You are on page 1of 8

Journal of Environmental Science and Health, Part A

Toxic/Hazardous Substances and Environmental Engineering

ISSN: 1093-4529 (Print) 1532-4117 (Online) Journal homepage: http://www.tandfonline.com/loi/lesa20

Modelling anaerobic digestion in an industrial


biogas digester: Application of lactate-including
ADM1 model (Part II)

Preseela Satpathy, Piotr Biernacki, Heribert Cypionka & Sven Steinigeweg

To cite this article: Preseela Satpathy, Piotr Biernacki, Heribert Cypionka & Sven Steinigeweg
(2016): Modelling anaerobic digestion in an industrial biogas digester: Application of lactate-
including ADM1 model (Part II), Journal of Environmental Science and Health, Part A, DOI:
10.1080/10934529.2016.1212559

To link to this article: http://dx.doi.org/10.1080/10934529.2016.1212559

Published online: 15 Aug 2016.

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=lesa20

Download by: [Cornell University Library] Date: 18 August 2016, At: 23:10
JOURNAL OF ENVIRONMENTAL SCIENCE AND HEALTH, PART A
2016, VOL. 0, NO. 0, 1–7
http://dx.doi.org/10.1080/10934529.2016.1212559

Modelling anaerobic digestion in an industrial biogas digester: Application of


lactate-including ADM1 model (Part II)
Preseela Satpathya,b, Piotr Biernackia, Heribert Cypionkab, and Sven Steinigewega
a
Emder Institut f€ur Umwelttechnik (EUTEC), University of Applied Sciences, Emden, Germany; bInstitute for Chemistry and Biology of the Marine
Environment (ICBM), University of Oldenburg, Germany

ABSTRACT ARTICLE HISTORY


A modified Anaerobic Digestion Model No. 1 (ADM1xp) including lactate was applied to a full-scale biogas Received 27 February 2016
plant. This model considers monosaccharides to degrade through lactic acid, which further degrades
KEYWORDS
majorly into acetate followed by propionate and butyrate. Experimental data were derived from the ADM1xp; lactate; food waste;
previous works in the same laboratory, and the proposed parameters were validated against batch cattle manure; biogas
experiments. After successful validation, the biogas plant bearing a fermenter size of 7 dam3 and operated
with food waste and cattle manure was simulated. The biogas production and methane content were
reliably simulated, and a good fit could be obtained against the experimental data with an average
difference of less than 1%. When compared to the original ADM1 model, the performance of the lactate-
incorporated model was found to be improved. Inclusion of lactate as a parameter in the ADM1xp model
is recommended for an increased sensitivity and enhanced prediction principally for systems dealing with
high carbohydrate and lactate loads.

Introduction
Carbohydrate-rich substrates are common for biogas production of input substrates and converts them into solid, liquid and gas-
due to their ability to produce higher biogas yield.[1] Food wastes eous forms as output, and also includes the transformation of
contain high biodegradable organic fraction and have been attrac- products from liquid phase to gas phase. Separate biomasses
tive substrates due to their abundance.[2] However, such rich and their decay fractions further strengthen the capability of
organic load comes with a risk of pH issues due to accumulation this model.[7]
of intermediate products like lactic acid and propionic acid,[3] Despite the wide application of ADM1 model as an anaerobic
thus affecting the stability of the reactor. Donoso et al.[4] report degradation model for several types of substrates, a limitation has
digester failures due to issues relating to rise in volatile fatty acids, been identified related to the model’s applicability to digesters
reduction in methane production, pH drops, etc., ultimately with carbohydrate overload.[8] Lactate is ignored in the original
destroying the balance of the system. In this regard, mathematical model, considering its insignificant impact on methanogenic and
models are beneficial in understanding the processes and predict- low loading systems.[9] However, various reports indicate the pro-
ing the digester conditions for extended periods of time, which duction of lactic acid as a major intermediate compound during
help prevent reactor breakdowns. These are cost and time effective the anaerobic digestion of kitchen wastes [3,10] and carbohydrates
[9,11]
and enable an effective forecast of the viability and efficiency of the as well. Considerable accumulation has also been reported
biogas digesters.[5] One such widely used model providing a com- after a sudden increase in the organic loading.[5,12,13] Lactate with
mon platform for modelling anaerobic digestion is the Anaerobic its low acid dissociation constant (pKa D 3.86) has a strong effect
Digestion Model No. 1 (ADM1) developed by the International on pH values,[11] ultimately destroying the balance of the system
Water Association’s (IWA) Task Group.[6] when accumulated. Such effects have been described in the works
of Cao et al.[14] and Ubalua [15] who describe lactic acid ceasing
fermentation. Despite such significant influences on biogas pro-
Anaerobic digestion model no. 1 duction, the ADM1 model does not consider pathways involving
The ADM1 model combines 19 biochemical and 2 physico- lactic acid.[9]
chemical processes to simulate the behaviour of various com- In our previous works,[16,17] we confirmed the influential
ponents in the anaerobic digesters. It includes the major role of lactate in causing an enhanced biogas production. After
processes involved in the bioconversion of complex organic this finding, the ADM1 model was further extended to include
materials into methane, carbon dioxide and inert by-products, lactate and its kinetic parameters to simulate a laboratory-based
with the main metabolic intermediates being hydrogen, acetic, standard reactor.[18] The aim of this work is to explore the
propionic and butyric acids.[6] The model conserves the mass applicability of this lactate-included modified ADM1xp model

CONTACT Preseela Satpathy preseela.satpathy@hs-emden-leer.de EUTEC, Constantiaplatz 4, Emden D-26721, Germany.


Color versions of one or more of the figures in this article can be found online at www.tandfonline.com/lesa.
© 2016 Taylor & Francis Group, LLC
2 P. SATPATHY ET AL.

to an existing full-scale high-rate biogas plant operated with a factor (fP), and this updated model was used for both SIMBAÒ 6
complex organic substrate like food waste. and in MATLAB R2006b, integrated by IFAK [21] and Biernacki
et al.[7] The advantage of using ADM1xp includes an ensured
incorporation of nutrient mineralization.[22] This ADM1xp
Methods
model was further modified by the authors [18] in the previous
System investigation to include the formation of lactate from carbohy-
drates and its further degradation. In this lactate-included
The system modelled here is the EWE biogas plant in Witt-
model, the monosaccharides were presented to degrade to ace-
mund, Lower Saxony, Germany [19] built in 1996. This plant
tate, lactate and butyrate; propionate was formed indirectly
operates on kitchen wastes from restaurants and households
from lactate. Lactate is further degraded to propionate, acetate
and agricultural waste like cattle and swine manure. The plant
and butyrate. The acid degraders and their decay in each step
and its description have been detailed in the works of Biernacki
were also included in the modified model. The COD flow
et al.[7]
describing the degradation pathways and the model matrix dis-
The EWE biogas plant has two parallel fermenters, each
playing the position of lactate in the model are presented in the
3.5 dam3, with an average hydraulic retention time of
first part of this study.[18]
20 days.[20] The biogas plant bears the facility of combined heat
and power (CHP) unit and produces electricity (2.5 MW) and
heat (3.4 MW) from biogas. The organic waste is collected Characterization of substrates
from farmers and food industries, and collected in an under-
ground tank (1.9 dam3) and is further transferred to a mixing Characterization of complex substrates by the Weende analysis
tank (620 m3), where it is mixed with manure in order to obtain and van Soest extension [23] were used for transferring the sub-
a homogeneous consistency and an optimal buffer/substrate strate characteristics into the ADM1 model. We considered the
ratio. Manure is significant because it provides an ideal buffer experimental data from Biernacki et al.[7] who calculated the nec-
system and a healthy bacterial composition for the production essary ADM1 fractions from the organic matter to raw lipids
of biogas. The biogas produced in the digester is then measured (RL), raw protein (RP), raw fiber (RF) and N-free extract (NfE) as
from the recorded cumulative gas flow, and its composition is described by Naumann and Basler.[24] The substrate characteriza-
monitored with the help of infrared sensors. tion was performed for the digester contents during the assess-
During the assessment period of the plant, it was operated ment period i.e. for both cattle manure and food waste (Table 1).
with an input of 180 m3 day¡1 of cattle manure and 100 m3
day¡1 of food waste on an average. The simulated data of bio-
Batch reactors setup and operation
gas production was collected for 28 days (19.03.2012 until
15.04.2012), and the cumulative biogas production was found Batch experimental data for both cattle manure and food
to be 4.57 dam3 day¡1. A schematic diagram of the representa- waste were derived from Biernacki et al.[7] They had pre-
tion of the plant and the setup in the model is shown (Fig. 1). pared the batch reactors with fresh starter inoculum from
the EWE biogas plant, and the experiments were performed
as per the German VDI 4630 [25] guidelines. Digital
The model
ANKOM (N1v0, 4RF2; RFS#194) wireless sensors (Mace-
The model used in this study has been stepwise developed by the don, NY, USA) were used in order to obtain automated
authors. The ADM1xp model updated by Biernacki et al. [7] was hourly measurements of the total biogas produced in the
used as the starting model. This model incorporates a new decay reactors. The blank data were subtracted, thus enabling the
product fraction (XP), which is formed from a decayed biomass validation of biogas produced only from the substrates

Figure 1. Model description of the EWE biogas plant in Wittmund, Germany, prepared using SIMBAÒ simulation software derived from Rojas et al.[20] and modified for
this research.
JOURNAL OF ENVIRONMENTAL SCIENCE AND HEALTH, PART A 3

Table 1. Characteristic of the examined substrates (taken from Biernacki et al.).[7] Optimization tool
Parameter Cattle Food
A numerical algorithm commonly referred as Simplex algo-
[Unit] manure waste rithm minimizes the difference between experimental and
simulation results with the help of the downhill simplex
Dry mass (DM) [%] 5.23 30.55
Organic dry mass (oDM) [%] 4.01 27.83 method algorithm from Nelder and Mead.[28] This algorithm
Raw protein (RP) [%] 0.74 7.47 is incorporated into MATLAB as a function fminsearch.[29]
Raw lipids (RL) [%] 0.17 9.41 For each substrate, the four adjustable parameters (kdis,
Raw fibre (RF) [%] 1.15 1.17
Neutral detergent fibre (NDF) [%] 2.24 9.53 khyd_ch, khyd_li, khyd_pr) were simultaneously fitted fol-
Acid detergent fibre (ADF) [%] 2.22 9.15 lowed by the other listed parameters (Table 1). The maximum
Acid detergent lignin (ADL) [%] 3.66 1.09 utilization constants (k) were found to be the parameters with
Nitrogen-free extracts (NfE)a [%] 1.96 9.78
Ash [%] 1.22 2.72 the most significant effects for the different groups of bacteria.
Composite fraction (Xc)a [kg COD/m3] 55.07 522.88 High sensitivity was also observed for flac_su, Ylac, KS_lac and
Protein content stoichiometric factor (fPr_xc)a [–] 0.185 0.268 fh2_lac, and these parameters were adjusted using the mathe-
Lipids content stoichiometric factor (fLi_xc)a [–] 0.042 0.338
Carbohydrates content stoichiometric factor (fCh_xc)a [–] 0.264 0.183 matical tool in order to reduce the errors between the experi-
Inerts content stoichiometric factor (fxi_xc)a [–] 0.509 0.210 mental and simulation data. After determination of the
a
kinetic constants for both the batch experiments, considering
Calculated value.
the same proportion as fed in the biogas plant, a fraction of
60% manure and 40% food waste was considered, and the final
during these batch experiments. Previous studies have values were calculated accordingly. This was the same propor-
emphasized on the need to determine hydrolytic constants tion in which the substrates were fed into the biogas digester
specifically for the substrates;[26] hence, the kinetic parame- during the assessment period.
ters and activity in each batch reactor were also accordingly
determined. Kinetic parameters in the model were individu-
ally determined for cattle manure and food waste. For each Analytics
substrate, a set of kinetic constants describing the phases of Dry solids and volatile solids contents were measured using
disintegration (kdis), hydrolysis of carbohydrates (khyd_ch), standard methods detailed in.[25]
hydrolysis of lipids (khyd_li) and hydrolysis of proteins The soluble nitrogen (NH4-N) was measured using the
(khyd_pr) was determined from Biernacki et al.[7] The other (LCK303 and LCK305) ammonium cuvettes from Hach Lange.
data sets of parameters were obtained from the previously COD was measured using the (LCK314) COD cuvettes from
described works of the authors.[18,27] These set of parame- Hach Lange as per the manufacturer’s instructions.
ters were calibrated with experimental data and were
adjusted further in order to obtain a desirable fit. An opti-
mization tool was utilized in order to obtain a sophisticated Determination of volatile fatty acids
calibration to simulate the batch results. Gas chromatography (GC) was performed in order to deter-
mine the intermediate organic acids content in the inoculum.
Agilent gas chromatograph (GC 7820) was used with an
Parameter estimation INNOPEG-FFAP column (length: 30 m, I.D.: 0.25 mm, film
thickness: 0.25 mm; CS, Germany). The flame ionization detec-
As previously mentioned, the hydrolytic constants were deter- tor (FID) with nitrogen gas as a carrier was utilized, and an iso-
mined from Biernacki et al.[7] since this set was found to provide a thermal temperature of 180 C was set for both the injector and
satisfactory optimized fit in combination with the other parame- the detector. Sample (1 mL) was injected with a flow splitting of
ters. The relevant lactate-related parameters were determined 1:50. The details are presented in the previous works of the
from the previous works of the authors.[18] The authors had esti- authors.[16]
mated the necessary parameters for inclusion of lactate into the
model from the steady-state experimental results directly available
from the culture on lactate. They also had outlined an approach Results and discussion
to parameter estimation and the same approach was utilized in
Initial state of the reactor
this study. After experimental determination of the lactate degra-
dation products, the dominant pathway and the parameters (Ylac, The starter inoculum considered in the batch experiments
KS_lac), the batch experiments were validated. A numerical optimi- was an extract from the EWE biogas plant’s digester, and
zation algorithm (described in the optimization tool) was used in the characterization of this inoculum was fed into the
order to determine the optimal set of parameters. Different ADM1 model as the reactor’s initial state. The dry mass of
approaches thus were used to ultimately obtain a successful fit. this inoculum, organic dry matter mass fraction, pH and
Table 1 outlines the considered set of parameters, which were ammonium content (represented as SNH4) in the model
determined from either of the following approaches: experimen- were determined to be 4.83%, 3.39%, 7.84 and 3.1 kg/m3,
tally determined from our laboratory, adjusted by the mathemati- respectively. The organic acid content too was fed into the
cal fminsearch tool for an improved fit, calibrated directly using model, which showed a positive impact on the simulation
the mathematical tool for an enhanced fit with the batch experi- results. The acetic acid, propionic acid, lactic acid, butyric
ments or from literature sources. acid and valeric acid contents were determined to be 0.5,
4 P. SATPATHY ET AL.

0.9, 0.64, 0.69 and 0.08 kg COD/m3, respectively, and were Model prediction by original ADM1
further updated into the model. It was encouraging to
At first, we tried to simulate the biogas plant without any
observe the presence of lactic acid already in the inoculum.
modification and considering the values specified by Batstone
et al.[6] A discrepancy was clearly observed between the exper-
imental data and the simulation outcome (Fig. 2). A difference
Validation of the kinetic parameters against batch of nearly 60,886 m3 (»48%) on an average was observed in the
experiments biogas production. We infer this difference to be due to the
lack of optimization of the parameters and the absence of
The batch results with kinetic constants describing the disinte-
decay products (XP), which provide an advanced explanation
gration and hydrolysis phase (shown in Table 1) were individu-
to the biological process dynamics. The basic model was thus
ally determined for cattle manure and food waste by Biernacki
found to be unable to simulate the data satisfactorily. How-
et al.[7] These values were found to be reliable; hence, no
ever, it was remarkable that the trends in the erratic biogas
changes were further necessary. Some parameters related to lac-
production could be well reflected by the original model.
tate were further optimized with the help of the optimization
An evident difference was also observed between the exper-
tool, and these sets of parameters (outlined in Table 1) were
imental and simulation data for the methane content.
found to provide a desirable fit with the batch experiments for
Although, unlike the total biogas production, the average
both the substrates. It was notable that the approach for charac-
methane content was well simulated by the original ADM1
terising the substrates described earlier (Characterization of
model, a difference of only ¡1.3% was observed between the
substrates) provided an optimal prediction along with a precise
experimental and simulation results. The trends however
description of the reaction kinetics during the batch experi-
were not found to be reliable (Fig. 3) and the predictions by
ments. After successful validation of the batch experiments, the
the original model appeared to be uncertain. While in some
large-scale biogas reactor of EWE company in Germany was
cases there was an overestimation of methane content, during
simulated.

Figure 2. Comparison of the (a) biogas production and (b) methane content between the experimental data collected from the EWE biogas plant and the model
simulations by the default ADM1xp model.
JOURNAL OF ENVIRONMENTAL SCIENCE AND HEALTH, PART A 5

Figure 3. Comparison of the (a) biogas production ant and (b) methane content in the EWE biogas plant between the lactate-including and lactate-devoid ADM1 model
developed in this study and by Biernacki et al.[7] respectively.

days 20–24, there seemed to be a clear underestimation in the Table 2. Parameters set in the modified ADM1xp model.
simulation results. Such difference (nearly 3% observed on
Sl no. Parameters Value Source
day 20) between the simulated and recorded data could be
misleading and substantially affect in practical scenarios. 1 kdis 1.704338 Biernacki et al. [7]
2 khyd_ch 0.433202 Biernacki et al. [7]
3 khyd_pr 0.697782 Biernacki et al. [7]
Modified ADM1 model predicts the biogas plant 4 khyd_li 0.3222691 Biernacki et al. [7]
5 km,lac 13 Hinken [27]
successfully 6 0.4202 Experimental data calibrated b
KS,lac
7 Ylac 0.06 Experimental data calibrated b
The lactate-included ADM1 model was used to verify both 8 0.02 Batstone et al. [6]
kdec, Xlac
biogas production and the methane content in the EWE bio- 9 kA/B, lac 108 Batstone et al. [6]
gas plant for 28 days. We found a satisfactory correlation 10 Ka, lac 1.259 £ 10¡4 Hinken [27]
11 KVFA, lac 0.5 Hinken [27]
between experimental and simulation results for both the sim- 12 Clac 0.03125 Hinken [27]
ulations. An integral objective of our present work included to 13 flac,su 0.456 Calibration c
investigate if the lactate-included ADM1 could further opti- 14 fh2, lac 0.3942 Calibration c
15 fac_lac 0.456 Experimental data calibrated b
mize the prediction capability of the model described by Bier- 16 0.1083 Experimentally determined a
fpro_lac
nacki et al.[7] Works of Biernacki et al.[7] in this laboratory 17 fbut_lac 0.0057 Experimentally determined a
had already shown simulation of this biogas plant after opti- 18 fCO2_lac 0.0358 Calibration c
mization of the kinetic constants (Fig. 3). They had success- a
Experimentally determined.
fully simulated the EWE biogas plant with a methane content b
Adjusted by the mathematical fminsearch tool for an improved fit.
underestimated only by 1.74% and the biogas production
c
Calibrated directly using the mathematical tool for an enhanced fit with the batch
experiments.
overestimated by 1.5%. The lactate-included ADM1 model
6 P. SATPATHY ET AL.

Table 3. Biogas production and methane content predicted for the EWE biogas plant with the different models: original ADM1 model, ADM1 modified by Biernacki
et al.[7] lacking lactate and ADM1xp with lactate.
Difference in Difference of methane
Biogas Methane content biogas production (m3) content compared to
(m3) (% vol.) compared to experimental data experimental data (%)

Experimental data 127,711 66.85


Original ADM1 66,824 67.91 60,886 ¡1.3
Modified ADM1without lactate 129,581 65.11 1,840 ¡1.7
Lactate-included ADM1 128,719 66.63 1,008 ¡0.2

applied in this study facilitated an even improved prediction Despite the capability and success of the model, some weak-
of the same plant reducing the differences between the experi- nesses could be with regard to the amount of lactate formed
mental and simulation data to nearly 0.2% and 0.8% for the from monosaccharides. Glucose and lactic acid breakdown has
methane content and total biogas production, respectively been reported to be dynamic and highly dependent on the
(Table 3). hydrogen pressure in the digesters.[8] We therefore believe that
The strength of both the modified models was clearly a more in-depth study of this sensitive aspect might be useful
observed through their capability to predict the biogas plant to improve the model’s prediction capability. It would be of
well despite the challenge of simulating the inconstant input great interest to analyse the operation of a biogas plant operat-
feed. Changes in input led to fluctuating biogas production in ing with heavy doses of lactate to evaluate the real strength and
the digester throughout the analysis period, and this could be capability of this model.
adequately reflected by the model. Inclusion of lactate in the
model not only produced a close and better fit with the experi-
Conclusion
mental data, but also showed an increased correlation to the
trend of the methane formation (Fig. 3b). Despite the successful The motivation of this paper came after successful validation of
simulation shown in the previous model by Biernacki et al.[7] the lactate-incorporated model against a laboratory-based reac-
the methane content was found to be predicted in an opposite tor operated with starch. In this study, we evaluated this modi-
trend (especially after 11 days). This shortcoming was however fied ADM1 model to describe a full-scale anaerobic reactor
overtaken by the lactate-included model clearly providing an treating food waste and cattle manure. With a difference of
improved prediction and a closer fit. We inferred these nearly 0.8% and 0.2%, respectively, for biogas yield and meth-
improvements to be due to the consideration of lactate as both ane content, the lactate-included ADM1xp model was found to
a substrate and an intermediate in the model, which was other- successfully improve the fit against the experimental data. The
wise neglected. We also anticipate that with inclusion of a trends of irregular methane formation during the assessment
strong acid like lactate, the impact of pH on the overall anaero- period were also well reflected by the model for this biogas
bic digestion process could be more accurately assessed by the plant developed by the coauthors.[7] Inclusion of lactate was
model. Influence of pH on affecting the anaerobic digestion clearly found to further enhance the prediction capability of the
process is widely reported.[8] We conclude these improved sim- ADM1xp model. This model was thus recommended to
ulations due to the models’ increased sensitivity to the real describe full-scale biogas plants operating with complex organic
overall process dynamics in the digester. In order to facilitate substrates like food waste. Further research focussing on reac-
an easier comparison of the different simulation results and the tors solely based on carbohydrate overload or in industries like
average differences overall between the original model and the dairy, pharmaceuticals, food and leather operated highly with
modified ADM1 models, the data are presented (Table 2). lactic acid might be useful in order to examine the model’s
Both modified ADM1xp by Biernacki et al.[7] and lactate- applicability and reliability.
included ADM1 models presented in this study show a good
correlation of biogas and methane yields when compared to the
Acknowledgments
experimental data. The correlation to the erratic tendencies of
the digester was also well predicted by the lactate-included We thank Prof. Eike Siefert and Dr. Frank Uhlenhut from the EUTEC
ADM1xp. The methane content also was very closely and reli- Institute for their support and guidance. We greatly acknowledge Mrs.
ably predicted with a difference of only 0.2% by the modified Manuela Bayer from the EWE biogas plant for providing us with data and
materials from the plant.
model incorporated with lactate.
The model considered in this study could thus successfully pre-
dict the biogas production in the large-scale biogas plant. Other Funding
researchers have also reported that the degradation of glucose
This research was funded by Deutscher Akademischer Austauschdienst
could be predicted better with inclusion of lactate,[30] and Soda (DAAD) and was conducted in Hochschule Emden/Leer in collaboration
et al.[9] clearly show lactate as the major product during monosac- with Carl von Ossietzky University, Germany.
charide degradation. We expect the lactate-incorporated model
used in this study to successfully predict large-scale biogas plants
operated with food waste over a longer duration of operation, References
thereby proving to be useful for biogas plant operators in prevent- [1] Weissbach, F. Gas production potential of fresh and ensiled sugar
ing possible digester stability issues. beets in biogas production. Landtechnik. 2009, 64(6), 394–397.
JOURNAL OF ENVIRONMENTAL SCIENCE AND HEALTH, PART A 7

[2] Wang, H.T.; Nie, Y.F. Municipal solid waste characteristics and man- [17] Satpathy, P.; Steinigeweg, S.; Siefert, E.; Cypionka, C. Effect of lactate
agement in China. J. Air &Waste Manage Assoc. 2001, 51, 251–272. and starter inoculum on biogas production from fresh maize and
[3] Zhang, B.; Wei-min, C.; Pin-jing, H. Influence of lactic acid on the maize silage (under review).
two-phase anaerobic digestion of kitchen wastes. J. Environ. Sci. [18] Satpathy, P.; Biernacki, B.; Uhlenhut, F.; Cypionka, H.; Steinigeweg,
2007, 19, 244–249. S. Modelling anaerobic digestion in a biogas reactor: ADM1 model
[4] Donoso- Bravo, A.; Mailier, J.; Martin, C.; Rodriguez, J.; Aceves-Lara, development with lactate as parameter (Part I). J. Environ. Sci.
C.; Wouwer, A.V. Model selection, identification and validation in Health A. 2016, (in press).
anaerobic digestion: A review. Water Res. 2011, 45, 5347–5364. [19] EWE Biogas GmbH & Co. KG. Wittmund biogas power plant:
[5] Schoen, M.; Sperl, D.; Gadermaier, M.; Goberna, M.; Franke- Whittle, I.; important data at a glance. Available at http://www.ewe-biogas.de/
Insam, H.; Ablinger, J.; Bermhard, W. Population dynamics at digester english/index_28.php. (accessed Jan 2016).
overload conditions. Bioresour. Technol. 2009, 100(23), 5648–5655. [20] Rojas, C.; Uhlenhut, F.; Schlaak, M.; Borchert, A.; Steinigeweg, S.
[6] Batstone D.J.; Keller, J.; Angelidaki, I.; Kalyuzhnyi, S.V.; Pavlostathis, Simulationsunterstuetzte Prozessfuehrung von Biogasanlagen.
S.G.; Rozzi, A.; Sanders, W. T.; Siegrist, H.; Vavilin, V. A. Anaerobic Nuernberg, Dechema Gesselschaft fuer Chemische Technik. Poster-
digestion model no.1. International Water Association, London, 2002. beitrag und Abstract, DECHEMA-Tagung: Bioprozessorientiertes
[7] Biernacki, P.; Steinigeweg, S.; Borchert, A.; Uhlenhut, F.; Brehm, A. Anlagendesign. Chem. Tech. 2010, 108–109.
Application of Anaerobic Digestion Model No. 1 for describing an [21] IFAK. SIMBA_ 5.1. User’s guide. Magdeburg: Ifak System GmbH,
existing biogas power plant. Biomass Bioenerg. 2013, 59, 441–447. 2005.
[8] Thamisiriroj, T.; Nizami, A.S.; Murphy, J.D. Why does mono-diges- [22] Koch, K.; L€ ubken, M.; Gehring, T.; Wichern, M.; Horn H. Biogas
tion of grass silage fail in long term operation? Appl. Energ. 2012, from grass silage measurements and modelling with ADM1. Biore-
95, 64–76. sour. Technol. 2010, 101(21), 8158–8165.
[9] Soda, S.; Wada, K.; Okuda, M.; Ike, M. Application of modified [23] van Soest, P.J.; Wine, R.H. Use of detergents in the analysis of fibrous
ADM1 to long-term experiments for methane/hydrogen production feeds IV. Determination of plant cell-wall constituents. J. Assoc. Off
from model organic waste. Water Pract. Tech. 2011, 6(1), Anal. Chem. 1967, 50, 50–59.
wpt2011009. [24] Naumann, C.; Bassler, R. Die Chemische Untersuchung von Futter-
[10] Wang, X.; Zhang, S.; Wang, J.; Yu, X.; Lu X. Exploring optimal feed mitteln. VDLUFA-Verlag: Darmstadt, 1993.
to microbes ratio for anaerobic acidogenic fermentation of cassava [25] Verein Deutscher Ingenieure (VDI). Fermentation of organic mate-
residue from brewery. BioResour. 2012, 7(1), 1111–1122. rial. Characterization of the substrate, sampling, collection of mate-
[11] Antonopoulou, G.; Gavala, H.N.; Skiadas I.V.; Lyberatos, G. Modeling rial data, fermentation tests. Verein Deutscher Ingenieure (VDI):
of fermentative hydrogen production from sweet sorghum extract Dusseldorf, 2006.
based on modified ADM. Int. J. Hydrogen Energ. 2012, 37, 191–208. [26] Wichern, M.; Gehring, T.; Fischer, K.; Andrade, D.; L€ ubken, M.;
[12] Mattiasson, B. Anaerobic digestion generates fatty acids. Ind. Biopro- Koch, K.; Gronauer, A.; Horn, H. Monofermentation of grass silage
ces. 2004, 26(6), 8–9. under mesophilic conditions: measurements and mathematical
[13] Pipyn, P.; Verstraete, W. Lactate and ethanol as intermediates in two- modelling with ADM1. Bioresour. Technol. 2009, 100(4), 1675–1681.
phase anaerobic digestion. Biotechnol. Bioeng. 1981, 28, 1145–1154. [27] Hinken, L. Mathematische Modellierung von anaeroben Abbaupro-
[14] Cao, Y.; Cai, Y.; Takahashi, T.; Yoshida, N.; Tohno, M.; Uegaki, R.; Non- zessen unterschiedlicher Substrate auf Basis des, Anaerobic Digestion
aka, K.; Terada, F. Effect of lactic acid bacteria inoculant and beet pulp Model No. 1”. PhD dissertation. Leibnitz University, Hannover, 2013.
addition on fermentation characteristics and in vitro ruminal digestion [28] Nelder, J.A.; Mead, R. A simple method for function minimization.
of vegetable residue silage. J. Dairy Sci. 2011, 94(8), 3902–3912. Comput. J. 1965, 7(4), 308–313.
[15] Ubalua, A.O. Cassava wastes: treatment options and value addition [29] The MathWorks, Inc. MATLAB – Optimization toolboxTM 6. user’s
alternatives. Afr J. Biotechnol. 2007, 6(18), 2065–2073. Guide. The MathWorks, Inc.: Natick, MA, USA, 2011.
[16] Satpathy, P.; Steinigeweg, S.; Cypionka, H.; Engelen, B. Different sub- [30] Peiris, B.R.H.; Rathnasiri, P.G.; Johansen, J.E.; Kuhn, A.; Bakke, R.
strates and starter inocula govern microbial community structures in ADM1 with modifications for Bio-hydrogen Simulations. Water Sci
biogas reactors. Environ. Technol. 2016, 37(11), 1441–1450. Technol. 2006, 53(8), 129–137.

You might also like