6th IFAC Conference on

6th
6th IFAC
IFAC Conference
Conference
Foundations on
on Biology in Engineering
of Systems
6th IFAC Conference
Foundations of on Biology in Engineering
Systems
Foundations
October 9-12,of Systems
2016. BiologyGermany
Magdeburg, in Engineering
Available online at www.sciencedirect.com
Foundations
October of Systems Biology in Engineering
October 9-12, 2016. Magdeburg, Germany
9-12, 2016. Magdeburg, Germany
October 9-12, 2016. Magdeburg, Germany
ScienceDirect
IFAC-PapersOnLine 49-26 (2016) 231–238
Mathematical
Mathematical Modelling
Modelling of
of Anaerobic
Anaerobic
Mathematical
Digestion with Modelling
Hydrogen of
andAnaerobic
Methane
Digestion with
Digestion with Hydrogen
Hydrogen  and
and Methane
Methane
Production
Production 
Production 
Milen Borisov ∗∗∗ Neli Dimitrova ∗∗ ∗∗ Ivan Simeonov ∗∗∗
∗∗ Ivan Simeonov ∗∗∗
Milen
Milen Borisov Neli Dimitrova ∗∗∗
Milen Borisov ∗ Neli Dimitrova ∗∗ Ivan
Borisov Neli Dimitrova Ivan Simeonov
Simeonov ∗∗∗
∗
∗ Institute of Mathematics and Informatics,
∗ Institute of Mathematics
Academy ofand Informatics,
∗
Institute of
of Mathematics
Bulgarian
Institute Mathematics and
and Informatics,
Sciences,
Informatics,
Acad. G. Bulgarian
Bulgarian
Bonchev
Bulgarian
Academy
Academy
Str. Bl.
Academy 8, of
of
1113
of
Sciences,
Sciences,
Sofia, Bulgaria,
Sciences,
Acad.
Acad. G.
G. Bonchev
Bonchev Str.
Str. Bl.
Bl. 8,
8, 1113 Sofia, Bulgaria,
Acad. (e-mail:
G. (e-mail: milen
Bonchevmilen
Str. Bl. 8, 1113 Sofia,
1113 Sofia, Bulgaria,
kb@math.bas.bg).
kb@math.bas.bg). Bulgaria,
∗∗ (e-mail:
Institute of
(e-mail: milen kb@math.bas.bg).
Mathematics
milen and Informatics,
kb@math.bas.bg).
∗∗
∗∗ Institute
∗∗ Institute of
of Mathematics
Academy ofand Informatics,
Institute of Mathematics
BulgarianMathematics and
and Informatics,
Sciences,
Informatics,
Acad. G. Bulgarian
Bulgarian
Bonchev
Bulgarian
Academy
Academy
Str. Bl.
Academy 8, of
of
1113
of
Sciences,
Sciences,
Sofia,
Sciences, Bulgaria,
Acad.
Acad. G.
G. Bonchev
Bonchev Str.
Str. Bl.
Bl. 8, 1113 Sofia, Bulgaria,
(e-mail:
Acad. G. Bonchev
(e-mail: Bl. 8, 1113 Sofia, Bulgaria,
8, 1113
nelid@math.bas.bg)
Str.
nelid@math.bas.bg)
Sofia, Bulgaria,
∗∗∗ (e-mail:
Institute of Microbiology, nelid@math.bas.bg)
Bulgarian Academy of Sciences,
(e-mail: nelid@math.bas.bg)
∗∗∗
∗∗∗ Institute of Microbiology,
∗∗∗ Institute of Bulgarian Academy of Sciences,
Acad.G. of Microbiology,
Acad.G. Bonchev
Institute Bonchev Str.,
Bulgarian
Str., Block
Microbiology, 26, 1113
Bulgarian
Block 26, 1113
Academy
Academy
Sofia,
of
of Sciences,
Sofia, Bulgaria,
Sciences,
Bulgaria,
Acad.G. Bonchev
Acad.G. Bonchev(e-mail:Str., Block 26, 1113
issim@microbio.bas.bg)
Str., Block 26, 1113 Sofia,
Sofia, Bulgaria,
Bulgaria,
(e-mail:
(e-mail: issim@microbio.bas.bg)
(e-mail: issim@microbio.bas.bg)
issim@microbio.bas.bg)
Abstract: We propose a new mathematical model describing a biotechnological process of
Abstract:
simultaneousWe
Abstract: We propose
propose a
a new
new mathematical model describing aa biotechnological
The processprocess of
Abstract:
simultaneous Weproduction
production
of
propose of new mathematical
hydrogen
a hydrogen and methane
mathematical
and methane
model
model
by
describing
by anaerobic a biotechnological
digestion.
describingdigestion.
anaerobic biotechnological
The process
process
process
is
of
is carried
of
carried
simultaneous
simultaneous production
out in two connected
production of hydrogen
continuously and methane
stirred
of hydrogenstirred by anaerobic
bioreactors.
and methane digestion.
The proposed
by anaerobic digestion. The
model process is
is developed
The process carried
by
is carried
out
out in
in two connected continuously bioreactors. The proposed model is developed by
in two
reducing
out
reducing
theconnected
two
the
well known
connected
well known
continuously
continuously
Anaerobic
stirred
stirred bioreactors.
Anaerobic Digester
Digester
Model No 1The
bioreactors.
Model No 11The
proposed
(ADM1).
proposed
(ADM1).
Themodel
The model is
is developed
mathematical
developed
mathematical
by
analysis
by
analysis
reducing
of the
reducing the
model
the well
well known
involves
known Anaerobic
computation
Anaerobic Digester
of
Digester Model
equilibrium
Model No
points,
No 1 (ADM1). The
investigation
(ADM1). The mathematical
of their local
mathematical analysis
stability
analysis
of
of therespect
the
with modeltoinvolves
model involves computation
computation
practically importantofofinput
equilibrium
equilibrium points,
points,
parameters, investigation
investigation
existence of maxima of their
of their local
of thelocal stability
stability
input-output
of therespect
with modeltoinvolves computation
practically important ofinput
equilibrium points,
parameters, investigation
existence of maximaof their
of the local stability
input-output
with
with respect
staticrespect to practically
characteristics. important
Numerical
to practically input
simulations
important parameters, existence
using a specially
input parameters, of maxima
elaborated
existenceelaborated of the input-output
web-based
of maxima web-based software
of the input-output
static
static characteristics.
characteristics.
environment Numerical
Numerical
are presented simulations
simulations
to demonstrate using
theusing a specially
aa specially
dynamic behaviorelaborated web-based
of the model solutions.software
software
static characteristics.
environment are Numerical
presented to simulations
demonstrate theusing
dynamic specially
behaviorelaborated
of the web-based
model solutions.software
environment
environment are
are presented
presented to
to demonstrate
demonstrate the
the dynamic
dynamic behavior
behavior of
of the
the model
model solutions.
solutions.
© 2016, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
Keywords: Biotechnological process; Hydrogen and methane production; Mathematical Model;
Keywords:
Keywords: Biotechnological
Equilibrium Biotechnological
points; Dynamics. process;
process; Hydrogen
Hydrogen and and methane
methane production;
production; Mathematical
Mathematical Model;
Model;
Keywords:
Equilibrium Biotechnological
points; Dynamics.process; Hydrogen and methane production; Mathematical Model;
Equilibrium points; Dynamics.
Equilibrium points; Dynamics.
1. INTRODUCTION A lot of models describing separately the fermentative
1.
1. INTRODUCTION
INTRODUCTION A
A lot
lot of models describing separately the fermentative
1. INTRODUCTION A lot of
hydrogen
hydrogen of models
production
models
production
describing
(see e. g.
describing
(see e.
separately
Nasr et al.
separately
g. Nasr et
the
al.
fermentative
the(2013), Wang
fermentative
(2013), Wang
Anaerobic digestion (AD) is a multi-step biotechnolog- hydrogen
et al.
hydrogen production
(2009)) and
production the (see
AD
(see e. g.
for
e.for Nasr
methane
g. Nasr et al.
et al. (2013),
(2013), Wang
production (see
Wang
Anaerobic
Anaerobic digestion (AD) is aahydrogen
multi-step biotechnolog- et
et al.
al. (2009))
(2009)) and
and the
the AD
AD for methane
methane production
production (see
(see
ical processdigestion
Anaerobic (Figure (AD)
digestion 1) with
(AD) is multi-step
is ahydrogen
multi-step (H2biotechnolog-
) as a non-
biotechnolog- e.
et
e.
g. Batstone
al. Batstone
g. (2009)) and et
et
al. (2002),
the(2002),
al.
Dochain
AD forDochain (2001),
methane(2001),
productionSimeonov
(see
Simeonov
ical
ical process
process (Figure
(Figure 1)
1) with
with hydrogen (H
(H )) g.
as
as aa non- e. g. Batstone et al. (2002), Dochain (2001), Simeonov
a non- (2010)) are known. However,
(2002), models of (2001),
the two-stage AD
accumulating intermediate product (see 2
e.
ical process (Figure 1) withproduct hydrogen (He.2
2 ) g.as Ahringnon- e. g. Batstone
(2010)) are et al.
known. However, Dochain
models of the Simeonov
two-stage
accumulating
accumulating
(2003), Deublein
accumulating
intermediate
intermediate product
(2008)). Recently,
intermediate product
(see
(see
the
(see e.
interest
e. g.
g.
Ahring
Ahring
in H2
Ahring
(2010))
process
(2010)) are
are known.
producing
known. However,
simultaneously
However, models
models H2of
of the
and
the CH4 are AD
two-stage
two-stage AD
not
AD
(2003), Deublein (2008)). Recently, the interest in H22 process
process producing
producing simultaneously H and CH are not
producing simultaneously H2 and CH4 are not
simultaneously H and CH are
(2003), 2 4
(2003), Deublein (2008)). Recently,
known the interest in
in H known.
production through AD, also as dark fermenta- process 2 4 not
production Deublein
through (2008)).
AD, Recently,
also known the
as interest
dark fermenta- H2 known.
known.
production
tive
production through
H2 production,
through AD,
has
AD, also
increased
also known
known(cf. as
Guo
as dark
et al.
dark fermenta-
(2010),
fermenta- known.
tive H22 production, has increased (cf. Guo In this paper we present a new mathematical model de-
tive
tive HH2 production,
Pakarinen has
has increased
et al. (2011)). This is due (cf.
(cf.to the et
Guo et
fact al.
al. (2010),
(2010),
al.that the In
In this
this paper we present aa new mathematical
H2 and CHmodel de-
Pakarinen
Pakarinen
production,
et
et al.
al.
rates of H2etproduction (2011)).
(2011)).
increased
This
This
are is
is
rather due
due to
to
high
Guo
the
the
et
fact
fact
andfact
(2010),
that
that
a variety the
the
of In this paper
scribing
scribing paper
the
we
we present
the process
process present
of a new
new mathematical
of simultaneous
simultaneous mathematical
H and CH
model
4 produc-
model de-
de-
produc-
Pakarinen
rates of H al. (2011)).
production This
are is
rather due to
high the
and a that
variety the
of scribing
tion by
scribing the
AD
the ofprocess
of organic
process of simultaneous
wastes
of wastes
simultaneousin a H 2
cascade
2 and
H2 and of CH
two
CH
4
4 produc-
continu-
produc-
rates
feedstockof H 2
can
2 production
be used asarea rather
substrate. highIn and a variety
traditional AD, of tion
tion by AD organic in a cascade of two 4 continu-
rates
feedstock
feedstock
of H 2
can
can
production
be
be used
used as
as
area
a
rather
substrate.
substrate.
highIn
In
and a variety
traditional
traditional AD,
AD,
of ouslyby
tion by AD
AD of
stirred organic
organic wastes
ofbioreactors. wastes in
in aa cascade
The biochemical of
of two
two continu-
cascade processes in the
continu-
H is not detected as it is consumed immediately e. g. by ously
ously stirred
stirred bioreactors.
bioreactors. The
The biochemical
biochemical processes
processes in
in the
feedstock can be used as a substrate. In traditional AD, in the
2 first bioreactor include disintegration of organic wastes,
H
H 2 is not detectedmethanogens
as it is consumed immediately e. g. by ously stirred bioreactors. The biochemical processes the
H22 is
is not
not detected
hydrogenotrophic as
as it
it is
detectedmethanogens
hydrogenotrophic is consumed
consumed to
immediately
to produce methane
immediately
produce methane
e.
e. g.CH
g.
CH
by
by4 first
first bioreactor
bioreactor
hydrolysis
first
include
include
and include
bioreactor acidogenesis disintegration
disintegration
with hydrogen
disintegration
of
of
of
organic
organic
organic
wastes,
wastes,
production.
wastes,
hydrogenotrophic methanogens to produce methane CH the444 hydrolysis and acidogenesis
and carbon dioxide
hydrogenotrophic
and
CO2 (cf. Gerardi
methanogens to produce (2003)).
methane OnCH hydrolysis
The methane and acidogenesis
production fromwith with
acetatehydrogen
hydrogen production.
production.
(methanogenesis) is
othercarbon
and
and hand, Hdioxide
carbon
carbon dioxide
2 can beCO
dioxide
CO
COproduced(cf. Gerardi
2 (cf.
2 Gerardi
separately
2 (cf. Gerardi
(2003)).
(2003)). On
On
On the
by engineering
(2003)).
the
the
hydrolysis
The
The methane
methane
separated
and
in the
acidogenesis
production
production from
from
second bioreactor.
with
acetate
acetate
hydrogen production.
(methanogenesis)
(methanogenesis)
The(methanogenesis)
proposed model is is
is
other
other
the hand,
hand,
process H
H 2 can
2 can
be
be
conditions. produced
produced separately
However,separately
the mainby by engineering
engineering
limitation of The methane
separated in production
the second from
bioreactor. acetate
The proposed model is
is
other
the hand,
process H 2 can be
conditions. produced
However, istheseparately
the main by engineering
limitation of separated
developed
separated in
by
in the
the second
reducing
second bioreactor.
the well
bioreactor. knownThe
The proposed
Anaerobic
proposed model
Digester
model is
is
the
dark process conditions.
fermentative H However,
production the main
rather limitation
low energy of developed
developed by reducing the the well known Anaerobic
Anaerobic Digester
the
dark
dark
process conditions.
fermentative
fermentative H
H
2
2
However, isthe
production
production is the
the
main
rather
rather
limitation
low
low energy
energy
of Model
developed No byby reducing
1 (ADM1)
reducing the basic well
well known
structure elaboratedDigester
known Anaerobic by the
Digester
recovery.
dark In order H
fermentative to2 completely
2 production utilize the organic
is the rather low energyacids Model
Model No
No
International 1
1 (ADM1)
(ADM1)
Water basic
basic
Association structure
structure elaborated
elaborated
(IWA), elaborated
see Batstone by
by the
the
recovery.
recovery. In
In order
order to
to completely utilize the organic acids Model No 1 (ADM1) basic structure byet al.
the
produced
recovery.
produced
during
In
during to completely
orderdark
dark
fermentation
completely
fermentation
utilize
utilize
and
the
the organic
and improve organic
improve
the acids
the
over-
acids
over-
International
International
(2002).
International
Water
Water
Water
Association
Association
Association
(IWA),
(IWA),
(IWA),
see
see
see
Batstone
Batstone
Batstone
et
et
et
al.
al.
al.
produced
all energy
produced during
during dark
conversion fermentation
efficiency,
dark fermentation a and improve
two-stage
and improve AD the over-
concept
the over- (2002).
(2002).
all
all energy
energy
consisting conversion
conversion
of hydrogenic efficiency,
efficiency,
process a
a two-stage
two-stage
followed by AD
AD concept
concept
methanogenic (2002).
The paper is organized in the following way. Section 2
all energy conversion efficiency, a two-stage AD concept The paper is organized in the following
consisting
consisting
process has
consisting
of hydrogenic
of hydrogenic
been suggested
of hydrogenic
process
process
process (seefollowed
followed by methanogenic
by methanogenic
e. g. Pakarinen
followed by methanogenic et al. The paper
is devoted
The paper to is organized
is process
organized in
in the
description. Section way.
the following
following way. Section
Section
3 presents
way. Sectionthe
2
2
2
process
process has
has been
been suggested
suggested (see
(see e.
e. g.
g. Pakarinen
Pakarinen et
et al. is
is devoted
devoted to
to process
process description.
description. Section
Section 3
3 presents
presents the
the
(2011)).
process
(2011)). has been suggested (see e. g. Pakarinen et al.
al. mathematical
is devoted
mathematical to model
process
model
of the
of the
dynamic
description.
the dynamic
processes
Section
processes 3 in the
presents
inare
two
the
thecom-
two
(2011)). mathematical
bioreactors.
mathematical Themodel of
of the dynamic
equilibrium
model points ofprocesses
dynamic the modelin
processes the
inare two
thecom-
two
(2011)). bioreactors. The equilibrium points of the model
 This research has been partially supported by the Bulgarian bioreactors.
puted in Section
bioreactors. The equilibrium
4 and their points
The equilibrium points of the
local stability model are com-
is investigated
of the model are com-

 This research has been puted in Section 4 and their local stability is investigated
been partially supported
supported by the Bulgarian puted in
 This
Science research
Fund under
This research
Science Fund
has contract
partially
has contract
under
No. DFNI–E02/13.
been partially
No. supported by
DFNI–E02/13.
by
The
The
thework
Bulgarian
thework of the
Bulgarian
of
in
in the in Section
the next
puted next
Section
Section
Section
44 5.
and
and
5.
their
their local
Section
Section local
66 is
stability
6 is devoted
stability
devoted
tois
to
investigated
isinvestigations
investigated
investigations
Science
second Fund
author under
has contract
been No.
partially DFNI–E02/13.
supported by The
Sofia work of the
the in the next Section 5. Section is devoted to investigations
Science
second Fund under
author has contract No.supported
DFNI–E02/13. The work of “St
University the of
in input-output
the next Section static
5. characteristics
Section 6 is with
devoted to respect to H2
investigations
second has been
author under been partially
partially supported by
by Sofia
Sofia University
University “St
“St of
of input-output static characteristics with respect to H
of input-output static characteristics with respect to
input-output static characteristics with respect to H
Kl. Ohridski” contract No. 58/06.04.2016. 2
second author under
Kl. Ohridski”
Kl. Ohridski” has been partially
contract
under contract No. supported
No.
by Sofia University “St
58/06.04.2016.
58/06.04.2016. H22
Kl. Ohridski” under contract No. 58/06.04.2016.
Copyright
2405-8963 © © 2016,
2016 IFAC
IFAC (International Federation of Automatic Control) 1 Hosting by Elsevier Ltd. All rights reserved.
Copyright
Copyright © 2016 IFAC 1
Peer review©
Copyright
2016 responsibility
©under
IFAC
2016 IFAC
1 Control.
of International Federation of Automatic
1
10.1016/j.ifacol.2016.12.131
2016 IFAC FOSBE
October 9-12, 2016. Magdeburg, Germany Milen Borisov et al. / IFAC-PapersOnLine 49-26 (2016) 231–238
232

Assume that the volumes V1 and V2 of the bioreactors are

constant. Let F1 and F2 be the inflows in the first and
second bioreactor respectively and let F1 = F2 = F be
valid. It is well known that the dilution rates D1 and D2
are defined as
F F
D1 = , D2 = .
V1 V2
Then
V1 V1
D2 = D1 := γD1 within γ := . (1)
V2 V2
It is known that the volume V2 of the second bioreactor
for methane production is larger than the volume V1 of the
first bioreactor. Therefore, γ < 1 should be valid. Later on
in the paper we shall determine the constant γ using the
proposed model equations.
Fig. 1. Biochemical scheme of the AD

and CH4 production; in particular existence of maxima is 3. MODEL DESCRIPTION

shown, which is important for the applications. Numerical
simulations are presented in Section 7 as illustration of
the theoretical results. The last Section 8 presents some
concluding remarks and ideas for future studies.
2. PROCESS DESCRIPTION
The application of a two-stage AD process for sequen-
tial H2 and CH4 production (Figure 2) has been pro-
posed as a promising technology for better process per-
formance and higher energy yields as compared to the
traditional one-stage CH4 production process. In the two-
stage AD system, relatively fast growing acidogens and
H2 producing microorganisms are developed in the first-
stage hydrogenic bioreactor (with working volume V1 )
and are involved in the production of volatile fatty acids
(VFA) and H2 . On the other hand, the slow growing
acetogens and methanogens are developed in the second-
stage methanogenic bioreactor (with working volume V2 )
in which the produced VFA are further converted to CH4
and CO2 .
It is known that in the two-stage H2 + CH4 system the
energy yields are up to 43% more, as compared to the tra-
ditional one-stage CH4 production process (cf. Schievano
et al. (2014)).
In this section we present the mathematical model describ-
ing the process of simultaneous H2 and CH4 production
by AD of organic wastes in a cascade of two continuously
stirred bioreactors. The model is derived on the basis of the
ADM1 basic structure as well as on our experience with
the two-phase AD process with hydrogen and methane
production, see Denchev et al. (2016). The following as-
sumptions have been accepted:
• The slow acetogenesis phase has been omitted and
consequently in the second bioreactor only methano-
genesis occurs.
• Balance equations of the hydrogen and methane in
the liquid phases have been neglected because they
are practically not dissolved in liquids.
• Hydrogenotrophic bacteria do not exist in this pro-
cess.
• Equations describing balances of inorganic compo-
nents and some biochemical equations have been ne-
glected in view of simplifying the model.
• The very important parameter pH is not included in
the model, however pH is kept in the interval 5.0−5.5
in the first bioreactor, and in the interval 6.5 − 8.5 in
the second bioreactor.
Following the above assumptions, the dynamics in the first
bioreactor is described by the following set of 10 nonlinear
ordinary differential equations

d in
Ssu (t) = D1 (Ssu − Ssu ) + khyd,ch Xch
dt
+ fsu,li khyd,li Xli − µsuaa,su Xsuaa (2)
d in
Saa (t) = D1 (Saa − Saa ) + khyd,pr Xpr
dt
− µsuaa,aa Xsuaa (3)
d
Sf a (t) = D1 (Sfina −Sf a )+ff a,li khyd,li Xli −µf a Xf a (4)
dt
d
Sac (t) = −D1 Sac + (1 − Ysuaa )fac,su µsuaa,su Xsuaa
dt
+ (1 − Ysuaa )fac,aa µsuaa,aa Xsuaa
Fig. 2. Two-phases process of AD with production of + 0.7 (1 − Yf a )µf a Xf a (5)
hydrogen (H2 ) and methane (CH4 )

2
 2016 IFAC FOSBE
October 9-12, 2016. Magdeburg, Germany Milen Borisov et al. / IFAC-PapersOnLine 49-26 (2016) 231–238 233

d Table 1.
Xc (t) = D1 (Xcin − Xc ) − kdis Xc (6)
dt Definition of the model variables
d Ssu concentration of monosacharides [gCOD/L]
Xch (t) = −(D1 + khyd,ch )Xch + fch,xc kdis Xc (7)
dt Saa concentration of amino acids (AA) [gCOD/L]
d Sf a concentration of fatty acids (LCFA) [gCOD/L]
Xpr (t) = −(D1 + khyd,pr )Xpr + fpr,xc kdis Xc (8) Sac concentration of total acetate [gCOD/L]
dt Sac,ch4 concentration of total acetate [gCOD/L]
d Xc concentration of composites [gCOD/L]
Xli (t) = −(D1 + khyd,li )Xli + fli,xc kdis Xc (9)
dt Xch concentration of carbohydrates [gCOD/L]
d Xpr concentration of proteins [gCOD/L]
Xsuaa (t) = (−D1 + Ysuaa µsuaa,su )Xsuaa Xli concentration of lipids [gCOD/L]
dt Xsuaa concentration of sugar and AA degraders [g/L]
+ Ysuaa µsuaa,aa Xsuaa (10) Xf a concentration of LCFA degraders [g/L]
d Xac concentration of acetate degraders [g/L]
Xf a (t) = −(D1 − Yf a µf a )Xf a , (11) D1 dilution rate in the first bioreactor [h−1 ]
dt D2 dilution rate in the second bioreactor [h−1 ]
with gaseous output
Table 2.
Qh2 = (Yh2,su µsuaa,su + Yh2,aa µsuaa,aa )Xsuaa Definition of the model parameters Values
in
Ssu input concentration of Ssu [gCOD/L] 0.01
+ Yh2,f a µf a Xf a , in
Saa input concentration of Saa [gCOD/L] 0.001
Sfina input concentration of Sf a [gCOD/L] 0.001
where Qh2 is the hydrogen flow rate. In (2)–(11) the Xcin input concentration of Xc [gCOD/L] 50
biomass specific growth rates are fch,xc stoichiometric parameter [–] 0.2
fpr,xc stoichiometric parameter [–] 0.2
µsuaa,su = µsuaa,su (Ssu , Saa ) fli,xc stoichiometric parameter [–] 0.3
fsu,li stoichiometric parameter [–] 0.05
Ssu Ssu ff a,li stoichiometric parameter [–] 0.95
= km,suaa · ,
ks,suaa + Ssu Ssu + Saa fac,su stoichiometric parameter [–] 0.41
µsuaa,aa = µsuaa,aa (Ssu , Saa ) fac,aa stoichiometric parameter [–] 0.4
Ysuaa stoichiometric parameter [–] 0.1
Saa Saa Yac stoichiometric parameter [–] 27.3
= km,suaa · ,
ks,suaa + Saa Ssu + Saa Yf a stoichiometric parameter [–] 0.06
Yh2,su physicochemical parameter [L2 /g] 0.7
Sf a Yh2,aa physicochemical parameter [L2 /g] 0.7
µf a = µf a (Sf a ) = km,f a .
ks,f a + Sf a Yh2,f a physicochemical parameter [L2 /g] 0.7
Ych4,ac physicochemical parameter [L2 /g] 75
We extend the model (2)–(11) by the following two equa- kdis biochemical parameter [h−1 ] 0.0208
tions describing the process in the second bioreactor khyd,ch biochemical parameter [h−1 ] 0.417
khyd,pr biochemical parameter [h−1 ] 0.417
khyd,li biochemical parameter [h−1 ] 0.417
d km,suaa biochemical parameter [h−1 ] 1.25
Sac,ch4 (t) = D2 (Sac − Sac,ch4 ) − Yac µac,ch4 Xac (12)
dt ks,suaa biochemical parameter [g/L] 0.5
km,f a biochemical parameter [h−1 ] 0.15
d km,ac biochemical parameter [h−1 ] 0.0167
Xac (t) = (−D2 + µac,ch4 )Xac (13)
dt ks,f a biochemical parameter [g/L] 0.67
with gaseous output ks,ac biochemical parameter [g/L] 0.4

Qch4 = Ych4,ac µac,ch4 Xac
where Qch4 is the methane flow rate, and
Sac,ch4
µac,ch4 = µac,ch4 (Sac,ch4 ) = km,ac
ks,ac + Sac,ch4
Using equation (6) we find the equilibrium component X̄c
with respect to the phase variable Xc :
D1 Xcin
. X̄c = . (14)
D1 + kdis

The definitions of the phase variables and parameters in Further, equations (7), (8) and (9) imply the equilibrium
the equations (2)–(13) are given in Table 1 and Table 2. components X̄ch , X̄pr and X̄li respectively:
The numerical coefficient values in the right most column
of Table 2 are taken from Rosen et al. (2006). In the fch,xc kdis X̄c
model variables and parameters the subscripts h2 and ch4 X̄ch = ;
D1 + khyd,ch
indicate hydrogen (H2 ) and methane (CH4 ) respectively.
fpr,xc kdis X̄c
X̄pr = ;
D1 + khyd,pr
4. EQUILIBRIUM POINTS OF THE MODEL
fli,xc kdis X̄c
X̄li = .
The equilibrium points of the model are solutions of D1 + khyd,li
the algebraic equations obtained from (2)–(13) by setting
the right-hand sides equal to zero. We are looking for Equations (4) and (11) are used to determine the steady
nonnegative solution due to obvious reasons. state components S̄f a and X̄f a with respect to the phase

3
2016 IFAC FOSBE
October 9-12, 2016. Magdeburg, Germany Milen Borisov et al. / IFAC-PapersOnLine 49-26 (2016) 231–238
234

variables Sf a and Xf a . Obviously, equation (11) suggests
the trivial solution Xf a = 0 which always exists, and
1
X̄f a = 0 ⇐⇒ S̄f a = Sfina + ff a,li khyd,li X̄li . (15)
D1
Assuming that Xf a = 0 we obtain from (11) the steady
state component S̄f a equal to
D1 ks,f a
S̄f a = . (16)
Yf a km,f a − D1
It is clear that S̄f a exists and S̄f a > 0 if and only if
D1 < Yf a km,f a .
Equation (4) then delivers the equilibrium component for
Xf a ,
D1 (Sfina − S̄f a ) + ff a,li khyd,li X̄li
X̄f a =
µf a (S̄f a )
Obviously, X̄f a from (18) satisfies X̄f a ≥ 0 if and only if
1 + (1 − Ysuaa )fac,aa µsuaa,aa (S̄su , S̄aa )X̄suaa (27)
S̄f a ≤ Sfina + ff a,li khyd,li X̄li .
D1
If equality holds true in (19) then X̄f a becomes equal
to zero, i. e. relation (15) is fulfilled. This means that
(1)
there exists a value D1 of the parameter D1 where
a (transcritical) bifurcation of the steady states occurs,
leading to wash-out of Xf a , i. e. X̄f a = 0 for all D1 > D1 .
(1)
This value D1 can be computed using (15) and (16).
Further, equations (2), (3) and (10) are used to compute
the steady state components S̄su , S̄aa and X̄suaa in case
positive solutions do exist.
It is clear from equation (10) that there are two possibil-
ities for the steady state component Xsuaa : Xsuaa = 0,
which always exists, and Xsuaa = 0.
Case 1. X̄suaa = 0
Equations (2), (3) and (5) deliver the following steady
states components related to Ssu , Saa and Sac respectively:
 in
D1 (Ssu − Ssu ) + khyd,ch X̄ch µsuaa,aa (Ssu , Saa )
+fsu,li khyd,li X̄li µsuaa,aa (Ssu , Saa )
 
in
− D1 (Saa − Saa ) + khyd,pr X̄pr µsuaa,su (Ssu , Saa ) = 0
D1 − Ysuaa [µsuaa,su (Ssu , Saa ) + µsuaa,aa (Ssu , Saa )] = 0.
(26)
If we can determine solutions S̄su > 0 and S̄aa > 0,
with S̄aa satisfying (25) then (24) will deliver X̄suaa > 0.
Comparing (21) and (25) it follows that there exists a value
(2) (2)
D1 such that for D1 = D1 a transcritical bifurcation
(17) of the steady states occurs leading to wash-out of Xsuaa ,
(2)
i. e. X̄suaa = 0 for all D1 > D1 .
Equation (5) is used to find the steady state component
. (18)
1 
S̄ac = (1 − Ysuaa )fac,su µsuaa,su (S̄su , S̄aa )X̄suaa
D1
(19) 
+ 0.7(1 − Yf a )µf a (S̄f a )X̄f a .
Obviously, X̄f a = 0 and X̄suaa = 0 lead to S̄ac = 0 for
some values of D1 .
(1)
The above calculations deliver the following sets of equi-
librium points of the model (2)–(11), where the zero com-
ponents are written explicitly, the other components are
positive:
{E00 (D1 )} = (S̄su , S̄aa , S̄f a , S̄ac = 0, X̄c , X̄ch , X̄pr , X̄li ,
(1) (2)
X̄suaa = 0, X̄f a = 0), if D1 > max{D1 , D1 };
{E01 (D1 )} = (S̄su , S̄aa , S̄f a , S̄ac , X̄c , X̄ch , X̄pr , X̄li ,
(1) (2)
X̄suaa = 0, X̄f a ), if D1 > D1 > D1 ;
{E10 (D1 )} = (S̄su , S̄aa , S̄f a , S̄ac , X̄c , X̄ch , X̄pr , X̄li ,
(2) (1)
X̄suaa , X̄f a = 0), if D1 > D1 > D1 ;

1   {E1 (D1 )} = (S̄su , S̄aa , S̄f a , S̄ac , X̄c , X̄ch , X̄pr , X̄li ,
in
S̄su = Ssu + khyd,ch X̄ch + fsu,li khyd,li X̄li , (20) X̄suaa , X̄f a ) if D1 < min{D1 , D1 }.
(1) (2)
D1
in 1
S̄aa = Saa + khyd,pr X̄pr , (21) All steady state components of {E00 (D1 )}, {E01 (D1 )},
D1
{E10 (D1 )}, and {E1 (D1 )} are parameterized on D1 . The
1 existence of either set of equilibrium points depends on the
S̄ac = 0.7(1 − Yf a )µf a (S̄f a )X̄f a . (22)
D1 (1)
relationship between the values D1 and D1 .
(2)

Case 2. X̄suaa = 0 Finally, equations (12) and (13) deliver the last equilibrium
components with respect to Xac and Sac,ch4 . Thereby we
In this case equation (10) reduces to assume that S̄ac is known, see (27).
D1 − Ysuaa [µsuaa,su (Ssu , Saa ) + µsuaa,aa (Ssu , Saa )] = 0.
(23) Equation (13) has the obvious solution Xac = 0 which is
Using equation (3) we express not of practical interest. Assuming that Xac = 0 equation
in (13) implies
D1 (Saa − Saa ) + khyd,pr X̄pr µac,ch4 (Sac,ch4 ) = D2 ,
Xsuaa = . (24)
µsuaa,aa (Ssu , Saa )
The equilibrium component Xsuaa should be positive and which gives the following steady state component for
this will be fulfilled if Sac,ch4
1 ks,ac D2
Saa < Saain
+ khyd,pr X̄pr . (25) S̄ac,ch4 = . (28)
D1 km,ac − D2
Substituting Xsuaa from (24) into equation (4) and using Then the corresponding equilibrium component X̄ac is
equation (23) we obtain the following nonlinear algebraic 1
system X̄ac = (S̄ac − S̄ac,ch4 ). (29)
Yac

4
 2016 IFAC FOSBE
October 9-12, 2016. Magdeburg, Germany Milen Borisov et al. / IFAC-PapersOnLine 49-26 (2016) 231–238 235

Obviously, the above steady state components are positive P1 (D1 ; λ) = (−D1 −λ)(−D1 −kdis −λ)(−D1 −khyd,li −λ)
if the following conditions are satisfied:
×(−D1 − khyd,pr − λ)(−D1 − khyd,li − λ)
km,ac > D2 , S̄ac > S̄ac,ch4 . (30)    
2 dµf a dµf a
The second inequality in (30) gives a relationship between × λ + D1 + (S̄f a )X̄f a λ + D1 (S̄f a )
dSf a dSf a
D1 and D2 , because S̄ac depends on D1 , see (27).
×Q(D1 ; λ);
It follows from (22) that the cases when X̄suaa = 0 and
X̄f a = 0 are simultaneously fulfilled, lead to S̄ac = 0; P10 (D1 ; λ) = (−D1 − λ)2 (−D1 − kdis − λ)
this is undesirable because it will lead to wash-out of ×(−D1 − khyd,li − λ)(−D1 − khyd,pr − λ)
the biomass Xac in the second bioreactor and thus of
breaking-down the methane production. Therefore, the set ×(−D1 − khyd,li − λ)(−D1 + Yf a µf a (S̄f a ) − λ)Q(D1 ; λ),
of equilibrium points {E00 (D1 )} will not be considered in
the following. where Q(D1 ; λ) is the characteristic polynomial of the
sub-matrix ∆ = (∆ij )3i,j=1 evaluated at the steady state
Assume that the steady state components X̄ac > 0 and
components (S̄su , S̄aa , X̄suaa ), within:
S̄ac,ch4 > 0 do exist. The extended model (2)–(13) pos-
sesses the following sets of equilibrium points, which we de- ∂µsuaa,su
note by {F01 (D1 , D2 )}, {F10 (D1 , D2 )} and {F1 (D1 , D2 )} ∆11 = −D1 − X̄suaa
∂Ssu
and present in the form
∂µsuaa,su
 ∆12 = − X̄suaa
∂Saa
F01 (D1 , D2 ) = {E01 (D1 )} {S̄ac,ch4 , X̄ac },
∆13 = −µsuaa,su

F10 (D1 , D2 ) = {E10 (D1 )} {S̄ac,ch4 , X̄ac }, ∂µsuaa,aa
 ∆21 = − X̄suaa
∂Ssu
F1 (D1 , D2 ) = {E1 (D1 )} {S̄ac,ch4 , X̄ac }.
∂µsuaa,aa
∆22 = −D1 − X̄suaa
∂Saa
5. LOCAL STABILITY OF THE EQUILIBRIUM ∆23 = −µsuaa,aa
 
POINTS ∂µsuaa,su ∂µsuaa,aa
∆31 = Ysuaa X̄suaa +
∂Ssu ∂Ssu
 
Using the numerical coefficient values in Table 2, numerical ∂µsuaa,su ∂µsuaa,aa
simulations suggest the following bifurcation values D1
(1) ∆32 = Ysuaa X̄suaa +
∂Saa ∂Saa
(2)
and D1 of the input D1 : ∆33 = 0.
(1) (2)
D1 ≈ 0.00843, D1 ≈ 0.095, The linear quantities in P1 (D1 ; λ) = 0 and P10 (D1 ; λ) = 0
(1) (2) produce negative zeroes, i. e. negative real eigenvalues of
i. e. D1 < D1 . This means that the steady states
the Jacobians. Further, in the first polynomial P1 (D1 ; λ),
{E01 (D1 )} and thus {F01 (D1 , D2 )} never exist. The in-
the quadratic equation
teresting (from practical point of view) steady states are  
{E10 (D1 )} and {E1 (D1 )}, and thus {F10 (D1 , D2 )} and 2 dµf a dµf a
λ + D1 + (S̄f a )X̄f a λ + D1 (S̄f a ) = 0
{F1 (D1 , D2 )}. First we shall investigate the local asymp- dSf a dSf a
totic stability of {E1 (D1 )} and {E10 (D1 )} for D1 ∈ possesses two real negative roots. The multiplier −D1 +
(1) (1) (2) Yf a µf a (Sf a ) − λ in the second polynomial P10 (D1 ; λ)
(0, D1 ) and D1 ∈ (D1 , D1 ) respectively.
delivers a negative root λ = −D1 + Yf a µf a (S̄f a ), because
It is well known that an equilibrium point is locally in this case D1 > Yf a µf a (S̄f a ) holds true.
asymptotically stable, if all eigenvalues of the Jacobian
matrix evaluated at this equilibrium have negative real It remains to show that Q(D1 ; λ) = 0 has negative
part, cf. e. g. Wiggins (1990). roots. This is demonstrated numerically below. By means
of the coefficient values in Table 2, we compute the
Denote by J1 (D1 ) and J10 (D1 ) the Jacobian of the right- equilibrium points {E1 (D1 )} and {E10 (D1 )} as functions
hand side functions of (2)–(11) evaluated at the equilib- of the parameter D1 . Then we construct a mesh of points
rium points {E1 (D1 )} and {E10 (D1 )} respectively, and let with respect to D1 and compute the eigenvalues of ∆. The
eigenvalues are presented graphically on Figures 3 and 4,
P1 (D1 ; λ) = det(J1 (D1 ) − λI) and and are obviously negative.
(1)
P10 (D1 ; λ) = det(J10 (D1 ) − λI) Therefore, the steady states {E1 (D1 )}, D1 ∈ (0, D1 ),
are locally asymptotically stable. After the transcritical
be their characteristic polynomials; in the above presenta- (1)
bifurcation at D1 = D1 the steady states {E1 (D1 )} do
tion I means the (10×10)–identity matrix. The eigenvalues not exist, but the equilibria {E10 (D1 )} become locally
of J1 (D1 ) and J10 (D1 ) coincide with the roots of the (1) (2)
stable for D1 ∈ (D1 , D1 ).
characteristic polynomials P1 (D1 ; λ) and P10 (D1 ; λ).
It is not difficult to see that The equilibrium points of the model (12)–(13) are known
to be locally asymptotically stable for any admissible

5
2016 IFAC FOSBE
October 9-12, 2016. Magdeburg, Germany Milen Borisov et al. / IFAC-PapersOnLine 49-26 (2016) 231–238
236

Qh2 = Qh2 (D1 ) = Yh2,su µsuaa,su (S̄su , S̄aa )X̄suaa

+ Yh2,aa µsuaa,aa (S̄su , S̄aa )X̄suaa
+ Yh2,f a µf a (X̄f a )X̄f a .
The function Qh2 (D1 ) depends on the input D1 and is
called input-output static characteristic with respect to
the hydrogen production.
The important (from practical point of view) question is
whether the function Qh2 (D1 ) is unimodal with respect
to D1 in the admissible interval for D1 . Qh2 (D1 ) is
called unimodal if there exist a unique (admissible) point
D1,max , such that Qh2 (D1 ) possesses maximum Qh2,max =
Qh2 (D1,max ), Qh2 (D1 ) is strongly increasing if D1 <
D1,max and Qh2 (D1 ) is strongly decreasing if D1 > D1,max .

Fig. 3. Eigenvalues of ∆ evaluated at {E1 (D1 )} as func- Figure 5 presents the graph of the input-output static
tions of D1 . The solid square on the horizontal axis characteristic Qh2 (D1 ) computed over the set {E1 (D1 )} ∪
(2)
denotes D1
(1) {E10 (D1 )}, D1 ∈ (0, D1 ). The function Qh2 (D1 ) pos-
sesses two maxima, the first one is achieved for small values
(1)
of D1 ∈ (0, D1 ) on the set {E1 (D1 )}. As a result of
(1)
bifurcation at D1 = D1 , the steady state component
X̄f a becomes equal to zero. The graph of Qh2 (D1 ) is
(1) (2)
further shown for D1 ∈ (D1 , D1 ) on the set of equilibria
{E10 (D1 )}. As one can see, the absolute maximum of
Qh2 (D1 ) is achieved on this set. We denote the maximum
value of D1 by D1,max ≈ 0.0439 and Qh2,max ≈ 0.169.

Fig. 4. Eigenvalues of ∆ evaluated at {E10 (D1 )} as func-

tions of D1 . The solid squares on the horizontal axis
(1) (2)
denote D1 and D1
Fig. 5. The graph of Qh2 (D1 ) as a function of D1
value of D2 (see e. g. Simeonov (2010)). Indeed, the Using the methane flow rate Qch4 we compute the input-
characteristic equation of the Jacobian matrix of (12)–(13) output static characteristic on the set of all steady states,
evaluated at the equilibrium point (S̄ac,ch4 , X̄ac ), has the namely:
form
Qch4 = Qch4 (D1 , D2 ) = Ych4,ac µac,ch4 (S̄ac,ch4 ) X̄ac .
λ2 − A λ + B = 0
The latter is parameterized on the two inputs D1 and D2
within (or equivalently, on D1 and γ, see (1)), since X̄ac depends
on both D2 and D1 . If we use the relation (1), then the
dµac,ch4 dependance is on D1 and γ.
A = −D2 − Yac (S̄ac,ch4 )X̄ac < 0,
dSac,ch4
With D1 = D1,max ≈ 0.0439 we compute the equilibrium
dµac,ch4
B = Yac µac,ch4 (S̄ac,ch4 ) (S̄ac,ch4 )X̄ac > 0. components S̄ac,ch4 and X̄ac according to (28) and (29),
dSac,ch4 as well as Qch4 , all of them as functions of D2 . An upper
The signs of A and B imply that the above quadratics pos- bound for the admissible values of D2 , obtained from (30)
sesses two roots with negative real parts, and this proves is
the local stability of the equilibrium points {F10 (D1 , D2 )} D2+ ≈ 0.014.
and {F1 (D1 , D2 )}. At this parameter value, X̄ac becomes equal to zero, or
equivalently we obtain S̄ac = S̄ac,ch4 . Figure 6 visualizes
the graph of Qch4 (D1,max , D2 ); the latter is a unimodal
6. THE INPUT-OUTPUT STATIC function, taking its maximum at D2,max ≈ 0.0099 with
CHARACTERISTICS Qch4,max = Qch4 (D1,max , D2,max ) ≈ 0.0388.
We compute the output Qh2 on the sets of steady states Using the presentation (1) we define and compute the
{E1 (D1 )} and {E10 (D1 )}: constant γ and in this way the relationship between the
volumes V1 and V2 of the two bioreactors:

6
2016 IFAC FOSBE

October 9-12, 2016. Magdeburg, Germany Milen Borisov et al. / IFAC-PapersOnLine 49-26 (2016) 231–238 237

Fig. 6. The graph of Qch4 (D1,max , D2 ) as a function of D2

Fig. 8. Dynamic behaviour of the model solutions with

variable D1 from Table 3

Fig. 7. Time evolution of Qh2 and Qch4 with variable D1

and D2 from Table 3
D2,max
γ= ≈ 0.226 =⇒ V2 ≈ 4.42 V1 . Fig. 9. Dynamic behaviour of the model solutions with
D1,max
variable D1 from Table 3

7. DYNAMIC BEHAVIOUR OF THE MODEL

SOLUTIONS The next Figures 10, 11 and 12 visualize the dynamics of
Qh2 (t), Qch4 (t) as well as of the model solutions, when D1
The dynamic model has been implemented in a web-based and D2 are fixed, D1 = 0.005 [1/h], D2 = 0.00113 [1/h],
simulation software environment, see and Xcin varies taking values from Table 4.
http://platform.sysmoltd.com/.
Figure 7 presents the time evolution of the outputs Qh2 (t)
and Qch4 (t) when D1 and D2 take values from Table 3.

Table 3. Values of D1
Time [h] 0 − 5000 5000− 10000− 15000−
10000 15000 20000
D1 [1/h] 0.005 0.01 0.015 0.02
D2 [1/h] 0.00113 0.00226 0.00339 0.00452

Fig. 10. Time evolution of Qh2 and Qch4 with variable Xcin
Figures 8 and 9 display the model solutions by variable D1 from Table 4
and D2 given in Table 3.

Table 4. Values of Xcin

Time [h] 0 − 5000 5000− 10000− 15000−
10000 15000 20000
8. CONCLUSION
Xcin [gCOD/L] 40 50 75 60
In this paper we propose a new mathematical model de-
scribing the process of simultaneous production of hy-

7
2016 IFAC FOSBE
238
October 9-12, 2016. Magdeburg, Germany Milen Borisov et al. / IFAC-PapersOnLine 49-26 (2016) 231–238
Fig. 11. Dynamic behavior of the model solutions with
variable Xcin from Table 4
Fig. 12. Dynamic behavior of the model solutions with
variable Xcin from Table 4
drogen and methane by anaerobic digestion of organic
wastes in a cascade of two connected continuously stirred
bioreactors with different volumes. The proposed model is
developed by reducing the universal but very complicated
Anaerobic Digester Model No 1 (ADM1). We compute
the equilibrium points of the model and study their local
stability with respect to the practically important input
parameters D1 and D2 . The investigation of the input-
output static characteristics Qh2 and Qch4 shows existence
of maxima with respect to hydrogen and methane. This
fact is important for the practical applications and will
be further used in optimising the bioreactors to achieve
maximal production of either hydrogen and methane. The
model also allows to find the optimal ratio between the
volumes (V1 and V2 ) of the two bioreactors subject to
the same optimisation goal. The future work envisages
extending the model of the second bioreactor by including
equations describing an acetogenesis phase, where other
VFA (propionate, butyrate, etc.) will be transformed into
acetate, as well as inhibitory kinetics of the microorgan-
isms. The elaborated web-based software will allow us
to check different model-based optimisation and control
strategies. Currently two-phase AD of lignocellulosic waste
with hydrogen and methane production in laboratory con-
ditions are in experimental studies using some theoretical
results obtained with the above presented model.
REFERENCES
Aceves-Lara, C.-A., Latrille, E., and Steyer, J.-P. (2010)
Optimal control of hydrogen production in a continuous
anaerobic fermentation bioreactor. Intern. Journal of
Hydrogen Energy, 35, I0710–I0718.
Ahring, B. (ed.) (2003) Biomethanation, I and II.
Springer-Verlag, Berlin-Heidelberg.
Batstone, D. J., Keller, J., Angelidaki, I., Kalyuzhnyi,
S. V., Pavlostathis, S. G., Rozzi, A., Sanders, W. T. M.,
Siegrist, H., and Vavilin, V. A. (2002) The IWA
Anaerobic Digestion Model No. 1 (ADM1). Water
Science and Technology, 45 (10), 65–73.
Denchev, D., Hubenov, V., Simeonov, I., Kabaivanova,
L. (2016) Biohydrogen production from lignocellulosic
waste with anaerobic bacteria. Proc. ICWEE’2016,
Burgas, Bulgaria, 01–03 June 2016, 7–12.
Deublein, D., and Steinhauser, A. (2008) Biogas from
waste and renewable resources. Wiley-VCH Verlag,
Weinheim.
Dochain, D., and Vanrolleghem, P. A. V. (2001) Dynam-
ical modeling and estimation in wasterwater treatement
process. IWA Publishing, London.
Gerardi, M. H. (2003) The microbiology of anaerobic
digesters. John Wiley&Sons, Inc. New Jersey.
Guo, X. M. et al. (2010) Hydrogen production from
agricultural waste by dark fermentation: a review.
Int. J. Hydrogen Energy, 36, 10660–10673.
Nasr, N. et al. (2013) Application of artificial neu-
ral networks for modeling of biohydrogen production.
Int. J. Hydrogen Energy, 38, 3189–3195.
Pakarinen, O. M. et al. (2011) Hydrogen and methane
yields of untreated, wter-extracted and acid (HCl)
treated maize in one- and two-stage batch assays.
Int. J. Hydrogen Energy, 36, 14401–144407.
Rosen, Ch., and Jeppsson, U. (2006) Aspects on ADM1
implementation within the BSM2 framework. CODEN:
LUTEDX, TEIE–7224, 1–35.
Schievano, A., Tenca, A., Lonati, S., Manzini, E., and
Adani, F. (2014) Can two-stage instead of one-stage
anaerobic digestion really increase energy recovery from
biomass?. Applied Energy, 124, 335–342.
Simeonov, I. (2010) Chapter 2. Dynamical modeling
and estimation in wasterwater treatement process. In
St. Tzonkov (ed.), Contemporary Approaches to Mod-
elling, Optimisation and Control of Biotechnological
Processes, Prof. M. Drinov Acad. Publ. House, Sofia.
Simeonov, I., Noykova, N., and Gyllenberg, M. (2007)
Identification and extremum seeking control of the
anaerobic digestion of organic wastes. Cybernetics and
information technologies, 7 (2), 73–84.
Simeonov, I. (1999) Mathematical modeling and parame-
ters estimation of anaerobic fermentation process. Bio-
process Engineering, 21 (4), 377–381.
Wang, J., and Wan, W. (2009) Kinetic models for fermen-
tative hydrogen production: A review. Int. J. Hydrogen
Energy, 34, 3313–3323.
Wang, H.-H., Krstic, M., and Bastin, G. (1999) Optimizing
bioreactors by extremum seeking. Int. J. Adapt. Control
Signal Process, 13, 651–669.
Wiggins, S. (1990) Introduction to applied nonlinear
dynamical systems and chaos. Springer Verlag, New
York.