Optimal operating conditions for maximum biogas production

in anaerobic bioreactors
W. Balmant
a
, B.H. Oliveira
a
, D.A. Mitchell
a
, J.V.C. Vargas
b,
*
, J.C. Ordonez
c
a
Departamento de Bioqumica e Biologia Molecular, Universidade Federal do Paran, C.P. 19011, Curitiba, PR 81531-980, Brazil
b
Departamento de Engenharia Mecnica, Ncleo de P&D em Energia Autossustentvel, Universidade Federal do Paran, C.P. 19011, Curitiba, PR 81531-980,
Brazil
c
Department of Mechanical Engineering, Energy & Sustainability Center, and Center for Advanced Power Systems, Florida State University, Tallahassee, FL
32310-6046, USA
h i g h l i g h t s
We introduce a general transient mathematical model for anaerobic biodigesters.
The model was experimentally validated.
A simulation and optimization study was conducted with the model.
The existence of optimal hydraulic retention time was investigated.
The model could be a tool for simulation, design, control and optimization of biodigesters.
a r t i c l e i n f o
Article history:
Received 9 July 2013
Accepted 19 September 2013
Available online 29 September 2013
Keywords:
Specic growth rates
Performance coefcients
Anaerobic bioreactor
a b s t r a c t
The objective of this paper is to demonstrate the existence of optimal residence time and substrate inlet
mass ow rate for maximum methane production through numerical simulations performed with a
general transient mathematical model of an anaerobic biodigester introduced in this study. It is herein
suggested a simplied model with only the most important reaction steps which are carried out by a
single type of microorganisms following Monod kinetics. The mathematical model was developed for a
well mixed reactor (CSTR e Continuous Stirred-Tank Reactor), considering three main reaction steps:
acidogenesis, with a m
max
of 8.64 day
1
and a K
S
of 250 mg/L, acetogenesis, with a m
max
of 2.64 day
1
and
a K
S
of 32 mg/L, and methanogenesis, with a m
max
of 1.392 day
1
and a K
S
of 100 mg/L. The yield co-
efcients were 0.1-g-dry-cells/g-pollymeric compound for acidogenesis, 0.1-g-dry-cells/g-propionic acid
and 0.1-g-dry-cells/g-butyric acid for acetogenesis and 0.1 g-dry-cells/g-acetic acid for methanogenesis.
The model describes both the transient and the steady-state regime for several different biodigester
design and operating conditions. After model experimental validation, a parametric analysis was per-
formed. It was found that biogas production is strongly dependent on the input polymeric substrate and
fermentable monomer concentrations, but fairly independent of the input propionic, acetic and butyric
acid concentrations. An optimisation study was then conducted and optimal residence time and sub-
strate inlet mass ow rate were found for maximum methane production. The optima found were very
sharp, showing a sudden drop of methane mass ow rate variation from the observed maximum to zero,
within a 20% range around the optimal operating parameters, which stresses the importance of their
identication, no matter how complex the actual bioreactor design may be. The model is therefore ex-
pected to be a useful tool for simulation, design, control and optimisation of anaerobic biodigesters.
2013 Elsevier Ltd. All rights reserved.
1. Introduction
One of the goals of anaerobic digestion is the production of
methane, which can be converted into electricity by its combustion.
The production of biogas, which is rich in methane, by such a
process involves a consort of microorganisms that degrade organic
substrates present in biological wastes. Not only is biogas a clean-
* Corresponding author. Tel.: 55 41 33613307; fax: 55 41 33613129.
E-mail addresses: wbalmant@gmail.com (W. Balmant), behaddad14@
yahoo.com.br (B.H. Oliveira), davidmitchell@ufpr.br (D.A. Mitchell), jvargas@
demec.ufpr.br, vargasjvcv@gmail.com (J.V.C. Vargas), ordonez@caps.fsu.edu
(J.C. Ordonez).
Contents lists available at ScienceDirect
Applied Thermal Engineering
j ournal homepage: www. el sevi er. com/ l ocat e/ apt hermeng
1359-4311/$ e see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.applthermaleng.2013.09.033
Applied Thermal Engineering 62 (2014) 197e206
burning fuel but it also has a lower heating value close to the lower
heating value of methane, i.e., 50.28 MJ kg
1
[1]. Such value de-
pends strongly on the proportion of methane present in the
mixture of gases, which is composed mostly by methane and car-
bon dioxide.
Hence the importance of biodigestion could be summarized by:
while being an efcient waste treatment method, reducing the
organic load of the waste stream, it also produces an environ-
mentally friendly fuel [2]. Moreover, the remaining residues, both
liquid and solid, can be used as biofertilisers [3].
Whenever anaerobic digestion is selected as the process to
generate electricity by the combustion of biogas, the production
rate of methane should be maximized. In that direction, Marcos
et al. [4] experimentally found optimal load rates for maximal
biodegradation rates and methane production obtained from the
anaerobic co-digestion of solid (e.g., fat, intestines, rumen, bowels,
whiskers) and liquid (e.g., blood, washing water, manure) wastes of
the meat industry, particularly the ones rising from the municipal
slaughterhouse of Badajoz, Spain.
Although mathematical models are efcient tools used to opti-
mise a process, it should be noted that to obtain reliable parameters
of an anaerobic digestion is very challenging. The complexity is
mainly due to the large number of microorganisms and compounds
(i.e. large number of parameters) [5], which are summarized in
Fig. 1, therefore, determining accurate values for a specic biore-
actor and substrate would be highly onerous. Therefore, for the
purpose of model development and concision, the organic material
can be classied into carbohydrates, proteins, fats and inert com-
pounds. Fig. 1 sketches the hydrolysis reactions for the rst three,
producing sugars, amino acids and long chain fatty acids,
respectively.
Several mathematical models based on each biochemical step
involved in the process and address variables that affect biogas
production (e.g., temperature) have been proposed. Valle-
Guadarrama [6] proposed a model, based on thermodynamic
principles, that predicts temperature changes in a pilot plant
thermophilic anaerobic digester. Fixed and variable overall heat
transfer coefcient values were used, in good agreement with
experimental data, concluding that temperature variation was
affected by the heterogeneity of the feeding and extraction pro-
cesses, by the heterogeneity of the digestate recirculation through
the heating system and by the lack of a perfect mixing inside the
biodigester tank. However, the use of a searching routine based on a
minimal square optimization criterion for parameters determina-
tion, model complexity and computational time are issues to be
considered when control and optimization are the objectives of the
model.
In general, the published models assume uniform temperature
within the reactor dening a class of models that the technical
literature is rich of. One of the most complete models that have
been proposed is the so called Anaerobic Digestion Model No. 1
(ADM1), which is a structured model that includes multiple steps
Nomenclature
CSTR continuous stirred-tank reactor
F total molar ow rate of gas out of the gas phase,
mol day
1
G
3
H
2
partial pressure, atm
G
4
CO
2
partial pressure, atm
G
7
methane partial pressure, atm
H
3
Henrys constant for H
2
, g L
1
atm
1
H
4
Henrys constant for CO
2
, g L
1
atm
1
H
7
Henrys constant for methane, g L
1
atm
1
HRT hydraulic retention time, V/Q, day
K
h
constant for rst order polymer hydrolysis, day
1
K
S1
saturation constant for acidogens, g L
1
K
S2
saturation constant for syntrophs A, g L
1
K
S3
saturation constant for hydrogenotrophic
methanogens regarding H
2
, g L
1
K
S4
saturation constant for hydrogenotrophic
methanogens regarding CO
2
, g L
1
K
S5
saturation constant for acetoclastic methanogens, g L
1
K
S6
saturation constant for syntrophs B, g L
1
Kla
3
gas-liquid mass transfer coefcient for H
2
, day
1
Kla
4
gas-liquid mass transfer coefcient for CO
2
, day
1/
Kla
7
gas-liquid mass transfer coefcient for methane, day
1
M
H2
molar mass of H
2
, g mol
1
M
CO2
molar mass of CO
2
, g mol
1
M
CH4
molar mass of methane, g mol
1
p
T
total pressure (gas phase), atm
Q substrates input volumetric ow rate, m
3
day
1
Q
g
biogas output volumetric ow rate, 22.4 F, L day
1
r reaction rate, day
1
R universal gas constant, atm L mol
1
K
1
S
j
concentration of substance j, g L
1
S
j
* liquid phase saturation concentration of substance j,
g L
1
S
j,ent
input concentration, g L
1
t time, days
T temperature, K
V liquid phase volume, L
V
g
gas phase volume, L
x
i
molar fraction of component i, G
i
/p
T
X
1
acetogen concentration, g L
1
X
2
syntroph A concentration, g L
1
X
3
hydrogenotrophic methanogen concentration, g L
1
X
4
acetoclastic methanogen concentration, g L
1
X
5
syntroph B concentration, g L
1
Y
Si/Sp
yield of S
i
from S
p
, g g
1
Y
Si/xp
yield of S
i
from X
p
, g g
1
Greek symbols
m specic growth rate, day
1
Subscript
ent inlet
j0 polymeric phase
j1 fermentable monomer
j2 propionic acid
j3 liquid phase H
2
j4 liquid phase CO
2
j5 acetic acid
j6 butyric acid
j7 liquid phase methane
k1 acidogens
k2 syntrophs A
k3 hydrogenotrophic methanogens
k4 acetoclastic methanogens
k5 syntrophs B
max maximum
opt optimal
W. Balmant et al. / Applied Thermal Engineering 62 (2014) 197e206 198
describing biochemical as well as physicochemical processes, but
includes 26 dynamic state concentration variables, and 8 implicit
algebraic variables per reactor vessel or element, and implemented
as differential equations only, there are 32 dynamic concentration
state variables [5].
The work of Vavilin et al. [7] is considered representative, and
was published as a rst user-friendly application made available for
general use, which was a new version of the simulation model
<METHANE> developed earlier [8,9]. The model was experimen-
tally validated by direct comparison to experimental data pub-
lished by Salminen et al. [10]. However, the model contained 31
ordinary differential equations (ODE) with respect to time with a
number of kinetic and physical parameters to be adjusted according
to the particular system under analysis, i.e., computational time
could still be considered an issue if multiple cases are to be run such
as in control and optimization procedures. More recently, Vavilin
et al. [11] added complexity to their previous models by presenting
a newmultidimensional (3 and 2-dimensional) anaerobic digestion
model for a cylindrical reactor with non-uniform inuent concen-
tration distributions, and successfully studied the way in which
mixing intensity affects the efciency of continuous-owanaerobic
digestion.
Simpler models have also been published, such as seen in the
work of Martinez et al. [12], in which the biodigestion process is
described by only seven ODEs, but the authors stated they used a
high time consuming genetic algorithm followed by a gradient
descendent procedure to adjust the model parameters in order to
match the results to experimental data. One of the possible reasons
of the need for such ne parameter adjustments is that the model
was dramatically simplied, not considering for example the two-
step kinetics of polymer hydrolysis, which is known to be impor-
tant since it accounts for the balance between the rates of polymer
hydrolysis and methanogenesis.
Based on the bibliographic review, it is recognized that there are
plenty of published models that have been experimentally vali-
dated and are very successful in representing in detail the physical
phenomena that occur in biodigestion processes. Also, in the other
extreme, mathematically very simple models were found, but that
require complex renement algorithms to be reliable.
Regarding biodigester optimization studies, very few studies
have been published, and most of them are experimental. As
pointed out earlier in the text, the work of Marcos et al. [4] is one
example. However, within the knowledge of the authors, there are
no published studies to nd general optimal residence time and
substrate inlet mass ow rate for maximum biogas production in
the technical literature.
Therefore, the aim of the present work is to demonstrate the ex-
istence of optimal residence time and substrate inlet mass ow rate
for maximum biogas production in biodigesters, through a simple
enough mathematical model which lies between the two extremes
discussed previously, i.e., that is neither too complex nor too simple,
so that practically meaningful results could be obtained directly. The
model should involve as few parameters as possible, and still be
representative of the basic physical phenomena driving the process.
2. Mathematical model
The reactions and processes occurring in the anaerobic digestion
system are simplied into the following general steps, which are
shown in Fig. 2:
1. The polymeric substrate (S
0
) is hydrolysed enzymatically, pro-
ducing fermentable monomers (S
1
);
2. The fermentable monomers (S
1
) are transformed into propionic
acid (S
2
), soluble hydrogen (S
3
), soluble carbon dioxide (S
4
),
acetic acid (S
5
) and butyric acid (S
6
) by acidogenic bacteria (X
1
);
3. The propionic acid (S
2
) is transformed into H
2
(S
3
), CO
2
(S
4
) and
acetic acid (S
5
) by syntrophic bacteria A (X
2
);
4. Acetic acid (S
5
) is transformed into methane (S
7
) and CO
2
(S
4
) by
acetoclastic methanogenic bacteria (X
4
);
5. The butyric acid (S
6
) is transformed into H
2
(S
3
) and acetic acid
(S
5
) by syntrophic bacteria B (X
5
);
6. CO
2
and H
2
are used by the hydrogenotrophic-methanogenic
bacteria (X
3
) to generate methane (S
7
), and
CO2
Complex
Organic
Material
GAS PHASE
LIQUID PHASE
Lipids
Proteins
LCFA
Simple
Sugars
Lactic Ac.
Butyric Ac.
Aminoacids
H
2
+CO
2
A
c
e
t
i
c
A
Valeric Ac.
Propionic Ac.
CH
4
H
2
+CO
2
Carbohidrates
NH
3
I I I
IV
1
1
2
2
2
2
2
2
2
2
3
2
4 4
3
3
3
III
A
c
e
t
i
c

A
c
i
d

V
F
A
Fig. 1. The complex web of reactions within the anaerobic digestion system. Steps of the biodigestion process: I) Hydrolysis; II) Acidogenesis; III) Acetogenesis; IV) Methanogenesis
e Microbial groups: 1) hydrolytics; 2) acidogens; 3) acetogens; 4) methanogens. VFA volatile fatty acids; LCFA long chain fatty acids.
W. Balmant et al. / Applied Thermal Engineering 62 (2014) 197e206 199
7. Transfer of CO
2
, H
2
and methane between the gas and liquid
phases of the bioreactor.
The model considers the following assumptions:
1. Gases behave as ideal gases;
2. The gas phase is well mixed;
3. The total pressure in the gas phase is constant;
4. The temperature of the systems is uniform and constant;
5. The biogas only contains CO
2
, H
2
and methane, and
6. The vapour pressure of water in the gas phase is negligible.
2.1. Rate expressions
The enzymatic hydrolysis of the polymeric substrate is consid-
ered to be a rst order process, as follows:
r
1
K
h
$S
0
(1)
The growth of each population of bacteria is directly propor-
tional to their concentration, according to the following
equations:
r
2
m
1
$X
1
acidogens (2)
r
8
m
2
$X
2
syntrophs A (3)
r
10
m
3
$X
3
hydrogenotrophic methanogens (4)
r
11
m
4
$X
4
acetoclastic methanogens (5)
r
14
m
5
$X
5
syntrophs B (6)
In each case the specic growth rate depends on the relevant
substrate concentration according to the Monod equation [13], as
follows:
m
1

m
max1
$S
1
K
S1
S
1
acidogens (7)
m
2

m
max2
$S
2
K
S2
S
2
syntrophs A (8)
m
3

m
max3
$S
3
K
S3
S
3
$
S
4
K
S4
S
4
hydrogenotrophic methanogens
(9)
m
4

m
max4
$S
5
K
S5
S
5
acetoclastic methanogens (10)
m
5

m
max5
$S
6
K
S6
S
6
syntrophs B (11)
The various products are treated as being growth-associated,
including production of propionic acid by acidogenic bacteria,
r
3
Y
S2=X1
$m
1
$X
1
(12)
production of H
2
by acidogenic bacteria,
r
4
Y
S3=X1
$m
1
$X
1
(13)
production of CO
2
by acidogenic bacteria,
r
5
Y
S4=X1
$m
1
$X
1
(14)
Fig. 2. Simplied diagram of reactions and processes occurring in the anaerobic digestion system.
W. Balmant et al. / Applied Thermal Engineering 62 (2014) 197e206 200
production of acetic acid by acidogenic bacteria,
r
6
Y
S5=X1
$m
1
$X
1
(15)
production of butyric acid by acidogenic bacteria,
r
7
Y
S6=X1
$m
1
$X
1
(16)
production of H
2
by syntrophic bacteria A,
r
9
Y
S3=X2
$m
2
$X
2
(17)
production of CO
2
, by the acetoclastic methanogens,
r
12
Y
S4=X4
$m
4
$X
4
(18)
production of acetic acid by syntrophic bacteria B,
r
13
Y
S5=X5
$m
5
$X
5
(19)
production of H
2
by syntrophic bacteria B,
r
15
Y
S3=X5
$m
5
$X
5
(20)
production of CO
2
by syntrophic bacteria A,
r
16
Y
S4=X2
$m
2
$X
2
(21)
production of acetic acid by syntrophic bacteria A,
r
17
Y
S5=X2
$m
2
$X
2
(22)
production of methane by the hydrogenotrophic methanogens,
r
18
Y
S7=X3
$m
3
$X
3
(23)
and production of methane by the acetoclastic methanogens,
r
19
Y
S7=X4
$m
4
$X
4
(24)
The rate of H
2
transfer from the liquid to the gas phase is given
by,
r
20
Kla
3
$

S
3
S
*
3

(25)
where
S
*
3
H
3
$G
3
(26)
The rate of CO
2
transfer from the liquid to the gas phase is given
by,
r
21
Kla
4
$

S
4
S
*
4

(27)
where
S
*
4
H
4
$G
4
(28)
The rate of CH
4
transfer from the liquid to the gas phase is given
by,
r
22
Kla
7
$

S
7
S
*
7

(29)
where
S
*
7
H
7
$G
7
(30)
2.2. Species conservation equations
Concentrations balances are written for the species present in
the bioreactor as follows:
a) polymeric substrate in the liquid phase of the bioreactor,
dS
0
dt

Q
V
$S
0ent
S
0
r
1
(31)
b) fermentable monomers in the liquid phase of the bioreactor,
dS
1
dt

Q
V
$S
1ent
S
1
Y
S1=S0
$r
1
Y
S1=X1
$r
2
Y
S1=S2
$r
3
Y
S1=S3
$r
4
Y
S1=S4
$r
5
Y
S1=S5
$r
6
Y
S1=S6
$r
7
(32)
c) acidogens in the liquid phase of the bioreactor,
dX
1
dt

Q
V
$X
1ent
X
1
r
2
(33)
d) propionic acid in the liquid phase of the bioreactor,
dS
2
dt

Q
V
$S
2ent
S
2
r
3
Y
S2=X2
$r
8
Y
S2=S3
$r
9
Y
S2=S4
$r
16
Y
S2=S5
$r
17
(34)
e) syntrophic bacteria A in the liquid phase of the bioreactor,
dX
2
dt

Q
V
$X
2ent
X
2
r
8
(35)
f) H
2
in the liquid phase of the bioreactor,
dS
3
dt

Q
V
$S
3ent
S
3
r
4
r
9
r
15
Y
S3=X3
$r
10
Y
S3=S7
$r
18
r
20
(36)
g) CO
2
in the liquid phase of the bioreactor,
dS
4
dt

Q
V
$S
4ent
S
4
r
5
r
12
r
16
Y
S4=X3
$r
10
Y
S4=S7
$r
18
r
21
(37)
h) hydrogenotrophic methanogens in the liquid phase of the
bioreactor,
dX
3
dt

Q
V
$X
3ent
X
3
r
10
(38)
W. Balmant et al. / Applied Thermal Engineering 62 (2014) 197e206 201
i) acetic acid in the liquid phase of the bioreactor,
dS
5
dt

Q
V
$S
5ent
S
5
r
6
r
13
r
17
Y
S5=X4
$r
11
Y
S5=S4
$r
12
Y
S5=S7
$r
19
(39)
j) acetoclastic methanogens in the liquid phase of the bioreactor,
dX
4
dt

Q
V
$X
4ent
X
4
r
11
(40)
k) butyric acid in the liquid phase of the bioreactor,
dS
6
dt

Q
V
$S
6ent
S
6
r
7
Y
S6=S5
$r
13
Y
S6=X5
$r
14
Y
S6=S3
$r
15
(41)
l) syntrophic bacteria B in the liquid phase of the bioreactor,
dX
5
dt

Q
V
$X
5ent
X
5
r
14
(42)
m) methane in the liquid phase of the bioreactor,
dS
7
dt

Q
V
$S
7ent
S
7
r
18
r
19
r
22
(43)
Since the overall pressure in the headspace is assumed to be
constant, the total molar ow rate of gas out of the gas phase is
equal to the amount of gas transferred from the liquid to the gas
phase, shown by the following expression:
F Kla
3
$

S
3
S
*
3

$V=M
H2
Kla
4
$

S
4
S
*
4

$V=M
CO2
Kla
7
$

S
7
S
*
7

$V=M
CH4
(44)
n) H
2
in the gas phase of the bioreactor, rearranged to be
expressed in terms of the partial pressure of H
2
in the gas
phase,
dG
3
dt

R$T
V
g
$

Kla
3
$

S
3
S
*
3

$V=M
H2
F$V$
G
3
P
T

(45)
o) CO
2
in the gas phase of the bioreactor, rearranged to be
expressed in terms of the partial pressure of CO
2
in the gas
phase,
dG
4
dt

R$T
V
g
$

Kla
4
$

S
4
S
*
4

$V=M
CO2
F$V$
G
4
P
T

(46)
p) methane in the gas phase of the bioreactor, rearranged to be
expressed in terms of the partial pressure of methane in the
gas phase,
dG
7
dt

R$T
V
g
$

Kla
7
$

S
7
S
*
7

V=M
CH4
F$V$
G
7
P
T

(47)
At this point, it is important to stress that the utilized hypoth-
eses and model equations grant exibility to the model, in the sense
that it could be used to describe a biodigester operating either as a
continuous stirred-tank reactor (CSTR) or a batch reactor. For
describing a batch reactor, it is only necessary to set Q 0 m
3
day
1
.
In fact, in this work, for performing the model experimental vali-
dation, biodigester batch reactor experimental data available in the
literature [10] are utilized, therefore the present model produces
results for Q0 m
3
day
1
. After model experimental validation, Q is
allowed to vary, therefore the optimization study is conducted for
biodigesters operating as continuous stirred-tank reactors (CSTR).
3. Numerical method
Eqs. (31)e(43), (45)e(47), for all species inthe bioreactor inliquid
and gas phases, and the specied initial conditions forma systemof
16 ordinary differential equations (ODEs) which need to be solved
together with 31 auxiliary algebraic equations, i.e., Eqs. (1)e(30) and
(44). The unknowns are the concentrations (liquidphase) andpartial
pressures (gas phase) of the species in the bioreactor.
The solution to the ODEs depict the transient behaviour of the
system, starting from a set of initial conditions, and marching in
time (checking for accuracy) until a steady state is achieved. The
equations were integrated in time implicitly using a Backward
Euler method [14], which was implemented in a programwritten in
FORTRAN language, using the subroutine DASSL (differential alge-
braic system solver) [15].
Table 1 lists the initial values of variables and the values of pa-
rameters that were used in the simulation. The kinetic parameters
were obtained from the technical literature [16]. The yield co-
efcients required by the mathematical model were determined by
undertaking a stoichometric balance on the overall reaction for
each step. For that, the substrate efciency coefcient was assumed
as 10%, and all other stoichiometric coefcients were calculated
from the balanced stoichiometric equations [17]. Table 2 lists the
calculated yield values.
4. Results and discussion
The numerical results obtained with the mathematical model
were compared to available batch reactor experimental data ob-
tained by Salminen et al. [10] in Figs. 3e7. Table 1 displays the list of
parameters and initial values used in the simulations, which were
selected according to the experimental data [10].
Fig. 3 depicts the transient evolution of the polymeric substrate
with respect to time. It is seen that the model was capable of
capturing the descending trend in time of the polymeric substrate
concentration, S
0
, but the behaviour was monotonic whereas the
experimental measurements showed that the concentration
initially decreased, then on the 10th day stabilized up to the 20th
day, and eventually decreased to reach steady state at complete
depletion around the 35th day, therefore only qualitative agree-
ment was observed in the polymeric substrate concentration
transient evolution. On the other hand, the numerical results
showed good quantitative and qualitative agreement with the
experimentally determined steady state, i.e., around the 35th day in
W. Balmant et al. / Applied Thermal Engineering 62 (2014) 197e206 202
both cases. Therefore, the simulated prole for the variation of
concentration of polymeric substrate in time represents an average
consumption rate as compared to the experimental data.
The unsteady simulation results for acetic and butyric acid show
good qualitative and quantitative agreement with the experimental
data reported by Salminen et al. [10] as shown in Figs. 4 and 5,
respectively. Fig. 4 also shows that the model could not predict
accurately the time to achieve steady state for the acetic acid, but
for butyric acid Fig. 5 demonstrates that the model predicted steady
conditions around the 20th day, therefore in good agreement with
the experiments.
For propionic acid, Fig. 6 illustrates that the model fails to
accurately represent the process from the 10th to the 30th day. The
transient numerical results showed good agreement with the ex-
periments until the 10th day. Experimentally it is seen that the
compound concentration was still increasing on the 10th day, in
which the simulation results started to drop, and continued to in-
crease. The experimentally measured concentration peak
(w3 g L
1
) occurred around the 25th day and was more than 20%
greater than the numerically simulated one (w2.4 g L
1
). Steady
state was achieved experimentally around the 40th day whereas
the numerical results stabilized around the 35th day, with the
compound concentration dropping to values close to zero.
In Fig. 7, it is noted that the numerically calculated molar frac-
tion of methane in the gas phase exhibits good qualitative agree-
ment with the experimental measurements both for the transient
and steady state evolution of the process. However, quantitatively,
the model underestimates the nal methane molar fraction ob-
tained in steady state conditions, which was close to 70% in the
experimental data and 50% in the simulated results.
The disagreement between the numerical results and some of
the experimental data is understandable, due to the model as-
sumptions. For example, it is well known that syntrophic bacteria,
which are responsible for propionic and butyric acid consumption,
are inhibited by hydrogen, but this effect was not taken into ac-
count in the model. Future improvements of the model could
include the incorporation of the hydrogen inhibitory effect.
Due to the qualitative and quantitative agreement of the nu-
merical results and experimental data in most of the transient and
steady state compounds concentrations and partial pressures, it is
reasonable to state that the model has been experimentally vali-
dated. This is mainly true for optimization purposes, since quali-
tative agreement was observed in all comparisons, i.e., the actual
system trends have been captured by the model, so that in spite of
quantitative discrepancies that might occur in numerically pre-
dicting maximum performance values (e.g., maximum methane
molar fraction output), the locations of the predicted optima are
expected to be accurate.
Additionally, the simplied reaction model of the anaerobic
digestion process that was developed in the present work is suit-
able for incorporation into models that describe more complex
systems in which the biodigester contents are not homogeneous.
For that, in order to account for reactor non-uniformities, spatial
dependence could be introduced in the model without further in-
crease in complexity, by the use of a volume element model [18]
which does not require the use of partial differential equations. In
such case, the reaction model will be only slightly modied, but the
expressions describing the hydrodynamic aspects of the system
will be more complex.
After model experimental validation, the analysis proceeded to
evaluate the response of the reactor to the variation of several
operating conditions, i.e., by performing a parametric analysis.
Although the model experimental validation has been conducted
for a batch reactor, generally, biodigesters operate continuously, so
that the effect of the continuous variation of system variables on
bioreactor production is important to be assessed.
Figs. 8 and 9 show the effects of the substrates concentrations
on the output total biogas volumetric ow rate. The input sub-
strates concentrations S
2ent
, S
5,ent
and S
6,ent
had no impact on the
biogas volumetric ow rate, as it is seen in Fig. 9a, b and c,
respectively. However, the input substrates concentrations S
0ent
and S
1ent
had a signicant effect on increasing biogas production, as
Table 1
Values of parameters [7,10] and initial values of variables.
Symbol Value Unit Symbol Value Unit
Q 250 day
1
K
S2
0.97 g L
1
V 2000 L K
S3
0.00005 g L
1
V
g
300 L K
S4
0.0019 g L
1
R 0.082 atm L mol
1
K
1
K
S5
0.019 g L
1
T 305 K K
S6
0.59 g L
1
G
3,4,7
0 at t 0 atm X
1
0.1 at t 0 g L
1
P
T
1 at t 0 atm X
2
0.1 at t 0 g L
1
S
0
60 g L
1
X
3
0.01 at t 0 g L
1
S
1
1 at t 0 g L
1
X
4
0.01 at t 0 g L
1
S
2,3,4,5,6,7
0 at t 0 g L
1
X
5
0.1 at t 0 g L
1
S
0ent
60 g L
1
X
1,2,3,4,5ent
0 g L
1
S
1,2,3,4,5,6,7ent
0 g L
1
Kla
3,4,7
48 day
1
K
h
0.5 day
1
S
3
* Eq. (26) g L
1
m
max1
0.2 day
1
S
4
* Eq. (28) g L
1
m
max2
0.00185 day
1
S
7
* Eq. (30) g L
1
m
max3
2 day
1
H
3
0.01 g L
1
atm
1
m
max4
0.0225 day
1
H
4
0.04 g L
1
atm
1
m
max5
0.01 day
1
H
7
0.07 g L
1
atm
1
K
S1
0.67 g L
1
M
H2
2 g mol
1
Table 2
Yield values (Y
a/b
, yield of a from b), g g
1
.
Y
S1/S0
1.11 Y
S2/S3
12.3 Y
S3/x2
0.811 Y
S4/x4
7.33 Y
S6/S3
22.0
Y
S1/S2
12.2 Y
S2/S4
1.68 Y
S3/x3
1.54 Y
S5/S4
1.36 Y
S6/x1
1.83
Y
S1/S3
40.0 Y
S2/S5
1.23 Y
S3/x5
0.455 Y
S5/S7
3.75 Y
S6/x5
10.0
Y
S1/S4
2.56 Y
S2/x1
0.822 Y
S4/S7
2.75 Y
S5/x1
2.83 Y
S7/x3
3.08
Y
S1/S5
3.53 Y
S2/x2
10.0 Y
S4/x1
3.91 Y
S5/x2
8.11 Y
S7/x4
2.67
Y
S1/S6
5.45 Y
S3/S7
0.5 Y
S4/x2
5.95 Y
S5/x4
10.0
Y
S1/x1
10.0 Y
S3/x1
0.25 Y
S4/x3
8.46 Y
S5/x5
13.6
0
5
10
15
20
25
0 20 40 60
S
0
(g L
-1
)
t (day)
Fig. 3. Transient evolution of the polymeric substrate concentration, S
0
(solid
line model prediction; squares experimental data [10]).
Fig. 4. Transient evolution of acetic acid concentration, S
5
(solid line model pre-
diction; squares experimental data [10]).
W. Balmant et al. / Applied Thermal Engineering 62 (2014) 197e206 203
it is shown in Fig. 8a and b, respectively. What is more striking is the
effect of S
1ent
on biogas production, which increases monotonically
and linearly as S
1ent
increases. Such effects demonstrate how much
the hydrolysis step of the polymeric substrate limits system ef-
ciency. Note that when the system is fed with hydrolysed substrate
(S
1
is a product of the S
0
hydrolysis), the biogas production in-
creases considerably. The analysis could be extended for higher
values of S
1ent
, what is done in Fig. 8c, in which S
1ent
was increased
up to 2000 g L
1
, and the data were tted through linear regression,
producing the following correlation:
Q
g
4:383 S
1ent
with R
2
0:9997 (48)
The correlation given by Eq. (48) is valid only for the set of data
used in this work. However, the output total biogas volumetric ow
rate, Q
g
, is expected to behave similarly for any continuous stirred-
tank reactor (CSTR) biodigester, i.e., linearly with respect to the
variation of S
1ent
.
Another important variable that affects continuous biodigesters
performance is the substrates input volumetric ow rate, Q, which
determines the system retention time, HRT (hydraulic retention
time). Fig. 10a shows that small or large HRT lead to poor perfor-
mance, so that there is an optimal hydraulic retention time
(HRT
opt
w 5 days) for maximum biogas production, which corre-
sponds to Q
opt
175 L day
1
for the set of data used in the simu-
lations conducted in this work, as it is seen in Fig. 10b. This optimal
condition is physically explained by analysing two extremes with
xed V: i) for small HRT, Q is high, therefore reaction kinetics is not
fast enough to process the substrates and biogas production is low,
and ii) for large HRT, Q is small, therefore substrates concentrations
are low, and biogas production is also low. Furthermore, the
maximum is quite sharp, i.e., there is a signicant drop in biogas
production that results from biodigester operation with HRT even
slightly different fromHRT
opt
. For example, if a narrowrange HRT
opt
5% is considered, a 50% drop is observed in Q
g
with respect to
Q
g,max
for the set of data used in this study.
The optimal values for HRT and Q found for the set of data used
in the numerical simulations conducted in this work are within
what is expected for a continuous stirred-tank reactor (CSTR) bio-
digester. However, depending on the input polymeric solids con-
centration, S
0ent
, the optimal substrates input volumetric ow rate,
Q
opt
, could change. As it is observed in Fig. 11, for high values of S
0ent
it is better to reduce Q, using HRT 30 days instead of HRT 6
days. The reason is that with higher HRT there is more time to
hydrolyse the polymeric substrate which in turn generates more
biogas.
0
1
2
3
4
5
6
0 20 40 60
t (day)
S
6
(g L
-1
)
Fig. 5. Transient evolution of butyric acid concentration, S
6
(solid line model pre-
diction; squares experimental data [10]).
0
0,5
1
1,5
2
2,5
3
0 10 20 30 40 50 60
S
2
(g L
-1
)
t (day)
Fig. 6. Transient evolution of propionic acid concentration (solid line model pre-
diction; squares experimental data [10]).
0
10
20
30
40
50
60
70
80
0 10 20 30 40 50 60
t (day)
x
7
(%)
Fig. 7. Transient evolution of methane molar fraction, X
7
(solid line model predic-
tion; squares experimental data [10]).
Fig. 8. The variation of the total biogas output volumetric ow rate, Q
g
, with respect
to: a) Input polymeric substrate concentration, S
0ent
; b) Input fermentable monomers
concentration, S
1ent
, up to 200 g L
1
, and c) Input fermentable monomers concentra-
tion, S
1ent
, up to 2000 g L
1
.
W. Balmant et al. / Applied Thermal Engineering 62 (2014) 197e206 204
At this point it is important to generalize the analysis. Funda-
mentally, the message is that no matter how complex the actual
biodigester design might be, there will be an optimal retention
time, HRT
opt
, and a corresponding optimal substrates input volu-
metric owrate, Q
opt
, worth to be found such that maximumbiogas
production, Q
g,max
, will be achieved.
5. Conclusions
In this paper, optimal hydraulic retention time, HRT
opt
(or input
volumetric ow rate, Q
opt
) was found that lead to sharp maximum
biogas production within a narrow range, stressing the importance
of its identication in any actual biodigester operation, i.e., for
HRT
opt
5%, a 50% drop was observed in Qwith respect to Q
g,max
, for
the biodigester conguration tested in this study. For that, a general
transient mathematical model for the species management of batch
and continuous stirred-tank reactor (CSTR) biodigesters has been
developed. The model was validated experimentally by direct
comparison of transient and steady state simulations results with
previously published experimental data for a batch reactor bio-
digester. Also, a sensitivity analysis was conducted for CSTR bio-
digesters identifying critical variables that affect output total biogas
volumetric ow rate. This was the case of the input fermentable
monomer substrate concentration, S
1
, which was shown to affect
signicantly biogas production that increases monotonically as S
1
increases.
The key conclusion of this study is that the simplied model
presented for the reactions that occur in anaerobic digestion has
the ability to describe both transient and steady state behaviour.
Therefore, the model could be applied in practice for the design of
small anaerobic bioreactors or large anaerobic bioreactors with
induced homogeneous mixtures.
The present model is also suitable for incorporation into
mathematical models that describe the operation of more complex
anaerobic bioreactors, in which the bioreactor contents are not
homogeneously mixed. Therefore it is expected that such applica-
tion could be used as an efcient tool for batch and CSTR bio-
digester design, control and optimization.
Acknowledgements
The authors acknowledge with gratitude the support of the
Brazilian National Council of Scientic and Technological Devel-
opment, CNPq (projects 401117/2004-9, 552867/2007-1, 574759/
2008-5 and 558835/2010-4), and of NILKO Tecnologia Ltda.
References
[1] A. Bejan, Advanced Engineering Thermodynamics, second ed., Wiley, New
York, 1997 (Chapter 7).
[2] H. Bouallagui, Mesophilic biogas production from fruit and vegetable waste in
a tubular digester, Bioresour. Technol. 86 (2003) 85e89.
Fig. 9. The variation of the total biogas output volumetric ow rate, Q
g
, with respect
to: a) Input propionic acid concentration, S
2ent
; b) Input acetic acid concentration, S
5ent
,
and c) Input butyric acid concentration, S
6ent
.
Fig. 10. The maximization of the total biogas output volumetric ow rate, Q
g
, with
respect to: a) Hydraulic retention time, HRT, and b) Substrates input volumetric ow
rate, Q.
Fig. 11. The variation of the total biogas output volumetric ow rate, Q
g
, with respect
to input polymeric substrate concentration, S
0ent
, and hydraulic retention time, HRT.
W. Balmant et al. / Applied Thermal Engineering 62 (2014) 197e206 205
[3] J.M. Lomas, C. Urbano, L.M. Camarero, Evaluation of a pilot scale downow
stationary xed lm anaerobic reactor treating piggery slurry in the meso-
philic range, Biomass Bioenergy 17 (1999) 49e58.
[4] A. Marcos, A. Al-Kassir, A.A. Mohamad, F. Cuadros, F. Lpez-Rodrguez,
Combustible gas production (methane) and biodegradation of solid and
liquid mixtures of meat industry wastes, Appl. Energy 87 (2010) 1729e
1735.
[5] D.J. Batstone, J. Keller, I. Angelidaki, S.V. Kalyuzhnyi, S.G. Pavlostathis, A. Rozzi,
W.T.M. Sanders, H. Siegrist, V.A. Vavilin, The IWA anaerobic digestion model
no 1 (ADM1), Water Sci. Technol. 45 (10) (2002) 65e73.
[6] S. Valle-Guadarrama, T. Espinosa-Solares, I.L. Lopez-Cruz, M. Domaschko,
Modeling temperature variations in a pilot plant thermophilic anaerobic
digester, Bioprocess Biosyst. Eng. 34 (4) (2011) 459e470.
[7] V.A. Vavilin, L.Y. Lokshina, S.V. Rytov, The <METHANE> simulation model as
the rst generic user-friend model of anaerobic digestion, Moscow Univ.
Chem. Bull. 41 (6) (2000) 22e26 (Supplement).
[8] V.B. Vasiliev, V.A. Vavilin, S.V. Rytov, A.V. Ponomarev, Simulation model of
anaerobic digestion of organic matter by a microorganism consortium: basic
equations, Water Resour. 20 (1993) 633e643.
[9] V.A. Vavilin, V.B. Vasiliev, A.V. Ponomarev, S.V. Rytov, Simulation-model
methane as a tool for effective biogas production during anaerobic conversion
of complex organic-matter, Bioresour. Technol. 48 (1) (1994) 1e8.
[10] E. Salminen, J. Rintala, L.Y. Lokshina, V.A. Vavilin, Anaerobic batch degradation
of solid poultry slaughterhouse waste, Water Sci. Technol. 41 (3) (2000) 33e41.
[11] V.A. Vavilin, L.Y. Lokshina, X. Flotats, I. Angelidaki, Anaerobic digestion of solid
material: multidimensional modeling of continuous-ow reactor with non-
uniform inuent concentration distributions, Biotechnol. Bioeng. 97 (2)
(2007) 354e366.
[12] E. Martinez, A. Marcos, A. Al-Kassir, M.A. Jaramillo, A.A. Mohamad, Mathe-
matical model of a laboratory-scale plant for slaughterhouse efuents bio-
digestion for biogas production, Appl. Energy 95 (2012) 210e219.
[13] J. Monod, The growth of bacterial cultures, Annu. Rev. Microbiol. 3 (1949) 371.
[14] D. Kincaid, W. Cheney, Numerical Analysis, Wadsworth, Belmont, CA, 1991.
[15] K.E. Brenan, S.L. Campbell, L.R. Petzold, Numerical Solution of Initial-value
Problems in Differential-algebraic Equations, Classics in Applied Mathe-
matics, Philadelphia, PA, 1996.
[16] G.B. Ryhiner, E. Heinzle, J.I. Dunn, Modeling and simulation of anaerobic
wastewater treatment and its application to control design: case whey, Bio-
technol. Prog. 9 (1993) 332e343.
[17] M.L. Shuler, F. Kargi, Bioprocess Engineering: Basic Concepts, second ed.,
Prentice Hall, 2001.
[18] J.V.C. Vargas, G. Stanescu, R. Florea, M.C. Campos, A numerical model to pre-
dict the thermal and psychrometric response of electronic packages, Trans.
ASME J. Electron. Packag. 123 (3) (2001) 200e210.
W. Balmant et al. / Applied Thermal Engineering 62 (2014) 197e206 206