1Þ
dt ah ah=b
ded. For modeling the liquid mixing behavior of a bio-
gas tower reactor Reinhold et al. (1996) applied a simple dXib X a  Xib
¼ i þ ðlbi  bi ÞXib ðA:2Þ
mixing model with characteristic mixing parameters dt ð1  aÞh=b
similar to those required by the two-region mixing
Insoluble substrate, Cis
model. The mixing parameters were calculated from
experimental tracer-response curves. This suggests the dCisa Cis;f  Cisa Cisb  Cisa
¼ þ  k a Cisa ðA:3Þ
possibility of calculating the parameters a and b using a dt ah ah=b
similar approach. Simple liquid mixing models, such as
dCisb C a  Cisba
the two-region model described here, appear to be useful ¼ is  k b Cisb ðA:4Þ
for the simulation of anaerobic reactors under non-ideal dt ð1  aÞh=b
mixing conditions. Imperfect mixing models may also be Soluble substrate, Cs
useful tools for reactor scale-up.
dCsa Cs;f  Csa Csb  Csa 162ye
¼ þ þ k a Cisa
dt ah ah=b 162 þ 17n
 12:858laA XAa ðA:5Þ
4. Conclusions
dCsb Csa Csb 162ye
¼ þ k b Cisb
A kinetic model describing the effects of substrate dt ð1  aÞh=b 162 þ 17n
inhibition, pH and thermodynamic considerations for
 12:858lbA XAb ðA:6Þ
anaerobic digestion of cattle manure was applied with a
two-region liquid mixing model to evaluate performance Total acetate, Cac
of non-ideal continuous flow reactors. The resulting a
dCac a
Cac;f  Cac C b  Cac
a
mathematical model could be used for simulation of ¼ þ ac þ 3:54laA XAa
dt ah ah=b
reactors with different degrees of mixing. Simulation
results showed that deviations from ideal mixing regime þ 8:006laAP XAP
a
þ 15:366laAB XAB
a

result in decreased performance of anaerobic reactors. It  24:135laM XM

a
ðA:7Þ
was also shown that methane yield is strongly dependent b a b
on pH of the reactor. In addition, methane yield was dCac Cac Cac
¼ þ 3:54lbA XAb þ 8:006lbAP XAP
b

shown to increase with greater HRTs and increased dt ð1  aÞh=b

degree of mixing in the reactor. Completely mixed re- þ 15:366lbAB XAB
b
 24:135lbM XM
b
ðA:8Þ
actors required a shorter HRT than incompletely mixed
reactors to achieve the same methane yield. On the other Total propionate, Cpr
hand, it was seen that whenever the hydraulic charac- a
dCpr a
Cpr;f  Cpr b
Cpr a
 Cpr
teristic time is significantly greater than the mixing ¼ þ þ 2:937laA XAa
dt ah ah=b
characteristic time, differences between methane yields b
for imperfectly mixed reactors decrease. Evaluation of  10:566laAP XAP ðA:9Þ
the impact of the characteristic mixing parameters a and b
dCpr a
Cpr b
Cpr
b on anaerobic digestion of cattle manure showed that ¼ þ 2:937lbA XAb  10:566lbAP XAP
b
dt ð1  aÞh=b
both liquid mixing parameters had significant effects on
the digestion process and that methane yield is a com- ðA:10Þ
plex function of both parameters. Total butyrate, Cbut
a
dCbut a
Cbut;f  Cbut C b  Cbut
a
¼ þ but þ 3:079laA XAa
dt ah ah=b
Acknowledgements
 11:919laBP XAB
a
ðA:11Þ
The authors would like to express their appreciation b b
for the financial support provided by Center of Re- dCbut C a  Cbut
¼ but þ 3:079lbA XAb  11:919lbBP XAB
b

newable Energies for Research and Application and Dr. dt ð1  aÞh=b

A. Ahmadi, the former head of the center. ðA:12Þ
 A. Keshtkar et al. / Bioresource Technology 87 (2003) 113–124 123

Total ammonium, Cam A.3. Thermodynamic equilibrium

a a b a
dCam Cam;f  Cam Cam 17ðn  mð1  ye ÞÞ a a
Cam
¼ þ k Cis The relation between free CO2 concentration in the a
dt ah ab 162 þ 17n
liquid phase and partial pressure of CO2 in the gas
 0:15ðlaA XAa þ laAP XAP
a
þ laAB XAB
a
þ laM XM
a
Þ phase, according to Henry’s law
ðA:13Þ
Pc
b
dCam  a
Cam b
17ðn  mð1  ye ÞÞ b b
Cam ½CO2 a ¼ ðA:24Þ
¼ þ k Cis Hc
dt ð1  aÞh=b 162 þ 17n
 0:15ðlbA XAb þ lbAP XAP
b
þ lbAB XAB
b
þ lbM XM
b
Þ A.4. Liquid phase equilibrium chemistry
ðA:14Þ
Ionic dissociation equations
Total carbon dioxide in the liquid phase, Cc
dCca Cc;f  Cca Ccb  Cca ½HCO þ
3 ½H 
¼  þ 2:413laA XAa CO2 þ H2 O $ HCO
3 þH
þ
ka1 ¼
dt ah ah=b ½CO2 
þ 1:01laAP XAP
a
 3:303laAB XAB
a ðA:25Þ
þ
Nca ½CO2
3 ½H 
þ 16:726laM XM
a
 ðA:15Þ HCO 2
3 $ CO3 þ H
þ
ka2 ¼ ðA:26Þ
aVl ½HCO 3
 þ
dCcb C a  Ccb ½AC ½H 
¼þ c þ 2:413lbA XAb HAc $ Ac þ Hþ ka3 ¼ ðA:27Þ
dt ð1  aÞh=b ½HAc
þ 1:01lbAP XAP
b
 3:303lbAB XAB
b ½Pr ½Hþ 
HPr $ Pr þ Hþ ka4 ¼ ðA:28Þ
½HPr
þ 16:726lbM XM
b
ðA:16Þ
½But ½Hþ 
Methane in the liquid phase, Cm HBut $ But þ Hþ ka5 ¼ ðA:29Þ
½HBut
Cmb
þ 1:509laAP XAP
a
þ 0:956laAB XAB
a ½NH3 ½Hþ 
ah=b NHþ
4 $ NH3 þ H
þ
ka6 ¼ ðA:30Þ
½NHþ 4
Nma
þ 6:082laM XM
a
 ¼0 ðA:17Þ H2 O $ OH þ Hþ kw ¼ ½OH ½Hþ  ðA:31Þ
aVl
dCmb Cmb Ionic balance equations for both a and b liquid
¼ þ 1:509lbAP XAP
b
phases
dt ð1  aÞh=b
þ 0:956lbAB XAB
b
þ 6:082lbM XM
b
ðA:18Þ ½Hþ  þ ½NHþ   2
4  ¼ ½OH  þ ½HCO3  þ 2½CO3 

where þ ½Ac  þ ½Pr  þ ½But  þ ½A Cþ 

Vl ðA:32Þ
h¼ ðA:19Þ
Qf where
Qe
b¼ ðA:20Þ Cam =17
Qf ½NHþ
4 ¼ ðA:33Þ
1 þ ka6 =½Hþ 
A.2. Gas phase ½OH  ¼ kw =½Hþ  ðA:34Þ
Carbon dioxide in the gas phase, Pc Cc =44
½HCO
3 ¼ ðA:35Þ
  1 þ ½H =ka1 þ ka2 =½Hþ 
þ
dPc RT Nca Pc
¼  Ft ðA:21Þ Cc =44
dt Vg 44 P ½CO2
2  ¼ ðA:36Þ
2
1 þ ½Hþ =ka2 þ ½Hþ  =ka1 ka2
Methane in the gas phase, Pm
  Cac =60
dPm RT Nma Pm ½Ac  ¼ ðA:37Þ
¼  Ft ðA:22Þ 1 þ ½Hþ =ka3
dt Vg 16 P
Cpr =74
Total material balance in the gas phase, Ft ½Pr  ¼ ðA:38Þ
1 þ ½Hþ =ka4
 
P Nma Nca Cbut =88
Ft ¼ þ ðA:23Þ ½But  ¼ ðA:39Þ
P  Pw 16 44 1 þ ½Hþ =ka5
124 A. Keshtkar et al. / Bioresource Technology 87 (2003) 113–124
References
Angelidaki, I., Ahring, B.K., 1993. Thermophilic anaerobic digestion
of livestock waste: The effect of ammonia. Appl. Microbiol.
Biotechnol. 38, 560–564.
Angelidaki, I., Ellegaard, L., Ahring, B.K., 1993. A mathematical
model for dynamic simulation of anaerobic digestion of complex
substrates: Focusing on ammonia inhibition. Biotech. Bioengng.
42, 159–166.
Archer, D.B., 1983. Enzyme Microbiol. Technol. 5, 162–167.
Bello-Mendoza, R., Sharratt, P.N., 1998. Modelling the effects of
imperfect mixing on the performance of anaerobic reactors for
sewage sludge treatment. J. Chem. Technol. Biotechnol. 71, 121–
130.
Dean, J.A., 1992. Lange’s Handbook of Chemistry, fourteenth ed.
McGraw-Hill, New York.
Dugba, P.N., Zhang, R., 1999. Treatment of dairy wastewater with
two-stage anaerobic sequencing batch reactor systems––thermo-
philic versus mesophilic operations. Bioresour. Technol. 68, 225–
233.
Hansen, K.H., Angelidaki, I., Ahring, B.K., 1999. Improving thermo-
philic anaerobic digestion of swine manure. Water Res. 33 (8),
1805–1810.
Hill, D.T., 1982. A comprehensive dynamic model for animal waste
methanogenesis. Trans. ASAE 25, 1374–1380.
Hills, D.J., Mehlschan, I.J., 1984. Plug flow digestion of dairy manure
of different solids concentration. Trans. ASAE 27 (3), 889–893.
Keshtkar, A., Ghaforian, H., Abolhamd, G., Meyssami, B., 2001.
Dynamic simulation of cyclic batch anaerobic digestion of cattle
manure. Bioresour. Technol. 80, 9–17.
Levenspiel, O., 1972. Chemical Reaction Engineering, second ed.
Wiley, New York.
Lin, K.C., Pearce, M.E.J., 1991. Effects of mixing on anaerobic
treatment of potato-processing wastewater. Can. J. Civ. Engng. 18,
504–514.
Montieth, H.D., Stephenson, J.P., 1981. Mixing efficiencies in full-
scale anaerobic digesters by tracer methods. J. WPCF 53, 78–84.
Nielsen, J., Villadesen, J., 1992. Modeling of microbial kinetics. Chem.
Engng. Sci. 47, 4225–4270.
Perot, C., Segent, M., Richards, P., Phan Tan Luu, R., Millot, N.,
1988. The effects of pH, temperature and agitation speed on sludge
anaerobic hydrolysis-acidification. Environ. Tech. Lett. 9, 741–752.
Reinhold, G., Merrath, S., Lennemann, F., Markl, H., 1996. Modeling
the hydrodynamics and the liquid mixing behavior of a biogas
tower reactor. Chem. Engng. Sci. 51 (17), 4065–4073.
Smith, L.C., Elliot, D.J., James, A., 1993. Characterization of mixing
patterns in an anaerobic digester by means of tracer curve analysis.
Ecol. Model. 69, 267–285.
Varel, V.H., Isaacson, H.R., Bryant, M.P., 1977. Thermophilic
methane production from cattle wastes. Appl. Microbiol. 29,
374–381.
Zeeman, G., Wiegant, W.M., Koster-Treffers, M.E., Lettinga, G.,
1985. The influence of a total ammonia concentration on the
thermophilic digestion of cow manure. Agric. Wastes 14, 19–35.