You are on page 1of 11

Bioresource Technology 96 (2005) 1832–1842

Application of the ADM1 model to advanced anaerobic digestion


Wayne J. Parker *

Department of Civil Engineering, University of Waterloo, Waterloo, Ont., Canada N2L 3G1

Received 20 November 2003; received in revised form 3 January 2005; accepted 5 January 2005
Available online 8 March 2005

Abstract

In this paper the ADM1 model that has been developed by the IWA Task Group for Mathematical Modelling of Anaerobic
Digestion Processes is summarized. The model was applied to a variety of anaerobic digestion scenarios that are presented in the
literature and for each data set the model predictions were compared to experimental values. Based upon the model applications
it was apparent that for accurate model simulations the influent sludge should be well characterized in terms of biodegradable
and recalcitrant COD and also nitrogenous compounds. In almost all cases the model was able to reflect the trends that were
observed in the experimental data however the concentrations of VFAs were consistently over-predicted in digesters with short
SRTs. It would appear that the inhibition functions associated with low pH values tend to overestimate the impact of pH on bioki-
netic rates for the acid-consuming bacteria. Application of the model with flow through of active biomass between digesters in series
in temperature-phased systems needs to be further evaluated in the future.
Ó 2005 Elsevier Ltd. All rights reserved.

Keywords: Anaerobic digestion; Two-phase; Temperature-phased; Mesophilic; Thermophilic; Model; Sludge

1. Introduction these processes. Given these factors, the use of models


for predicting process performance over a range of
Owners and operators of wastewater treatment design and operating conditions becomes attractive.
plants are increasingly considering the use of advanced Over the years a range of models have been develo-
digestion technologies for producing pathogen-free bio- ped for modeling anaerobic digestion processes. Early
solids and for enhancing sludge stabilization. Some models were steady state and assumed a rate-limiting
examples of such technologies include staged thermo- step (Lawrence, 1971). However, the increasing com-
philic (Krugel et al., 1998), temperature-phased (TPAD) plexity of the advanced digestion technologies requires
(Han et al., 1997), two-phase (Ghosh, 1987) and three- more complex models that can represent the impacts
phase digestion (Drury et al., 2002). With the increasing of changing environments on chemical and microbial
complexity of these processes it is difficult to evaluate species. Based on reports in the literature there is evi-
the impact of all process variables on the performance dence of a number of multi-species models that are
of the digesters. Hence, it is difficult to optimize the based upon differing assumptions and have differing
design and operation of these processes. Pilot testing configurations (Angelidaki et al., 1999; Pavlostathis
for the purposes of optimization is challenging due to and Gossett, 1986; Siegrist et al., 1993). Relatively re-
the extended time periods that are required to operate cently there has been a move by the International Water
AssociationÕs (IWA) Task Group for Mathematical
Modelling of Anaerobic Digestion Processes to develop
*
Tel.: +1 519 888 4567x6324; fax: +1 519 888 4349. a common model that can be used by researchers and
E-mail address: wjparker@uwaterloo.ca practitioners (IWA, 2002). This model (ADM1) has a

0960-8524/$ - see front matter Ó 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.biortech.2005.01.022
W.J. Parker / Bioresource Technology 96 (2005) 1832–1842 1833

structure that is similar to the IWA activated sludge the major processes that are involved in the conversion
models that have received acceptance by practitioners of complex organic substrates into methane and carbon
over the last 10 years. The application of a version of dioxide and inert byproducts. In Fig. 1 an overview of
the model to municipal sludge digestion has been de- the substrates and conversion processes that are ad-
scribed by Siegrist et al. (2002). dressed by the model is presented. From Fig. 1 it can
The objective of this study was to examine the appli- be seen that the model includes disintegration of com-
cation of the ADM1 model to advanced digestion tech- plex solids into inert substances, carbohydrates, proteins
nologies. This paper presents an overview of the model and fats. The products of disintegration are hydrolyzed
structure and assumptions and defines important model to sugars, amino acids and long chain fatty acids
inputs. A description of the model application to exist- (LCFA) respectively. Carbohydrates and proteins are
ing data sets for a variety of anaerobic digester configu- fermented to produce volatile organic acids (acido-
rations will be presented. The impact of modifying genesis) and molecular hydrogen. LCFA are oxidized
process parameters on process performance, as pre- anaerobically to produce acetate and molecular hydro-
dicted by the model, will be summarized. Difficulties gen. Propionate, butyrate and valerate are converted
encountered in model use and recommendations for to acetate (acetogenesis) and molecular hydrogen. Meth-
modifications will be presented. ane is produced by both cleavage of acetate to methane
(aceticlastic methanogenesis) and reduction of carbon
dioxide by molecular hydrogen to produce methane
2. Model description (hydrogenotrophic methanogenesis).
To address these mechanisms, the model employs
The ADM1 model is described in considerable detail state variables to describe the behaviour of soluble and
in the report prepared by the IWA Task Group for particulate components. All organic species and molecu-
Mathematical Modeling of Anaerobic Digestion Pro- lar hydrogen are described in terms of chemical oxygen
cesses (IWA, 2002). The following provides a brief over- demand (COD). Nitrogenous species and inorganic car-
view of the model for the purposes of this discussion. bon species are described in terms of their molar concen-
The ADM1 model is a structured model that reflects trations. Soluble components are those that can pass

Complex Particulate
Organic Matter (Xc)
Inert
Particulates
(XI)

Carbohydrates (Xch) Proteins (Xpr) Fats (Xli)


Inert
Soluble
(SI)
Amino
Acids Long Chain
Sugars Fatty Acids
(Ssu) (Saa)
(Sfa)

Propionate (Spro) Butyrate (Sbu)


Valerate (Sva)

Acetate (Sac) Hydrogen (Sh2)

Methane (Sch4)

Fig. 1. Conceptual model for ADM1 model.


1834 W.J. Parker / Bioresource Technology 96 (2005) 1832–1842

through microbial cellular walls and include the mono- processes are subject to inhibition by accumulation of
mers of complex polymers (sugars, amino acids, long molecular hydrogen and aceticlastic methanogenesis is
chain fatty acids), volatile organic acids (propionate, inhibited at elevated free ammonia concentrations. Inhi-
butyrate, valerate, acetate), hydrogen, and methane. In bition that is caused by molecular hydrogen and free
Fig. 1, soluble species are represented with a capital ammonia is implemented in the model by employing
‘‘S’’. In addition to the organic species, the model ad- rate multipliers that reflect non-competitive inhibition.
dresses inorganic carbon (carbon dioxide and bicar- An empirical correlation is employed as a process rate
bonate) and nitrogenous species (ammonia and multiplier to reflect the effects of extreme pH.
ammonium). All of the species that dissociate as a func- Liquid–gas mass transfer of gaseous components
tion of pH (VFAs and ammonia) have variables defined (methane, carbon dioxide and molecular hydrogen) is de-
for both the protonated and non-protonated species. scribed by mass transfer relationships. Hence the appli-
The model maintains a charge balance among ionic spe- cation of the model equations requires separate mass
cies and hence there are variables for inorganic anions balances for the liquid and gas phases of the components.
and cations including the hydrogen ion. The model
solves for the hydrogen ion concentration, and thereby
the pH, by ensuring chemical neutrality in solution. 3. Model application
Particulate species consist of either active biomass
species or particulate substances that are incapable of In this study a selected number of data sets were
directly passing through bacterial cell walls. In Fig. 1 chosen from previously published reports on anaerobic
particulate species are those with a capital ‘‘X’’. The digestion of municipal wastewater sludges. Data sets
microbial species that are considered in the model in- were selected to encompass a range of digester configu-
clude sugar fermenters, amino acid fermenters, LCFA rations and on the basis of the completeness of the data
oxidizers, butyrate and valerate oxidizers, propionate sets that would be employed for model inputs and for
oxidizers, aceticlastic methanogens and hydrogeno- comparison with model predictions. In all cases, studies
trophic methanogens. Non-microbial particulate species that employed actual sludges from municipal waste-
include complex organics that either enter the process in water treatment plants were selected. The data sets that
the influent or that result from the death and decay of were employed in this study are described in Table 1.
microbial species and the products of disintegration of The ADM1 model employs a large number of con-
the complex organics. This latter group consists of car- stants and coefficients. Given the model complexity it
bohydrates, proteins and LCFAs. was impossible to calibrate the model parameters with
Substrate conversion processes are described by a any of the data sets that were available. In the report
number of kinetic expressions that describe the conver- describing the ADM1 model the authors reviewed the
sion rates in terms of substrate concentrations and rate previously published reports on anaerobic digestion pro-
constants. The disintegration of Xc and hydrolysis of cesses and presented recommended values for model
Xch, Xpr and Xli are described by first order rate expres- parameters. For the purposes of this study the recom-
sions. Substrate conversion processes have Monod-type mended model parameters were employed unless addi-
kinetic expressions while endogenous decay processes tional information was provided by the original
are first order in biomass concentration. It should be researchers that allowed for an improved estimate of
noted that the ADM1 model differs from the ASM mod- the model parameters.
els in that microbially mediated processes are defined in In order to achieve accurate model predictions it
terms of substrate conversion as opposed to microbial is important to define the properties of the sludge stream
growth. For each of the above-mentioned processes entering the digester. For organic substances, the
the rate of generation of products is related to the pro- ADM1 model defines these inputs in terms of soluble
cess rate through stoichiometric coefficients. For exam- and particulate COD. For municipal sludges a majority
ple the rate of growth of an organism is related to the
rate of substrate consumption through the yield coeffi-
cient for the organism on the substrate. This format is Table 1
Data sets referenced in this study
consistent with the approach that is employed in the
ASM models. Digester configuration Sludge source References
It is recognized that a number of the conversion pro- Single stage Mixed PS and WAS Cacho Rivero et al.
cesses that are active in anaerobic digestion of municipal mesophilic digestion (2002)
Acid phase digestion PS Eastman and
sludges can be inhibited by the accumulation of interme- Ferguson (1981)
diate products such as molecular hydrogen, ammonia or Temperature-phased PS Han and Dague
by extremes of pH. In the model, all microbially medi- anaerobic digestion (1995)
ated substrate conversion processes are subject to inhibi- Mixed PS and WAS Han et al. (1997)
tion by extremes of pH. All anaerobic oxidation Two-phase digestion Mixed PS and WAS Ghosh (1987)
W.J. Parker / Bioresource Technology 96 (2005) 1832–1842 1835

of the organic loading is associated with the particulate lease of elevated concentrations of ammonia that can be
COD. The particulate COD entering the digester is de- inhibitory to aceticlastic methanogens (IWA, 2002). It is
fined in terms of biodegradable (Xc) and non-biodegrad- therefore important to characterize the concentration of
able components. Estimation of these parameters is ammonia/ammonium in the digester inlet as well as the
often challenging for many data sets as in many cases nitrogen content of the sludges. It should be noted that
the sludge COD is not reported and in almost all cases the ADM1 model does not maintain a perfect mass
the biodegradable fraction is not independently mea- balance on nitrogen (Blumensaat and Keller, 2005).
sured. In most cases the sludge is characterized in terms Ammonium that is taken up by microbial growth is
of its volatile solids content. not completely released during subsequent decay.
The relationship between volatile solids content and Hence, it can be expected that the model will underesti-
COD will depend upon the relative contribution of pri- mate the concentrations of ammonium.
mary (PS) and waste activated sludge (WAS) to the The data sets employed in this study did not contain
sludge composition (Parkin and Owen, 1986). Primary all of the information that was previously described.
sludges typically contain approximately 2.0 kg COD/kg Where necessary, typical values were assumed. The im-
VS while WAS typically has a value of 1.4 kg COD/kg pact of these assumptions on model predictions will be
VS for this parameter. The inlet COD can therefore be subsequently discussed.
estimated on the basis of these typical values if the rela-
tive contributions of PS and WAS are known. 3.1. Single stage mesophilic digestion
The biodegradable fraction of the sludge particulate
COD will also be a function of the sludge make-up. Cacho Rivero et al. (2002) reported a study that as-
Primary sludges have been estimated to have a COD sessed the impact of digester SRT on mesophilic an-
‘‘ultimate’’ biodegradability of 69% (Parkin and Owen, aerobic digestion of mixed PS and WAS. A series of
1986). The biodegradable fraction of WAS is dependent digesters were operated over SRTs ranging from 5 to
upon the sludge age that is employed in the aeration 40 days. In their paper the sludge COD, ammonia and
process (Gossett and Belser, 1982). Sludges that have ex- TKN content and VFA composition were detailed.
tended solids residence times (SRT) in the aeration basin For this study, the biodegradable COD was estimated
will have been highly oxidized and hence will be rela- by extrapolating the results that were obtained for ex-
tively recalcitrant to biodegradation in anaerobic diges- tended SRTs. In their study COD removal, ammonia
tion. The ultimate biodegradability of WAS has been and TKN content as well as VFA concentrations in
found to range from 30% to 50% over the range of SRTs the digested sludges were reported and were employed
typically employed in wastewater treatment processes. for comparison with the model predictions.
Hence, it is apparent that accurate application of the The comparison of the model predictions for effluent
model requires a detailed characterization of the inlet COD, NH4/NH3-N, and VFAs is summarized in Fig. 2.
sludge composition. The sludge composition should be The error bars in Fig. 2 represent 1 standard deviation
determined in terms of COD and the biodegradable of the experimental data. From Fig. 2 it can be seen that
fraction should be determined. This latter parameter the model was able to predict the effluent COD with
could be determined through the use of a long term considerable accuracy. Nitrogen concentrations were
batch digestion test to identify the maximum biodegrad- accurately predicted for the shorter SRTs and while
ability of the sludge. While there are no standard proto- the trend of increasing concentrations with increasing
cols for such a test, existing anaerobic biodegradability SRT was reproduced, the absolute values that were pre-
protocols could presumably be adapted for this purpose. dicted at longer SRTs were somewhat lower that the ob-
If the contribution of PS and WAS to the digester feed served values. The differences in nitrogen concentrations
were to vary substantially with time, then this testing may have been due to the lack of mass balance on nitro-
should be performed on the PS and WAS streams sepa- gen in the ADM1 model. It would be expected that
rately. The properties of the composite sludge as a func- under conditions where there is substantial solids
tion of time could subsequently be estimated. destruction that the model would underestimate the con-
The ADM1 model also estimates the behaviour of centrations of ammonium-nitrogen.
nitrogen compounds in anaerobic digestion. In the cases The differences between the model predictions and the
of municipal sludges the presence of ammonia nitrogen observed results may also have resulted from differences
in the inlet and the release of ammonia from decay of between the assumed and the actual protein content of
solids has a substantial influence on the buffering of the sludge. The reference did not provide any informa-
pH. As will be demonstrated later in this paper the con- tion on the distribution of carbohydrates, proteins and
centration of ammonia/ammonium in the inlet can have lipids in the sludge and hence the default model values
a substantial impact upon the pH of acid-phase digesters were employed for this parameter.
that have a relatively short SRT. In addition, the diges- The model predictions for VFA concentrations were
tion of highly concentrated sludges can result in the re- relatively accurate for SRTs greater than or equal to
1836 W.J. Parker / Bioresource Technology 96 (2005) 1832–1842

30 2500

25
2000

20

NH4/NH3-N (mg/L)
1500
COD (g/L)

Model Model
15
Experimental Experimental
1000
10

500
5

0 0
5 10 20 40 5 10 20 40
SRT (d) SRT (d)

1800 300

1600
250
1400

Propionic Acid (mg/L)


Acetic Acid (mg/L)

1200 200

1000
Model Model
150
Experimental Experimental
800

600 100

400
50
200

0 0
5 10 20 40 5 10 20 40
SRT (d) SRT (d)

100 100

80
80

60
Butyric Acid (mg/L)

60
Valeric Acid (mg/L)

40

Model Model
40 20
Experimental Experimental

0
20 5 10 20 40
-20

0
5 10 20 40 -40

-20 -60
SRT (d) SRT (d)

Fig. 2. Comparison of model predictions with data of Cacho Rivero et al. (2002).

10 days. However the model clearly overpredicted the inhibition of this activity by ammonia. The model pre-
concentration of acetate while underpredicting the con- dicted a 40% reduction in the activity of these organisms
centrations of propionate, butyrate and valerate. These due to the presence of ammonia. The impact of reduced
results suggest that the rates of oxidation of propionate, rates of aceticlastic activity on model predictions would
butyrate and valerate were somewhat overestimated by be greatest at the lower SRTs.
the model and this would partially, but not completely,
explain the elevated acetate concentrations. It would 3.2. Acid phase digestion
appear that the rate at which acetate was converted to
methane at the lower SRT was somewhat under- Eastman and Ferguson (1981) performed one of the
estimated. This may have resulted from either underesti- first detailed studies on the acid-phase digestion of mu-
mation of the substrate consumption coefficients for nicipal sludges. In their study, the impact of HRT was
aceticlastic methanogenesis or an overestimation of the assessed over a range from 9 to 36 h. The impact of seed
W.J. Parker / Bioresource Technology 96 (2005) 1832–1842 1837

culture was also evaluated. The model does not have the concentrations at 9 h is in agreement with the overesti-
capability to address this parameter and hence only the mation of the effluent pH in this test. It would appear
tests that were conducted with raw sludge as the seed that the model underestimated the rates of disintegra-
were employed for this analysis. The model predictions tion, hydrolysis and acidification under these relatively
for ammonia/ammonium-N, pH and total volatile acids extreme conditions of SRT and pH. It should be noted
(as acetate) were compared with the observed values in that the model does not correct any of the disintegration
Fig. 3. or hydrolysis rates for pH. Ghosh (1987) has demon-
From Fig. 3 it can be seen that the model somewhat strated that the rate of hydrolysis is influenced by pH.
underpredicted the organic acid concentrations at the An improvement of the model for addressing acid phase
lowest SRT of 9 h and overpredicted these values for digesters would be to include a rate correction term for
the longest SRT of 72 h. The underprediction of acid hydrolysis processes.
While not presented in Fig. 3 it must be noted that
12 although Eastman and Ferguson (1981) observed meth-
ane production at the longer SRTs the model did not pre-
10 dict the generation of appreciable quantities of methane
under these conditions. The conversion of VFAs to meth-
8 ane in the experimental data may explain the higher model-
VFA (g COD/L)

Model
predicted VFA concentrations relative to the observed
6
Experimental values. The results suggest that methanogens are less sen-
sitive to pH than the pH inhibition functions suggest.
4
The ammonia-nitrogen concentrations were under-
2
predicted at the lowest SRTs and overpredicted at the
highest SRTs. These results tend to confirm the model
0
predictions of VFA concentrations since an underpre-
9 18 36 72 diction of solids destruction and hydrolysis, as indicated
SRT (hrs)
by reduced VFAs, would also result in a reduced release
7 of ammonium.

6
3.3. Temperature-phased anaerobic digestion (TPAD)
5
TPAD processes consist of reactors operating at ther-
4
Model
mophilic and mesophilic temperatures in series. While
pH

Experimental either process may be first, the most common orientation


3
has the thermophilic digester ahead of the mesophilic di-
2 gester. For the purposes of this study two papers on
TPAD digestion were referenced; one that studied diges-
1
tion of PS alone (Han and Dague, 1995) and one that
0
studied a mix of PS and WAS (Han et al., 1997). In the
9 18 36 72 former paper the ratio of the volumes of the first and sec-
SRT (hrs)
ond digesters was 1:2. In the latter paper two systems
800 were operated with system A having a ratio of volumes
of 1:2.5 while system B had a ratio of volumes of 1:5.
700
In all of the systems the mesophilic temperature was
600 35 °C while the thermophilic temperature was 55 °C. A
500
comparison of the model predictions with the data pre-
NH4-N (mg/L)

Model
sented in the paper of Han and Dague (1995) is summa-
400
Experimental rized in Fig. 4. The comparison of model predictions
300 with the results of Han et al. (1997) are presented for sys-
tems A and B in Figs. 5 and 6 respectively.
200
It should be noted that the model does not explicitly
100 predict volatile solids removal (VSR). For the purposes
0
of this paper it was assumed that the removal of volatile
9 18 36 72 solids was proportional to the removal of overall
SRT (hrs)
COD. This assumes that all of the COD remaining
Fig. 3. Comparison of model predictions with data of Eastman and after digestion have the same ratio of volatile solids
Ferguson (1981). concentration:COD. This undoubtedly introduces some
1838 W.J. Parker / Bioresource Technology 96 (2005) 1832–1842

16 60

14
50

12
CH4 Production (L/d)

40
10

VSR (%)
Model Model
8 30
Experimental Experimental

6
20

10
2

0 0
10 11.5 12.5 13.6 15 10 11.5 12.5 13.6 15
SRT (d) SRT (d)

9000 2500

8000
1st Stage
2000
2nd Stage
7000

6000
VFA (mg/L)

1500

VFA (mg/L)
5000
Model Model
Experimental Experimental
4000
1000
3000

2000
500
1000

0 0
10 11.5 12.5 13.6 15 10 11.5 12.5 13.6 15
SRT (d) SRT (d)

Fig. 4. Comparison of model predictions with data of Han and Dague (1995).

error in the estimates however, there was generally insuf- The results suggest that for thermophilic conditions
ficient data on the composition of the digester effluent to the model overpredicts the generation of volatile fatty
perform a more refined conversion of COD to solids acids and that this is accentuated at shorted SRTs. It
concentrations. should be noted that at low SRTs the model predicted
The patterns with respect to the model predictions substantial inhibition of the acetoclastic methanogens
and observed data that are presented in Figs. 4–6 are due to low pH (IWA, 2002). It may be that the inhibi-
consistent. In all three cases, the model overpredicted tion functions for this process were too severe.
the production of methane by the temperature-phased Although the model predicted high VFA concentra-
processes. It should be noted that in the papers only tions in the thermophilic phase reactor, it predicted that
total methane production was reported and hence it essentially all of the VFAs could be converted in the sec-
was not possible to compare methane production from ond phase mesophilic reactor. The predicted effluent
the two reactors separately. In all cases the extent of concentrations were actually lower than those that were
overprediction was greatest for the lower SRTs and pre- observed. It should be noted that in this modeling effort
dictions improved for the longer SRTs. The predictions the biomass that was present in the first phase reactor
for VSR were best for the results presented in Fig. 4 was allowed to flow into the second phase reactor and
while in Figs. 5 and 6 the model consistently overpre- remain active at the new temperature. This may have re-
dicted the VSR. The overprediction of VSR was consis- sulted in the overprediction of activity in the latter reac-
tent with the overprediction of methane generation. In tor as it is unlikely that all of the thermophilic biomass
all three cases the model substantially overpredicted leaving the first digester would remain viable in the sec-
the concentrations of VFAs in the thermophilic reactor. ond stage digester. It may be more appropriate to as-
The greatest overprediction was associated with the sume that the biomass entering the second digester
shortest SRTs and the predictions improved at longer should be considered as biodegradable particulate or-
SRTs. With the exception of the 10 day SRT in Fig. 4 ganic matter.
the model tended to underpredict the concentrations The differences between the model predictions and
of VFAs in the mesophilic second stage digester. the observed values of the VFA concentrations may also
W.J. Parker / Bioresource Technology 96 (2005) 1832–1842 1839

12 60

10 50
CH4 Production (L/d)

8 40

VSR (%)
Model Model
6 30
Experimental Experimental

4 20

2 10

0 0
14 20 28 14 20 28
SRT (d) SRT (d)

6000 250
2nd Stage
5000 1st Stage
200

4000
VFA (mg/L)

150
VFA (mg/L)
Model Model
3000
Experimental Experimental
100
2000

1000 50

0 0
14 20 28 14 20 28
SRT (d) SRT (d)

Fig. 5. Comparison of system a model predictions with data of Han et al. (1997).

have been due to the procedure used to adjust the bioki- the 3 day SRT the first stage had an SRT of 0.9 days
netic coefficients for temperature. In the model docu- while the second stage had an SRT of 2.1 days.
mentation (IWA, 2002) it is suggested that a constant With the 7 day SRT the first stage had an SRT of
correction factor be employed for all of the microbial 2 days and the second stage had an SRT of 5 days. In
species. Implementing this strategy tends to result in Tables 2 and 3 a comparison of some of the model
an accumulation of VFAs at the higher temperatures. predictions and the reported experimental values are
It may be that differing temperature correction factors presented.
should be employed for the different microbial From Table 2 it can be see that with the exception of
species. the first stage of the 7 day SRT digesters the model pre-
dictions for VFAs were relatively close to the observed
values and the pH values for the second stage digesters
3.4. Two-phase anaerobic digestion were also well predicted. The paper did not report the
first stage pHs and hence it was not possible to use this
In two-phase anaerobic digestion the first digester is parameter to evaluate the predictions for VFAs. The
operated at a short SRT to wash out methanogenic bac- overprediction of VFAs for the short SRT reactors
teria and promote the establishment of an acidic envi- was consistent with that observed in the previously de-
ronment. In the second stage digester the VFAs that scribed temperature phased digestion studies.
are generated in the first stage are converted to methane. The model significantly under-predicted the NH4-N
In the study reported by Ghosh (1987) a number of concentrations for the 3 day SRT system while this re-
experiments were performed with digesters operating sponse was relatively well predicted for the 7 day SRT
at both mesophilic and thermophilic conditions. For system. It should be noted that there appeared to be
the purposes of this paper only the tests that were per- an inconsistency in the data for this response since the
formed under mesophilic conditions were examined. observed values for the 3 day SRT system were substan-
Testing was conducted with total SRTs of 3 days and tially higher than the 7 day system. This seems to be
7 days and an influent TS concentration of 7%. With inconsistent with the VSR data that will be subsequently
1840 W.J. Parker / Bioresource Technology 96 (2005) 1832–1842

10 60

9
50
8
CH4 Production (L/d)

7
40
6

VSR (%)
Model Model
5 30
Experimental Experimental
4
20
3

2
10
1

0 0
12 17 24 12 17 24
SRT (d) SRT (d)

7000 250
1st Stage 2nd Stage
6000
200
5000
VFA (mg/L)

VFA (mg/L)
150
4000
Model Model
Experimental Experimental
3000
100

2000

50
1000

0 0
12 17 24 10 11.5 12.5
SRT (d) SRT (d)

Fig. 6. Comparison of system B model predictions with data of Han et al. (1997).

Table 2 Table 3
Comparison of volatile acids, pH And NH4-N for two phase digestion Comparison of VSR for two phase digestion
Response SRT = 3 days SRT = 7 days Response VSR (%)
Stage 1 Stage 2 Stage 1 Stage 2 SRT = 3 days SRT = 7 days
Vol. acids (mg/l) Model 34.0 42.0
Exper. NA 1680 1610–1810 109
Experimental
Model 3811 1393 6711 180
MOPa 26.5 33.6
pH Weight of gasb 35.5 51.5
Exper. NA 7.2 NA 7.3 Theor. gas yieldc 28.3 43.4
Model 5.8 7.0 5.2 7.3 a
VS reduction was calculated as: VSR = 100 * (VS1 VS0)/
NH4-N (mg/l) [VS1 (VS1 * VS0)].
b
Exper. NA 1820 NA 1049 VS reduction was calculated as VSR = 100 * (weight of gas/weight
Model 472 899 766 961 of VS fed).
c
VS reduction was calculated as VSR = 100 * (observed gas yield/
theoretical gas yield of 1.078 SCFM/kg VS added).
described which indicated higher solids reduction for the
longer SRT system.
The predicted VSR values along with three different
measures of VSR for the experimental data that were re- 4. Discussion
ported in the original paper are presented in Table 3.
From Table 3 it can be seen that the model predictions In this paper the predictions of the ADM1 model
were within the range of values that were reported in using the default values for most of the model coeffi-
the papers. It should however be noted that the range of cients were able to reflect most of the trends that were
values reported in the paper was quite wide and hence reported for a variety of digester configurations. There
the assessment of the model predictions could not be very were however consistent deviations between the model
rigorous. predictions and observed values for VFAs when the
W.J. Parker / Bioresource Technology 96 (2005) 1832–1842 1841

model was employed to predict the behaviour of low industry. The ammonia and TKN concentrations
SRT systems. In the two phase systems the model was present in the feed need to be well characterized because
often able to perform reasonably in predicting second of their impact on pH buffering and inhibition func-
stage concentrations of VFAs as with the longer SRTs tions.
in these stages the rates of VFA conversion were able The model tended to overpredict VFA concentrations
to compensate for the high inlet concentrations of for reactors that were operated at reduced SRTs. This
VFAs. It would appear that there could be improve- was observed for both mesophilic and thermophilic
ments made to the model in the estimation of VFA con- digesters. The results suggest that the inhibition function
centrations under these conditions. It may be necessary for pH may over emphasize the impact of reduced pH
to more closely examine the relationship between pH on biological activity. In addition, the model does not
and rate coefficients in this regard. incorporate a pH function for the disintegration and
For the purposes of this study it was often necessary hydrolysis processes. This will have some impact on
to estimate the values that were input into the model for low SRT systems that tend to operate at reduced pHs.
sludge characteristics such as COD, biodegradable frac- The relationship between COD and VS concentra-
tion of the COD, TKN and NH4-N. These have a sub- tions for digested sludges should be established. This
stantial influence on model predictions. If the model is would improve the estimates of VSR since the model
to be used as an analysis and design tool it would benefit only predicts COD concentrations.
from more careful characterization of these parameters. Implementation of the model for temperature-phased
A standardized protocol for determining the anaerobi- systems should be further examined since the current
cally biodegradable fraction of the sludge COD would implementation assumes that the biomass leaving the
assist in this regard. upstream digesters can become active in the downstream
The model predictions for VSR that were reported in digesters at the downstream temperature. It does not
this paper assumed that the reductions in volatile solids seem that this would be likely for thermophiles entering
are proportional to the reductions in COD. However, it a mesophilic digester. The contribution of the incoming
is known that the COD content of volatile solids de- biomass to the activity of the digesters should be further
pends upon the sludge source and its degree of stabiliza- quantified.
tion. Hence, the estimated values for VSR likely contain
error. The extent of this error has not been quantified
for this paper. For more accurate predictions of VSR References
the COD contents of volatile solids in the feed sludge
should be accurately characterized. In addition, the Angelidaki, I., Ellegard, L., Ahring, B.K., 1999. A comprehensive
use of typical values for the COD content of digested model of anaerobic bioconversion of complex substrates to biogas.
Biotechnol. Bioeng. 63, 363–372.
sludge should be employed to convert predicted COD Blumensaat, F., Keller, J., 2005. Modelling of two-stage anaerobic
concentrations to VS concentrations. digestion using the IWA Anaerobic Digestion Model No. 1
In this implementation of the model it was assumed (ADM1). Water Res. 39, 171–183.
that for digesters in series the biomass which moved from Cacho Rivero, J.A., Suidan, M.T., Ginestet, P., Audic, J.-M., 2002.
Effect of SRT on the anaerobic digestion of excess municipal
one digester to another would be active in the down-
sludge. Proceedings of WEFTEC 2002, Chicago, IL.
stream reactor. This assumption should be valid for Drury, D.D., Lee, S.A., Baker, C., 2002. Comparing three-phase
two-phase systems where the digester temperatures are thermophilic continuous feed system to semi-batch feed/hold/draw
the same in both digesters. Implementation of the model system. Proceedings of the 16th Annual WEF Residuals and
in this manner for temperature-phased configurations Biosolids Management Conference, Austin, Texas.
requires more analysis as it is likely that the biomass Eastman, J.A., Ferguson, J.F., 1981. Solubilization of particulate
organic carbon during the acid phase of anaerobic digestion. J.
entering the second stage digester will be somewhat less WPCF 53, 352–366.
active than the model predicts. Ghosh, S., 1987. Improved sludge gasification by two-phase anaerobic
digestion. ASCE J. Environ. Eng. 113, 1265–1284.
Gossett, J.M., Belser, R.L., 1982. Anaerobic digestion of waste
5. Conclusions activated sludge. ASCE J. Environ. Eng. 108, 1101–1120.
Han, Y., Dague, R.R., 1995. laboratory studies on the temperature-
phased anaerobic digestion of domestic wastewater sludges. Pro-
The ADM1 model is a powerful tool for predicting ceedings of WEFTEC 1995, Miami Beach, FL.
the behaviour of anaerobic digesters treating municipal Han, Y., Sung, S., Dague, R.R., 1997. Temperature-phased anaerobic
sludges. However, for successful simulation the feed digestion of wastewater sludges. Water Sci. Technol. 36, 367–374.
stream should be well characterized with respect to itÕs IWA 2002. Anaerobic Digestion Model No. 1 (ADM1), International
Water Association Scientific and Technical Report No. 13, IWA
COD content and the biodegradable fraction of this Publishing, London, UK.
material. A standardized protocol for measuring the Krugel, S., Nemeth, L., Peddie, C., 1998. Extending thermophilic
latter parameter would further use of the model by the anaerobic digestion for producing class a biosolids at the greater
1842 W.J. Parker / Bioresource Technology 96 (2005) 1832–1842

vancouver regional districts annacis island wastewater treatment Pavlostathis, S.G., Gossett, J.M., 1986. A kinetic model for anaerobic
plant. Water Sci. Technol. 38, 409–416. digestion of biological sludge. Biotechnol. Bioeng. 28, 1519–1530.
Lawrence, A.W., 1971. Application of process kinetics to design of Siegrist, H., Renngli, D., Gujer, W., 1993. Mathematical modeling of
anaerobic processes. In: Gould, R.F. (Ed.), Anaerobic Biological anaerobic mesophilic sewage sludge treatment. Water Sci. Technol.
Treatment Processes, Advances in Chemistry Series No. 105. 27, 25–36.
American Chemical Society, Washington, DC. Siegrist, H., Vogt, D., Garcia-Heras, J., Gujer, W., 2002. Mathemat-
Parkin, G.F., Owen, W.F., 1986. Fundamentals of anaerobic digestion ical model for meso and thermophilic anaerobic sewage sludge
of wastewater sludges. ASCE J. Environ. Eng. 112, 867–920. digestion. Environ. Sci. Technol. 36, 1113–1123.

You might also like