Anaerobic Digestion Modelling

CHAPTER TWO
Anaerobic Digestion Modelling

Karthik R. Manchala, Yewei Sun, Dian Zhang and Zhi-Wu Wang1
Virginia Tech/Occoquan Laboratory, Manassas, VA, United States
1
Corresponding author: E-mail: wzw@vt.edu
Contents
1. Introduction 71
2. Basic Models Used in Anaerobic Digestion 72
2.1 Description of Anaerobic Digestion Processes 72
2.1.1 Hydrolysis 73
2.1.2 Acidogenesis 74
2.1.3 Acetogenesis 75
2.1.4 Methanogenesis 76
2.2 Kinetic and Equilibrium Models 77
2.2.1 Microbial Growth and Substrate Utilization Kinetics 77
2.2.2 Decay Kinetics 80
2.2.3 Hydrolysis Kinetics 81
2.2.4 AcideBase Equilibrium and Kinetics 82
2.2.5 GaseLiquid Equilibrium 83
2.2.6 Diffusion Kinetics 84
2.2.7 Temperature Effect on Reaction Kinetics and Thermodynamics 84
3. Common Model Frameworks 86
3.1 Rate-Limitation Step 86
3.2 Anaerobic Digestion Model No. 1 86
4. Modelling Major Factors Affecting Anaerobic Digestion 90
4.1 Substrate Composition 91
4.1.1 Substrate 91
4.1.2 Hydrolysis 93
4.1.3 Codigestion 94
4.1.4 Biogas Production Rate 95
4.2 Inhibition 96
4.2.1 Haldane Approach 97
4.2.2 Substrate Utilization Competition 97
4.2.3 Andrews Approach 97
4.2.4 Inhibition due to Toxicity 98
4.2.5 Substrate Inhibition 99
4.2.6 Ammonia Inhibition 100
4.2.7 Hydrogen as a Control Parameter 101
4.2.8 Influence of pH on Microbial Growth 102
4.2.9 Computational Fluid Dynamics Applications 103
Advances in Bioenergy, Volume 2

© 2017 Elsevier Inc.
j
ISSN 2468-0125
http://dx.doi.org/10.1016/bs.aibe.2017.01.001 All rights reserved. 69
70 Karthik R. Manchala et al.
4.3 Forms of Living 105

4.3.1 Modelling Biofilm 105
4.4 Total Solids 109
4.4.1 Theoretical Models 111
4.4.2 Empirical Models 117
4.4.3 Statistical Models 119
5. Model Procedure 120
5.1 Model Concept and Structure Selection 122
5.2 Sensitivity Analysis 122
5.3 Parameter Estimation 123
5.4 Data Collection 125
5.5 Parameter Optimization 125
5.5.1 Local Methods 126
5.5.2 Global Methods 127
5.5.3 Bayesian Approach 127
5.6 Model Validation 128
5.6.1 Direct Validation 128
5.6.2 Cross-Validation 128
5.7 Model Assessment 129
6. Modelling Challenges 131
6.1 Initial Conditions 131
6.2 Complexity 131
6.3 Data Availability 132
6.4 Parameter Interpretation 133
6.5 Modelling Species Unaccounted for 133
7. Conclusions 134
References 135
Abstract
Anaerobic digestion is a well-established technology for stabilization and conversion of
various organic wastes into bioenergy. Typical sources of organic wastes suitable for
anaerobic digestion are mainly from municipal, agricultural and industrial producers.
Because anaerobic digestion is such a complex process, mathematical modelling could
assist in providing better understanding, design, optimization and prediction of the
performance of anaerobic digesters. State-of-the-art models have been developed to
simulate the methane yield improvement, reactor stability enhancement, operational
issue identification and waste codigestion in anaerobic digestion systems. In this
chapter, a summary of existing anaerobic digestion models is presented. These models
were derived for describing substrates characteristics, rate-limiting conditions, process
inhibition, operating conditions, methane potential and production rate, liquidegas
interface mass transfer, as well as the application of computational fluid dynamics. A
discussion with regard to their different assumptions, structures, applications and
limitations is included. Modelling procedure adopted in the course of model structure
selection, parameter estimation and model validation methods is also described.
Anaerobic Digestion Modelling 71
Finally, future work identifying the research needs in modelling anaerobic digestion
is proposed. The intent of the authors of this chapter is to provide a comprehensive
understanding of the fundamental concepts used in anaerobic digestion modelling
and to provide ideas for further research work.
1. INTRODUCTION
Anaerobic Digestion (AD) is a biological process that converts
organics into biogas, in which methane and carbon dioxide are primary
constituents (Herrmann et al., 2016; Loehr, 1968). Methane, if not
captured by anaerobic digesters, creates nearly 25 times greater impact as
a greenhouse gas in comparison to carbon dioxide (Chakraborty et al.,
2011). Biogas can be combusted to generate heat and electricity (Leon
and Martin, 2016). Methane, produced through AD processes, from
different sources such as crop stover, animal waste, food waste, municipal
solid waste (MSW), industrial waste and biosolids, has drawn significant in-
terest across the world in recent years for its energy value and greenhouse
gas effect mitigation potential (Li et al., 2011; Lyberatos and Skiadas, 1999;
Zhang et al., 2014).
Although AD is a well-established process, the optimal design of
anaerobic digesters for maximum methane production is still a challenge
(Mulka et al., 2016). This is due to the substrate variability, microbial
consortia complexity, as well as the complicated biochemical, physical
and chemical interactions involved in the AD processes (Donoso-Bravo
et al., 2011). Mathematical models are useful tools we can leverage to
improve the design and efficiency of AD systems (Gavala et al., 2003). It
is generally accepted that well-developed models should describe
the main aspects of a biological process, help to better understand the
underlying phenomenon and provide accurate prediction of the AD
performance as well as optimize operational parameters (Batstone et al.,
2002a; Yu et al., 2013).
To give an overview on various models that have been broadly used
in many AD studies, this chapter is designated to summarize most of the
existing models in the field of AD in regard to their assumptions, structures,
applications and limitations. A detailed description of systematic modelling
procedure including model structure selection, parameter estimation and
model validation is also included. Finally, future work identifying the
research needs in AD modelling is also proposed.
2. BASIC MODELS USED IN ANAEROBIC DIGESTION

2.1 Description of Anaerobic Digestion Processes
In general, AD of complex organic matter to methane can be regarded
as a sequence of four reaction steps as illustrated in Fig. 1, which include
hydrolysis, acidogenesis, acetogenesis and methanogenesis (Gavala et al.,
2003). Simple stoichiometric equations were developed to estimate the
amount of biogas produced from organic matter. The use of these equations
Complex organic matter
1. Hydrolysis by extracellular enzymes
Carbohydrates Proteins Lipids or fats
Monosaccharides Amino acids Long chain fatty acids
2. Acidogenesis by acidogens
Valerate,
Butyrate
Acetate Propionate
3. Acetogenesis by acetogens Hydrogen
4. Methanogenesis by methanogens Methane,

Carbon
dioxide
Figure 1 Reaction steps in AD, in which Brown dotted line represents the breakdown of
lipids into monosaccharides; Violet dotted line represents breakdown of monosaccharides
to acetate, propionate, butyrate and valerate; Green broken line represents breakdown of
amino acids into acetate, propionate, valerate and butyrate, and hydrogen; Black dotted
line represents the breakdown of the long chain fatty acids into acetate and hydrogen.
was for estimating the theoretical yields of methane and carbon dioxide
(Gerber and Span, 2008). In 1952, Buswell and Mueller (1952) proposed
a general stoichiometric equation (Eq. 1) for AD stoichiometry. The uncer-
tainty in the estimation was reported to be around 5% (Buswell and Mueller,
1952),

b c a b c a b c
Ca Hb Oc þ a H2 O/ þ CH4 þ þ CO2
4 2 2 8 4 2 8 4
(1)
The theoretical methane yield can be determined from Eq. (1) as a molar
ratio of methane to the organic matter expressed in g CH4 (g CaHbOc)1. A
recent study by Nguyen et al. (2014) has used this ‘Buswell’s equation’
(Eq. 1) to develop a model for biogas estimation using food waste. Boyle
(1976) modified Eq. (1) to incorporate nitrogen and sulphur components
in the organic matter into Eq. (2), which allows the final products such as
ammonia and hydrogen sulfide to be estimated in addition to methane
and carbon dioxide as in Eq. (1). Although the feedstock elemental
composition can be easily determined using modern analytical equipment
such as CHNS elemental analyzers, a challenge still remains in using
Eqs. (1) and (2) for theoretical methane yield estimation, i.e., the biodegrad-
able fraction of the feedstock is usually unknown. Hence, Eqs. (1) and (2)
should have better applicability to readily biodegradable carbohydrates
such as glucose, sucrose, lactose and readily hydrolysable proteins such as
casein and albumin (Dahlquist, 2013),

b c 3d e a b c 3d e
Ca Hb Oc Nd Se þ a þ þ H2 O/ þ
4 2 4 2 2 8 4 8 4

a b c 3d e
CH4 þ þ þ þ CO2 þ dNH3 þ eH2 S
2 8 4 8 4
(2)
2.1.1 Hydrolysis
In this first step, particulate material is converted to soluble compounds that
can then be further hydrolyzed to simple monomers used by bacteria
performing fermentation. The chemical composition of those complex
organic matters can be simplified as carbohydrates, lipids and proteins.
Accordingly, carbohydrates are degraded to simple sugars; proteins are
hydrolyzed to amino acids; lipids are broken down to long chain fatty acids
(LCFAs) (Fig. 1). These three parallel hydrolysis processes are carried out by
extracellular enzymes secreted from hydrolytic bacteria. These bacteria
get directly benefitted from these soluble products (Baeyens et al., 2015).
The enzymatic hydrolysis is performed by enzymes secreted by organisms
into the bulk liquid and then adsorbed onto soluble substrate surface.
It can also be performed by enzymes produced by organisms directly
attached on to the insoluble substrate surface in the vicinity. It is generally
believed that organisms growing on the substrate surface play a major role
towards the insoluble substrate hydrolysis (Batstone et al., 2006; Vavilin
et al., 1996; Wang and Chen, 2009). The stoichiometry of carbohydrate
hydrolysis can be described as Eq. (3) (Angelidaki et al., 1999; Yu et al.,
2013),
ðC6 H10 O5 Þis /Yc ðC6 H10 O5 Þs þ ð1 Yc ÞðC6 H10 O5 Þin (3)
in which (C6H10O5)is represents complex carbohydrates (insoluble)
including both biodegradable and inert carbohydrates; (C6H10O5)s
and (C6H10O5)in represents soluble and inert carbohydrates respectively;
Yc represents the fraction of carbohydrates that is biodegradable. The
stoichiometry of lipid hydrolysis can be described as Eq. (4) (Angelidaki
et al., 1999; Yu et al., 2013),
C57 H104 O6 þ 3H2 O/C3 H8 O3 þ 3C18 H34 O2 (4)
in which C57H104O6 represents lipids (glycerol-trioleate); C3H8O3 repre-
sents glycerol trioleate; C18H34O2 represents LCFA. The stoichiometric
yield of glycerol-trioleate (assumed as standard lipid) calculated from Eq. (4)
is 0.104 g lipid (g glycerol-trioleate)1. The stoichiometry of protein
hydrolysis can be described as Eq. (5) (Angelidaki et al., 1999; Yu et al.,
2013),

ðProteinÞis /Yp ðAminoacidsÞ þ 1 Yp ðProteinÞin (5)
in which (Protein)is represents complex protein; (Protein)in represents inert
protein; Yp represents the fraction of the protein that is biodegradable; only
the biodegradable fraction of the protein is hydrolyzed to amino acids and
the inert fraction is unchanged.
2.1.2 Acidogenesis
Acidogenesis, also termed as fermentation, is generally defined as an anaerobic
acideproducing microbial process without an additional electron acceptor
or donor (Gujer and Zehnder, 1983b). In this fermentation process,
substrates serve as both the electron donors and acceptors. The principle
fermentation products from the sugars and amino acids produced from
hydrolysis are volatile fatty acids (VFAs) (e.g., acetate, propionate, butyrate,
etc.), CO2 and hydrogen (Fig. 1). Since the degradation of LCFA is an
oxidation reaction with an external electron acceptor, it is usually included
in acetogenesis to be introduced in the next section. The stoichiometry of
simple carbohydrates conversion to acetate, propionate and butyrate can
be described as Eq. (6) (Yu et al., 2013),
C6 H10 O5 þ 0:1115NH3 /0:1115C5 H7 NO2 þ 0:744C2 H4 O2
þ 0:5C3 H6 O2 þ 0:4409C4 H8 O2 þ 0:6909CO2 þ 0:0254H2 O (6)
in which C6H10O5 represents simple carbohydrates; C5H7NO2 represents
microbial cell formula; C2H4O2 represents acetate; C3H6O2 represents
propionate and C4H8O2 represents butyrate. The stoichiometry of amino
acids (CH2.03O0.6N0.3S0.001) conversion to VFAs can be described as Eq. (7)
(Angelidaki et al., 1999; Yu et al., 2013),
CH2:03 O0:6 N0:3 S0:001 þ 0:3006H2 O/0:017013C5 H7 NO2
þ 0:29742C2 H4 O2 þ 0:02904C3 H6 O2 þ 0:022826C4 H8 O2
þ 0:013202C5 H10 O2 þ 0:07527CO2 þ 0:28298NH3 þ 0:001H2 S
(7)
in which C5H10O2 is valerate. The microorganisms responsible for
acidogenesis consist of obligatory and facultative anaerobic bacteria like
Arachnia propionica and Propionibacterium acidipropionici (Cibis et al., 2016).
The stoichiometry of glycerol conversion to propionate can be described as
Eq. (8) (Yu et al., 2013),
C3 H8 O3 þ 0:04071NH3 þ 0:0291CO2 /0:04071C5 H7 NO2
þ 0:9418C3 H6 O2 þ 1:09305H2 O (8)
in which C3H8O3 is glycerol.
2.1.3 Acetogenesis
Acetogenesis refers to further fermentation of VFAs to acetate, CO2 and
hydrogen, which are the precursors of methane formation (Fig. 1). LCFAs
generated in the hydrolysis step will be decomposed in this step to acetate,
CO2 and hydrogen, as well. The free energy change associated with the
conversion of propionate and butyrate to acetate and hydrogen requires
hydrogen to be at low concentrations in the system (H2 < 104 atm).

Otherwise, the reaction will not proceed (McCarty and Smith, 1986).
Most of the hydrogen produced comes from the oxidation of LCFAs and
intermediate VFAs to acetic acid.
Eqs. (9)e(11) show the stoichiometric transformation of VFAs to acetate
(Angelidaki et al., 1999). The stoichiometry of propionate to acetate
conversion can be described as Eq. (9) (Angelidaki et al., 1999; Yu et al.,
2013),
C3 H6 O2 þ 1:764H2 O þ 0:0458NH3 /0:0458C5 H7 NO2

þ 0:9345C2 H4 O2 þ 2:804H2 þ 0:902CO2 (9)
The stoichiometry of butyrate to acetate conversion can be described as
Eq. (10) (Angelidaki et al., 1999; Yu et al., 2013),
C4 H8 O2 þ 1:7818H2 O þ 0:0544NH3
þ 0:0544CO2 /0:0544C5 H7 NO2 þ 1:8909C2 H4 O2 þ 1:8909H2
(10)
The stoichiometry of valerate to acetate conversion can be described as
C5 H10 O2 þ 0:8045H2 O þ 0:0653NH3

þ 0:5543CO2 /0:0653C5 H7 NO2 þ 0:8912C2 H4 O2
þ 0:02904C3 H6 O2 þ 0:4454CH4 (11)
The stoichiometry of LCFAs to acetate conversion can be described as
C18 H34 O2 þ 15:2398H2 O þ 0:1701NH3
þ 0:25CO2 /0:1701C5 H7 NO2 þ 8:6998C2 H4 O2 þ 14:5H2 (12)
2.1.4 Methanogenesis
Methanogenesis is carried out by a group of Archaea organisms collectively
known as methanogens. There are two groups of methanogens. One group
is termed aceticlastic (or acetotrophic) methanogens, who split acetate
into methane and CO2. The second group is termed hydrogen-utilizing
(or hydrogenotrophic) methanogens, who use hydrogen as the electron
donor and CO2 as the electron acceptor to produce methane (Fig. 1)
(Liu et al., 2016). Hydrogenotrophic methanogens are responsible for

methane generation from hydrogen and CO2 as shown in the stoichiometric
Eq. (13),
2:804H2 þ 0:01618NH3 þ 0:7413CO2 /0:001618C5 H7 NO2
þ 0:6604CH4 þ 1:45H2 O (13)
Acetotrophic methanogens are responsible for the methane and carbon
dioxide generated from acetate as described by the stoichiometric Eq. (14)
(Hill, 1982; Yu et al., 2013),
C2 H4 O2 þ 0:022NH3 /0:022C5 H7 NO2 þ 0:945CH4 þ 0:945CO2
þ 0:066H2 O
(14)
2.2 Kinetic and Equilibrium Models

Kinetic models provide expression for the rate of a reaction that transforms
the substances in AD. The basic kinetic models described in this section
estimate the rates of microbial growth, hydrolysis, acidebase reaction,
mass transfer, mass diffusion, as well as the temperature effect on AD rate.
2.2.1 Microbial Growth and Substrate Utilization Kinetics

Typically, there are five phases of bacterial growth as described in Fig. 2 and
below:
1. The lag phase when there is no increase in microbial growth rate and
acclimatization occurs.
2. Acceleration growth phase when there is a monotony increase in micro-
bial growth rate.
3. Constant growth where there is a constant rate increase.
4. Decelerating growth phase where there is a monotony decrease in
growth rate.
5. Finally, declining phase when the death rate of old cells exceeds the
growth rate of new cells.
Substrate utilization in AD process is through catabolism and anabolism.
In catabolism, substrate is consumed and energy is released in the form
of adenosine triphosphate (ATP). Anabolism is microbial growth due to
consumption of the energy generated from catabolism. The amount of
energy generated per unit of substrate is represented as ATP yield factor.
The product yield and growth yield are defined as the amount of product
formed in terms of DPðg=LÞ; and the amount of biomass generated in terms
2,500
1. Lag phase
2. Accelerating growth phase
3. Constant growth phase
2,000
4. Decelerating growth phase
Colony formation units·mL-1
5. Declining phase
1,500 4 5
1,000
3
500
2
1
-
0 5 10 15 20 25 30 35 40
Incubation time (h)
Figure 2 Phases of bacterial growth (net growth). Adapted from the study by Zhang,
H.Q., Liu, Y.Q., Liu, B., Gao, P.J., 2006. A novel approach for estimating growth phases and
parameters of bacterial population in batch culture. Science in China Series C-Life Sciences
49, 130e140.
of DBðg=LÞ per substrate consumed DSðg=LÞ in Eqs. (15) and (16), respec-
tively. The determination of growth and product yield coefficient (YB/S and
YP/S) is usually through experimentation or by stoichiometry calculation
using equations such as the ones in Eqs. (6)e(12),
DB
YB=S ¼ (15)
DS
DP
YP=S ¼ (16)
DS
The yield coefficients are usually assumed to be constants in mathe-
matical modelling, and can be used to construct correlation between
the rates of microbial growth, product generation and substrate utiliza-
tion. For example, by rearranging Eq. (16), the rate of substrate reduction
can be expressed in terms of the rate of product formation using Eq. (17),
dS 1 dP
¼ $ (17)
dt YP=S dt
The growth rate of microorganism is defined by Eq. (18)

rg ¼ m$B (18)
in which rg is microbial growth rate (g /L d); m is microbial specific growth
rate (d1) and B is the microbial cell concentration (g/L). m is an essential
model parameter defined by Monod equation Eq. (19) as a function of
the limiting substrate concentration, i.e., the specific growth rate of
microorganisms is determined by the concentration of the substrate that
limits their growth.
S
m ¼ mmax $ (19)
KS þ S
in which mmax is the maximum specific growth rate (d1); KS is half-
saturation coefficient (g/L); namely the substrate concentration at which
m ¼ 0.5 mmax. Accordingly, the growth rate of microorganisms can be
expressed in Eq. (20) by substituting Eq. (19) into Eq. (18),
SB
rg ¼ mmax $ (20)
KS þ S
in which the growth rate (rg, g/L d) of a given microbial species is
related to the concentration of its substrate (S, g/L) and microbial cell mass
(B, g/L).
It is noteworthy that Monod equation (Eq. 19) is only applicable to the
description of microbial growth on substrate in dissolved form such as sugars,
amino acids, VFAs and H2/CO2 generated during AD processes (Wang and
Li, 2014). Eq. (19) also can be modified to incorporate various types of
inhibition effect to be discussed in Section 4 of this chapter (Lauwers
et al., 2013).
The rate of substrate utilization (rS, g/L d) by AD microorganisms can
be correlated to that of the microbial growth in the Monod equation
(Eq. 20), with a growth yield coefficient (YB/S, g/g) quantifying the
new microbial cell mass that can grow from a unit of substrate mass as
described in Eq. (21) (Xu et al., 2015). Hence, the production rate of a
product (rP, g/L d) can be modelled in proportion to the substrate utilization
rate with a product yield coefficient (YP/S, g/g) defining the mass of this
product that can be produced from a unit of substrate mass consumed,
rg rP
rS ¼ ¼ (21)
YB=S YP=S
Monod equation in Eq. (19) indeed provides a basis for the development
of many other microbial growth kinetics related equations. Parameters have
been added to Monod equation to account for other factors affecting the
growth rate including permeation capacity of the substrate, microorganism
adoption, enzyme activity, nature of the culture and growth phases of
microorganisms. A brief summary of these models is presented in Table 1.
Readers are advised to refer to the corresponding literature for additional
information.
Briefly, as shown in Table 1, Moser’s equation (Eq. 22) incorporates a
parameter ‘n’ to integrate effects of microbial adaptation to stationary
processes by mutation. Fig. 3 shows an exemplary comparison of the growth
rate as predicted by Monod (Eq. 19) and Moser (Eq. 22) growth rate
equations. As it can be observed, these two models predict similar trends
with moderate difference governed by additional parameters incorporated
into the models.
2.2.2 Decay Kinetics

The decrease of the biomass concentration shown in Fig. 2 can be modelled
as a biomass decay process (Gavala et al., 2003). Cell lysis and endogenous
respiration are responsible for this decrease (Gavala et al., 2003). The dead
Table 1 Summary of anaerobic digestion models using Monod-like kinetics

Model Kinetics/Equation Notes
Monod (1949) m ¼ mmax $KSSþS

(19)
n
Moser (1958) m ¼ mmax $KSSþSn Parameter n accommodates
(22) microorganism adoption to
a new substrate
Contois (1959) m ¼ mmax $KCS=B
þS=B
Cell concentration
S (23) influences m
Chen and m ¼ mmax $ S So
Initial substrate concentration
þKCH $ð 1SSo Þ
Hashimoto So
(24) (So) affects m
(1979) h i
t
Bergter (1983) m ¼ mmax $KSSþS$ 1 e tL m decreases with the increase
(25) of lag phase time (tL)
n
Mitsd€
orffer m ¼ mmax $Sn $ð1þGSS $Kb $Sn Þ m depends on the gas
(1991) (26) production GS
(L/g organic matter)
in which KC is half-saturation coefficient for S/B; and KCH is kinetic parameter.
μ
Specific growth rate
Monod
Moser (n=2)
Substrate concentration
Figure 3 Specific growth rate as a function of substrate concentration. Adapted from
Kythreotou, N., Florides, G., Tassou, S.A., 2014. A review of simple to scientific models for
anaerobic digestion. Renewable Energy 71, 701e714.
biomass is assumed to contribute to the amount of insoluble substrate for

microbial uptake (Angelidaki et al., 1999). Hence, biomass decay is usually
modelled in the same fashion as the substrate hydrolysis which is simplified as
a first-order reaction,
rd ¼ kd $B (27)
in which rd is biomass decay rate; and kd is the decay coefficient (d1).
The value of decay coefficient is about 5% of the maximum specific growth
rate (Gavala et al., 1999). It is common to ignore decay in modelling
methanogen growth due to the relatively low value of decay coefficient (1%
of the maximum specific growth rate) (Gavala et al., 2003).
2.2.3 Hydrolysis Kinetics

The rate of hydrolysis of a certain type of substrate is related to hydrolytic
bacteria activity, pH, temperature, substrate concentration, biodegradability,
total solids (TS) content, etc (Gujer and Zehnder, 1983a). Similar to biomass
decay rate described earlier, hydrolysis rate is commonly modelled using
first-order reaction kinetics represented by Eq. (28) (Vavilin et al., 2008),
rh ¼ kh S (28)
in which rh (g/L d) represents the hydrolysis rate of substrate; kh represents
the hydrolysis rate coefficient (d1). Since hydrolysis is a surface reaction,
another hydrolysis model proposed by Contois was developed by taking

into consideration the saturation of substrate and biomass which is expressed
by Eq. (29) (Valentini et al., 1997),
ðS=BÞ
rh ¼ kmax $B$ (29)
ðKC þ S=BÞ
in which kmax is the maximum specific hydrolysis rate (d1); KC is the half-
saturation coefficient for ratio S/B. The study by Ramirez et al. (2009) has
shown better correlation results by modelling hydrolysis using Contois
function on waste activated sludge as substrate.
MichaeliseMenten kinetics, a well-known model of enzyme kinetics, is
also applied to describe the hydrolysis rate as a function of the enzyme and
substrate concentrations as expressed in Eq. (30) (Vavilin et al., 2008),
S
rh ¼ kmax $Se $ (30)
KS;m þ S
in which S and Se are the substrate and enzyme concentrations (g/L),
respectively; KS,m is the half-saturation rate coefficient (g/L).
2.2.4 AcideBase Equilibrium and Kinetics

Besides the biochemical processes described previously, there are physico-
chemical acidebase reactions in the liquid phase of anaerobic digesters
(Batstone et al., 2002b). As the acidebase association/dissociation processes
proceed so rapidly relative to the other reactions in AD, they are commonly
considered to be equilibrium processes where the change of Gibbs free
energy ðDG; J=moleÞ is zero. This equilibrium state can be represented
by an implicit set of algebraic equations. To model the acidebase equilib-
rium, charge balance of the cations and anions is used. In an anaerobic
digester, the cation equivalents are the same as the anion equivalents
(Eq. 31) (Batstone et al., 2002a),
X X
SCþ SA ¼ 0 (31)
P P
in which SCþ and SA represent the equivalent concentration of
cations and anions, respectively (g/L). Equivalent concentration is
determined as a product of valence and molar concentration. The acide
base pairs commonly considered in AD include CO2 =HCO3 ,
NH4 þ =NH3 , H2S/HS, H2O/(OHþHþ), HAc/Ac, HPr/Pr and
other organic acids/the corresponding bases. Therefore, the charge balance
in AD taking into account the equivalent concentrations for all the acid/base
pairs is expressed as,
SAc SPr SBu SVa
SCatþ þ SNH4 þ þ SHþ SHCO3
64 112 160 208 (32)
SOH SAn ¼ 0

in which SCatþ and SAn are concentrations (k mole/m3) of metal ions

associated with strong acids and bases. The denominator for the organic acid
concentration ðSAc ; SPr ; SBu and SVa Þ indicates the g COD content per
charge (Batstone et al., 2002a)
2.2.5 GaseLiquid Equilibrium

Almost all AD models consider the concentration equilibrium between
gas and liquid inside the anaerobic digester. Volatile gases generated include
carbon dioxide, hydrogen, hydrogen sulfide and ammonia. Gas and liquid
phases in contact will reach steady state with respect to each other. When
the liquid phase is relatively dilute, Henry’s law can be used to describe
the equilibrium relationship in Eq. (33) (Batstone et al., 2002a).
Sliq;i;ss
KH ¼ (33)
pgas;i;ss
in which KH is the Henry’s volatility constant (g/L bar); Sliq,i,ss is the liquid
phase steady-state concentration (g/L) of gas i; and pgas,i,ss is the partial
pressure of gas i (bar). Again, once this equilibrium is disrupted, for instance,
by additional gas production from AD, gas will transfer from liquid to gas
phases according to the kinetics described in Eq. (34) (Batstone et al., 2002a),

rT ;i ¼ kLa;i Sliq;i KH;i pgas;i (34)
in which rT,i is the dynamic transfer rate for gas i; kLa,i is the gaseliquid mass
transfer coefficient (d1) for gas i; KH,i is Henry’s constant (g/L bar) for gas i;
pgas,i is gas partial pressure (bar) for gas i; and Sliq,i is dissolved gas concen-
tration of gas i (g/L). The rate of the liquid/gas transfer for methane,
hydrogen and CO2 are expressed in Eqs. (35)e(37) (Batstone et al., 2002a),

rT ;CH4 ¼ kLa $ Sliq;CH4 64$KH;CH4 $ pgas;CH4 (35)

rT ;H2 ¼ kLa $ Sliq;H2 16$KH;H2 $pgas;H2 (36)

rT ;CO2 ¼ kLa $ Sliq;CO2 KH;CO2 $pgas;CO2 (37)
Hence, the dynamic gas concentration in the headspace of an anaerobic

digester can be estimated by,
dSgas qgas $Sgas rT $Vliq

¼ þ (38)
dt Vgas Vgas
in which Sgas is the concentration of gas (g/L); Vgas and Vliq are the volume of
the gas and liquid in the digester (L); qgas is the constant gas flow rate (L/d); rT
is the liquid/gas transfer rate (g/L d).
2.2.6 Diffusion Kinetics

Diffusion limitation can impede AD performance when high solid concen-
tration or high bacterial density present such as those commonly seen in solid
state AD and biofilm AD. Fick’s first law in Eq. (39) is commonly used to
describe the rate of substrate, enzyme or bacteria flux along a concentration
gradient through a given surface area,
vS
JD ¼ De $ (39)
vx
in which JD is the diffusion flux (g/m2 d); De is diffusion coefficient (m2/d);

S is substrate concentration (g/L); and x is diffusion distance (m). The rate of
substrate concentration change can be as described by Fick’s second law in
Eq. (40),

vS v vS
¼ De $ (40)
vt vx vx
When modelling solid-state anaerobic digestion (SS-AD) or biofilm

mediated AD, Eqs. (39) or (40) are indispensable.
2.2.7 Temperature Effect on Reaction Kinetics and Thermodynamics

Influence of temperature on bacterial growth is represented using the
Arrhenius equation (Batstone et al., 2002a). Typically, there are three
operating ranges considered for AD (Batstone et al., 2002a), namely
Psychrophilic (4e15 C), Mesophilic (20e40 C) and Thermophilic (45e
70 C). Although the reactors can operate over the given range, it is
observed that the optimum temperature for mesophilic AD is about
35 C and 55 C for thermophilic digestion. Following are the biochemical
reaction phenomena generally observed due to temperature change

(Batstone et al., 2002a):
1. Increase in reaction rates with increase in temperature.
2. Decrease in reaction rates above an optimal temperature.
3. Decrease in yield and increase in half-saturation coefficient with temper-
ature decrease.
4. Change in the reaction pathway due to changes in thermodynamic yield.
5. Increase in death rate with temperate increase due to increase in cell lysis
and maintenance.
The dependence of the specific reaction rate constants on temperature is
important to know in order to adjust the constant values at other tempera-
tures. The oldest equation for describing temperature effect is that of Arrhe-
nius in Eq. (36) (Kythreotou et al., 2014),

ERa $T
k ¼ A$e (41)
in which Ea is activation energy (J/mole); T is temperature (K); R is a gas
constant (J/mole K); k is a rate constant (d1); and A is a prefactor (d1).
If a rate constant is known at a given temperature, it may be calculated
at another temperature through rearrangement of the Arrhenius equation
as van’t Hoff-Arrhenius relationship in Eq. (42) (Tchobanoglous et al.,
2003),

k2 Ea
ln ¼ $ðT2 T1 Þ (42)
k1 R$T1 $T2
in which k1 and k2 are reaction rate constants at temperature T1 and T2.
Because most biochemical reaction is carried out in a relatively
narrow temperature range, the term E=ðR$T1 $T2 Þ in Eq. (42) may be
assumed to be a constant for all practical purposes. Then, Eq. (43) can be
written as,

k2
ln ¼ C$ðT2 T1 Þ (43)
k1
in which C ¼ Ea =ðR$T 1$T 2Þ. Eq. (43) can be rewritten as (Tchobanoglous
et al., 2003),
k2
¼ qT2 T1 (44)
k1
q is usually assumed to be a constant even though it can vary considerably
with temperature. Values for q for biological treatment systems usually vary
from 1.02 to 1.10.
3. COMMON MODEL FRAMEWORKS

One of the methods predominantly used for AD modelling in 1970s
was using the rate-limiting step to simplify the complex AD processes
(Donoso-Bravo et al., 2011). Not until 1990s, Anaerobic Digestion Model
No. 1 (ADM1) was developed as a common modelling framework. A brief
description of the rate-limitation and ADM1 are presented in this section.
3.1 Rate-Limitation Step

Earlier models were based on the rate-limiting step, as it simplifies the
modelling of a complex system. Usually, the overall rate of the AD process
can be assumed to be governed by the rate of the slowest step. For example,
assuming the rate limitation occurs in the first-order hydrolysis step of a
continuously stirred tank reactor operated at steady state (Vavilin et al.,
2008), the effluent substrate concentration and specific methane production
can be estimated using the following equations (Eqs. 45 and 46).
Sin
Sout ¼ (45)
1 þ kh ðSRT Þ
kh $ðSRT Þ
Ysm ¼ Ysm;max $ (46)
1 þ kh $ðSRT Þ
in which Sin and Sout are the influent and effluent concentrations (g/L); Ysm is
specific methane production (g/g); Ysm, max is maximum specific methane
production (g/g); kh is the hydrolysis rate coefficient (d1); and SRT is solids
retention time (d).
Some models were developed considering acetogenic methanogenesis as
the rate-limiting step, while some others considered the VFAs conversion as
the rate-limiting step (Andrews and Graef, 1971; Appels et al., 2008).
Depending on the organic material characteristics, hydraulic loading,
temperature and etc (Speece, 1983a, 1983b), the rate-limiting step may
change with time during AD. Although a rate-limiting step can simplify
AD models, models developed using this approach only give a snapshot
of the dynamic processes occurring in an anaerobic digester because the
rate-limiting step may also shift with time.
3.2 Anaerobic Digestion Model No. 1

As described earlier, although there are mainly four reaction steps in an
AD process, many parallel reactions and intermediate products are indeed
generated in each step (Gavala et al., 2003). International Water Association

Task Group developed ADM1 as a unified modelling framework to simulate
the reactions in Eqs. (48)e(52) in a structured way (Appels et al., 2008;
Donoso-Bravo et al., 2011). The biochemical processes considered in
ADM1 are shown in Fig. 1, which include disintegration/hydrolysis of
complex organic matter, acidogenesis, acetogenesis and methanogenesis.
ADM1 considers microorganisms as part of the complex organic matter.
Hence, the dead microorganisms are utilized as substrate after they decay
into carbohydrates and proteins (Angelidaki et al., 1999). As described
earlier, acidogenesis or fermentation is the uptake of sugars, amino
acids and LCFAs; likewise, acetogenesis is the uptake of valerate,
butyrate and propionate; Methanogenesis is the uptake of acetate and
hydrogen. The four steps in ADM1 are identified as 19 different processes
as listed below:
• Disintegration of complex organic matter (1 process).
• Hydrolysis of carbohydrates, proteins and lipids (3 processes).
• Uptake of sugars, amino acids, LCFAs, valerate, butyrate, propionate,
acetate and hydrogen (8 processes).
• Decay of sugars, amino acids, LCFAs, valerate and butyrate, propionate,
acetate and hydrogen (7 processes)
The mass balance based on the microbial cell mass in a continuous flow
reactor can be simulated by taking into account the input, output, growth
and death as described in Eq. (47) (Gerber and Span, 2008),
dB
¼ D$B0 D$B þ rg þ kd $B (47)
dt
in which B0 is the initial microbial cell mass concentration in (g/L); kd is death
rate (d1); and D is the dilution rate (d1) defined as the inverse of hydraulic
retention time (HRT 1). The disintegration/hydrolysis reaction in ADM1 is
assumed to follow the first-order kinetics as described previously in Eq. (28).
The remaining steps are assumed to follow Monod-type kinetics as described
in Eq. (19). Monod-type kinetics is used instead of MichaeliseMenten
growth kinetics, because ADM1 uses substrate uptake kinetics to describe
variable change and inhibition of microorganisms (Batstone et al., 2002a).
The physicochemical steps considered in ADM1 include:
• Liquideliquid processes (acidebase or ion association/disassociation)
• Liquidegas processes (liquidegas transfer) nonequilibrium reactions
Liquidesolid processes like precipitation and solubilization are not
considered (Batstone et al., 2002a).
Glucose (C6H12O6) is considered as the primary monosaccharide

substrate produced from carbohydrate hydrolysis for the acidogenesis step.
The stoichiometric Eqs. (48)e(50) show the transformation of glucose to
VFAs (acetate, propionate and butyrate) (Batstone et al., 2002a). Although
different pathways for propionate generation are possible, in ADM1 it is
assumed that any reaction generating propionic acid also generates acetic
acid (Eq. 51).
The stoichiometry of glucose to acetate conversion can be described as,
C6 H12 O6 þ 2H2 O/2CH3 COOH þ 2CO2 þ 4H2 (48)

The stoichiometry of glucose to propionate conversion can be described as,
3C6 H12 O6 /4CH3 CH2 COOH þ 2CH3 COOH þ 2CO2 þ 2H2 O

(49)
The stoichiometry of glucose to butyrate conversion can be described as,
C6 H12 O6 /CH3 CH2 CH2 COOH þ 2CO2 þ 2H2 (50)

The stoichiometry of propionate conversion to acetate can be described as,
CH3 CH2 COOH þ 2H2 O/CH3 COOH þ CO2 þ 3H2 (51)

The stoichiometry of acetate to methane conversion can be described as,
CH3 COOH/CH4 þ CO2 (52)

Methane can be generated from carbon dioxide reduction by hydrogen.
This conversion can be described as,
CO2 þ 4H2 /CH4 þ 2H2 O (53)

In ADM1, a matrix format is used to represent the stoichiometric yield
coefficients for aforementioned set of equations. The applicable kinetics
(first order, Monod-type) and inhibition functions (Table 2) associated
with each of the 19 processes are considered. The noncompetitive inhibi-
tion (ammonia and hydrogen) is applicable for inhibition applied to all
the processes except hydrolysis in AD. Empirical inhibition (pH) and
secondary substrate inhibition are applied to all the processes. The compet-
itive inhibition is applied only for butyrate and valerate uptake processes. A
detailed description of modelling inhibition is presented in Section 4 of this
chapter.
Table 2 Form of inhibition considered in anaerobic digestion model no. 1

Inhibition type Equation
Noncompetitive I ¼ 1þS1I =KI (54)

inhibition (ammonia
and hydrogen)
1þ2$ð100:5ðpHLL pHUL Þ Þ
Empirical I¼ 1þ10ðpHpHUL Þ þ10ðpHLL pHÞ
(55)
(pH inhibition)
8 2
>
>
< 3
ðpHpHUL Þ
ðpHUL pHLL Þ
I¼ e for pH < pHUL (56)
>
>
:
1 for pH > pHUL
Competitive uptake I ¼ 1þS1I =S (57)

(butyrate and valerate)
Secondary substrate I ¼ 1þK1I =SI (58)
(uptake inhibition
at zero inorganic
nitrogen concentration)
KI, Inhibition parameter (g/L); S, Substrate concentration (g/L); SI, Inhibitor concentration (g/L);
pHUL, pHLL, Upper and lower limits of pH where the growth rate is inhibited by 50%.
As shown in Fig. 4, the model concept of a typical single-tank anaerobic

digester includes the inputs, reaction kinetics and output (Batstone et al.,
2002b).
The mass balance on a given substrate for a typical anaerobic digester as
shown in Fig. 4 incorporating all the four steps of AD can be written as,
dSliq qin $Sin qout $Sliq X
¼ þ rh;f ;g;G (59)
dt V V
Input Gas phase Output
B – Biomass B – Biomass
X – Solid substrate
Liquid phase S – Soluble substrate
K – Rate constant G – Gas
µmax – Specific growth rate
Reaction rate
Ks – Half saturation coeff. rx – Solid substrate
Y – Yield coefficient rs – Soluble substance
rb – Biomass
rp – Product
Figure 4 Illustration of kinetics in an anaerobic digester.

in which Sliq is soluble substrate concentration (g/L); P Sin is the initial

concentration qin is inflow (L/d); qout is outflow (L/d); rh;f ;g;G is the sum
of the rates of hydrolysis, fermentation (acidogenesis), growth and gas
formation processes (g/L d) and V is volume of the digester (L).
The differential equation system representing the mass balance of
inorganic carbon (as CO2) incorporating growth kinetics, gas transfer and
acidebase reactions of carbonate system is given by Eqs. (60) and (61),
dSCO2 qin $Si;CO2 qout $Si;CO2 X
¼ þ þ rh;f ;g;G rT þ rAB (60)
dt Vliq Vliq
dSHCO3 qout $SiHCO3

¼ rAB (61)
dt Vliq
in which rT is the liquid/gas transfer rate (g/L d); rAB is the production rate
of base from acid (CO2 from HCO3) (g/L d); Si;CO2 and Si;HCO3 are the
initial concentrations of CO2 (g/L) and HCO3 , respectively.
About 32 differential equations can be developed to determine the
dynamic states of different components in AD using ADM1 (Gerber and
Span, 2008). Some of the advantages of ADM1 provided for anaerobic
digester design include increased model application for full-scale designs
and process optimization, as well as the opportunity to implement AD
research to industrial applications, which provides a good basis for other
model development and validation studies (Batstone et al.). So far, ADM1
application to different types of substrates has been successful (Batstone
et al., 2002b; Mendes et al., 2015). However, modelling the liquidesolid
transformations like the precipitation and solubilization of ions is not consid-
ered in the ADM1 (Batstone et al., 2002b).
4. MODELLING MAJOR FACTORS AFFECTING

ANAEROBIC DIGESTION
Modelling a complex process such as AD requires a good understand-
ing of factors associated with the system performance and process stability.
Some of these factors include substrate composition, inhibitory compounds,
form of microbial living (attached and suspended growth) and TS. The
composition of the substrate affects the disintegration/hydrolysis and
methane yield. Biodegradability of the substrate determines the methane
production rate. Inhibition causes the reduction in reaction rate of the mi-
croorganisms due to unfavourable reactor conditions like low pH,
accumulation of VFAs, hydrogen and ammonia. Accumulation of interme-

diate compounds formed during the AD process can also impede biological
reaction. The form of living of microorganisms in a reactor, whether they
are in suspension in the bulk liquid or attached to the substrate, makes a
significant difference in the AD efficiency. As discussed earlier, the attached
growth systems (biofilms) are more effective for hydrolysis enhancement and
synergistic cooperation between AD microbial communities. Hence,
specific models have also been established for biofilm-based AD processes.
Moreover, the mode of AD reactor operation also plays an important role
in the performance of an AD process. Different models have been developed
for continuous, plug flow and batch reactor configurations. Last but not
the least, the TS concentration in an AD process significant affects the AD
effectiveness because at high TS concentration the diffusion limitation will
become evident, contributing to the model complexity. A summary of
mathematical models in the description of those factors is presented in this
section.
4.1 Substrate Composition

4.1.1 Substrate
The substrate in AD models is expressed as the chemical oxygen demand
(COD) or total organic carbon (TOC). As described earlier, the composi-
tion of the substrate in terms of the carbohydrates, proteins and lipids will
affect the intermediate product formation. For instance, the intermediate
products generated due to AD of substrates with high protein content
will result in high amino acids (through hydrolysis), acetate and hydrogen
(through acidogenesis) (Batstone et al., 2002a). In contrast, substrates with
high carbohydrates such as potato waste generate high VFAs through
acidogenesis of simple sugars (Jacob and Banerjee, 2016). In AD process
there is strong correlation between substrate composition and the microbial
composition (Gavala et al., 1999). For instance, dairy waste has high
percentage of acidogens, likewise piggery waste has high percentage of
methanogens, and olive-mill waste will have high percentage of acetogens
(Gavala et al., 1999). The maximum methane yield of a substrate depends
on biodegradability of the substrate, therefore higher percentage of readily
biodegradable fraction of the substrate yields high methane (Galí et al.,
2009). To improve the methane production, wastes with high percentage
of slowly biodegradable ingredients like lignocellulosic biomass are
pretreated (Mata-Alvarez et al., 2014). Carbon to nitrogen (C/N) ratio is
another indicator of the performance of AD process. Substrates (depending
on biodegradability) with high C/N and poor buffering capacity may

produce high VFAs. Substrates with low C/N and good buffering capacity
may result in free ammonia inhibition to methanogenesis (Mata-Alvarez
et al., 2014).
Poggio et al. (2016) developed a model to determine the composition of
carbohydrates, proteins and lipids in complex organic wastes such as food,
green or kitchen waste. It was assumed that the COD of the substrate is
equal to the theoretical CODth which is determined using the following
stoichiometric equation (Eq. 62) assuming the substrate is fully oxidized to
carbon dioxide and water,

a b 3 a 3
Cn Ha Ob Nc þ n þ c O2 /nCO2 þ c H2 O þ cNH3
4 2 4 2 2
(62)
From Eq. (63) the theoretical specific COD (g CODth/g VS) is estimated
using the following expression,

n þ 4a 2b 34 c
CODth ¼ 32 (63)
Substrate molecular weight
Molecular weight was determined using molecular formula assigned for
the carbohydrate (C6H10O5), protein (C51H10O6), lipid [kitchen waste
(C3.85H7.64NO2.17) and food waste (C3.95H7.74NO2.06)] to determine the
COD and specific oxygen demand for these components. The biochemical
fractions for the three components (carbohydrates, proteins and lipids) are
determined using the COD, nitrogen and mass balance. The COD balance
adopted in this model is expressed as,
fc þ fp þ fl ¼ 1 (64)
in which fc, fp and fl are the fractions of carbohydrate, protein and lipid COD
in the total COD, respectively. Once the nitrogen content is known, the
protein fraction can be estimated by,
gNsubstrate gCODp gp
fp ¼ (65)
gCODsubstrate gp gNp
in which gNsubstrate (gCODsubstrate)1 is the nitrogen content in the substrate
(g/g); gCODp gp1 is the COD content in protein (gCOD/g); and gNp (gp)1
is the nitrogen content in protein (g/g).
The mass balance on the three components is expressed as,

1

1 1 1

1 1
fc $ gCODc gc þ fp $ gCODp gp þ fl $ gCODl gl
(66)
gCODsubstrate
¼1
gVSsubstrate
in which gCODsubstrate (gVSsubstrate)1 is specific COD of the substrate;
gCODc gc1 ; gCODp gp1 and gCODl gl1 are the COD contents (gCOD/g)
of carbohydrate, protein and lipid, respectively. Thereby, fp, and fl can be
determined from Eqs. (64)e(66).
The particulate fraction of the substrate includes the readily and slowly
biodegradable fractions. Parametric estimation using ADM1 model structure
was adopted to determine the readily biodegradable and slowly biodegrad-
able fractions. Another similar biochemical method for determination of
substrate composition was developed by Zaher et al. (2009a) based on the
ADM1, using the mass balance of stoichiometric transformations of the
carbohydrates, protein and lipids expressed in the form of a transformation
matrix.
4.1.2 Hydrolysis
Hydrolysis is usually regarded as the rate-limiting step of AD because others
steps proceed relatively faster (Vavilin et al., 2008). Vavilin et al. (2008)
considered the rate of hydrolysis to be dependent on the concentration of
the biodegradable organic matter. Eq. (67) expresses the hydrolysis rate in
terms of substrate concentration (Vavilin et al., 1996).
1
= 2
=
rh ¼ kh Sfb3 S 3
(67)
in which rh is the hydrolysis rate (g/L d); kh is the hydrolysis rate coefficient
(d1); Sfb is the initial concentration (g/L) of biodegradable substrate; S is
the current substrate concentration (g/L). The magnitude of hydrolytic
coefficient is dependent on the substrate biodegradability, solubility, substrate
concentration, temperature and mass transfer (Li et al., 2016). It was further
determined that the hydrolysis rate coefficient is related to the particle size and
density of the substrate (Sanders et al., 2000), and the density of biofilm
covering substrate surface according to Eq. (68) (Vavilin et al., 1996),
rB d
kh ¼ 6rmS ; (68)
rS d
in which rmS is the maximum specific hydrolysis rate (d1); rB and rS

are densities of biofilm and substrate (g/m3), respectively. d is the depth
of biofilm layer on the substrate surface (m); and d is the diameter of
the substrate particle (m). It was shown by Valentini et al. (1997) that
the hydrolysis rate coefficient is exponentially related to the particle
diameter as,
d
kh ¼ ko e d0 (69)
in which kh is hydrolytic constant (d1), d is particle diameter, ko and do are
constants. Assuming hydrolysis is the rate-limiting step, the cumulative
methane production can be estimated with equation Eq. (70) using
hydrolysis rate coefficient (Veeken and Hamelers, 1999),

CH4 ðtÞ ¼ CH4max 1 eðkh tÞ (70)
in which CH4(t) represents the cumulative methane production at time t

(in STP mL, i.e., mL at standard temperature of 0 C and standard pressure of
1 atm); CH4max represents the maximum methane yield of the substrate (STP
mL), kh represents the hydrolysis rate constant (d1). Detailed derivation of
Eq. (70) can be referred to the work by Veeken and Hamelers (1999).
4.1.3 Codigestion
The concept of codigestion involves treating different types of
wastes in the same reactors (Vavilin and Angelidaki, 2005). Advantages
of codigestion include maintaining a balance of pH and C/N ratio
(García-Gen et al., 2015). Use of ADM1 will provide a composition of
complex organic matter in terms of carbohydrates, proteins and lipids
(Zaher et al., 2009b). Benchmark Simulation Model No.2 was improvised
for codigestion application to include features like addition of cosubstrates
and inhibition of LCFAs (Arnell et al., 2016). This model applies ADM1
with additional input of biodegradable fraction (soluble and particulate
fractions) of the substrate. The biodegradable fraction fD was determined
using Eq. (71).
YSM
fD ¼ VS (71)
350CODT
in which YSM is the ultimate methane potential (N m3 CH4 ton/VS),
CODT is total COD (g/L), and VS is the volatile solids (g/L). The inhibition
due to LCFAs modelled in ADM1 was expressed by Eq. (72), which is

similar to Eq. (56),
8 2
>
>
< 2:77259 KI;fa;high KI;fa;low
Sfa KI;fa;low
Ifa ¼ e for Sfa > KI;fa;low (72)

>
>
:
1 for Sfa KI;fa;low
in which Ifa is the inhibition function applied on maximum specific growth
rate, Sfa is fatty acids concentration (g/L), and KI,fa,low (g/L) and KI,fa,high (g/L)
are inhibition parameters.
An ANN (artificial neural networks) type model was developed by Jacob
and Banerjee (2016) to determine the methane yield for a codigestion
application. The independent variables used in this model were substrate
concentration, inoculum and cosubstrate. The model used Eq. (73) to
optimize the model coefficients,
YBM ¼ bo þ b1 S þ b2 Bi þ b3 C þ b11 S2 þ b22 B2i þ b33 C 2 þ b12 SBi
þ b23 Bi C þ b13 SC
(73)
in which YBM is methane yield [L/(kg VS)], S is the substrate concentration
(g/L), Bi is the inoculum concentration [% VS/(VS)], C is cosubstrate
proportion (% TS, w/w), and bo, b1, . b13 are model coefficients.
4.1.4 Biogas Production Rate

The biogas production rate is correlated to the hydraulic retention time of
anaerobic digesters in continuous flow processes (Yu et al., 2013). Applying
the steady-state conditions to the differential equation in Eq. (74) gives the
production rate in terms of the substrate utilization (Karim et al., 2007),
Ysm $ðSo SÞ
rCH4 ¼ (74)
HRT
in which rCH4 is the methane production rate (L/L d); HRT is the hydraulic
retention time (d) defined as a ratio of reactor volume V (L) to the flow rate
q (L/d); So and S are influent and effluent substrate concentrations (g/L); and
Ysm is specific methane productivity (L/g).
Nature of the substrate affects biogas production rate. High volatile
fraction in the substrate correlates to higher biogas production. Volume of
methane generated from AD of municipal sludge is dependent on the

volatile solids reduction and can be approximated as (Appels et al., 2008),
VCH4 ¼ 0:35 VSred (75)

in which VCH4 is methane volume (L/d); VSred is the volatile solids
destroyed in AD (g/d). Volatile solids destroyed can be expressed in terms of
the substrate utilization and biomass generated as,

YB=S $E$ðBODÞ
VSred ¼ E BOD 1:42$ (76)
1 þ kd SRT
in which YB/S is a microbial growth yield coefficient (g/g); E is the
efficiency of substrate utilization; BOD is ultimate biochemical oxygen
demand of substrate (g/d); kd is an endogenous coefficient (d1); and SRT is
solids retention time (d). Solids retention time is defined as the ratio of the
biomass in the reactor to the biomass wastage rate.
4.2 Inhibition
Inhibition will reduce microbial activity in an AD process. This could be
due to specific compounds or reactor conditions (e.g., pH, weak acid/
base, product, cations) (Batstone et al., 2002a). During inhibition, the
conditions in the reactor are toxic which limit the activity of enzymes,
cell activity and diffusion of chemical substances in the organisms
(Kythreotou et al., 2014). Inhibition is expressed using Eq. (77) to allow
for easy substitution or addition of inhibition functions (Batstone et al.,
2002a),

mmax S
rg ¼ $I1 $I2 .In B (77)
KS þ S
in which the first part of the equation is uninhibited Monod-type uptake
and I1.In are the inhibition functions. An inhibition term (inhibition
function) is multiplied by the Monod expression for specific growth rate as
shown in Eq. (77) to estimate the inhibited growth rate (rg). The different
types of inhibition used in modelling include, but not limited to the
following:
• Haldane approach for substrate inhibition.
• Inhibition due to competition in substrate utilization.
• Andrews approach for substrate inhibition.
4.2.1 Haldane Approach

Haldane derived the approach for enzyme inhibition at high concentration
of substrate (Armstrong, 1930). Inhibition function in Eq. (77) is used to
model inhibition from all intracellular processes with different parameters
for acetogens, acidogens, hydrogen-utilizing methanogens and aceticlastic
methanogens. Hydrogen inhibition of acetogens and free ammonia
inhibition of aceticlastic methanogens can also be modelled using Eq. (54)
(Batstone et al., 2002a) in which SI is the substrate concentration of
inhibitor; KI is inhibition parameter (g/L).
4.2.2 Substrate Utilization Competition

Although competition in utilization of the substrate does not has a direct
impact on the microbial activity, it can be expressed as an inhibition function
shown in Eq. (78) similar to Eq. (57) which shows the inhibition due to
competitive uptake (Batstone et al., 2002a),
1
I ¼ (78)
1 þ SSI
in which S and SI are the concentrations of the substrate and inhibitor (g/L).
4.2.3 Andrews Approach

Andrews (1968) developed another approach for modelling inhibition by
the substrate, instead of adding an inhibition function, an inhibition term
was added in the denominator of the Monod’s growth rate expression as
shown in Eq. (79),
mmax ðSÞ
m¼ (79)
S þ KS þ S KSI
Hill and Barth (1977) modified Andrews growth rate inhibition by

incorporating an additional term for other inhibitors,
mmax S
m¼ (80)
S þ KS þ S KSI1I;1 þ KSI2I;2
in which SI1, SI2 are concentration of two different inhibitors (g/L); and KI1,
KI2 are inhibition parameters (g/L). Other inhibition models considered are
summarized in Table 3.
Table 3 Summary of models incorporating growth rate inhibition

Model Equations Notes
Webb (1963) m ¼ mmaxðSÞð1þbS=K

IÞ (81) b represents the
2
SþKS þKS
I ‘allosteric’ effect of
reaction rate
mmax ðSÞ
m¼
Pn S i
Yano et al. (1966) (82) Accounts for the
KS þS 1þ i¼1 KI;i
influence of ‘n’
inhibitors on the
specific growth rate
mmax
Grant (1967) m ¼ ðSþK IÞ
(83) Assumes a linear
decrease in the
growth rate due to
inhibition at high
substrate
concentration
KS
Aiba et al. (1968) m ¼ ðKmmax S
S þSÞ
e I (84) Derived based on
empirical correlation
of substrate inhibition
Hill et al. (1983) kd ¼ kdmax
KI (85) Microbial decay
1þVFA
inhibition due to total
VFA
kd, decay rate (d1); kd-max, maximum decay rate (d1); KI, inhibition parameter (g/L); VFA,
concentration of total volatile fatty acids (g/L).
4.2.4 Inhibition due to Toxicity

Toxicity affects specific targets on microbial cells and usually its effect is
irreversible. Toxicity could be due to LCFA, detergents, aldehydes, nitro-
compounds, cyanide, azide, antibiotics and electrophiles (Batstone et al.,
2002a). From the modelling standpoint, toxicity affects biomass decay
rate, while inhibition influences kinetic uptake and growth (maximum
uptake, yield, half-saturation parameters) (Batstone et al., 2002a). LCFA
inhibition was modelled using secondary noncompetitive inhibition as
aceticlastic methanogens are affected by the LCFA (Zonta et al., 2013).
Eq. (86) shows the inhibition function which is similar to the Haldane’s
approach in Eq. (54),
KI;LCFA
I¼ (86)
ðKI:LCFA þ SLCFA Þ
in which KI,LCFA is the inhibitor parameter concentration (g/L); and SLCFA
is LCFA concentration (g/L).
4.2.5 Substrate Inhibition

When a maximum specific growth rate is reached, a further increase of the
substrate concentration may result in a decrease of the specific growth rate.
According to the study by Gerber and Span (2008), this substrate inhibition
effect can be attributed to a high osmotic pressure of the medium or a
specific toxicity of the substrate. Substrate inhibition can also be due to
intermediate products formed like VFAs, ammonia and hydrogen in AD pro-
cess (Angelidaki et al., 1993; Hill et al., 1983; Palatsi et al., 2010). Increase in
VFA concentration usually occurs when the rates of acetic acid and hydrogen
consumption by the aceticlastic and hydrogenotrophic methanogens
decrease. Hydrolysis rate is inhibited by the sum of VFAs (acetate,
propionate and butyrate) on molar basis expressed in terms of acetate concen-
tration generated in the acidogenesis step, the rate of this inhibition is
expressed similar to the Haldane’s approach (Angelidaki et al., 1993),

KI;VFA
r h ¼ kh S P (87)
VFA þ KI;VFA
P rate (g/L d); kh is uninhibited hydrolysis first-order

in which rh is the hydrolysis
rate coefficient (d1); VFA is sum concentration of VFAs (acetate,
propionate and butyrate) generated in the acidogenesis phase expressed in
terms of acetate (g/L); KI,VFA is the inhibition parameter (g/L); and S is the
substrate concentration (g/L). As VFAs accumulate, the pH of the reactor
drops and in turn the methane production will be reduced. A study by
Weedermann et al. (2015) presented a model of growth rate of acetoclastic
methanogens, the inhibition due to pH and undisassociated acetic acid con-
centration is given by Eq. (88) in analogy to Andrews inhibition in Eq. (81),
mmax $ðAcHÞ
m¼ (88)
AcH þ KS;Ac þ ðAcHÞ
2
KI;AcH
in which mmax is maximum specific growth rate (d1); AcH is acetic acid
concentration (g/L); KS,Ac is half-saturation coefficient (g/L); and KI,AcH is
inhibition parameter (g/L).
In the same study, the growth of acetogenic bacteria and hydrogenotro-
phic methanogens inhibited by the hydrogen and acetate were given by
Eqs. (89) and (90), respectively.
mmax $ðBuHÞ
m¼ (89)
KS;Bu þ KI;H $H þ BuH
mmax $ðHÞ
m¼ (90)
H þ KS;H þ KI;Ac $AcH
in which mmax is maximum specific growth rate (d1); BuH is butyric
acid concentration (g/L); H is hydrogen concentration (g/L); KS,Bu is half-
saturation coefficient for butyric acid (g/L); KI,H is inhibition coefficient
for hydrogen; AcH is acetic acid concentration (g/L); H is hydrogen
concentration (g/L); KS,H is half-saturation coefficient (g/L); and KI,Ac is
inhibition parameter for acetate.
4.2.6 Ammonia Inhibition

Ammonia-bicarbonate system plays a crucial role in the stabilization of
the pH in AD processes (Angelidaki and Ahring, 1993). However, when
free ammonia accumulation in the reactor exceeds a certain limit, it will
result in toxic conditions in the reactor and inhibit methanogenesis. As
acetogenesis is inhibited, they are accumulated in the reactor, which
contributes to the drop of pH. Angelidaki et al. (1999) modelled this
scenario using livestock manure as substrate. The inhibition is due to the
NH4 þ =NH3 equilibrium and its impact on the pH (hydrogen ion concen-
tration). Eq. (91) shows the relation between the free ammonia and the
hydrogen ion concentration.
ðNH3 ÞT
ðNH3 Þf ¼ þ (91)
1 þ HKa
in which (NH3)f is free ammonia concentration (g/L), (NH3)T is total

ammonia concentration (g/L), Hþ is the hydrogen ion concentration
(mole), and Ka is the dissociation coefficient (mole). The free ammonia
inhibition to the growth rate of the aceticlastic methanogenic organisms
using Haldane’s approach is shown in Eq. (92). The pH function in Eq. (55)
in Table 2 is also included as the growth rate is inhibited in a certain range of
pH (upper and lower limits).
! 0 1 0 1
!
mmax ðT ÞB @ 1 A$B 1 C 1
rg ¼ KS;AcH $ @ ðNH3 Þf
A$
1 þ AcH
KS;NH3
1 þ ðNH3 Þ 1 þ 1 þ KLCFA
I;LCFA
T KI;NH3 (92)
0:5ðpH pH Þ !
1 þ 2$ 10 LL UL
$
1 þ 10ðpHpHUL Þ þ 10ðpHLL pHÞ
in which rg is the growth rate (g/L d), KI;NH3 and KI,LCFA are inhibition
parameters (g/L), mmax(T ) is temperature dependent maximum specific
growth rate pHUL and pHLL are the lower and upper limits of pH where
microorganism growth rates will be inhibited. The growth rate expressed in
terms of the VFAs which inhibits acidogenesis and methanogenesis is given
by Eq. (93) (Batstone et al., 2002a). Hill and Barth’s approach (Eq. 80) is
used to develop this equation,
mmax
m¼ (93)
1 þ AcH þ KAcH
KS
I;AcH
þ NH3
KI;NH
3
in which m is specific growth rate, mmax is maximum specific growth rate;

AcH is the undisassociated acetic acid concentration (g/L) that is also
inhibitive, NH3 is the concentration of free ammonia as an inhibitor; KI,AcH
is inhibition parameter (g/L).
4.2.7 Hydrogen as a Control Parameter

Hydrogen is a key parameter to regulate the production of fatty acids from
glucose (Gavala et al., 2003). The methane yield of AD process also depends
on hydrogen as a portion of the methane generated by the hydrogen utiliz-
ing methanogens. Variation in the hydrogen partial pressure also affects the
pH of the reactor and may result in the VFAs accumulation. Presence of
dissolved hydrogen in anaerobic digester and its impact on microbial growth
rate was studied by Mosey (1983). As hydrogen is the byproduct of the
acetogenesis, when the gas phase concentration is high, the acidogenesis
of glucose reaction moves towards the generation of propionate and
butyrate rather than acetate.
The model assumes that the relative availability of the reduced form
(NADH) and oxidized form (NADþ) of nicotinamide adenine dinucleotide
inside the microbial cell wall or the redox potential will affect the VFA
generation (Batstone et al., 2002a) under the following conditions:
1. NADH/NADþ ratio (Eq. 94) inside the microbial cell is assumed
constant irrespective of the reactor condition.
2. Free hydrogen can easily diffuse in and out of the cell wall
½NADH
¼ 1; 500pH2 ¼ 1:5 103 H (94)
½NADþ
in which pH2 is partial pressure of hydrogen, and H is concentration of
hydrogen in digester gas (mg/L by volume).
The substrate utilization rate is affected by the NADH/NADþ ratio. The

rate of substrate consumption (glucose) can be expressed in terms of
NADH/NADþ ratio as (Gavala et al., 2003),
d½gluc rgluc
¼ (95)
dt 1 þ ½NADH
½NAD
þ
kG $Bgluc $½gluc
where rgluc ¼
Km; gluc þ ½gluc
in which [gluc] is the concentration of glucose (mM), rgluc is the unregulated
rate of the uptake of glucose (m moles/L d), kG is the maximum rate
constant (ion moles/g d), Km,gluc is the Michaelis-type constant (mM), and
Bgluc is the concentration of glucose fermenters (mg/L).
4.2.8 Influence of pH on Microbial Growth

The pH of the digester is one of the key performance indicators. The growth
rate of microorganism changes quite drastically with pH. The equilibrium of
H2S/HS, NH3 =NH4 þ and CO2 =HCO3 =CO3 2 is dependent on the
pH of the reactor. In most of the models, the pH is modelled based
on the ionic equilibrium in the AD process. The reactor pH has a major
influence on the microbial growth. pH is expressed as negative logarithm
of hydrogen ion concentration in Eq. (96),
pH ¼ log10 Hþ (96)
The pH inhibition is a combination of disruption of homoeostasis and
increased weak acids concentration. At low pH, inhibition occurs due to
weak bases, on the other hand at high pH the transport limitations are
observed affecting all organisms to some degree. Impact of pH on
methanogenesis is of particular interest as methanogens can only use
undissociated acetic acid (Gerber and Span, 2008). Several studies have
shown that buffering the anaerobic digester to neutral resulted in improved
methane generation (Fox et al., 1992; Kaseng et al., 1992; Lay et al., 1997).
The kinetics of the substrate degradation is affected by inhibitor concentra-
tions and ambient conditions (pH, ion equilibrium, gaseliquid equilibrium
and temperature). To model the impact of pH inhibition on acetogenesis,
the specific growth rate expressed in Eq. (97) was used (Angelidaki and
Ahring, 1993),
! ! !
mmax 1 1 þ 2 100:5ðpHLL pHUL Þ
m ¼ KS $ $ (97)
S þ1
1 þ AcH
KI 1 þ 10ðpHpHUL Þ þ 10ðpHLL pHÞ
in which m is specific growth rate (d1), S is either propionate or butyrate

concentration (g/L), AcH is undisassociated acetic acid concentration (g/L),
and pHUL, pHLL are the lower and upper limits of pH at which microor-
ganism growth will be inhibited.
Eqs. (98)e(101) were used by different researchers (Gerber and Span,
2008; Kythreotou et al., 2014; Moesche and Joerdening, 1999) to factor
in the pH effect on the microbial growth rate,
m ¼ KO þ K1 $pH þ K2 $pH2 (98)
mmax
mmax ðpHÞ ¼ (99)
1 þ Hþ þ K2 $Hþ
K1
KH
m ¼ mmax $ (100)
K H þ Hþ
KOH
m ¼ mmax $ (101)
KOH þ OH
in which KH and KOH are half-saturation coefficients, Hþ and OH are the
hydrogen and hydroxyl ion concentrations, and Ko, K1 and K2 are kinetic
parameters.
4.2.9 Computational Fluid Dynamics Applications

The biochemical and kinetic aspects of the AD have been studied
extensively. In most of the cases for simplicity of modelling, it is assumed
that the reactors are well mixed (L opez-Jiménez et al., 2015). However,
in reality there are mixing issues resulting in decreased performance of the
digesters. Traditionally, to determine the actual solids retention time, known
concentration of tracers were added to the digester feed to correlate it with
mass balance. An alternative approach is by using computational fluid
dynamics (CFD) models which apply the fluid dynamics principles to solve
flow problems. Conservation principles of mass (Eq. 102), momentum
(Eq. 103) and energy (Eq. 104) are applied to determine the fluid mechanic
characteristics. The advancement in the computers with respect to the

computational power has facilitated the application of complex numerical
methods for solving these flow problems. Following equations are essential
for CFD modelling (L opez-Jiménez et al., 2015).
Continuity or mass conservation equation (L
opez-Jiménez et al., 2015):
þ V $Vr$!
vr
V$ v ¼M (102)
vt
in which r is the fluid density (kg/m3); !
v is velocity (m/s); M is mass within
the control volume (kg); V is the volume (m3); V is vector differential

operator vx þ vy þ vz .
v v v
Momentum equation:
vðr$! vÞ
þ Vr$ð! v $!v Þ ¼ Vp þ Vs þ r$! g þ! F (103)
vt
! !
in which p is static pressure (Pa); g is acceleration due to gravity (m/s2); F is
the outer force (N/m ); and s is stress tensor (Pa).
3
Energy equation:
vðr$T Þ
þ rVð!
m$VT
v $T Þ ¼ þf (104)
vt Pr
in which, T is temperature, m is dynamic viscosity, Pr is Prandtl number, and
f is source heat flux (W/m2). CFD methods have been employed to study
different aspects of the anaerobic digesters with draft tube aerators. For
example, Karim et al. (2004) has applied the CFD methods to determine the
mixing energy, velocity profiles and impact of gas flow rate on mixing.
Research by Azargoshasb et al. (2015) developed a simulation based on
three-dimensional CFD coupled with population balance equations of
syntrophic (acetogenesis and methanogenesis) reactions in a continuous stirred
reactor. The velocity vectors calculated by the model are shown in Fig. 5.
The reaction rates for acidogenesis and methanogenesis were defined as,
dSi
¼ k0 $Sin (105)
dt
in which species i represents acetic acid, butyric acid and propionic acid.
Parameters n and k0 were determined using batch experiments operated at
mesophilic (37 1 C) temperature conditions. Mixing law was used to
calculate the density of the biogas from a mixture of hydrogen, methane and
Figure 5 Velocity vectors on a plane at (A) z ¼ 0 and (B) y ¼ 4.5 cm. Adapted from the
study by Azargoshasb, H., Mousavi, S.M., Amani, T., Jafari, A., Nosrati, M., 2015. Three-phase
CFD simulation coupled with population balance equations of anaerobic syntrophic
acidogenesis and methanogenesis reactions in a continuous stirred bioreactor. Journal of
Industrial and Engineering Chemistry 27, 207e217.
carbon dioxide. Fig. 6 shows the molar concentration profiles for VFAs
generated in the reactor.
4.3 Forms of Living

Microorganisms grow either in the form of bioflocs in free suspension of
reactor liquid medium or by forming a dense layer of biofilms attaching on
the solid surface of the reactor medium. The mass transportation of
substrates in the biofilms is driven by mass diffusion gradient. The interactions
of the mass transport and substrate utilization processes were studied to under-
stand the performance of biofilms. Mathematical models of three types of bio-
films are presented in the following section. For further reading on biofilms,
readers are referred to the ‘Mathematical modeling of Biofilms’ a report pre-
pared by IWA Task group on Biofilm Modeling by Wanner et al., 2006.
4.3.1 Modelling Biofilm

Biofilm will form two interfaces, i.e., one with the substratum (also known as
carrier or surface on which biofilm is developed) and the other with the bulk
liquid. The surface of the biofilm is considered as a boundary layer (Fig. 7). To
model a biofilm, mass balance need to be constructed on the substrate con-
centration in biofilms, in the boundary layer and in the bulk liquid.
Figure 6 The molar concentration profiles of (A) butyrate, (B) acetic acid, (C) propio-
nate, (D) carbon dioxide, (E) hydrogen, (F) methane and (G) water during anaerobic
digestion at steady-state condition on a plane at z ¼ 0. Adapted from the study by
Azargoshasb, H., Mousavi, S.M., Amani, T., Jafari, A., Nosrati, M., 2015. Three-phase CFD
simulation coupled with population balance equations of anaerobic syntrophic acido-
genesis and methanogenesis reactions in a continuous stirred bioreactor. Journal of
Industrial and Engineering Chemistry 27, 207e217.
Bulk liquid
JL Boundary layer SB
JF Biofilm SF
Substratum S
Figure 7 Schematic of biofilm, boundary layer, and substratum.
The mass flux is a result of mass diffusion (substrate gradient) or convec-

tion (fluid velocity). The mass balance can be generally written using
Eq. (106) (Wanner et al., 2006),
¼ Vð!
vSF
v $SF Þ þ VðDe $VSF Þ þ rsF (106)
vt
in which SF is the concentration of substrate in the biofilm (g/L); ! v is the
fluid velocity vector (m/d); rsF is substrate utilization rate (g/L d); De is the
diffusion coefficient (m2/d); and V is vector differential operator

v þ v þ v . Depending on the model assumptions, Eq. (106) can be
vx vy vz
simplified to one- or two-dimension biofilm model. For instance, the mass

transfer rate in one-dimension boundary layer is given by Eq. (107) (Wanner
et al., 2006),
dS De
JF ¼ De ¼ ðSB SF Þ (107)
dx LF
in which SB is the concentration of the substrate in bulk fluid (g/L); SF is
the concentration of the substrate in the biofilms (g/L); LF is the
biofilm thickness (m); and JF is the mass flux in the boundary layer of the
biofilms (g/m2 d). To solve the partial differential equation in Eq. (106),
boundary conditions are required. Some of the typical boundary conditions
applied are given as follows:

1. Substrate flux is zero vx ¼ 0 at the surface of the inert substratum
vSF
that supports the biofilm (x ¼ 0 in Fig. 7).

2. Substrate concentration is given at the surface of the biofilm (x ¼ LF in
Fig. 7).
3. The diffusive flux at the surface of the biofilm is equal to the reaction rate
within biofilms.
4.3.1.1 Modelling Biofilm on a Flat Surface

The dynamic accumulation of the substrate inside the biofilms is dependent
on the diffusion through the biofilm and the substrate utilization rate. The
mass balance equation (Eq. 108) (Morgenroth et al., 2000) used for model-
ling the diffusion and reaction in biofilm is based on the Fick’s law Eqs. (39)
and (40),
vSF De v2 SF
¼ rsF (108)
vt vx2
in which rsF and SF are the substrate utilization rate (g/L d); and the substrate
concentration (g/L) at a given depth (x) inside biofilms (m); De is the

diffusion coefficient in the biofilm (m /d). At steady state vt ¼ 0 , the
2 vSF
partial differential equation (Eq. 108) can be used along with the boundary

conditions vx ¼ 0 at x ¼ 0; SF ¼ So at x ¼ LF to determine the
vSF
substrate concentration as a function of biofilm thickness.
4.3.1.2 Modelling Biofilm on a Spherical Surface

The mass balance on the spherical substrate carrier illustrated in Fig. 8 is
given by Eq. (109) (Odriozola et al., 2016).
2
vSF v SF 2vSF
¼ De þ rsF (109)
vt vr 2 rvr
Boundary Conditions:
For r ¼ RP ; SF ¼ SoF
dSF
For r ¼ RS ; ¼0
dr
in which De is the diffusion coefficient (m2/d); SF is the substrate
concentration in the biofilm (g/m3); SoF is the substrate concentration on
the surface of the biofilm (g/L); r is the distance from the centre of the
spherical biomass carrier (m); rsF is rate of substrate utilization and biofilm
formation.
Biofilm
RP Carrier
Lf RS
Figure 8 Schematic of biofilms on spherical carrier surfaces, in which Rp is the radius of

the biofilm particle (m), RS is the radius of the biofilm carrier (m), and Lf is the biofilm
thickness (m).
4.3.1.3 Modelling Anaerobic Granules

Modelling biofilm in the form of anaerobic granules is a special case of
modelling biofilm on spherical carrier with carrier radius equal to zero.
The mass transfer occurs only in the radial direction of the granules. The
mass balance inside the granule for a particular process is expressed in the
same way as Eq. (109) (Odriozola et al., 2016). Boundary condition for
anaerobic granules will be as follows:
For r ¼ RP ; SF ¼ SoF
dSF
For r ¼ 0; ¼0
dr
The reaction rate in the granule (rsF) is usually described with Monod
equation (Eq. 19). For further reading on modelling anaerobic sludge
granules, readers are referred to the work by Odriozola et al. (2016).
4.4 Total Solids

TS concentration is an essential factor influencing the performance of AD
process. The methane production rate was shown to decrease when the
TS concentration is above 15% (Xu et al., 2014). At high TS, the internal
mass diffusion limitation causes the hydrolytic products accumulation and
in turn results in the inhibition of hydrolysis (Xu et al., 2014). The normal
operating range for liquid anaerobic digestion (L-AD) is between 0.5%
and 15% TS (Xu et al., 2014). If the TS exceeds 15%, it is usually considered
as SS-AD (Bollon et al., 2011). Some of the wastes subjected to LS- and SS-
AD include the following:
1. low solids (0%e5% TS) e wasted activated sludge, food wastes,
2. medium solids (5%e15% TS) e dairy manure, swine manure, municipal
sludge and
3. high solids (>15% TS) e organic fraction of municipal solid waste
(OFMSW), agricultural wastes and pulp-paper sludge
SS-AD is advantageous over L-AD in solid waste handling as it allows
much higher loading in a smaller volume with less energy input and water
addition (Karthikeyan and Visvanathan, 2013). Moreover, the compost-
like digestate remaining after SS-AD is easier and less costly to transport
(Li and Wang, 2011). The fibrous biomass that may cause floating and
stratification problems in L-AD can also be easily handled by SS-AD
(Xu et al., 2013). In spite of these advantages, SS-AD is also subjected to
severe mass transfer limitations associated with the high TS concentration.
Unlike L-AD in which a universally acceptable framework such as ADM1

has been well developed, there is not yet a consensus on SS-AD modelling
due to the complexity in operational conditions, mass transfer mechanism,
reaction kinetics and rate-limiting steps. Table 4 presents a general
comparison of the differences in the characteristics associated with L-AD
and SS-AD.
In SS-AD due to high TS concentration, the overall microbial reaction
rate is lower. This was attributed to the unfavourable conditions
for acetoclastic methanogens due to VFA accumulation and drop in
pH (Abbassi-Guendouz et al., 2013), limited mass transfer in the reactor
(Liotta et al., 2014), lower hydrolysis rate constant (Xu et al., 2014) and
higher half-saturation coefficient (Bollon et al., 2011). Three different
categories of models are reviewed in this section which include models
derived from theoretical principles, empirical approaches and statistical
approaches.
Table 4 Comparison of solid- and liquid-state anaerobic digestion (Xu et al., 2015)
L-AD SS-AD
Feedstock Organic wastes like sewage Solid organic wastes like

sludge, manure and food OFMSW, yard trimmings,
wastes with smaller particle crop residues usually
sizes and higher characterized with higher
biodegradability. recalcitrance.
Reactor design Effective mixing, shorter HRT Longer HRT, without mixing
except leachate recirculation.
Phases of mass Two phases (liquidegas) Three phases (solideliquid
transfer egas)
Rate of mass Reactor contents can be Mass transfer is severely limited.
transfer assumed homogenous with Diffusion or convection via
high mass transfer rate. leachate flow is assumed as
the major forms of mass
transfer.
Rate-limiting Both hydrolysis and Hydrolysis is usually the rate-
step methanogenesis can be the limiting step.
rate-limiting step.
Dispersion The inhibitors formed will be Dispersion is less due to limited
of inhibitors dispersed easily. Shock or no mixing.
loading will negatively affect
the performance.
HRT, hydraulic retention time; L-AD, liquid anaerobic digestion; OFMSW, organic fraction of
municipal solid waste; SS-AD, solid-state anaerobic digestion.
4.4.1 Theoretical Models

Models developed based on the theoretical principles discussed in this
section include two-particle model, reaction front model, distributed model,
spatialetemporal model, modified ADM1 model and diffusion limitation
model. A graphical illustration of the theoretical models is shown in Fig. 9.
4.4.1.1 Two-Particle Model

The two-particle model developed by Kalyuzhnyi et al. (2000) considers the
SS-AD to be heterogeneous with two kinds of particles in one reactor,
namely ‘seed’ and ‘waste’ based on the biodegradability and methanogenic
activity. Seed is the inoculum with low biodegradability and high methano-
genic activity, waste is the substrate with low methanogenic activity and
high biodegradability. An illustration of this model is shown in Fig. 9.
The model assumptions are as follows:
• Seed particles and waste particles are adjacent to each other.
ADM1 derived model

VFA
Biogas
Homogeneous
substrate
Inoculum Depleted
zone
Methanogenic zone
Buffer zone
Acetogenic zone
Solid substrate
SS-AD reactor
Diffusion
Substrate Inoculum
particle particle
Inoculum
Solid substrate
VFAs
Diffusion limitation model

“Two-particle” model
Figure 9 Theoretical solid-state anaerobic digestion models developed. Adapted from

the study by Xu, F.Q., Li, Y.B., Wang, Z.W., 2015. Mathematical modeling of solid-state
anaerobic digestion. Progress in Energy and Combustion Science 51, 49e66.
• Diffusion of intermediate particles occurs from the waste to seed for

biogas production.
• Fick’s law is used to express diffusion rate.
• No leachate flow will occur.
The diffusion of intermediate products (VFAs) depends on the concen-
tration gradients, size of seed and waste particles. The diffusion rate is
expressed using Eq. (110).
2De $ðPs Pw Þ
rD ¼ (110)
Ls2 þ Lw2
in which, rD (g/L d) is the diffusion rate; De (m2/d) is the diffusion
coefficient; Ps and Pw (g/L) are the concentrations of the product in the seed
and waste particles; Ls and Lw (m) are the diameters of seed and waste
particles. The factors that influence the stability of SS-AD predicted by this
model include the rate of solute transport and biodegradability of the waste.
The diffusion limitations will limit the rate of VFAs entering ‘seed’ particles,
which mitigates the inhibition of VFAs to methanogenesis.
4.4.1.2 Reaction Front Model

The reaction front model was developed to interpret the slow transport
mechanism in the reactor. The reaction front as described by the model is
a solid substrate composed of multiple layers of an acetogenesis zone, a buffer
zone and a methanogenesis zone, followed by a depleted zone (Martin et al.,
2003). There is a possibility of mass transfer of the solute limiting the
methanogenesis process.
The model assumptions include (Martin et al., 2003) the following:
1. The inoculum distribution is even at the start of the reaction.
2. Initial dimension of the inoculum particle is zero.
3. The thickness of the reaction front is constant.
4. Rate of reaction front advancement within the digester matrix is a
constant.
5. The volume of the reaction fronts is related to the AD reaction rate.
An illustration of the inoculum distribution in solids substrate is shown in
Fig. 10. The initial distance between the ‘reaction fronts’ is 2R (m) and
the radius of the ‘reaction front’ is r (m) at given time (t). G is the methane
production (g/L) from the system.
The fundamental concept of this model is the SS-AD reaction rate varies
along with increase in the total surface area of the ‘reaction fronts’. The
interaction between the reaction fronts is shown in Fig. 11. The surface
Inoculum
Input 2r 2R Output
R, r, F G
Solid substrate
Figure 10 Reaction front model illustration. Adapted from the study by Martin, D.J.,
Potts, L.G.A., Heslop, V.A., 2003. Reaction mechanisms in solid-state anaerobic digestion: 1.
The reaction front hypothesis. Process Safety and Environmental Protection 81, 171e179.
area of the reactor front (A), which is calculated as 4pr2, will decrease as it
merges with six others. This process can be expressed as,

A ¼ 4pr 2 6p$ r 2 R2 (111)
The findings of this model are that the distribution pattern of the
reaction fronts would determine the reaction kinetics. To avoid VFA
accumulation, model simulation suggested the thickness of the reaction
front should be over 7 cm. Although the presence of the zones as described
in the reactor front has never been experimentally proven, this model
attempts to explain the mass diffusion mechanism within SS-AD reactor.
Figure 11 Illustration of interaction between reaction fronts. Adapted from the study by
Martin, D.J., Potts, L.G.A., Heslop, V.A., 2003. Reaction mechanisms in solid-state anaerobic
digestion: 1. The reaction front hypothesis. Process Safety and Environmental Protection
81, 171e179.
4.4.1.3 Distributed Model

This is a simple one-dimensional model that involves the mass transfer due
to diffusion and leachate flow in the SS-AD reactor (Vavilin et al., 2003). It
is assumed that the solid matrix inside the reactor is homogenous in the
horizontal direction. It is to be noted that the acidogenesis and acetogenesis
steps are combined as one step based on the assumption that the fermenta-
tion process is faster to convert the hydrolytic products to acetate. The
microorganisms and acetate are modelled to move in the vertical direction
of the reactor. The illustration of the distributed model is shown in Fig. 12.
The following equations Eqs. (112)e(114) represent the four steps
considered in the AD. Acidogenesis and acetogenesis were considered as a
single step,
vX
¼ kh $X$f ðSÞ (112)
vt
in which X (g/L) is the concentration of solid substrate; kh (d1) is the first-
order hydrolysis rate constant; f(S) is the inhibition function of acetate to
hydrolysis. Eqs. (113) and (114) take into consideration the mass diffusion,
microorganism growth and substrate utilization with space and time,
vS v2 S vS S$B
¼ DS $ 2 e qsa $ þ c$kh $X$f ðSÞ rmax $ $gðSÞ (113)
vt vZ vZ KS þ S
vB v2 B vB S$B
¼ DB $ 2 qsa $a$ þ YB=S $rmax $gðSÞ kd $B (114)
vt vZ vZ Ks þ S
in which S (g/L) and B (g/L) are the concentrations of the soluble substrate
and microbes respectively; g (S) is the inhibition function of microbial
growth; Z (m) represents the vertical coordinate of the reactor, with the
Product Substrate
inhibition inhibition
Solid Intermediate
substrate products CH4
Microbial cells (Bi)
Hydrolysis/acidogenesis
Figure 12 Distributed model illustration. Adapted from study by Xu, F.Q., Li, Y.B., Wang,
Z.W., 2015. Mathematical modeling of solid-state anaerobic digestion. Progress in Energy
and Combustion Science 51, 49e66.
boundary condition defined based on the total length of the effective

volume L (0 Z L); qsa (L/m2 d) is the volumetric liquid flow rate per
unit surface area; c is the stoichiometric coefficient of different hydrolytic
products to acetate; DS and DB (m2/d) are the diffusion coefficients of
the soluble substrate (acetate in this research) and microbes, respectively;
KS (g/L) is the half-saturation coefficient of acetate; a is the fraction of
microbial cell mass transferred by liquid flow; YB/S (g/g) is the methanogen
growth yield coefficient over acetate consumption; rmax (d1) is the
maximum acetate utilization rate; kd (d1) is the microbial cell mass decay
coefficient; and a is the fraction of the microbial cell mass transferred by
liquid flow. The methane production (G) is related to the growth rate,
substrate concentration and yield coefficient. The methane production is
given by Eq. (115),
vG SB
¼ g$ 1 YB=S $rmax $ $gðSÞ (115)
vt Ks þ S
in which G (g/L) is the concentration of gaseous products; and g is the mass
fraction of methane in biogas. This model considered the diffusion and
leachate flow in one dimension (along the vertical axis) of the reactor which
provides a good understanding with simplified approach.
4.4.1.4 SpatialeTemporal Model

The purpose of spatialetemporal model (Eberl, 2003) is to get a better
understanding of heterogeneous mass distribution in SS-AD. The model
concept is an extension of the distribution model to a special regime along
with using the ‘reaction front’ mechanism. To reduce the complexity of the
model, Monod kinetics was not used. The following model equations
(Eqs. 116e119) are developed to determine the instantaneous microbial
growth rate, substrate utilization and methane production,
Xt ¼ De;X $DX a1 $qc $VX b1 $f ðSÞ$X (116)
St ¼ De;s $DS qc $VS þ k2 $X$f ðSÞ b3 $gðSÞ$B (117)
Bt ¼ De;B $DB a2$ q$VB þ b4 $gðSÞ$B b5 $B (118)
Gt ¼ De;G $DG qc $VG þ b6 $gðSÞ$B (119)

in which X, S, B, and G (g/L) are the concentrations of solid substrate,
soluble substrate, microbes and biogas, respectively; t is time; qc is the
convective velocity of leachate flow (L/m2 d); Xt, St, Bt and Gt are functions
of time; a1, a2, b1, b2, b3, b4 and b5 are model parameters; f(S) and g(S) are
inhibition of acetate to hydrolysis and microbial growth, respectively; De,X,
De,s, De,B and De,G are diffusion coefficients (m2/d). This model includes the
diffusion and leachate flow for mass transfer with the reactor. Inhibition due
to VFAs is also considered.
4.4.1.5 Modified ADM1

Abbassi-Guendouz et al. (2012) used the ADM1 to study the impact of SS-
AD on cardboard. TS of the substrate employed in this study ranges from
10% to 35%. The ADM1 was modified by calibrating the model using
the methane generation data at TS concentration of 10%. It is assumed
that the hydrolysis rate coefficient will decrease with the increase in TS
concentration. It was shown that at high TS concentration, the mass
transfer coefficient was reduced due to the pasty texture of the reactor
content, and in turn reduced the release of gases (methane, carbon dioxide
and hydrogen), thereby inhibited methanogenesis.
4.4.1.6 Diffusion Limitation Model

The purpose of diffusion limitation model (Xu et al., 2014) was to
determine the impact of TS on lignocellulosic substrates. The underlying
concept in this model is that AD is initiated when ‘pin-point microflora’
inoculated into the substrate. The hydrolytic products formed are diffused
towards microflora due to substrate gradient. Therefore, the rate at which
the substrate is utilized also depends on the mass diffusion. Fick’s law was
used to predict the rate of diffusion, rd, through a layer of substrate.
Eq. (120) represents substrate diffusion rate,
De $A$ðS S 0 Þ
rd ¼ (120)
L$VSL
in which De is diffusion coefficient for the substrate (m2/d); A is the surface
area of the microorganisms (m2); L is the thickness of the effective zone of
enzymatic hydrolysis (m); VSL is the volume of the substrate layer around the
microflora (m3); S 0 is the substrate concentration (g/L) inside the microflora
and S is the concentration of the soluble substrate (g/L). The hydrolytic
soluble substrate formed (glucose) during hydrolysis was assumed to inhibit
the hydrolysis with an empirical expression for inhibition is given by,
ri ¼ ki $S (121)
in which ri (g/L d) is the reduction of hydrolysis rate due to inhibition, and ki

is the hydrolysis inhibition coefficient.
4.4.2 Empirical Models

4.4.2.1 Logistic Model
A logistic model was developed by Pommier et al. (2007) to predict the
impact of water on the solid waste mechanization. The purpose of this study
was to simplify the model so as to estimate the methane generation from the
landfill. The landfill is considered as an SS-AD in this situation.
The assumptions on which the logistic model is based on are as follows:
1. Methane generation is only due to the microorganisms.
2. Microorganisms growth is subject to the hydrolysis rate and substrate
availability.
3. No inhibition is considered in an attempt to simply the model.
The rate of the change of microbial concentration (B) is given by
Eq. (122),

dB X0max X
¼ mmax $B 1 (122)
dt Xo
in which mmax is the maximum specific growth rate (d1); X (gCOD/ginitial TS)
is the solid substrate concentration; X0max (gCOD/g initial TS) is the initial
value of X; X0 (gCOD/ginitial dry matter) is the microbial accessible solid substrate
in X0max .
It is assumed that the substrate saturated with water is available
for methane production. The model assumes two parameters are linearly
related to moisture content, namely microbial maximal specific growth
rate mmax and the amount of accessible organic matter X0. The water content
(u) was empirically correlated to mmax and X0 represented by the following
expressions Eqs. (123) and (124):
mmax ¼ s$mR
max (123)
Xo ¼ s$Xomax (124)
in which s is a function such that if u < umin, s ¼ 0; if umin < u < uR,
s ¼ (u umin)/(uR umin); if u > uR, s ¼ 1; and where uR is the water
1
holding capacity; mRmax (d ) is the optimal maximum specific growth rate;
and umin is the minimal water content required for the initiation of
bioconversion. These equations suggest that if the actual water content u
exceeds the water holding capacity uR, all the solid substrate S0 is accessible
to microbes and reaches the optimal maximum specific growth rate mR max .
While, below the minimal water content umin, no bioconversion is possible
(S0 ¼ 0 and mmax ¼ 0). When u ranges between umin and uR, both S0 and
mmax cannot reach their optimal values and are directly proportional to the
water content.
The model assumes that solids in the digester need to be in saturated
condition for microbial activity to happen. This is a conservative assumption
to simplify SS-AD model. Substrate used in this model consists of paper and
cardboard with TS content ranging from about 6% to 80%. Model results
showed that the water holding capacity (uR) of the substrate was at 34%
water content and the minimal water content required for bioconversion
(umin) was 60%.
4.4.2.2 General Kinetic Model

General kinetic model was implemented in a study by Fernandez et al.
(2010) and Fdez-G€
uelfo et al. (2012) on SS-AD using OFMSW as substrate.
The model is based on the biochemical rate law given by Eq. (125),
k
B þ X/GBðG > 1Þ (125)
in which k is the rate constant of the process; X (g/L) is the concentration
of dissolved organic carbon in solid substrate; B (g/L) is the microbial
concentration in dissolved organic carbon; and G is a stoichiometric
constant. Based on the rate law, Eqs. (126) and (127) are derived,
dB
¼ ðG 1Þ$k$X$B (126)
dt
dX
¼ k$X$B (127)
dt
With this concept, a general kinetic equation was derived as Eq. (128),
dX ðh XÞ$ðX XNB Þ
¼ mmax (128)
dt ðXo XNB Þ
in which X0 (g DOC/L) is the initial solid substrate concentration (DOC
stands for dissolved organic carbon); XNB (g DOC/L) is the concentration
of nonbiodegradable substrate; and h (g DOC/L) is the maximum microbial
cell mass concentration that can be reached. Expression for methane
generation developed in this study is as shown in Eq. (129). Eq. (129) is a
modified Gompertz equation (Xu et al., 2015), which is used for bioenergy
production involving microbial growth (AD, fermentation and biohydrogen

production).
" #
eðBtÞ 1
G ¼ YG=X
(129)
ð1=ðh X0 ÞÞ þ eBt ðX0 XNB ÞÞ
in which B ¼ mmax $XhX NB

0 XNB
; G (L/L) is the total methane produced; t is
the time (day); X0 (g DOC/L) is the initial solid substrate concentration;
XNB (g DOC/L) is the concentration of nonbiodegradable substrate;
h (g DOC/L) is the maximum microbial cell mass concentration that can be
reached, and YG/X (LCH4/g DOCconsumed) is the methane yield per unit
substrate consumed.
4.4.3 Statistical Models

The statistical models are helpful when the exact mechanistic explanation
is not available. Statistical models may be relatively easy to use in case of
SS-AD where the concepts of mass transfer are not completely studied.
4.4.3.1 Linear Regression

A study by Le Hyaric et al. (2012) has shown linear relationship of the
specific methane activity to the moisture content during mesophilic AD
of MSW. It was observed that the specific activity decreased significantly
when the moisture content reduced from 82% to 65%. Liew et al. (2012)
in a different study has shown linear relationship of the total methane yield
to the lignin content. Although a good correlation (R2 ¼ 0.95) has been
observed, application of the coefficients determined from this linear regres-
sion model to other studies may not be possible because the data used to
develop the model are limited. However, if the experiments are well
controlled with limited number of variables, simple linear regression will
be effective to correlate the model prediction to experimental results.
4.4.3.2 Multiple Linear Regression

The study by Motte et al. (2013) developed a quadratic multiple linear
regression model to determine the impact of TS content, inoculation ratio
and particle size of lignocellulosic biomass on the SS-AD performance.
The response variables used in the model were methane production, pH
and VFAs. A three level BoxeBehnken experimental design was imple-
mented in an SS-AD batch study using wheat straw as a model substrate.
The response variable expressed in terms of the explanatory variable is given

by Eq. (130),
X
k X
k X
k
y ¼ a0 þ a m xi þ amm x2i þ amn xi xj (130)
i¼1 i¼1 i¼1
where y is a response variable standing for methane yield; VFA

concentration stands for pH; xi and xj are explanatory variables, and a0,
am, amm, and amn are model coefficients.
5. MODEL PROCEDURE
The typical procedure adopted in model application for AD processes
includes defining the objective followed by the model structure selection,
conceptualization, calibration/sensitivity analysis, model validation and
evaluation (Donoso-Bravo et al., 2011). Examples of the model objectives
can be methane production prediction, rate of substrate degradation,
optimum organic loading rate determination and suitable hydraulic reten-
tion time, etc.
According to the study by Donoso-Bravo et al. (2011), the procedure for
modelling AD includes the following steps as illustrated in Fig. 13.
1. Based on the model objectives and prior knowledge about the biological
processes, a model structure or framework is selected to build the model
concept.
2. The next step is model assumptions and formulation, which includes
selection of appropriate equations for each model component based on
the assumptions made. All possible parameters that will affect the AD
processes are incorporated during this step.
3. Sensitivity analysis is performed to identify the key parameters showing
more influence on the model output. Priority is given to those key
parameters in the course of calibration for the parameter determination.
Model selection is revisited depending on the outcome of the sensitivity
analysis.
4. Depending on the objective and existing data availability, a new set of
experimental data may be required to estimate the parameters. Data
analysis is performed to compare model output against experimental
data.
Objective
Model Experimental/Literature data

selection
Model
formulation
Calibration and
sensitivity Experimental/Literature data
analysis
Objective function
selection and
parameter
optimization
Direct validation
Comparison with
experimental data
Cross validation
Figure 13 Model procedure.
5. Model is validated using the experimental data used in step 4 or a new

data set.
Model parameter estimation and validation are important aspects in
which lots of effort needs to be dedicated. Experimental studies are required
to determine some, but not all, of specific parameter values (Donoso-Bravo
et al., 2011). In this step, the model prediction is adjusted to fit the reality
(experimental data). Typically, the parameter estimation is a major step in
model calibration. Trial and error method to fit model output to the
observed real experimental data is quite tedious. Therefore, computational
model regression is helpful for determination of the parameter values. The
interpretation of the model structure and the ability of the model to predict
reality depend not only on how well the model is formulated but also how
well it is calibrated.
5.1 Model Concept and Structure Selection

As recommended by Spriet (1985), an effective model structure selected
needs to have the following four characteristics:
1. simple,
2. relevant,
3. include available parameters and
4. applicable to different conditions.
In general, identifying the appropriate model structure will require a
good understanding of (1) the experimental data obtained from experiment;
(2) the level of detail the model prediction is intended for and (3) available
methods to reduce the number of unknown parameters in the model
development. The model structure in the case of AD process is characterized
by a series of nonlinear differential equations with unknown parameters.
Only some, but not all, of those unknown parameters can be determined
through experimentation. Some of the model structures adopted program-
ming languages such as C, and Matlab (Donoso-Bravo et al., 2011). The
model development effort is an iterative step in which additional parameters
may need to be deleted or refined to provide a simple model structure but
still with mechanistic meanings.
5.2 Sensitivity Analysis

Sensitivity analysis is an effort to determine which parameters play more
influence on the model objectives. This analysis usually consists of identifi-
ability analysis and uncertainty analysis. Identifiability is a procedure adopted
to determine the unique model parameters from the available data to
estimate the uncertainty of those parameter estimates. There are four types
of sensitivity functions used to identify the unique model parameters
(Reichert, 1998).
1. Absoluteeabsolute sensitivity function which is defined as the absolute
change in variable y per unit change in parameter a, expressed as,
vy
y;a ¼
da;a (131)
va
2. Relativeeabsolute sensitivity function which is defined as the relative
change in variable y per unit change in parameter a, expressed as,
1 vy
y;a ¼
dr;a (132)
y va
3. Absoluteerelative sensitivity function which is defined as the absolute

change in variable y for a 100% change in parameter a, expressed as,
vy
y;a ¼ a
da;r (133)
va
4. Relativeerelative sensitivity function which is defined as the relative
change in variable y for a 100% change in parameter a, expressed as,
a vy
y;a ¼
dr;r (134)
y va
The absoluteerelative and relativeerelative sensitivity functions are
commonly used as they are independent of the unit of the parameter.
For most of the AD processes, the derivatives associated with the sensitivity
functions are calculated using the finite difference approximation (Donoso-
Bravo et al., 2011) expressed as,
vy yðai þ Dai Þ yðai Þ
z (135)
vai Dai
in which Dai is 1% of the standard deviation sai of the parameter ai.
The calculated sensitivity (change in variable caused by change in
the parameter) can be correlated to the estimation of uncertainty of that
parameter (Reichert, 1998). If the sensitivity is smaller for parameter a1 in
comparison to parameter a2, the uncertainty of the estimate of parameter
a1 is larger in comparison to a2.
5.3 Parameter Estimation

Unknown parameters need to be estimated through model calibration. The
collection of the experimental information is very crucial in this step of
the modelling study. Increase in the number of processes and microbial
pathways will add more parameters. Previous research work (Batstone and
Keller, 2003; Derbal et al., 2009; Fedorovich et al., 2003; Galí et al.,
2009; Lee et al., 2009; Ramirez et al., 2009) has successfully applied AD
models to different type of wastewaters. In all these instances, the experi-
mental data were able to match the model simulated data by calibrating
specific model parameters. A typical model calibration procedure (Dewil
et al., 2011) adopted includes the following:
1. Data collection from the anaerobic experiment.
2. Identification of key parameters from sensitivity analysis for model
calibration.
3. Use of literature suggested values as initial values of these parameters.

4. Adjustment of the selected parameters to fit model simulation with
experiential data.
As mentioned earlier AD processes involve multiple parallel reactions
therefore an objective function is used to determine the deviation of
the model value to the experimental data. The objective function is a
model parameter vector, which defines the deviation between the model
prediction and real experimental data (Donoso-Bravo et al., 2011). Sum
of least squares (Eq. 136) is one of the objective functions commonly used
in AD studies, where the standard deviation is assumed constant (Noykova
and Gyllenberg, 2000),
X
N 2
JðaÞ ¼ min vexp ðiÞ vsim ði; aÞ (136)
i¼1
in which a is the parameter which is to be determined; vexp and vsim are

the experimental and model (simulated) values, respectively; vexp(i)
is the ith measurement; N is the number of measurements. This
objective function was employed to perform sensitivity analysis on
the parameters for simulating anaerobic mesophilic sludge digestion
using ADM1 (Mendes et al., 2015). When the error (difference of
the simulated and experimental values) does not have a constant
standard deviation, weighting factors wt have to be used, then the objective
function is defined using Eq. (137) (Donoso-Bravo et al., 2011; Palatsi
et al., 2010),
X
N 2
JðaÞ ¼ min wi vexp ðiÞ vsim ði; aÞ (137)
i¼1
If the measurements are in vector form the objective function can be

expressed as Eq. (138),
X
N
JðaÞ ¼ min vexp ðiÞ vsim ði; aÞ W vexp ðiÞ vsim ði; aÞ (138)
i¼1
in which W is an N N weighting matrix.

Weighting factor can be estimated as a difference between the maximum
and minimum values (Palatsi et al.) or as deviation from the mean (Flotats
et al., 2003). According to Donoso-Bravo et al. (2011), if the errors are
unknown and relative in nature the following objective function Eq. (139)
can be used,
X N
vexp ðiÞ vsim ði; aÞ 2
JðaÞ ¼ min (139)
i¼1
vexp ðiÞ
Batstone et al. (2003) assumed that the logarithm of the error is relative
and therefore objective function as shown in Eq. (140) was used,
X
N 2
JðaÞ ¼ min ln vexp ðiÞ lnðvsim ði; aÞÞ (140)
i¼1
After an appropriate objective function is selected, it is minimized using

numerical methods along with the unknown parameter determination.
5.4 Data Collection

The nature of the experimental data plays a crucial role in model formula-
tion and calibration. As data collection is quite expensive, the complexity of
the model and the number of parameters to be included in the model will
drive the extent of data collection. There are two issues associated with the
experimental data collection for AD modelling, one issue is that most of the
systems are mixed culture and therefore the ‘culture history’ is dependent on
the reactor operating condition (Donoso-Bravo et al., 2011); the other
concern is that parameters of a Monod’s equation cannot be uniquely deter-
mined using batch experiments (Baltes et al., 1994), which is mostly
commonly used reactor configuration for collecting experimental data.
5.5 Parameter Optimization

According to the study by Donoso-Bravo et al. (2011), the model optimi-
zation is achieved by optimizing the objective function to determine the pa-
rameters for model prediction improvement. Parameter optimization is
based on the trial and error method. This can be cumbersome; therefore,
lots of effort has been invested in the development of optimization tech-
niques. Numerical algorithms are used to determine the optimum values
of parameters with the assistance of objective function. Two algorithms
are available for parameter optimization, namely local and global methods
(Donoso-Bravo et al., 2011). The local methods assume objective function
to be a convex function (Fekih-Salem et al., 2012). An alternative method of
optimization is Bayesian approach, which treats the model parameters as
random variables.
5.5.1 Local Methods

Local optimization methods are common in AD modelling (Donoso-Bravo
et al., 2011). The fundamental assumption of this method is convexity of
objective function. For a convex function, value at the midpoint of
two conditions is less than the mean of the two conditions. If the objective
function is not convex, a random search of the initial parameter values called
‘multi-start strategy’ is implemented (Gyorgy and Kocsis, 2011). Some of the
local methods are described in the following subsections.
5.5.1.1 Simple Unconstrained Optimization

Steepest descent method uses the first-order information to approach the
gradient. A very large number of iterations are required to achieve conver-
gence with this method. Another method used is the GausseNewton
method, which is based on the minimization of sum of squares of errors.
The GausseNewton method uses linear approximation of the errors
using Taylor’s theorem. LevenbergeMarquardt method (Marquardt, 1963)
combines both the steepest descent method and GausseNewton method.
This is most commonly applied to AD models for treating livestock manual,
distillery waste and animal wastes (Martin et al., 2002). The solution deter-
mined by this method depends on the initial and current values estimated
during the subsequent iterations.
5.5.1.2 Nonlinear Constrained Optimization

Sequential quadratic programming is used as a numerical solution to the
nonlinear optimization problems. In this method the objective function is
approximated as a quadratic problem with constraints. If there are no
constraints, this method reduces to Newton method (Donoso-Bravo
et al., 2011).
5.5.1.3 Multiple Shooting

In this method, the objective function is evaluated repeatedly by solving the
dynamic equations numerically. Another method employed is by setting up
an optimization algorithm with model differential equations as constraints
(Carbonell et al., 2016).
5.5.1.4 Direct Search Methods

The family of direct search methods includes pattern search methods,
simplex methods and adaptive sets of search directions (Donoso-Bravo
et al., 2011). Simplex methods were used for optimizing AD models. These
methods come in handy where the numerical derivatives of the objective

function are unknown (Lewis et al., 2000). The objective function is
minimized by doing search in a multidimensional parameter space.
A nondegenerate simplex [as combination of nþ1 points (vertices) in an
N dimensional space] is first developed and then reflection is done by
replacing a vertex to determine the minimum value. The simplex is not
sensitive to the local minima in comparison to the other gradient-based
methods (LevenbergeMarquardt, GausseNewton). The convergence to
the optimum solution is slower and closer, and there are no means to deter-
mine the sensitivity of the solution. This method was used to estimate
parameters on substrate and product inhibition in AD models (Moesche
and Joerdening, 1999).
5.5.2 Global Methods

In nonlinear parameter identification, problems exist when there are various
local minimum present for the objective function. Global methods will
provide solutions in these situations. Some of the global methods used in
AD process are as follows:
• simulated annealing,
• genetic algorithms and
• particle swarm optimization.
Simulated annealing is a stochastic optimization procedure where the
objective function is treated analogous to internal energy of the system
and the undetermined parameters will be optimized in an imaginary system
(Kirkpatrick, 1984). Generic algorithms are based on the theory of
evolution, the solutions of which are developed by natural evolution
techniques like inheritance, mutation, selection and crossover. This method
is useful when sample space is large and difficult for analytical treatment.
Particle swarm optimization is a stochastic optimization method inspired
from social behaviour of animals (Bai et al., 2015). This method is similar
to the genetic algorithms. Some of the advantages of this method include
less number of assumptions and no differentiation of objective function is
required. Bai et al. (2015) used this method for parameter estimation to
model effective VFA generation from waste activated sludge.
5.5.3 Bayesian Approach

Bayesian approach is a probabilistic approach used in the calibration
of model parameters. These model parameters are considered as random
variables with probability density function. A ‘joint posterior distribution’
of parameters is used to define the ‘subjective beliefs’ during calibration

(Omlin and Reichert, 1999). The probability function of a particular
parameter is expressed using Bayes’ theorem as shown in Eq. (141),
pðy=aÞ$pðaÞ
pða=yÞ ¼ (141)
pðyÞ
pðy=aÞ$pðaÞ
pða=yÞ ¼ R fpðy=aÞ$pðaÞ (142)
a pðy=aÞ$pðaÞ$da
in which a is model parameters vector; y represents experimental data;

p(a/y) is posterior probability function (posterior beliefs after having eval-
uated the model residuals); p(a) is prior probability distribution (modeller
prior beliefs); p(y/a) is likelihood function of the observations; p(y) is
probability of observations (expected value of the likelihood function over
the parameter space). Readers are referred to the work by Donoso-Bravo
et al. (2011) for more information on parametric optimization.
5.6 Model Validation

The purpose of the model validation is to verify if the model developed is
able to simulate the experimental data. The model validation step will
improve the confidence with which it can predict the reality. There are
two types of validation usually employed, one is direct validation and the
other is cross-validation.
5.6.1 Direct Validation

For direct validation, the experimental data used for parameter estimation
are employed to validate the model. Residual analysis can be performed
to determine the fit. Goodness of fit is expressed in terms of coefficients
like R2 coefficient, estimation of variance or analysis of randomness.
5.6.2 Cross-Validation
Cross-validation is required to confirm the validity of the complex models.
A new data set, other than the one used for parameter estimation, needs to
be used for the cross-validation of data. If a new data source is not available,
two data sets are created from the available data source, and one is used
for calibration and the other is used for cross-validation (Fig. 14). Cross-
validation of data will require modification to the initial conditions.
Additional calibration is required when there are changes to substrate and
microorganism population.
Calibration Data source -1

Data set -1
If a second data source

Data set -2 not available
Direct validation
(different from the data
(check for
goodness of fit) used for calibration)
Cross validation Data source -2
Figure 14 Cross-validation.
5.7 Model Assessment

Model assessment is the last step employed in the AD modelling. Setting up
the acceptable margin of error in the model prediction and the observed data
is very important. Model assessment methods are dependent on the model
complexity and data availability. Some of the methods used to assess how
well the model predicted data are to that of the observed data are introduced
below.
1. Average relative error (ARE)
ARE is expressed as the ratio of difference between model prediction and
experimental data,
vsim vexp
AREð%Þ ¼ 100 (143)
vsim
in which vexp is the experimental value, and vsim is the model predicted value.
2. Relative root mean square error (RMSE)
RMSE is an average of the square of the errors (difference of model and
experimental values or observed values) (Jacob and Banerjee, 2016),
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
uP
un 2
u vsim ðiÞ vexp ðiÞ
t i
RMSE ¼ (144)
n
in which n is the number of observations; vexp(i) and vsim(i) are the
experimental value and model predicted value of the ith measurement,
respectively.
3. Model efficiency (ME)

ME is a measurement of the deviation of the model to the observed data
(Eq. 145) (Hosaini et al., 2009),
n
P 2 P
n 2
vexp vexp vsim vexp
i i
ME ¼ n
P 2 (145)
vexp vexp
i
in which vexp is the experimental value; vsim is the model predicted value;
vexp and vsim are the average of the experimental and model values.
4. Coefficient of residual mass (CRM)
CRM coefficient compares the residual (difference between experi-
mental and model values) to the experimental value (Eq. 146) (Hosaini
et al., 2009).
P
n P
n
vexp ðiÞ vsim ðiÞ
i i
CRM ¼ P
n (146)
vexp ðiÞ
i
5. Sum of residual squared error (SRSE)

SRSE is another metric to measure the goodness of fit (Karim et al.,
2007). A less value of SRSE metric indicates a better fit,
X n
vsim ðiÞ vexp ðiÞ 2
SRSE ¼ (147)
i
vexp ðiÞ
6. Correlation coefficient
Correlation Coefficient (r2) is a measure of the closeness of model fit to
the observed data (Poggio et al., 2016). Unlike aforementioned methods
where closer to zero is indicative of better-fit, r-squared value closer to
one indicates a better fit,
n
P 2
vsim ðiÞ vexp ðiÞ
r 2 ¼ 1 iP
n 2 (148)
vsim ðiÞ vexp
i
in which r2 is correlation coefficient; vexp is the experimental value; vsim is the

model predicted value; n is the number of observations; and vexp is the
average of the experimental value.
6. MODELLING CHALLENGES
Although the elements for a good model such as model objective,
model assumptions and model validation should be included (Jakeman
et al., 2006), there are challenges faced by AD modeler. In some situations,
uncertainty analysis may be required to provide mechanistic reason to justify
the model results. The typical challenges faced during the modelling of AD
process include judicious selection of initial conditions, understanding the
complexity of substrates, availability of data for model validation, model
overfitting and the need for modelling species generally not accounted for
in the typical models.
6.1 Initial Conditions

One of the aspects, which is not given significant importance in model
optimization, is the initial conditions of the variables (Donoso-Bravo
et al., 2011). Many people think that the initial conditions can be
determined based on the experimentation and existing literature. However,
in many cases, this can be very challenging. For example, it is laborious and
even impossible to experimentally determine the initial concentrations of
various species of microorganisms in the seed sludge of AD. A common
practice is to predict the steady state of an AD process from which the
seed sludge was obtained, so that the seed sludge microbial constitutes can
be estimated.
6.2 Complexity
There has been significant progress on AD modelling. However, mecha-
nisms in AD processes like disintegration and hydrolysis, are quite complex,
and are usually simplified. For instance, most of the AD models assume that
this process follows the first-order kinetics and the calibration is performed
by estimating the rate constant (Dewil et al., 2011). According to Batstone
et al. (2002b), modelling the hydrolysis through surface-based kinetics could
provide more accurate results. Likewise, mass transfer limitation in SS-AD is
another factor which adds complexity to the model development, for
instance, the increase in TS content causes a steeper gradient of hydrolytic

production which can inhibit hydrolysis in substrate layers (Xu et al., 2014).
The loading rate of the substrate affects the accuracy of the model
development. Research from Vavilin and Angelidaki (2005) on AD of
MSW has shown that, at high organic loading rate, low mixing was found
to increase the degradation as intensive mixing could disperse the ‘methano-
genic centers’, and in turn impede the degradation. Modelling rate of waste
degradation for complex substrate increases the model complexity, as the
parameters like moisture content, particle size, waste density, leachate
recirculation and neutralization need to be accounted for in addition to
the common parameters which are associated with the AD process (Vavilin
and Angelidaki, 2005). Although focussing on all the processes in the AD
will result in large set of parameters to deal with, depending on the model
objectives, adjusting the most important parameters while keeping the other
parameters constant will simplify the modelling efforts.
6.3 Data Availability

The accuracy and robustness of the AD models will improve if more
independent data sets are available for model calibration and validation.
One of the problems AD modellers are facing is the data availability.
Although most of the reactor state variables are measurable, their determina-
tion takes significant time and efforts (Steyer et al., 2006). To overcome this
data limitation, all unknowns are set as parameters and then experimental
data are used to calibrate the model to give reasonable predictions of the
remaining unknowns. Many online monitoring techniques can be
employed to generate massive real-time data in terms of TOC, alkalinity,
dissolved carbon dioxide/hydrogen, VFAs and nitrogen, phosphorus
species, etc. (Dewil et al., 2011).
The latest trend in AD modelling is the use of software sensors (Dewil
et al., 2011), which is an indirect measurement of the components required
for performing the model calibration and validation so that this can be
applied to predict the new scenarios. Software sensors are software that is
capable of predicting nonmeasured process variables based on a mathemat-
ical model. The model-based soft sensors used in the current AD modelling
applications include the following:
1. Extended Kalman filters (Dochain and Perrier, 1998): In this method an
initial estimate of the unknown parameters is made and correction
is applied in subsequent steps based on the actual measurements. The
accuracy of this method lies in the judicious selection of the initial

estimate.
2. Extended Luenberger observers (Mendez-Acosta et al., 2010): This
method is similar to the extended Kalman filter except that the uncertain
terms associated with the influent composition of the substrate are deter-
mined by the extended Luenberger observer.
3. Adaptive observers (Bastin, 1990): In this method, the microorganism
growth rate represents a time-varying parameter, which simplifies the
model to improve convergence.
4. Asymptotic observers (Dochain and Perrier, 1998): This global method is
an intermediate of the extended Kalman filter/extended Luenberger
observer and adaptive observer.
For more detailed information on the soft sensors, readers are referred to
the review paper by Dewil et al. (2011). To improve the AD reactor perfor-
mance, mathematical modelling of the heterogeneity of the reactor
contents, mass transfer/diffusion limitations and operational parameters is
required. Data sharing and coordination amongst the research groups will
help solve many data availability issues (Xu et al., 2015).
6.4 Parameter Interpretation

The process of the parameter calibration requires a lot of trial and error effort
(Donoso-Bravo et al., 2011). One may find increasing the number of
parameters will increase the freedom of the model fitness to the experi-
mental data (Donoso-Bravo et al., 2011). However, it should be noted
that increase in the number of parameters with limited data will result in
overfitting where model validation will be challenged. Multiple iterations
are required to find reasonable solutions to the problem. As a good practice,
any parameters added to a model should have a solid mechanistic meaning.
More experimental data are necessary to reasonably determine the parameter
values in the course of parameter calibration.
6.5 Modelling Species Unaccounted for

Although AD modelling does not consider all products or microbial species
involved in the AD processes, it does not mean those unaccounted factors
are unimportant. Take ADM1 for instance, it does not take into account
the products like lactic acid and ethanol generated by the fermentation of
glucose. However, literature has shown the presence of lactic acid during
high loading conditions where the reactor pH is significantly low. Model-
ling these species needs to be considered depending on modelling objectives
and model complexity. It has been pointed out that, even though ADM1 is
the most comprehensive AD modelling framework developed so far, many
processes are not included. These include the following:
• alternative products to glucose (lactate and ethanol),
• diffusion limitation,
• sulphate reduction and sulfide inhibition,
• weak acid and base inhibition to methanogenesis,
• long chain fatty acid inhibition to methanogenesis,
• acetate oxidation competition with aceticlastic methanogens,
• homoacetogenesis competition with methanogens and
• solids precipitation.
The simultaneous sulphate and iron reduction process is associated with
microbial interaction between acidogens, acetogens, sulphate-reducing
bacteria and iron-reducing bacteria. Modelling this concept into a structured
modelling framework has not been fully established. To study the impact of
the structure of microbial species on AD performance and determine the
microbial populations in AD, advance molecular techniques such as
polymerase chain reaction (Karthikeyan et al., 2016), DNA sequencing,
fluorescent in situ hybridization (Liu et al., 2012), DNA stable isotope
probing, temperature and denaturing gradient gel electrophoresis have
been applied (Mata-Alvarez et al., 2014). Depending on the objective
of the model, the information determined from these techniques can be
useful to quantify the change when there is a change in AD conditions
(Mata-Alvarez et al., 2014).
7. CONCLUSIONS
State-of-the-art models have been developed to describe methane
yield improvement, reactor stability enhancement, operational issue
identification and waste codigestion in AD systems. These mathematical
models are capable of providing better understanding, design,
optimization and prediction of the performance of AD systems. Theoretical,
empirical and statistical approaches were taken to derive these AD models
for describing substrates characteristics, rate-limiting conditions, process
inhibition, operating conditions, methane potential and production rate,
liquidegas interface mass transfer, as well as the application of computational
fluid dynamics. An effective modelling procedure should be adopted in
the course of model structure selection, parameter estimation and model
validation. Software sensors can be used in the field of AD as an indirect

measurement for model calibration and validation. Online monitoring is
encouraged to generate real-time data to avoid the data limitation for AD
model improvement.
REFERENCES
Abbassi-Guendouz, A., Brockmann, D., Trably, E., Dumas, C., Delgenes, J.P., Steyer, J.P.,
Escudie, R., 2012. Total solids content drives high solid anaerobic digestion via mass
transfer limitation. Bioresource Technology 111, 55e61.
Abbassi-Guendouz, A., Trably, E., Hamelin, J., Dumas, C., Steyer, J.P., Delgenes, J.P.,
Escudie, R., 2013. Microbial community signature of high-solid content methanogenic
ecosystems. Bioresource Technology 133, 256e262.
Aiba, S., Shoda, M., Nagatani, M., 1968. Kinetics of product inhibition in alcohol
fermentation. Biotechnology and Bioengineering 10, 845e864.
Andrews, J.F., 1968. A mathematical model for the continuous culture of microorganisms
utilizing inhibitory substrates. Biotechnology and Bioengineering 10, 707e723.
Andrews, J.F., Graef, S.P., 1971. Dynamic Modeling and Simulation of the Anaerobic
Digestion Process.
Angelidaki, I., Ahring, B.K., 1993. Thermophilic anaerobic digestion of livestock waste- the
effect of ammonia. Applied Microbiology and Biotechnology 38, 560e564.
Angelidaki, I., Ellegaard, L., Ahring, B.K., 1993. A mathematical model for dynamic
simulation of anaerobic digestion of complex substrates - focusing on ammonia
inhibition. Biotechnology and Bioengineering 42, 159e166.
Angelidaki, I., Ellegaard, L., Ahring, B.K., 1999. A comprehensive model of anaerobic
bioconversion of complex substrates to biogas. Biotechnology and Bioengineering 63,
363e372.
Appels, L., Baeyens, J., Degreve, J., Dewil, R., 2008. Principles and potential of the anaerobic
digestion of waste-activated sludge. Progress in Energy and Combustion Science 34,
755e781.
Armstrong, E.F., 1930. Enzymes. By J.B.S. Haldane, M.A. Monographs on Biochemistry.
Edited by R.H.A. Plimmer, D.Sc., and Sir F. G. Hopkins, M.A., M.B., D.Sc., F.R.S.
Pp. viiþ235. London: Longmans, Green & Co., 1930. Price 14s. Journal of the Society
of Chemical Industry 49, 919e920.
Arnell, M., Astals, S., Åmand, L., Batstone, D.J., Jensen, P.D., Jeppsson, U., 2016. Modelling
anaerobic co-digestion in Benchmark Simulation Model No. 2: parameter estimation,
substrate characterisation and plant-wide integration. Water Research 98, 138e146.
Azargoshasb, H., Mousavi, S.M., Amani, T., Jafari, A., Nosrati, M., 2015. Three-phase CFD
simulation coupled with population balance equations of anaerobic syntrophic acidogen-
esis and methanogenesis reactions in a continuous stirred bioreactor. Journal of Industrial
and Engineering Chemistry 27, 207e217.
Baeyens, J., Kang, Q., Appels, L., Dewil, R., Lv, Y., Tan, T., 2015. Challenges and
opportunities in improving the production of bio-ethanol. Progress in Energy and
Combustion Science 47, 60e88.
Bai, J., Liu, H., Yin, B., Ma, H.J., 2015. Modeling of enhanced VFAs production from waste
activated sludge by modified ADM1 with improved particle swarm optimization for
parameters estimation. Biochemical Engineering Journal 103, 22e31.
Baltes, M., Schneider, R., Sturm, C., Reuss, M., 1994. Optimal experimental design for
parmeter estimation in unstructured growth-models. Biotechnology Progress 10,
480e488.
Bastin, G., 1990. Adaptive Control of Bioreactors.
Batstone, D.J., Keller, J., 2003. Industrial applications of the IWA anaerobic digestion model
no. 1 (ADM1). Water Science and Technology 47, 199e206.
Batstone, D.J., Keller, J., Angelidaki, I., Kalyuzhnyi, S.V., Pavlostathis, S.G., Rozzi, A.,
Sanders, W.T.M., Siegrist, H., Vavilin, V.A., 2002a. IWA Anaerobic Digestion Model
No. 1 (ADM1). IWA Publishing, London, U.K.
Batstone, D.J., Keller, J., Angelidaki, I., Kalyuzhnyi, S.V., Pavlostathis, S.G., Rozzi, A.,
Sanders, W.T.M., Siegrist, H., Vavilin, V.A., 2002b. The IWA anaerobic digestion
model no. 1 (ADM 1). Water Science and Technology 45, 65e73.
Batstone, D.J., Keller, J., Steyer, J.P., 2006. A review of ADM1 extensions, applications, and
analysis: 2002-2005. Water Science and Technology 54, 1e10.
Batstone, D.J., Pind, P.F., Angelidaki, I., 2003. Kinetics of thermophilic, anaerobic oxidation
of straight and branched chain butyrate and valerate. Biotechnology and Bioengineering
84, 195e204.
Bergter, F., 1983. Wachstum von Mikroorganismen: Experimente und Modelle. 2. Auflage,
Jena.
Bollon, J., Le-hyaric, R., Benbelkacem, H., Buffiere, P., 2011. Development of a kinetic
model for anaerobic dry digestion processes: focus on acetate degradation and moisture
content. Biochemical Engineering Journal 56, 212e218.
Boyle, W.C., 1976. Energy recovery from sanitary landfills. In: Schlegel, H.G., Barnea, J.
(Eds.), Microbial Energy Conversion.
Buswell, A.M., Mueller, H.F., 1952. Mechanism of methane fermentation. Industrial &
Engineering Chemistry 44, 550e552.
Carbonell, F., Iturria-Medina, Y., Jimenez, J.C., 2016. Multiple shooting-local linearization
method for the identification of dynamical systems. Communications in Nonlinear
Science and Numerical Simulation 37, 292e304.
Chakraborty, M., Sharma, C., Pandey, J., Singh, N., Gupta, P.K., 2011. Methane emission
estimation from landfills in Delhi: a comparative assessment of different methodologies.
Atmospheric Environment 45, 7135e7142.
Chen, Y.R., Hashimoto, A.G., 1979. Kinetics of Methane Fermentation.
Cibis, K.G., Gneipel, A., K€ onig, H., 2016. Isolation of acetic, propionic and butyric acid-
forming bacteria from biogas plants. Journal of Biotechnology 220, 51e63.
Contois, D.E., 1959. Kinetics of bacterial growth - relationship between population density
and specific growth rate of continuous cultures. Journal of General Microbiology 21,
40e50.
Dahlquist, E., 2013. Technologies for Converting Biomass to Useful Energy: Combustion,
Gasification, Pyrolysis, Torrefaction and Fermentation. CRC Press.
Derbal, K., Bencheikh-lehocine, M., Cecchi, F., Meniai, A.H., Pavan, P., 2009. Application
of the IWA ADM1 model to simulate anaerobic co-digestion of organic waste
with waste activated sludge in mesophilic condition. Bioresource Technology 100,
1539e1543.
Dewil, R., Lauwers, J., Appels, L., Gins, G., Degreve, J., Impe, J.V., 2011. Anaerobic
digestion of biomass and waste: current trends in mathematical modeling. In:
International Federation of Automatic Control, August 28-September 2, 5024 5033.
Dochain, D., Perrier, M., 1998. Monitoring and adaptive control of bioprocesses.
Advanced Instrumentation, Data Interpretation, and Control of Biotechnological
Processes 347e400.
Donoso-Bravo, A., Mailier, J., Martin, C., Rodriguez, J., Aceves-Lara, C.A., Vande
Wouwer, A., 2011. Model selection, identification and validation in anaerobic digestion:
a review. Water Research 45, 5347e5364.
Eberl, H.J., 2003. Simulation of chemical reaction fronts in anaerobic digestion of solid waste.
In: Kumar, V., Gavrilova, M.L., Tan, C.J.K., Lecuyer, P. (Eds.), Computational Science
and its Applications - Iccsa 2003, Pt 1, Proceedings.
Fdez-G€
uelfo, L.A., Alvarez-Gallego, C., Sales, D., Romero García, L.I., 2012. Dry-
thermophilic anaerobic digestion of organic fraction of municipal solid waste: methane
production modeling. Waste Management 32, 382e388.
Fedorovich, V., Lens, P., Kalyuzhnyi, S., 2003. Extension of anaerobic digestion model
no. 1 with processes of sulfate reduction. Applied Biochemistry and Biotechnology
109, 33e45.
Fekih-Salem, R., Abdellatif, N., Sari, T., Harmand, J., 2012. On a three step model of
anaerobic digestion including the hydrolysis of particulate matter. IFAC Proceedings
Volumes 45, 671e676.
Fernandez, J., Perez, M., Romero, L.I., 2010. Kinetics of mesophilic anaerobic digestion of
the organic fraction of municipal solid waste: influence of initial total solid concentration.
Bioresource Technology 101, 6322e6328.
Flotats, X., Ahring, B.K., Angelidaki, I., 2003. Parameter identification of thermophilic
anaerobic degradation of valerate. Applied Biochemistry and Biotechnology 109, 47e62.
Fox, E.J., Clanton, C.J., Goodrich, P.R., Backus, B.D., Morris, H.A., 1992. Liming an
anaerobic cheese whey digester. Transactions of the ASAE 35, 269e274.
Galí, A., Benabdallah, T., Astals, S., Mata-Alvarez, J., 2009. Modified version of ADM1
model for agro-waste application. Bioresource Technology 100, 2783e2790.
García-Gen, S., Sousbie, P., Rangaraj, G., Lema, J.M., Rodríguez, J., Steyer, J.-P.,
Torrijos, M., 2015. Kinetic modelling of anaerobic hydrolysis of solid wastes, including
disintegration processes. Waste Management 35, 96e104.
Gavala, H.N., Angelidaki, I., Ahring, B.K., 2003. Kinetics and modeling of anaerobic
digestion process. Biomethanation I 81, 57e93.
Gavala, H.N., Skiadas, I.V., Lyberatos, G., 1999. On the performance of a centralised
digestion facility receiving seasonal agroindustrial wastewaters. Water Science and
Technology 40, 339e346.
Gerber, M., Span, R., 2008. An analysis of available mathematical models for anaerobic
digestion of organic substances for production of biogas. In: International Gas Union
Research Conference.
Grant, D.J.W., 1967. Kinetic aspects of the growth of Klebsiella aerogenes with some
Benzenoid carbon sources. Microbiology 46, 213e224.
Gujer, W., Zehnder, A.J.B., 1983a. Conversion processes in anaerobic-digestion. Water
Science and Technology 15, 127e167.
Gujer, W., Zehnder, A.J.B., 1983b. Conversion processes in anaerobic digestion. Water
Gyorgy, A., Kocsis, L., 2011. Efficient multi-start strategies for local search algorithms.
Journal of Artificial Intelligence Research 41, 407e444.
Herrmann, C., Kalita, N., Wall, D., Xia, A., Murphy, J.D., 2016. Optimised biogas
production from microalgae through co-digestion with carbon-rich co-substrates.
Hill, D.T., 1982. A comprehensive dynamic model for animal waste methanogenesis.
Transactions of the ASAE 1374e1380.
Hill, D.T., Tollner, E.W., Holmberg, R.D., 1983. The kinetics of inhibition in methane
fermentation of swine manure. Agricultural Wastes 5, 105e123.
Hill, D.T., Barth, C.L., 1977. A dynamic model for simulation of animal waste digestion.
Journal Water Pollution Control Federation 2129e2143.
Hosaini, Y., Homaee, M., Karimian, N., Saadat, S., 2009. Modeling vegetative
stage response of canola (Brassica napus L.) to combined salinity and boron tresses. Inter-
national Journal of Plant Production 3, 91e104.
Jacob, S., Banerjee, R., 2016. Modeling and optimization of anaerobic co-digestion of potato
waste and aquatic weed by response surface methodology and artificial neural network
coupled genetic algorithm. Bioresource Technology 214, 386e395.
Jakeman, A.J., Letcher, R.A., Norton, J.P., 2006. Ten iterative steps in development
and evaluation of environmental models. Environmental Modelling & Software 21,
602e614.
Kalyuzhnyi, S., Veeken, A., Hamelers, B., 2000. Two-particle model of anaerobic solid state
fermentation. Water Science and Technology 41, 43e50.
Karim, K., Klasson, K.T., Drescher, S.R., Ridenour, W., Borole, A.P., Al-Dahhan, M.H.,
2007. Mesophilic digestion kinetics of manure slurry. Applied Biochemistry and
Biotechnology 142, 231e242.
Karim, K., Varma, R., Vesvikar, M., Al-Dahhan, M.H., 2004. Flow pattern visualization of a
simulated digester. Water Research 38, 3659e3670.
Karthikeyan, O., Visvanathan, C., 2013. Bio-energy recovery from high-solid organic
substrates by dry anaerobic bio-conversion processes: a review. Reviews in Environ-
mental Science and Bio/Technology 12, 257e284.
Karthikeyan, O.P., Selvam, A., Wong, J.W.C., 2016. Hydrolysis-acidogenesis of food waste
in solid-liquid-separating continuous stirred tank reactor (SLS-CSTR) for volatile
organic acid production. Bioresource Technology 200, 366e373.
Kaseng, K., Ibrahim, K., Paneerselvam, S.V., Hassan, R.S., 1992. Extracellular enzymes and
acidogen profiles of a laboratory scale 2-phase anaerobic digestion system. Process
Biochemistry 27, 43e47.
Kirkpatrick, S., 1984. Optimization by simulated annealing - quantitative studies. Journal of
Statistical Physics 34, 975e986.
Kythreotou, N., Florides, G., Tassou, S.A., 2014. A review of simple to scientific models for
anaerobic digestion. Renewable Energy 71, 701e714.
Lauwers, J., Appels, L., Thompson, I.P., Degreve, J., Van Impe, J.F., Dewil, R., 2013.
Mathematical modelling of anaerobic digestion of biomass and waste: power and
limitations. Progress in Energy and Combustion Science 39, 383e402.
Lay, J.-J., Li, Y.-Y., Noike, T., 1997. Influences of pH and moisture content on the methane
production in high-solids sludge digestion. Water Research 31, 1518e1524.
Le Hyaric, R., Benbelkacem, H., Bollon, J., Bayard, R., Escudie, R., Buffiere, P., 2012.
Influence of moisture content on the specific methanogenic activity of dry mesophilic
municipal solid waste digestate. Journal of Chemical Technology and Biotechnology
87, 1032e1035.
Lee, M.-Y., Suh, C.-W., Ahn, Y.-T., Shin, H.-S., 2009. Variation of ADM1 by using
temperature-phased anaerobic digestion (TPAD) operation. Bioresource Technology
100, 2816e2822.
Leon, E., Martin, M., 2016. Optimal production of power in a combined cycle from manure
based biogas. Energy Conversion and Management 114, 89e99.
Lewis, R.M., Torczon, V., Trosset, M.W., 2000. Direct search methods: then and now.
Journal of Computational and Applied Mathematics 124, 191e207.
Li, H.L., Wang, Y., 2011. Influence of total solid and stirring frequency on performance of dry
anaerobic digestion treating cattle manure. In: Zhouzhou, Y., Luo, Q. (Eds.), Chemical,
Mechanical and Materials Engineering. Stafa-Zurich: Trans Tech Publications Ltd.
Li, Y., Jin, Y., Li, J., Li, H., Yu, Z., 2016. Effects of thermal pretreatment on the biomethane
yield and hydrolysis rate of kitchen waste. Applied Energy 172, 47e58.
Li, Y.B., Park, S.Y., Zhu, J.Y., 2011. Solid-state anaerobic digestion for methane production
from organic waste. Renewable & Sustainable Energy Reviews 15, 821e826.
Liew, L.N., Shi, J., Li, Y., 2012. Methane production from solid-state anaerobic digestion of
lignocellulosic biomass. Biomass and Bioenergy 46, 125e132.
Liotta, F., Chatellier, P., Esposito, G., Fabbricino, M., Frunzo, L., van Hullebusch, E.D.,
Lens, P.N.L., Pirozzi, F., 2014. Modified anaerobic digestion model no.1 for dry and
semi-dry anaerobic digestion of solid organic waste. Environmental Technology 36,
870e880.
Liu, C., Li, H., Zhang, Y.Y., Chen, Q.W., 2016. Characterization of methanogenic activity
during high-solids anaerobic digestion of sewage sludge. Biochemical Engineering
Journal 109, 96e100.
Liu, Y., Zhang, Y., Quan, X., Li, Y., Zhao, Z., Meng, X., Chen, S., 2012. Optimization of
anaerobic acidogenesis by adding Fe-0 powder to enhance anaerobic wastewater
treatment. Chemical Engineering Journal 192, 179e185.
Loehr, R.C., 1968. Anaerobic treatment of wastes. Developments in Industrial Microbiology
9, 160e174.
L
opez-Jiménez, P.A., Escudero-Gonzalez, J., Montoya Martínez, T., Fajardo
Monta~ nana, V., Gualtieri, C., 2015. Application of CFD methods to an anaerobic
digester: the case of Ontinyent WWTP, Valencia, Spain. Journal of Water Process
Engineering 7, 131e140.
Lyberatos, G., Skiadas, I.V., 1999. Modelling of anaerobic digestion - a review. Global Nest:
the International Journal 1, 63e76.
Marquardt, D.W., 1963. An algorithm for least-squares estimation of nonlinear parameters.
Journal of the Society for Industrial and Applied Mathematics 11, 431e441.
Martin, D.J., Potts, L.G.A., Heslop, V.A., 2003. Reaction mechanisms in solid-state
anaerobic digestion: 1. The reaction front hypothesis. Process Safety and Environmental
Protection 81, 171e179.
Martin, M.A., Raposo, F., Borja, R., Martin, A., 2002. Kinetic study of the anaerobic
digestion of vinasse pretreated with ozone, ozone plus ultraviolet light, and ozone plus
ultraviolet light in the presence of titanium dioxide. Process Biochemistry 37, 699e706.
Mata-Alvarez, J., Dosta, J., Romero-G€ uiza, M.S., Fonoll, X., Peces, M., Astals, S., 2014. A
critical review on anaerobic co-digestion achievements between 2010 and 2013.
Renewable and Sustainable Energy Reviews 36, 412e427.
McCarty, P.L., Smith, D.P., 1986. Anaerobic wastewater treatment. Environmental Science
& Technology 20, 1200e1206.
Mendes, C., Esquerre, K., Queiroz, L.M., 2015. Application of anaerobic digestion model
no. 1 for simulating anaerobic mesophilic sludge digestion. Waste Management 35,
89e95.
Mendez-Acosta, H.O., Palacios-Ruiz, B., Alcaraz-Gonzalez, V., Gonzalez-Alvarez, V.,
Garcia-Sandoval, J.P., 2010. A robust control scheme to improve the stability of
anaerobic digestion processes. Journal of Process Control 20, 375e383.
Mitsd€orffer, R., 1991. Charakteristika der zweistufigen thermophilen/mesophilen Schlamm-
faulung unter Ber€ ucksichtigung kinetischer Ans€atze. Berichte aus Wasserg€ ute- und
Abfallwirtschaft, Technische Universit€at M€ unchen 109.
Moesche, M., Joerdening, H.-J., 1999. Comparison of different models of substrate and
product inhibition in anaerobic digestion. Water Research 33, 2545e2554.
Monod, J., 1949. The growth of bacterial cultures. Annual Review of Microbiology 3,
371e394.
Morgenroth, E., van Loosdrecht, M.C.M., Wanner, O., 2000. Biofilm models for the
practitioner. Water Science and Technology 41, 509e512.
Moser, H., 1958. The dynamics of bacterial populations maintained in the chemostat.
Carnegie Institution of Washington Publication 614, 1e136.
Mosey, F.E., 1983. Mathematical modeling of the anaerobic digestion process - regulatory
mechanisms for the formation of short-chain volatile fatty acids from glucose. Water
Motte, J.C., Escudié, R., Bernet, N., Delgenes, J.P., Steyer, J.P., Dumas, C., 2013. Dynamic
effect of total solid content, low substrate/inoculum ratio and particle size on solid-state
anaerobic digestion. Bioresource Technology 144, 141e148.
Mulka, R., Szulczewski, W., Szlachta, J., Mulka, M., 2016. Estimation of methane
production for batch technology - a new approach. Renewable Energy 90, 440e449.
Nguyen, H.H., Heaven, S., Banks, C., 2014. Energy potential from the anaerobic digestion
of food waste in municipal solid waste stream of urban areas in Vietnam. International
Journal of Energy and Environmental Engineering 365e374.
Noykova, A.N., Gyllenberg, M., 2000. Sensitivity analysis and parameter estimation in a
model of anaerobic waste water treatment processes with substrate inhibition. Bioprocess
Odriozola, M., L opez, I., Borzacconi, L., 2016. Modeling granule development and reactor
performance on anaerobic granular sludge reactors. Journal of Environmental Chemical
Omlin, M., Reichert, P., 1999. A comparison of techniques for the estimation of model
prediction uncertainty. Ecological Modelling 115, 45e59.
Palatsi, J., Illa, J., Prenafeta-Boldu, F.X., Laureni, M., Fernandez, B., Angelidaki, I.,
Flotats, X., 2010. Long-chain fatty acids inhibition and adaptation process in anaerobic
thermophilic digestion: batch tests, microbial community structure and mathematical
modelling. Bioresource Technology 101, 2243e2251.
Poggio, D., Walker, M., Nimmo, W., Ma, L., Pourkashanian, M., 2016. Modelling the
anaerobic digestion of solid organic waste e substrate characterisation method for
ADM1 using a combined biochemical and kinetic parameter estimation approach. Waste
Management 53, 40e54.
Pommier, S., Chenu, D., Quintard, M., Lefebvre, X., 2007. A logistic model for the
prediction of the influence of water on the solid waste methanization in landfills.
Biotechnology and Bioengineering 97, 473e482.
Ramirez, I., Mottet, A., Carrere, H., Deleris, S., Vedrenne, F., Steyer, J.P., 2009.
Modified ADM1 disintegration/hydrolysis structures for modeling batch thermophilic
anaerobic digestion of thermally pretreated waste activated sludge. Water Research 43,
3479e3492.
Reichert, P., 1998. AQUASIM 2.0duser Manual. Swiss Federal Institute for Environmental
Science and Technology, Dubendorf, Switzerland.
Sanders, W.T.M., Geerink, M., Zeeman, G., Lettinga, G., 2000. Anaerobic hydrolysis
kinetics of particulate substrates. Water Science and Technology 41, 17e24.
Speece, R., 1983a. Anaerobic biotechnology for industrial wastewater treatment.
Environmental Science and Technology 17, 416Ae426A.
Speece, R.E., 1983b. Anaerobic biotechnology for industrial wastewater treatment.
Environmental Science & Technology 17, A416eA427.
Spriet, J.A., 1985. Structure characterization e an overview. In: Barker, H.A., Young, P.C.
(Eds.), Identification and System Parameter Estimation. Pergamon Press, Oxford.
Steyer, J.P., Bernard, O., Batstone, D.J., Angelidaki, I., 2006. Lessons learnt from 15 years of
ICA in anaerobic digesters. Water Science and Technology 53, 25e33.
Tchobanoglous, G., Burton, F.L., Stensel, H.D., Metcalf, Eddy, I., 2003. Wastewater
Engineering: Treatment and Reuse. McGraw-Hill Education.
Valentini, A., Garuti, G., Rozzi, A., Tilche, A., 1997. Anaerobic degradation kinetics
of particulate organic matter: a new approach. Water Science and Technology 36,
239e246.
Vavilin, V.A., Angelidaki, I., 2005. Anaerobic degradation of solid material: importance of
initiation centers for methanogenesis, mixing intensity, and 2D distributed model.
Biotechnology and Bioengineering 89, 113e122.
Vavilin, V.A., Fernandez, B., Palatsi, J., Flotats, X., 2008. Hydrolysis kinetics in anaerobic
degradation of particulate organic material: an overview. Waste Management 28,
939e951.
Vavilin, V.A., Rytov, S.V., Lokshina, J.Y., Pavlostathis, S.G., Barlaz, M.A., 2003. Distributed
model of solid waste anaerobic digestion - effects of leachate recirculation and pH
adjustment. Biotechnology and Bioengineering 81, 66e73.
Vavilin, V.A., Rytov, S.V., Lokshina, L.Y., 1996. A description of hydrolysis kinetics in
anaerobic degradation of particulate organic matter. Bioresource Technology 56,
229e237.
Veeken, A., Hamelers, B., 1999. Effect of temperature on hydrolysis rates of selected
biowaste components. Bioresource Technology 69, 249e254.
Wang, Z.-W., Li, Y., 2014. A theoretical derivation of the Contois equation for kinetic
modeling of the microbial degradation of insoluble substrates. Biochemical Engineering
Journal 82, 134e138.
Wang, Z.W., Chen, S.L., 2009. Potential of biofilm-based biofuel production. Applied
Microbiology and Biotechnology 83, 1e18.
Wanner, O., Eberl, H.J., Morgenroth, E., Noguera, D.R., Picioreanu, C., Rittmann, B.E.,
van Loosdrecht, M.C.M., 2006. Mathematical Modeling of Biofilms. IWA Publishing.
Webb, J.L., 1963. Enzyme and Metabolic Inhibitors. Academic Press, New York.
Weedermann, M., Wolkowicz, G.S.K., Sasara, J., 2015. Optimal biogas production in a
model for anaerobic digestion. Nonlinear Dynamics 81, 1097e1112.
Xu, F., Wang, Z.-W., Tang, L., Li, Y., 2014. A mass diffusion-based interpretation of the
effect of total solids content on solid-state anaerobic digestion of cellulosic biomass.
Xu, F.Q., Li, Y.B., Wang, Z.W., 2015. Mathematical modeling of solid-state anaerobic
digestion. Progress in Energy and Combustion Science 51, 49e66.
Xu, F.Q., Shi, J., Lv, W., Yu, Z.T., Li, Y.B., 2013. Comparison of different liquid anaerobic
digestion effluents as inocula and nitrogen sources for solid-state batch anaerobic
digestion of corn stover. Waste Management 33, 26e32.
Yano, T., Nakahara, T., Kamiyama, S., Yamada, K., 1966. Kinetic studies on microbial
activities in concentrated solutions. Agricultural and Biological Chemistry 30, 42e48.
Yu, L., Wensel, P.C., Ma, J., Chen, S., 2013. Mathematical modeling in anaerobic digestion
(AD). Journal of Bioremediation & Biodegradation S4, 003.
Zaher, U., Buffiere, P., Steyer, J.P., Chen, S., 2009a. A procedure to estimate proximate
analysis of mixed organic wastes. Water Environment Research 81, 407e415.
Zaher, U., Li, R.P., Jeppsson, U., Steyer, J.P., Chen, S.L., 2009b. GISCOD: general
integrated solid waste Co-Digestion model. Water Research 43, 2717e2727.
Zhang, C., Su, H., Baeyens, J., Tan, T., 2014. Reviewing the anaerobic digestion of food
waste for biogas production. Renewable and Sustainable Energy Reviews 38, 383e392.
Zhang, H.Q., Liu, Y.Q., Liu, B., Gao, P.J., 2006. A novel approach for estimating
growth phases and parameters of bacterial population in batch culture. Science in China
Series C-Life Sciences 49, 130e140.
Zonta, Z., Alves, M.M., Flotats, X., Palatsi, J., 2013. Modelling inhibitory effects of long
chain fatty acids in the anaerobic digestion process. Water Research 47, 1369e1380.

Anaerobic Digestion Modelling

Uploaded by

Copyright:

Available Formats

Anaerobic Digestion Modelling

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Anaerobic Digestion Modelling

Uploaded by

Copyright:

Available Formats

CHAPTER TWO

Anaerobic Digestion Modelling

Advances in Bioenergy, Volume 2

4.3 Forms of Living 105

2. BASIC MODELS USED IN ANAEROBIC DIGESTION

Complex organic matter

1. Hydrolysis by extracellular enzymes

Carbohydrates Proteins Lipids or fats

Monosaccharides Amino acids Long chain fatty acids

3. Acetogenesis by acetogens Hydrogen

4. Methanogenesis by methanogens Methane,

hydrogen to be at low concentrations in the system (H2 < 104 atm).

C3 H6 O2 þ 1:764H2 O þ 0:0458NH3 /0:0458C5 H7 NO2

C5 H10 O2 þ 0:8045H2 O þ 0:0653NH3

(Liu et al., 2016). Hydrogenotrophic methanogens are responsible for

2.2 Kinetic and Equilibrium Models

2.2.1 Microbial Growth and Substrate Utilization Kinetics

The growth rate of microorganism is deﬁned by Eq. (18)

2.2.2 Decay Kinetics

Table 1 Summary of anaerobic digestion models using Monod-like kinetics

Monod (1949) m ¼ mmax $KSSþS

biomass is assumed to contribute to the amount of insoluble substrate for

2.2.3 Hydrolysis Kinetics

another hydrolysis model proposed by Contois was developed by taking

2.2.4 AcideBase Equilibrium and Kinetics

in which SCatþ and SAn are concentrations (k mole/m3) of metal ions

2.2.5 GaseLiquid Equilibrium

Hence, the dynamic gas concentration in the headspace of an anaerobic

dSgas qgas $Sgas rT $Vliq

2.2.6 Diffusion Kinetics

in which JD is the diffusion ﬂux (g/m2 d); De is diffusion coefﬁcient (m2/d);

When modelling solid-state anaerobic digestion (SS-AD) or bioﬁlm

2.2.7 Temperature Effect on Reaction Kinetics and Thermodynamics

reaction phenomena generally observed due to temperature change

3. COMMON MODEL FRAMEWORKS

3.1 Rate-Limitation Step

3.2 Anaerobic Digestion Model No. 1

generated in each step (Gavala et al., 2003). International Water Association

Glucose (C6H12O6) is considered as the primary monosaccharide

C6 H12 O6 þ 2H2 O/2CH3 COOH þ 2CO2 þ 4H2 (48)

3C6 H12 O6 /4CH3 CH2 COOH þ 2CH3 COOH þ 2CO2 þ 2H2 O

C6 H12 O6 /CH3 CH2 CH2 COOH þ 2CO2 þ 2H2 (50)

CH3 CH2 COOH þ 2H2 O/CH3 COOH þ CO2 þ 3H2 (51)

CH3 COOH/CH4 þ CO2 (52)

CO2 þ 4H2 /CH4 þ 2H2 O (53)

Table 2 Form of inhibition considered in anaerobic digestion model no. 1

Noncompetitive I ¼ 1þS1I =KI (54)

Competitive uptake I ¼ 1þS1I =S (57)

As shown in Fig. 4, the model concept of a typical single-tank anaerobic

Input Gas phase Output

Figure 4 Illustration of kinetics in an anaerobic digester.

in which Sliq is soluble substrate concentration (g/L); P Sin is the initial

dSHCO3 qout $SiHCO3

4. MODELLING MAJOR FACTORS AFFECTING

accumulation of VFAs, hydrogen and ammonia. Accumulation of interme-

4.1 Substrate Composition

on biodegradability) with high C/N and poor buffering capacity may

The mass balance on the three components is expressed as,

in which rmS is the maximum speciﬁc hydrolysis rate (d1); rB and rS

in which CH4(t) represents the cumulative methane production at time t

Webb (1963) m ¼ mmaxðSÞð1þbS=K