See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/6943466

Implementation of the IWA anaerobic digestion model No.1 (ADM1) for

simulating digestion of blackwater from vacuum toilets

Article  in  Water Science & Technology · February 2006

DOI: 10.2166/wst.2006.273 · Source: PubMed

CITATIONS READS

28 266

4 authors, including:

Joachim Behrendt Claudia Wendland

Technische Universität Hamburg Technische Universität Hamburg
55 PUBLICATIONS   497 CITATIONS    26 PUBLICATIONS   292 CITATIONS   

SEE PROFILE SEE PROFILE

Ralf Otterpohl
Technische Universität Hamburg
121 PUBLICATIONS   3,005 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Treatment of textile industry wastewater View project

Treatment of mature landfill leachate View project

All content following this page was uploaded by Yucheng Feng on 09 January 2016.

The user has requested enhancement of the downloaded file.

Implementation of the IWA anaerobic digestion model

Water Science & Technology Vol 53 No 9 pp 253–263 Q IWA Publishing 2006

No.1 (ADM1) for simulating digestion of blackwater
from vacuum toilets
Y. Feng, J. Behrendt, C. Wendland and R. Otterpohl
Institute of Wastewater Management, Hamburg University of Technology, Germany

Abstract The IWA anaerobic digestion model No.1 (ADM1) is applied to the blackwater anaerobic
digestion (BWAD) plant in this work. In order to verify the biochemical kinetics, batch experiments were
executed. According to the Monod type kinetics, the maximum uptake rates (km) of butyric acid (HBu),
propionic acid (HPr) and acetic acid (HAc) are testified as 18, 14, 13 d21, and their half saturation
concentrations (KS) are 110, 120, 160 g COD/m3, respectively. Afterwards, the model was calibrated based
on the performance of a laboratory scale BWAD plant (under mesophilic conditions) by three scenario
studies, i.e. the reference conditions, different feeding frequencies and high NHþ4 concentration. The model
successfully simulated three scenarios. The further two virtual scenario studies were achieved based on the
calibrated model. First, the performance of BWAD plant was predicted with different hydraulic retention
times (HRT); second, the kitchen refuse (KR) was added into the BWAD plant as additional organic loading.
The model predicted the perspective of BW plus KR digestion and generated valuable suggestions for the
operation of the real BWAD plant.
Keywords ADM1; blackwater anaerobic digestion (BWAD); ecological sanitation (ECOSAN); kitchen
refuse (KR); mesophilic

Introduction
Anaerobic digestion (AD) is among the oldest processes used for the stabilisation of
solids and biosolids (Metcalf and Eddy, 2003). Both the European community and United
States consider the anaerobic treatment as the most promising approach for future sustain-
able development (NRC, 1995; Lema and Omil, 2001). Meanwhile, AD treatment can be
the indispensable element in the sustainable sanitation concepts (Otterpohl et al., 1997).
Based on the principle of separating different flows of domestic wastewater according
to their characteristics, ecological sanitation (ECOSAN) aims at establishing an efficient
domestic water system, including nutrients recycling. With the increasing awareness of its
ecological and economic value, more and more ECOSAN projects are currently being built
world wide. In Lübek-Flintenbreite, Germany, an innovative decentralised sanitation con-
cept has been realised in a peri-urban area, where blackwater (BW) from vacuum toilets
and greywater from kitchens and bathrooms are collected and treated separately (Otterpohl
et al., 2004). Our work is based upon the AD treatment of BW from this project.
The mathematical anaerobic digestion model (ADM) has been extensively investigated
and developed during the last three decades (Gavala et al., 2003). Being one of the most
sophisticated and complex ADM, the IWA Anaerobic Digestion Model No. 1 (ADM1)
was published by the IWA Task Group for Mathematical Modelling of Anaerobic Diges-
tion Processes in 2002 (Batstone et al., 2002). ADM1 involves in total 19 biochemical
processes, as well as two types of physiochemical processes. Processes can easily be
added to, or subtracted from, the ADM structural model. Therefore, ADM1 is chosen for
the basis of our work.
doi: 10.2166/wst.2006.273 253
 The targets of this work are: implement and develop a mathematical anaerobic diges-
tion model; verify kinetics parameters; calibrate the model by the laboratory scale black-
water anaerobic digestion (BWAD) reactors; and improve the performance of BWAD
plant by the model.

Materials and methods

Y. Feng et al.

Three parallel laboratory scale BWAD reactors were constructed as Continual Stirred
Tank Reactors (CSTR). Each reactor was a 10.0 L PVC cylinder with dimensions of
19.0 cm inner diameter and 35.5 cm height. Each reactor was filled with 8.0 L sludge and
BW, while 2.0 L was retained as the headspace. The reactors were operated at 38 8C
(mesophilic conditions) with discontinuous BW feeding, which are the same conditions
as a full scale on-site AD plant (Wendland et al., 2004). Due to the very low flushing
water consumption (0.7 –1.0 L water per flush) and the separated sewerage system, BW
from vacuum toilets has relative high COD, which is normally in the range of 4,500 to
13,000 g COD/m3, with an average level of 6,500 g COD/m3. Before feeding, BW was
heated up to 38 8C. The biogas production and pH were measured on-line. Total COD,
total suspended solids (TSS), volatile solids (VS), N-NH4 and total inorganic carbon
(TIC) were checked once every week. The short chain fatty acids (SCFA) and biogas
components were measured by gas chromatography (GC) aperiodically. The laboratory
scale AD reactors had been steadily operated for 2 years. Our model was calibrated and
validated based on the performance of these reactors.
ADM1 uses Monod type kinetics to describe the uptake of substrates. In order to
verify the biochemical kinetics parameters, the maximum uptake rate (km) and the half
saturation concentration (KS), batch experiments were additionally carried out. In each
experiment, the defined amounts of SCFA were added into the 600 ml bottle together
with 500 ml sludge from the BWAD reactors. The headspace was flushed by pure N2 gas
for 2 min before the experiment started. Afterwards, the experiments were executed under
the same conditions as laboratory scale BWAD reactors. The reduction of SCFA was
checked by GC. Two experiments were carried out, i.e. by adding 10.0 ml 5,500 g/m3
butyric acid (HBu); and by adding 10.0 ml 19,219 g/m3 sodium acetate and 10.0 ml
7,928 g/m3 propionic acid (HPr) together.
For simulation, ADM1 was implemented with the software AQUASIM 2.0, which is a
computer program for data analysis and simulation of aquatic systems. AQUASIM 2.0
also provides two powerful tools: linear sensitivity analysis and parameter estimation,
where the latter tool was used for verifying the kinetics parameters later on.

Results and discussion

Verification
The results of batch experiments estimated by AQUASIM 2.0 are shown in Figure 1(a)
and (b). km of HBu, HPr and acetic acid (HAc) were recorded as 18, 14, 13 d21, and
their KS were 110, 120, 160 g COD/m3, respectively. Except km of HAc and KS of HPr,
the experimental values were close to the recommended values by ADM1. Meanwhile,
the same kinetics parameters were used for both butyrate and valerate in ADM1. How-
ever, after testing it was found that the process of valerate uptake could be skipped in our
case. The kinetics parameters for the uptake rates of sugars, amino acids, fats and hydro-
gen were set up by the recommended values of ADM1. All employed values of par-
254 ameters can be found in Feng (2005).
 Y. Feng et al.

Figure 1 Kinetics parameters verification of HBu, HPr and HAc

Calibration and validation

Three scenario studies were performed for calibrating and validating the model based on
the experimental data from our laboratory scale AD plant, i.e. reference conditions,
different feeding frequencies and with high NHþ 4 concentration. In all three scenarios,
before simulating the real data, the model had been running continuously for 60 days
with the average-level input in order to reach the steady state. In general, the model fitted
in with three scenarios reasonably well.

Scenario one: reference conditions. One reactor was always discontinuously fed with
1.0 L BW three times per week. The hydraulic retention time (HRT) and the sludge
retention time (SRT) were the same at 20 days. This reactor was regarded as the
reference reactor. The parameters and coefficients of the model were calibrated based on
the performance of this reactor, and then used for all scenarios.
Figure 2(a) shows the simulation of specific biogas production rate (BPR), which is the
absolute biogas production rate divided by the volume of reactor. After each feeding, one
half-bell shape was formed for BPR, which can be classified into three periods. The first
period (indicated by P1 in Figure 2(a)) had the highest BPR that, relatively slowly, 255
Y. Feng et al.

Figure 2 Simulation results of the reference reactor

decreased. P1 occurred because a high amount of acetate already existed in BW. The anal-
ysis of BW showed that the acetate was normally around 20% of total COD. Total SCFA in
BW could be up to 30%. P1 lasted approximately 10–12 h. Afterwards, in second period
(P2), BPR decreased rapidly. Apparently, BPR was increasingly limited by the previous pro-
cesses (e.g. hydrolysis, LCFA degradation, etc.). P2 persisted for approximately 13–17 h.
Finally, it enters the third period (P3), where BPR was very low and decreased tardily. Our
model fitted these three periods very well. Meanwhile, BPR fell to zero after each feeding, as
BW has lower CO2 concentration than in the liquid phase in the reactor, so CO2 in the head-
space is dissolved back into the liquid phase after feeding. As we can see, the highest BPR
appeared at around 2 h after feeding. This is a common delay phase as microorganisms need
time to adapt new conditions. Nevertheless, the model can not simulate this type of delay, so
the simulated peaks always appear around 1 h earlier than the reality.
Because many parameters, e.g. COD, NHþ 4 and TIC, etc., were measured only once per
week, more circles and diamonds presenting experimental data are generated within each
week in the graphs in order to make the graphs easier to read. The simulation of total
COD, soluble COD, NHþ 4 and TIC in the reactor fitted with experimental data (Figure 2(b)
and (c)). As to pH, the measurement data were distributed between 7.3 and 7.6, which is
well simulated by a relatively straight line (Figure 2(d)). The percentage of CH4 in dried
biogas was around 72%, which was appropriately simulated and not shown here.

Scenario two: different feeding frequencies. In this scenario, the reactors were
investigated with feeding BW once every 24 h and once every 12 h, respectively. Two
reactors executed these two cases separately in the same period. All other conditions were
the same as in the reference reactor. The model simulated two cases correctly, so only the
simulation curves of BPR are shown in this paper (Figure 3). In both cases, BPR is
basically modelled well, while from day 0 to day 2.5 it does not fit entirely (Figure 3(a)
and (b)). In view of all reactors’ performance, the error might come from the measurement.
256 On the contrary, BPR increase after feeding because of the higher CO2 content in BW. The
 Y. Feng et al.

Figure 3 Simulation results of the reactors with different feeding frequencies

delay phase still can not be simulated and even amplified, so it makes the whole simulation
curves shift the to slightly left side referring to the measurement data.

Scenario three: with high NHþ þ

4 concentration. In BW, NH4 concentration is held at a
high level due to the existence of urine. Therefore, the impact of NHþ 4 on AD processes
needs to be understood. The NHþ 4 concentration in the reactor was artificially increased
from 1,000 to 2,000 g NHþ 4 -N/m 3
step by step, by adding certain ammonium salts. The
NHþ 4 concentration in the model was also gradually raised.
In order to obtain the right BPR curve (Figure 4(a)), the smaller half inhibitory con-
centration of ammonia to methanogens had to be used to compare to the scenario. The
free ammonia concentrations are around 40 –45 and 65– 70 g NH3-N/m3 (calculated by
the model) in the reference reactor and the reactor with 2,000 g NHþ 3
4 -N/m , respectively.
It is assumed that there is a threshold of ammonia inhibition, which could be around 50 g
NH3-N/m3. Lay et al. (1998) also found that the NH3 inhibition is increased up to 50 to
500 g NH3-N/m3. This could illustrate what happened in our simulation. pH was kept
at approximately 7.5 which was simulated properly (Figure 4(b)). Other parameters, 257
Y. Feng et al.

Figure 4 Simulation results of the reactor with high NHþ

4 concentration

e.g. COD and NHþ 4 2 N, etc., were also appropriately modelled and are therefore not
shown again. So far, our model achieves all scenarios successfully, which justifies that
the model can simulate the reality properly.

Prediction of BWAD plant by the model

Building up a rational and controllable model to simulate a real situation is just the first
step. Using this model to predict the experimental data is one aspect of the true purpose of
mathematical models. For the BWAD plant, we are interested in several questions, e.g. the
maximum capacity of our reactors, and the optimal loading of the reactors. According to
these requirements, two virtual scenarios were carried out based on the calibrated model.

Virtual scenario one: with different HRT. In this scenario, the capacity of the reactors
was checked by varying HRT. The reactor was still fed three times per week. Besides 20
days, 10, 7.5 and 6 days of HRT were tested. All other conditions were the same as in
the reference reactor. The average COD level of BW 6,500 g COD/m3 was used as input.
258 The simulation shows that pH, IC, IN and biogas components, etc. were kept at the same
level, except BPR and the total COD, although the HRT were quite different. Clearly, the
more input, the more biogas products results, as well as the more COD in the effluent
(Figure 5). The biogas product was increased 2.4 times with decreasing HRT from 20 to
6 d (Figure 5(a)), where the total COD removal rate was decreased from 60–65% to
55–60% (Figure 5(b)). This indicates that the reactor had a quite large capacity when it
treats BW. So, HRT of 10 days is suggested, as the shock loading has to be dealt with.
From this scenario, it is concluded that the reactor has a large unused capacity.

Y. Feng et al.
Subsequently, the next scenario was elaborated.

Virtual scenario two: plus kitchen refuse. The second virtual scenario with a higher
organic loading was tried out. Biowastes can be placed into reactors in order to increase
the organic loading. Kitchen refuse (KR) is the ideal biowaste to be digested, together
with BW. This type of sanitation approach can be a good option for ECOSAN (Otterpohl
et al., 1999). In recent years, the decentralised sanitation and reuse (DESAR) concept
have also developed (Lens et al., 2001), and digesting BW plus KR has been investigated
within the DESAR concept (Kujawa-Roeleveld et al., 2003).

Biogas production rate (BPR)

biogas

Figure 5 Simulation results of the reactor with different HRT 259

 The biodegradability of KR is one of the key issues. If it is classified collected, it has sta-
bly high biodegradability (e.g. in Germany domestic refuse is strictly collected in assorted
tanks). In this scenario, 0.8 is chosen as the biodegradable ratio. The hydrolysis rate of KR
should also be prudently deduced. Veeken and Hamelers (1999) discovered that the hydroly-
sis rates of six different biowastes are in the range 0.10–0.35 d21. Hereby, the hydrolysis
rate 0.30 d21 is chosen for KR in our model. Meanwhile, it is of concern that only the dis-
solved (i.e. hydrolysed) part of KR will be accounted for in the total COD in reactor, and the
Y. Feng et al.

undissolved part will be not flushed out with the effluent. In real cases, the above key par-
ameters should be adjusted according to the reality, or tested by experiments. All other oper-
ation conditions remained the same as the reference reactor. Due to its quite low hydrolysis
rate comparing to BW, KR was fed less frequently. Two KR feeding methods were pro-
posed, simulated and compared, i.e. once per week and once every 2 weeks.
KR feed once per week. Three kinds of additional organic loading were tested, 1.5,
1.0 and 0.75 kg COD KR/(m3 reactor £ week). The simulation of several important par-
ameters, which were generated based on the 90 days pre-running of the model, are shown
in Figure 6. 1.5, 1 kg and 0.75 kg/(m3 week), indicate the cases with their specified

BPR (once per week)

(a) 0.50
F1 F1 F1
m3 norm/(m3 reactor vol .d)

0.40

0.30

0.20

0.10

0.00
0 2 4 6 8 10 12 14 16 18 20 22
Time (day)
1.5 kg/(m3.week) 1 kg/(m3.week) 0.75 kg/(m3.week) no KR

totCOD in reactor (once per week)

(b) 4500

F1 F1 F1
4000
g COD/m3

3500

3000

2500

2000
0 2 4 6 8 10 12 14 16 18 20 22
Time (day)

1.5 kg/(m3.week) 1 kg/(m3.week) 0.75 kg/(m3.week) no KR Exp.

260 Figure 6 Simulation results of the reactor with KR feed once per week
organic loadings; no KR and Exp. indicate the simulation and the experimental data
of the reference reactor, respectively. F1 with an arrow in the graphs indicates the KR
feeding time.
BPR is certainly increased (Figure 6(a)). In a 1-week period, the total biogas
product was increased around 1.9 times by raising the KR loading from 0 to 1.5 kg COD
KR/(m3 reactor week). The total COD in the reactor is evidently increased, which means
the effluent contains more COD (Figure 6(b)). The effluent worsens, especially after the

Y. Feng et al.
last BW þ KR feeding. The CH4 ratio in biogas product is decreased a little bit (the
curve is not shown), and other parameters have no significant variations.
KR feed once every 2 weeks. In the same way, the organic loading 1.5, 1.0 and
0.75 kg COD KR/(m3 reactor £ week) were simulated. The results are shown in
Figure 7. Likewise, F2 with the arrow indicates the KR feeding time too. This time, the
simulation was generated based on 120 days pre-running of the model.
After KR feeding, more biogas is apparently obtained in the first week, but it decreases
to almost the same level at the end of the 2 week period (Figure 7(a)). The effluent is

BPR (once per 2 weeks)

(a) 0.50

F2 F2
m3 norm/(m3 reactor vol .d)

0.40

0.30

0.20

0.10

0.00
0 2 4 6 8 10 12 14 16 18 20 22
Time (day)
1.5 kg/(m3.week) 1 kg/(m3.week) 0.75 kg/(m3.week) no KR

totCOD in reactor (once per 2 weeks)

(b) 4500
F2 F2
4000
g COD/m3

3500

3000

2500

2000
0 2 4 6 8 10 12 14 16 18 20 22
Time (day)

1.5 kg/(m3.week) 1 kg/(m3.week) 0.75 kg/(m3.week) no KR Exp.

Figure 7 Simulation results of the reactor with KR feed once every 2 weeks 261
 very changeable in all three kinds of organic loading (Figure 7(b)), which might not be
expected.
In both KR feeding cases, the variation trends are observed from 0 to 1.5 kg COD
KR/(m3 reactor £ week). The relations among organic loading, biogas products and
COD are clearly depicted by the simulation. Furthermore, the comparison between the
two KR feeding methods was worked out in order to give some instructional idea for the
operation of the BWAD plant.
Y. Feng et al.

Comparison between two KR feeding methods. The organic loading is fixed to 1.0 kg
COD KR/(m3 reactor £ week) for this study. The comparison of BPR shows that more
stable BPR is obtained with F1 than F2. Similar amounts of biogas products are gained
from both KR feeding methods within a 2-week period, which is approximately 1.6 times
higher than the reference reactor. Using the 2-week period as the frame of reference, the
total COD in effluent with F1 is lower than it is with F2 after KR feeding in the first
week, but it is reversed in the following week. The effluents of both cases are around 800
to 1,000 g COD/m3 higher than the reference reactor. So F1 should be preferable in most
cases.

Conclusion
The mathematical model is an effective, efficient and economic method to design and
control systems. ADM1 was successfully implemented and applied to the BWAD plant,
and predicted the results of desired experiments. Our targets were commendably
achieved.
As a generic ADM platform, ADM1 attempts to comprise both information and items
as much as possible. Therefore, the ADM1 should be modified (either simplification or
complication) for each instance in order to obtain the optimum model. ADM1 is very
easy to add or subtract processes due to its structure. In our case the valerate was skipped
from the model. Based on our studies, it was found that the distribution ratio from input
composites to carbohydrates and proteins and lipids are not so sensitive, whereas the
amount of SCFA in the input is more important. Fortunately, SCFA are much easier to
characterise than the composition of input. We also found that the disintegration and
hydrolysis are not the rate-limiting steps (at least not the sole rate-limiting step) in meso-
philic digestion of BW. It is worth testing km and KS of acetate, which is more convin-
cing and reliable. Meanwhile, it proves that BW and KR digestion together is a good
option for both ECOSAN and DESAR.

References
Batstone, D.J., Keller, J., Angelidaki, I., Kalyuzhnyi, S., Pavlostathis, S.G., Rozzi, A., Sanders, W., Siegrist,
H. and Vavilin, V. (2002). (IWA Task Group on Modelling of Anaerobic Digestion Processes).
Anaerobic Digestion Model No. 1 (ADM1), IWA Publishing, London, UK.
Feng, Y. (2005). Calibration and verification of a mathematical model for the simulation of blackwater/
biowaste digestion. M.Sc. thesis, Hamburg University of Technology (TUHH), Germany.
Gavala, H.N., Angelidaki, I. and Ahring, B.K. (2003). Kinetics and modeling of anaerobic digestion process.
Adv. Biochem. Eng. Biotechnol., 81, 57 – 93.
Kujawa-Roeleveld, K., Elmitwalli, T., Gaillard, A., van Leeuwen, M. and Zeeman, G. (2003). Co-digestion
of concentrated black water and kitchen refuse in an accumulation system within the DESAR
(decentralized sanitation and reuse) concept. Wat. Sci. Tech., 48(4), 121 – 128.
Lay, J.J., Li, Y.Y. and Noike, T. (1998). The influence of pH and ammonia concentration on the methane
production in high-solids digestion processes. Wat. Environ. Res., 70, 1075– 1082.
Lema, J.M. and Omil, F. (2001). Anaerobic treatment: a key technology for a sustainable management of
262 wastes in Europe. Wat. Sci. Tech., 44(8), 133 – 140.
Lens, P., Zeeman, G. and Lettinga, G. (2001). Decentralised Sanitation and Reuse: Concepts, Systems and
Implementation, IWA Publishing, London, UK.
Metcalf & Eddy (2003). Wastewater Engineering: Treatment and Reuse, 4th edn, McGraw-Hill, Inc.
NRC (1995). The Role of Technology in Environmentally Sustainable Development, National Academy Press,
Washington D.C., U.S.
Otterpohl, R., Grottker, M. and Lange, J. (1997). Sustainable water and waste management in urban areas.
Wat. Sci. Tech., 35(9), 121 – 133.

Y. Feng et al.
Otterpohl, R., Albold, A. and Oldenburg, M. (1999). Source control in urban sanitation and waste
management: ten systems with reuse of resources. Wat. Sci. Tech., 39(5), 153 – 160.
Otterpohl, R., Braun, U. and Oldenburg, M. (2004) Innovative technologies for decentralised wastewater
management in urban and peri-urban areas. Wat. Sci. Tech., 48(10 –11), 23 – 32.
Wendland, C., Deegener, S., Behrendt, J. and Otterpohl, R. (2004) Anaerobic digestion of blackwater from
vacuum toilets. In Preprints of the 10th World Congress on Anaerobic Digestion (AD10), proceedings,
Montreal, 29 August – 3 September.
Veeken, A. and Hamelers, B. (1999). Effect of temperature on hydrolysis rates of selected biowaste
components. Biores. Technol., 69(3), 249 –254.

263

View publication stats