A novel mathematical modelling of waste biomass ...

University of Southern Denmark

A novel mathematical modelling of waste biomass decomposition to facilitate rapid methanepotential prediction

Vazifehkhoran, Ali Heidarzadeh; Triolo, Jin Mi

Published in:Journal of Cleaner Production

DOI:10.1016/j.jclepro.2019.01.161

Publication date:2019

Document version:Accepted manuscript

Document license:CC BY-NC-ND

Citation for pulished version (APA):Vazifehkhoran, A. H., & Triolo, J. M. (2019). A novel mathematical modelling of waste biomass decomposition tofacilitate rapid methane potential prediction. Journal of Cleaner Production, 220, 1222-1230.https://doi.org/10.1016/j.jclepro.2019.01.161

Go to publication entry in University of Southern Denmark's Research Portal

Terms of useThis work is brought to you by the University of Southern Denmark.Unless otherwise specified it has been shared according to the terms for self-archiving.If no other license is stated, these terms apply:

• You may download this work for personal use only. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying this open access versionIf you believe that this document breaches copyright please contact us providing details and we will investigate your claim.Please direct all enquiries to [email protected]

Download date: 24. Jun. 2022

https://doi.org/10.1016/j.jclepro.2019.01.161

https://doi.org/10.1016/j.jclepro.2019.01.161

https://portal.findresearcher.sdu.dk/en/publications/d90c3b66-83b0-4b9b-ab8c-e5da1d940db9

Accepted Manuscript

A novel mathematical modelling of waste biomass decomposition to facilitate rapid methane potential prediction

Ali Heidarzadeh Vazifehkhoran, Jin M. Triolo

PII: S0959-6526(19)30180-5

DOI: 10.1016/j.jclepro.2019.01.161

Reference: JCLP 15548

To appear in: Journal of Cleaner Production

Received Date: 02 March 2018

Accepted Date: 14 January 2019

Please cite this article as: Ali Heidarzadeh Vazifehkhoran, Jin M. Triolo, A novel mathematical modelling of waste biomass decomposition to facilitate rapid methane potential prediction, Journal

(2019), doi: 10.1016/j.jclepro.2019.01.161of Cleaner Production

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

1

1 A novel mathematical modelling of waste biomass decomposition to facilitate rapid methane

2 potential prediction

3 Ali Heidarzadeh Vazifehkhoran, Jin M. Triolo*

4 University of Southern Denmark, Faculty of Engineering, Institute of Chemical Engineering,

5 Biotechnology and Environmental Technology, Campusvej 55, DK-5230 Odense M, Denmark

6 *Corresponding author. Tel: +4531794600; E-mail: [email protected]

7 Abstract

8 Biogas production is known to be the most sustainable bioenergy technology available owing to

9 its utilization of waste biomass. Due to the widely ranging energy contents of major organic

10 carbons and their diverse degradation pathways, as well as their highly varying hydrolysis

11 capacities, predicting methane yield is not as simple as predicting the production of other

12 biofuels. This study investigated the hydrolysis behaviour of organic compounds using the

13 operational conditions of real-scale biogas plants (hydraulic retention time (HRT) of 15, 20, 30,

14 and 45 days) and explored a new approach to determine the biochemical methane potential

15 (BMP), which can be used to predict the methane yield of a wide range of agro-industrial

16 wastes. The degradation of hemicelluloses and cellulose increased gradually, closely following

17 first-order degradation kinetics, at longer retention times. In spite of their longer retention times,

18 only 45% of hemicelluloses and 34% of cellulose decomposed. The newly proposed BMP

19 model was validated using 65 internal and external datasets. The model error (RMSEP) was in

20 the range of 37.4 to 87.7 NL CH4 kg VS–1 while the relative model error (rRMSEP) was in the

21 range of 12.1%–47.8%. The model fits best to gently lignified biomass, but the overestimated

22 results obtained with woody biomass and winter harvested grass indicate that further

23 investigation of the hydrolysis of well-lignified biomass is required to enhance the precision of

24 the model.

25 Keywords: Anaerobic digestion; BMP model; lignocellulose; lignin; CSTR; methane

mailto:[email protected]

ACCEPTED MANUSCRIPT

2

1 Nomenclature

VS Volatile solids

DM Dry matter

TKN Total Kjeldahl nitrogen

TAN Total ammoniacal nitrogen

NDF Neutral detergent fiber

ADF Acid detergent fiber

ADL Acid detergent lignin

BMP (NL CH4 kg VS–1) Biochemical methane potential

TBMP (NL CH4 kg VS–1) Theoretical biochemical methane potential

𝑑𝑓𝜃 (%) The decay of an organic compound i

𝑑𝑓∞ (%) The ultimate decay of an organic compound i

k (day-1) The decay rate constant for a fraction

θ (day) Hydraulic retention time

(NL CH4 L–1 day-1)𝑟𝐶𝐻4 the specific rate of CH4 production

Mi0 (NL CH4 kg VS–1) TBMP of organic compounds (i) in the input feedstock

RMSEP Residual mean square error in prediction

rRMSEP Relative residual mean square error in prediction

VSED Easily degradable VS

VSSD Slowly degradable VS

VSND Non-degradable VS

XP Crude proteins

XL Crude lipids

(%)𝑉𝑆𝑋𝑃 Concentration of crude protein in VS

(%)𝑑𝑓𝑋𝑃 Ultimate degradation factor of 𝑉𝑆𝑋𝑃

(%)𝑉𝑆𝐻𝑀 Concentration of hemicellulose in VS

(%)𝑑𝑓𝐻𝑀 Ultimate degradation factor of 𝑉𝑆𝐻𝑀

(%)𝑉𝑆𝐶𝐸 Concentration of cellulose in VS

(%)𝑑𝑓𝐶𝐸 Ultimate degradation factor of 𝑉𝑆𝐶𝐸

(NL CH4 kg VS–1)𝐵𝑀𝑃𝑃 Predicted BMP

(%)𝑉𝑆𝑉𝐹𝐴 Concentration of VFA in VS

(%)𝑉𝑆𝐸𝑡. Concentration of ethanol in VS

(%)𝑉𝑆𝑋𝐿 Concentration of crude lipid in VS

(%)𝑉𝑆𝐶𝐵 Concentration of non-fibrous carbohydrates in VS

(NL CH4 kg VS–1)𝑇𝐵𝑀𝑃𝑉𝐹𝐴 Stoichiometric methane potential of 𝑉𝑆𝑉𝐹𝐴

ACCEPTED MANUSCRIPT

3

(NL CH4 kg VS–1)𝑇𝐵𝑀𝑃𝐸𝑡. Stoichiometric methane potential of 𝑉𝑆𝐸𝑡.

(NL CH4 kg VS–1)𝑇𝐵𝑀𝑃𝑋𝑃 Stoichiometric methane potential of 𝑉𝑆𝑋𝑃

(NL CH4 kg VS–1)𝑇𝐵𝑀𝑃𝑋𝐿 Stoichiometric methane potential of 𝑉𝑆𝑋𝐿

(NL CH4 kg VS–1)𝑇𝐵𝑀𝑃𝐻𝑀 Stoichiometric methane potential of 𝑉𝑆𝐻𝑀

(NL CH4 kg VS–1)𝑇𝐵𝑀𝑃𝐶𝐿 Stoichiometric methane potential of 𝑉𝑆𝐶𝐿

(NL CH4 kg VS–1)𝑇𝐵𝑀𝑃𝐶𝐵 Stoichiometric methane potential of 𝑉𝑆𝐶𝐵

1 1. Introduction

2 Biogas production by anaerobic digestion (AD) is known to be the most sustainable bioenergy

3 technology available owing to its utilization of waste biomass as a substrate, greenhouse gas

4 reduction efficacy, and production of digestate, which is used as a fertilizer (Saidu et al., 2013).

5 The flexibility of using biomass as a source of energy is another benefit. For example, while

6 bioethanol and biodiesel require specific types of biomass as substrates, biogas can be generated

7 from any organic compound, either solid or liquid. Therefore, a wide range of biomasses,

8 including animal slurry, agricultural residues, municipal solid waste, and food processing waste

9 can be used to produce biogas (Tuesorn et al., 2013).

10 Owing to the diverse AD pathways of organic carbons, predicting the actual methane yield is not

11 as simple as predicting the yield of other biofuels. Furthermore, owing to the flexibility in the

12 biomasses that can be used for biogas production and the fact that AD of single substrates

13 (mono-digestion) suffers from drawbacks related to substrate properties, co-digestion of different

14 biomasses has been implemented in industrial biogas plants (Mata-Alvarez et al., 2014). The co-

15 digestion of organic waste materials with animal manure is appealing for biogas production from

16 food, food processing, slaughterhouse, fruit, and vegetable wastes (Llaneza Coalla et al., 2009).

17 However, the full energy content of major organic carbons ranges widely from 375 NL CH4 kg–1

18 (organic acids) to 1014 NL CH4 kg–1 (lipid). Moreover, their hydrolysis behaviour, especially of

19 lignocellulose-combined carbohydrates, is strongly affected by the physicochemical bonding

20 within and between lignocellulosic macromolecules, i.e. the lignification degree, amorphous and

21 crystalline contents in cellulose, and the degree of crystallinity are affected by the nature of

22 bonding. Furthermore, the hydrolysis mechanism of these biopolymers and their structural

ACCEPTED MANUSCRIPT

4

1 barriers against enzymatic hydrolysis are not clearly understood, which in turn hinders the

2 extensive use of alternative potential biomasses (Wilson, 2011).

3 Many researchers have attempted to derive the kinetic constants of specific biomasses to predict

4 biogas yield in real scale biogas reactors using the data from batch AD experiments (Ibrahim,

5 2014; Syaichurrozi et al., 2015). However, the decomposition environment in real scale

6 continuous biogas reactors is not similar to that in a batch test. Differences exist in terms of the

7 limited microbial population and their acclimation issues, agitation effect, etc. With regard to the

8 energy contents, biochemical methane potential (BMP) is the most commonly used parameter to

9 assess the methane potential of a biomass and numerous studies have dealt with analysing the

10 BMPs of a wide range of biomasses. Due to the time-consuming and labour-intensive

11 determination procedure as well as the high possibility of errors in the reproducibility of BMP

12 results, many researchers have suggested alternative measurements (Bayard et al., 2018; Bekiaris

13 et al., 2015; Cu et al., 2015; Doublet et al., 2013; Raju et al., 2011; Strömberg et al., 2015; Triolo

14 et al., 2014, 2011; Xu et al., 2014). Simple or multiple regression statistical approaches were

15 developed (Bayard et al., 2016; Cu et al., 2015; Kafle and Chen, 2016; Thomsen et al., 2014) and

16 spectroscopic methods, such as applying near infrared spectroscopy (Doublet et al., 2013; Raju et

17 al., 2011; Triolo et al., 2014) or Fourier transform mid-infrared photoacoustic spectroscopy

18 (Bekiaris et al., 2015), were suggested. However, the state-of-the art technology still requires

19 sample drying prior to obtaining the spectra, which makes real time analysis of BMP impossible.

20 The aim of the present study is to provide a new insight on the hydrolysis behaviour of major

21 organic compounds during co-digestion in the semi-continuous mode and to attempt at

22 developing a new BMP model from the results.

23 2. Materials and methods

24 2.1. Feedstock and inoculum

ACCEPTED MANUSCRIPT

5

1 Energy-dense and easily degradable agro-industrial wastes and industrial wastewater as co-

2 substrates (25 wt.%) and pig manure (75 wt.%) as prime feedstock were examined in both batch

3 and semi-continuous modes. The co-digestion parameters, i.e. the mixing ratio and choice of co-

4 substrate, were similar to those used commonly in Danish central biogas plants; in these plants,

5 25 wt.% of diverse energy-dense wastes (slaughterhouse waste, expired canned ham, fish and

6 dairy waste, etc.) are typically co-digested with animal manure. Brewery wastewater was

7 collected from a local brewery (Odense, Denmark) and processed meat (canned pork and ham)

8 were purchased from supermarkets. Inoculum, pig manure and slaughterhouse waste were

9 obtained from the Fangel biogas plant (Odense, Denmark). The feed for AD consisted of 75

10 wt.% pig manure and 25 wt.% of the co-substrate mixture. The final composition of the feed was

11 75 wt.% pig manure, 11.25 wt.% processed meat waste, 11.25 wt.% brewery waste, and 2.5 wt.%

12 slaughterhouse waste. The dry matter (DM) and VS content of inoculum were 47.8 (±2.3) and

13 29.6 (±1.8) g kg–1 respectively.

14 2.2. Anaerobic digestion in the semi-continuous mode

15 Four stirred tank reactors (CSTRs) with a total volume of 20 L were utilized for the AD process.

16 Each reactor was made of two stainless steel plates and an acrylic cylinder (diameter 360 mm

17 and height 360 mm) was fixed between the two plates. The engine was installed at the bottom of

18 the reactor to mix the medium using two radial flow-impellers of 0.2 m diameter. In order to

19 supply the heat required for the reaction, four 50 W elements were attached to the bottom plate

20 and were controlled by a temperature sensor inside the reactor. The gas exhaust and feeding inlet

21 were supported by the top plate, while the two digestate outlets were fixed on the bottom plate at

22 two different levels.

23 The inoculum was obtained from the Fangel biogas plant, which processes 75 wt.% of animal

24 manure as the prime feedstock and 25 wt.% of industrial food processing waste at mesophilic

25 conditions. After an acclimation period of 14 days, the four reactors were subjected to different

26 organic loading rates (OLRs) and hydraulic retention times (HRTs) of 15, 20, 30, and 45 days

ACCEPTED MANUSCRIPT

6

1 under mesophilic conditions at 37°C. The volatile solid (VS) content of the feed mixture was

2 58.26 g kg–1 while the OLR values were 3.88 g VS L–1 day–1, 2.91 g VS L–1 day–1, 1.94 g VS L–1

3 day–1, and 1.29 g VS L–1 day–1. The reactors were fed on a daily basis and the gas production was

4 monitored. Volatile fatty acids (VFAs) were measured twice a week to monitor the stability of

5 the anaerobic digestion process. The AD operation was continued for a period of three times as

6 long as the HRTs. At the end of the operation time, the digestion was terminated and the

7 digestates were removed and placed in a freezer for further analysis.

8 2.3. BMP assay

9 The German standard method (VDI 4630, 2006) was used to determine the BMP of the

10 substrates and digestates from each reactor using batch reactors of 1 L capacity. The inoculum-

11 to-substrate ratio (I/S ratio) was set at 2.5 based on the VS content. All the experiments were

12 conducted in triplicate. The biogas yield of inoculum as a blank test for the correction of gas

13 production and methane yield of microcrystalline cellulose (Avicel® PH-101, Sigma-Aldrich) as

14 a reference material was determined. After mixing the substrate and inoculum, which had already

15 been degassed for two weeks, 150 mL of a nitrogen-flushed buffer medium was added to the

16 inoculum and substrate mixture according to the ISO standard method. The VS of BMP tests was

17 4.7 % (<10 %) ( ISO, 1995; Pham et al., 2013; Holliger et al., 2016). The batch digesters were

18 flushed with nitrogen gas to provide a fully anaerobic system. The BMP test was performed

19 under mesophilic conditions at 37°C for 60 days. The digesters were shaken manually to prevent

20 layering and to boost degassing. A syringe (Hamilton Super Syringe) was used to measure the

21 wet biogas volume and later, these gas volumes were corrected to dry biogas as normal litres.

22 CH4 concentration was determined using a gas chromatograph (GC; 7890A, Agilent

23 Technologies, USA) equipped with a thermal conductivity detector and a 30 m x 0.320 mm

24 column (J&W 113-4332, Agilent Technologies, USA).

25 2.4. Sample and digestate analysis

ACCEPTED MANUSCRIPT

7

1 The DM was measured according to a standard procedure (APHA, 2005) by drying the samples

2 at 105°C. The ash content of the dry matter was determined after placing the dried samples in a

3 muffle furnace at 550°C for 2 h and the VS content was calculated by the difference in the DM

4 and ash content. Total Kjeldahl nitrogen (TKN) and total ammoniacal nitrogen (TAN) were

5 determined according to a standard method (APHA, 2005). The crude lipid content was

6 measured by Soxhlet extraction.

7 The obtained DM results were corrected because during DM measurement, easily volatile

8 compounds evaporate. It is important to measure their concentration in the untreated sample and

9 add this concentration to the oven-dried samples to obtain the corrected DM (Weissbach and

10 Strubelt, 2008). VFA concentration from C2–C5 was determined using a gas chromatograph

11 (Hewlett Packard 6890, Italy) with a flame ionization detector and a 30 m x 0.25 mm x 0.25 μm

12 column (HP-INNOWax, USA) (Larsen et al., 2017). Ethanol content was measured by high-

13 performance liquid chromatography (HPLC; Agilent 1100, Germany). For alcohol and VFA

14 analysis, the samples were filtered through a 0.2 μm nylon membrane filter prior to injection into

15 the GC and HPLC columns. For VFA determination, the pH of the samples was adjusted to about

16 2 by adding phosphoric acid (Larsen et al., 2017). Lignin, cellulose, and hemicellulose were

17 determined according to Van Soest’s characterization for fibre analysis (Van Soest, 1963).

18 2.5. Data analysis

19 Crude protein was determined by multiplying the difference between TKN and TAN by 6.25

20 (Triolo et al., 2011). α-NDF was used to measure the total fibrous fractions consisting of

21 hemicellulose, cellulose, and lignin. Acid detergent fibre (ADF) includes cellulose and lignin

22 whereas acid detergent lignin (ADL) represents the lignin content. The hemicellulose content

23 was calculated from the difference between α-NDF and ADF and cellulose content was

24 calculated from the difference between ADF and ADL (Triolo et al., 2011).

25 Bushwell’s formula was used to calculate the theoretical biochemical methane potential (TBMP)

26 of a specific compound with defined molecular formulae under standard conditions (273 K and

ACCEPTED MANUSCRIPT

8

1 101.325 kPa). TBMP is a term estimated by assuming the total conversion of a specific

2 compound to methane and carbon dioxide ( Symons and Buswell, 1933; Raposo et al., 2012):

3 (1)𝐶𝑛𝐻𝑎𝑂𝑏𝑁𝐶 + (𝑛 ―𝑎4 ―

𝑏2 +

3𝑐4 )𝐻2𝑂→(𝑛

2 +𝑎8 ―

𝑏4 ―

3𝑐8 )𝐶𝐻4 + (𝑛

2 ―𝑎8 +

𝑏4 +

3𝑐8 )𝐶𝑂2 + 𝑐𝑁𝐻3

4 (2)𝐵𝑡ℎ =(𝑛 2 + 𝑎 8 ― 𝑏 4 ― 3𝑐 8)22400

12𝑛 + 𝑎 + 16𝑏 + 14𝑐 𝑁𝐿 𝐶𝐻4 𝑘𝑔 𝑉𝑆 ―1

5 TBMP of the substrate and digestate was calculated using the TBMP of each VS compound; the

6 TBMPs of VFAs, ethanol, crude proteins, crude lipids, lignin, hemicellulose, cellulose, and

7 carbohydrates were 373, 730, 496, 1014, 727, 415, 415 and 415 NL CH4 kg VS–1, respectively.

8 2.6. Degradation kinetics and CH4 production model

9 The hydrolysis of organic matter for biogas production is often described by first-order kinetics

10 (Angelidaki et al., 2009; Dandikas et al., 2018; Ebrahimi-Nik et al., 2018; Hashimoto, 1989;

11 Triolo et al., 2014). As the degradation rate of organic compounds controlled by hydrolysis, first-

12 order kinetics is used to derive the kinetics equation. However consideration should be given to

13 the fact that each compound is composed of both degradable and non-degradable fractions.

14 (3)𝑆 = 𝑆𝐷 + 𝑆𝑁𝐷

15 (4)𝑑𝑆𝐷

𝑑𝑡 = ―𝑘𝑆𝐷

16 Therefore:

17 (5)𝑆𝐷 = 𝑆0𝐷.𝑒 ―𝑘𝜃

18 and/or (6)𝑑𝑓∞ =𝑆0𝐷

𝑆0𝑑𝑓𝜃 =

𝑆0𝐷 ― 𝑆𝐷 𝑆0

19 From the combination of Eq. (5) and (6), the degradation kinetics would be as follows:

20 (7)𝑑𝑓𝜃 = 𝑑𝑓∞ * (1 ― 𝑒 ―𝑘 𝜃)

ACCEPTED MANUSCRIPT

9

1 where S is the total concentration of an organic compound, SD is the concentration of the

2 degradable part of an organic compound, SND is the concentration of the non-degradable (inert)

3 part of an organic compound, dfθ is the decay of an organic compound (%) at a certain time, df∞

4 is the ultimate decay of an organic compound (%), k is the decay rate constant for an organic

5 compound (day–1) and θ is the HRT (days).

6 From Eq. (7), the specific rate of CH4 production can be predicted as shown below.

7 (8)𝑟𝐶𝐻4 =∑𝑛

𝑖 = 1(𝑑𝑓𝑖𝜃 ·𝑀𝑖0)

𝜃

8 Here, is the specific rate of CH4 production (NL CH4 L–1 day–1), dfiθ is the decay of an 𝑟𝐶𝐻4

9 organic compound i (%), and Mi0 is the TBMP of the organic compounds i in the input feedstock

10 (NL CH4 kg wet weight (ww)–1).

11 During the validation of the BMP prediction model, the precision of the model was tested using

12 external datasets by applying the statistical parameters of residual mean square error in

13 prediction (RMSEP) and relative residual mean square error in prediction (rRMSEP), which is

14 normalized RMSEP using the mean value of the measured BMP. Least-squares non-linear

15 regression was performed by minimising the residual sum of squares (RSS):

16 (9)𝑟𝑒𝑠𝑖𝑑𝑢𝑎𝑙 𝑠𝑢𝑚 𝑜𝑓 𝑠𝑞𝑢𝑎𝑟𝑒 (𝑅𝑆𝑆) = ∑𝑛𝑖 = 1(𝑦𝑖 ― 𝑦𝑖)2

17 where is the predicted value and the measured value. 𝑦𝑖 𝑦𝑖

18 3. Results and discussion

19 3.1. Characteristics of the substrates and digestates

20 The physicochemical characterization results together with the concentrations of major organic

21 compounds are presented in Table 1. Lignocelluloses were not present in any of the co-

22 substrates, but in contrast, the pig manure was mainly composed of lignocellulose. The brewery

23 wastewater was diluted to 7.8 g DM kg–1 and was mostly composed of ethanol. The

ACCEPTED MANUSCRIPT

10

1 slaughterhouse waste and processed meat waste, which were two types of semi-solid wastes,

2 contributed to a concentration of 70 g DM kg–1 in the feed mixture. The processed meat waste

3 was rich in both proteins and lipids while the slaughterhouse waste contained mostly lipids.

4 (Table 1 near here)

5 The characteristics of the pig manure used in this study (DM content and organic composition)

6 were similar to those used in previous studies (Møller et al., 2004; Tsapekos et al., 2017). The

7 high content of lignocelluloses in the pig manure clearly indicates that the pig manure is only a

8 recalcitrant substrate in the feed mixture. Although plant biomass, typically wood or agricultural

9 residues, such as straw and bagasse, are generally known as lignocellulosic biomasses, the

10 lignocellulosic characteristics of animal manure have been pointed out in recent studies due to its

11 large concentration (Bruni et al., 2010; Cu et al., 2015).

12 The organic composition of the mixture of these four substrates resulted in a fairly good balance

13 of lipids, proteins, and lignocelluloses with a small content of dissolved VS sources, i.e. ethanol,

14 VFAs, etc., which reduces the concentration of the recalcitrant lignocelluloses; meanwhile, the

15 proteins and lipids became the dominant organic compounds in the feed mixture (Table 1).

16 3.2. Methane potentials

17 Table 1 shows that the organic compositions of the four substrates agree well with the BMP and

18 biodegradability results. The BMP of pig manure was considerably lower than that of other co-

19 substrates. In fact, the BMP of slaughterhouse waste, which mostly contains lipids, was

20 approximately three times higher than the BMP of pig manure. The BMP calculated for pig

21 manure is in good agreement with previously reported values (Triolo et al., 2013). The

22 differences in the BMPs of the substrates were considerably more evident when the specific

23 methane potential was considered (NL CH4 kg ww –1). The values of the specific methane

24 potential values of the substrates were 9 NL CH4 kg ww –1 (pig manure), 145 NL CH4 kg ww –1

ACCEPTED MANUSCRIPT

11

1 (processed meat waste), 258 NL CH4 kg ww –1 (slaughterhouse waste), and 3.6 NL CH4 kg ww –1

2 (brewery wastewater).

3 Such large differences in the specific methane potentials of the substrates consequently led to

4 diverse levels of contribution to the overall methane potential of the feed mixture. Among

5 others, pig manure, which was 75 wt.% of the feed mixture, contributed to only 22% of the

6 overall methane potential. Moreover, the brewery wastewater contributed to only 1% of the

7 methane potential of the feed mixture. On the other hand, despite its content being only 2.5

8 wt.% in the feed mixture, slaughterhouse waste contributed to 23% of the overall methane

9 potential due to its high solid and lipid concentration. In the feed mixture, the contribution of

10 VFAs, ethanol, XP, XL, hemicellulose, cellulose, lignin and carbohydrate to its TBMP was 5, 1,

11 26, 47, 4, 6, 7 and 4 % respectively. Therefore, a large fraction of methane sources in the feed

12 mixture consisted of lipids.


14 The results highlight that the contribution of brewery wastewater to the overall methane potential

15 during the co-digestion of different biomasses was nearly negligible; however, it would be

16 effective as a buffer medium to dilute so-called inhibitors, especially when AD is carried out

17 with other protein and lipid substrates.

18 3.3. CH4 production and CSTR performance

19 Specific biogas and CH4 production rate (NL CH4 L–1 day–1) and VFAs concentration during AD

20 in semi-continuous mode in each reactor are plotted against time in Figure 1. It can be seen that

21 the CH4 production was steady and there was no clear inhibition in the gas yield. The CH4

22 content in the biogas was very stable and varied from 69% to 71%. The high CH4 concentration

23 in the exhausted biogas was probably due to the presence of lipids in the feed.

24 (Figure 1 near here)

ACCEPTED MANUSCRIPT

12

1 The VFA concentration gradually diminished at greater HRTs: (7.69 ± 0.17) g kg–1 (HRT 15

2 days), (2.96 ± 0.09) g kg–1 (HRT 20 days), and (0.65 ± 0.05) g kg–1 (HRT 30 days).

3 Accumulation of VFAs in the reactor was observed clearly at a HRT of 15 days; however, the

4 considerably high level of VFAs was stable throughout HRT durations of 20 days and 30 days.

5 Aguilar et al. (1995) showed that there was no inhibition in acetate degradation when the VFA

6 concentration was less than 4 g L–1; when it exceeded 10 g L–1, the AD process was inhibited

7 (Aguilar et al., 1995). Fortela et al. (2016) found that activated sludge microbial consortia were

8 highly inhibited at total VFA concentrations of 10 g L–1 and 20 g L–1 (Fortela et al., 2016).

9 TAN was moderately high, ranging from 3.77 g kg–1 to 4.02 g kg–1. Despite the high TAN

10 content in the reactors, there was no clear inhibition of ammonia; for example, the VFA

11 concentration varied widely with variations in the HRT and in the reactor with an HRT of 45

12 days, it was nearly zero in the steady state phase with no evidence of the inhibited methanogenic

13 activity caused by ammonia. The unclear inhibition due to TAN can be attributed to the fact that

14 the inoculum had adapted well after it was obtained from the biogas plant operated using manure,

15 food processing waste, and slaughterhouse waste.

16 Previous studies on ammonia inhibition reported a wide range of ammonia inhibition activities

17 (mostly expressed as TAN concentration) from 1.7 g L–1 to 14 g L–1; the large variations in the

18 TAN level have been discussed in terms of the different substrates used, inocula, pH,

19 temperature, other operational conditions, and acclimation (Chen et al., 2008; Yenigün and

20 Demirel, 2013). Regarding the adaptation of methanogens and acclimation to ammonia, Yenigün

21 and Demirel (2013) reported that AD occurred efficiently at concentration of TAN up to 9 g L–1

22 (Yenigün and Demirel, 2013).

23 3.4. Anaerobic digestion of organic compounds

24 The contents of crude lipids, crude proteins, cellulose, hemicelluloses, lignin, VFAs, and ethanol

25 were measured before and after digestion (Figure 2). As expected, ethanol was completely

ACCEPTED MANUSCRIPT

13

1 decomposed in all the reactors. It can be observed in Figure 2 that the degradation of lipids and

2 proteins was similar in all the reactors; it is noteworthy that crude lipid was not present in the

3 digestate in any of the reactors; 27% of the feed protein was present in the digestate, which

4 corresponds to a decomposition of 73% ± 3% of the crude protein. The remaining organic

5 nitrogen is probably non-degradable, i.e. lignin-combined nitrogen, or it is transformed to

6 microbial communities.


8 Due to the presence of relatively high levels of VFAs at shorter HRTs, it is not clear whether the

9 lipid was completely digested or not within 15 days of AD; however, it can still be confirmed

10 that the entire lipid content was hydrolysed within 15 days and in addition, it can be stated that

11 the extent of hydrolysis of the lipid and protein was not affected by the operational time, given

12 that the retention time is long enough (≥15 days). It can be inferred from the biodegradability

13 index (BMP/TBMP) in Table 1 that the results obtained on the hydrolysis of lipids and proteins

14 were in good agreement with the ratio between BMP and TBMP. Consequently, the BMP of

15 lipid- and protein-rich biomass can be analyzed by measuring the protein and lipid content.

16 On the other hand, the degradation of the two lignocelluloses (hemicelluloses and cellulose)

17 increased gradually, closely following first-order degradation kinetics. Furthermore, the

18 degradation of hemicellulose was faster than that of cellulose. Interestingly, in spite of a long

19 retention time (HRT of 45 days), only 45% of the hemicelluloses and 34% of cellulose

20 decomposed, implying that 55% and 66% of hemicelluloses and cellulose, respectively, were

21 available in the feed mixture as compared to the initial feed.

22 The reason for the relatively easier hydrolysis of hemicellulose as compared to cellulose is due to

23 its amorphous structure. It is known that lignin does not degrade in the absence of oxygen and

24 our results also show that its concentration in all the digestates is nearly the same as its

25 concentration in the input feed mixture.

ACCEPTED MANUSCRIPT

14

1 Table 3 presents the first-order decay kinetic constants of cellulose and hemicellulose and their

2 statistical parameters calculated using Eq. (7). Figure 3 visualizes the degradation of cellulose

3 and hemicellulose in terms of the measured and predicted values and the simulation results of

4 200 days extended retention time.



7 Residual analysis using Eq. (7) with 4 dataset points for both hemicellulose and cellulose

8 indicates that hemicellulose was composed of 59.3% VSSD while the remaining 40.7% was inert

9 (non-degradable VS (VSND)). In the case of cellulose, the fraction of VSSD was even smaller at

10 36.3% (Figure 3). Furthermore, when the first-order decay kinetics model was fitted using the

11 lignocellulose data points, it was found that 72.6% of the mass corresponded to VSND. Therefore,

12 it can be roughly assessed that the biodegradability of lignocellulose was ~37.4%. This result is

13 probably comparable with the biodegradability data of 55 lignocellulosic biomass samples

14 reported previously by Triolo et al. (2012), who reported the following values) hedge cuttings,

15 39.9(7.6)%; woody cuttings 32.7(5.2)%; wild plants 44.9(12.5)% (Triolo et al., 2012).

16 3.5. Prediction of BMP

17 With the aim of providing a new and alternative method for analysing the BMP, we categorized

18 organic compounds into three pools: easily degradable VS (VSED), VSSD, and VSND. VSED

19 includes non-lignocellulosic carbohydrate, organic acids, alcohols including all the intermediate

20 products produced during hydrolysis, and methanogens; crude lipid was also categorized as VSED

21 due to its complete decomposition within 15 days. Crude protein was divided into VSED and

22 VSND, taking the remaining protein in the digestate into account. Hemicellulose and cellulose

23 were defined as VSSD and lignin VSND.

24 = + (1 – ) (10)𝑉𝑆𝑋𝑃 𝑑𝑓𝑋𝑃 𝑉𝑆𝑋𝑃 𝑑𝑓𝑋𝑃 𝑉𝑆𝑋𝑃

ACCEPTED MANUSCRIPT

15

1 Here, is the concentration of crude protein in VS and is the ultimate degradation 𝑉𝑆𝑋𝑃 𝑑𝑓𝑋𝑃

2 factor of , which has a value of 0.73. 𝑉𝑆𝑋𝑃

3 Eq. (11) and (12) were deduced from Table (3).

4 = + (1 – ) (11)𝑉𝑆𝐻𝑀 𝑑𝑓𝐻𝑀 𝑉𝑆𝐻𝑀 𝑑𝑓𝐻𝑀 𝑉𝑆𝐻𝑀

5 is the concentration of hemicellulose in VS and is the ultimate degradation factor of 𝑉𝑆𝐻𝑀 𝑑𝑓𝐻𝑀

6 , which has a value of 0.59.𝑉𝑆𝐻𝑀

7 = + (1 – ) (12)𝑉𝑆𝐶𝐸 𝑑𝑓𝐶𝐸 𝑉𝑆𝐶𝐸 𝑑𝑓𝐶𝐸 𝑉𝑆𝐶𝐸

8 is the concentration of cellulose in VS and is the ultimate degradation factor of , 𝑉𝑆𝐶𝐸 𝑑𝑓𝐶𝐸 𝑉𝑆𝐶𝐸

9 which has a value of 0.36.

10 Finally, the BMP of the biomass can be estimated by Eq. (13).

11 𝐵𝑀𝑃𝑃 = 𝑇𝐵𝑀𝑃𝑉𝐹𝐴·𝑉𝑆𝑉𝐹𝐴 + 𝑇𝐵𝑀𝑃𝐸𝑡.·𝑉𝑆𝐸𝑡. + 𝑑𝑓𝑋𝑝·𝑇𝐵𝑀𝑃𝑋𝑃 ·𝑉𝑆𝑋𝑃 + 𝑇𝐵𝑀𝑃𝑋𝐿·𝑉𝑆𝑋𝐿 +12 (13)𝑑𝑓𝐻𝑀·𝑇𝐵𝑀𝑃𝐻𝑀·𝑉𝑆𝐻𝑀 + 𝑑𝑓𝐶𝐿·𝑇𝐵𝑀𝑃𝐶𝐿·𝑉𝑆𝐶𝐿 + 𝑇𝐵𝑀𝑃𝐶𝐵 ∙ 𝑉𝑆𝐶𝐵

13 refers to the predicted BMP (NL CH4 kg VS–1); , , 𝐵𝑀𝑃𝑃 𝑉𝑆𝑉𝐹𝐴, 𝑉𝑆𝐸𝑡. 𝑉𝑆𝑋𝑃, 𝑉𝑆𝑋𝐿 , 𝑉𝑆𝐻𝑀, 𝑉𝑆𝐶𝐿

14 and are VS compounds referring to VFA, ethanol, crude protein, crude lipid, hemicellulose, 𝑉𝑆𝐶𝐵

15 cellulose, and non-fibrous carbohydrates, respectively (kg kg VS–1). , ,𝑇𝐵𝑀𝑃𝑉𝐹𝐴 𝑇𝐵𝑀𝑃𝐸𝑡.

16 are stoichiometric methane potentials 𝑇𝐵𝑀𝑃𝑋𝑃 , 𝑇𝐵𝑀𝑃𝑋𝐿, 𝑇𝐵𝑀𝑃𝐻𝑀, 𝑇𝐵𝑀𝑃𝐶𝐿, and 𝑇𝐵𝑀𝑃𝐶𝐵

17 (NL CH4 kg VS–1) corresponding to , , and , 𝑉𝑆𝑉𝐹𝐴, 𝑉𝑆𝐸𝑡. 𝑉𝑆𝑋𝑃, 𝑉𝑆𝑋𝐿 , 𝑉𝑆𝐻𝑀, 𝑉𝑆𝐶𝐿 𝑉𝑆𝐶𝐵

18 respectively; , , and are the ultimate degradation factors of , and 𝑑𝑓𝑋𝑃 𝑑𝑓𝐻𝑀 𝑑𝑓𝐶𝐿 𝑉𝑆𝑋𝑃, 𝑉𝑆𝑋𝑃

19 , respectively.𝑉𝑆𝑋𝑃

20 Different studies have used various approaches to predict BMP. These have mostly focused on a

21 statistical model using regression and correlation between BMP and some variables, especially

22 lignin, cellulose and hemicellulose (Bayard et al., 2016; Kafle and Chen, 2016; Thomsen et al.,

23 2014; Triolo et al., 2011; Xu et al., 2014). In the BMP model suggested in this paper, the

ACCEPTED MANUSCRIPT

16

1 coefficient for each compound was derived from their degradation kinetics, in which only

2 degradable components were taken into consideration. However further study is needed to

3 include the lignin effect on the degradation kinetics of cellulose and hemicellulose. As most of

4 the components were included in the present model, it can be used for a wide variety of

5 biomasses for which models based on the lignocellulosic fraction alone might not be very

6 effective or useful for proteinous and fatty substrates.

7 3.6. BMP model validation

8 The precision of the suggested model in predicting the BMP was tested using the samples from

9 this study. External validation was performed using the BMP results and compositions published

10 elsewhere as the reference data variables. The results obtained using the internal data were quite

11 precise with very low model error. The model error in prediction (RMSEPE) was only 25 (NL CH4

12 kg VS–1); thus, the normalized model error was low as well (4.2% of rRMSE). Even if the BMP

13 prediction model was not built by directly using the BMP values from the testing set, the

14 satisfactory validation results would however still obtained because the same samples were used

15 for validation. Nonetheless, the very high precision seen in Table 4 is most interesting, which

16 indicates that the calibrated model exhibits high precision.

17 For the validation of the suggested model, external data from 10 maize samples (Oslaj et al., 2010)

18 from Oslaj et al. (2010) and 51 lignocellulosic samples from the authors’ previous study (Triolo et

19 al., 2012) were used as the testing sets. In total, 65 external data sets were tested. The

20 lignocellulosic data sets were divided into two categories: 1) grasses and agricultural residues and

21 2) woody samples, and they were tested separately. In the case of the grass and agricultural residues,

22 the model was tested with and without four well-crystallized winter-harvested grasses with

23 critically low methane potentials. This was done to test the responses of the model to winter-

24 harvested grasses.


ACCEPTED MANUSCRIPT

17

1 The results of the validation exercise were found to be very interesting as the model had a tendency

2 to overestimate the BMP of woody biomass and winter-harvested grasses (Figure 4; Table 5). On

3 the other hand, it exhibited relatively fine precision when tested on green grasses with maize. The

4 obtained rRMSEPE values for maize and grasses were 12.1% and 15.8%, respectively, which

5 corresponds to a model error of 37.4 (NL CH4 kg VS–1) for maize and 49.9 (NL CH4 kg VS–1) for

6 straw. When the winter-harvested grasses were included, the relative model error increased to

7 19.6%. R2 decreased to 0.47 when winter grasses were excluded (Table 5). The higher correlation

8 level compared to the lower model precision when including winter harvested grasses is typically

9 seen in wider ranges of independent variables when including samples with very low methane

10 potentials (Triolo et al., 2011).

11 The BMP model was most precise in the case of the maize samples. This was probably because of

12 the highly homogeneous characteristics of the maize samples, when compared to a wide range of

13 lignocellulosic biomasses harvested through different seasons. Despite a low model error for maize,

14 there was no correlation found, probably due to low variation of methane potentials included in

15 the model validation. Except for the woody and winter-harvested grass samples, the predicted and

16 measured BMP values are not clearly biased, as can be seen in Figure 4.


18 Only a few BMP prediction studies on organic compounds carried out external validation;

19 therefore, a comparison was not possible to analyze the precision of the BMP model. Comparing

20 this model to the other alternative methods used to predict BMP (for e.g. near infrared spectroscopy

21 (NIR)), it was found that the results were almost equivalent (Bekiaris et al., 2015).

22 4. Conclusions

23 This study emphasized that the majority of lipids and proteins in agro-industrial waste are

24 transformed into methane in modern industrial biogas plants (HRT ≥ 15 days). The methane yield

25 obtained by the degradation of the remaining hemicelluloses and cellulose increases at longer

ACCEPTED MANUSCRIPT

18

1 retention times. The model in this paper includes both lignocellulosic and non-lignocellulosic

2 compounds, therefore it can be applied to a wide range of biomasses where the regression models

3 based on the lignocellulose fraction alone might not be applicable for non-lignocellulosic

4 biomasses. The developed BMP model fits best to lightly lignified biomass but the overestimated

5 results obtained with woody biomass and winter-harvested grass indicate that further

6 investigation of the hydrolysis of well-lignified biomass is required to improve the model

7 precision. The model can easily be applied to energy-intense lipid- or protein-rich biomass with

8 high precision.

9 Acknowledgements

10 This study was supported by a grant from the Danish Council for Strategic Research (12-132631)

11 under the work programme “Optimisation of Value Chains for Biogas Production in Denmark”

12 (BioChain). The authors are responsible for the content of this publication.

13 References

14 Aguilar, A., Casas, C., Lema, J.M., 1995. Degradation of volatile fatty acids by differently

15 enriched methanogenic cultures: Kinetics and inhibition. Water Res. 29, 505–509.

16 https://doi.org/10.1016/0043-1354(94)00179-B

17 Angelidaki, I., Alves, M., Bolzonella, D., Borzacconi, L., Campos, J.L., Guwy, A.J.,

18 Kalyuzhnyi, S., Jenicek, P., Van Lier, J.B., 2009. Defining the biomethane potential

19 (BMP) of solid organic wastes and energy crops: A proposed protocol for batch assays.

20 Water Sci. Technol. 59, 927–934. https://doi.org/10.2166/wst.2009.040

21 APHA, 2005. Standard Methods for the Examination of Water and Wastewater, 21st. ed.

22 American Public Health Association, Washington, DC.

23 Bayard, R., Benbelkacem, H., Gourdon, R., Buffière, P., 2018. Characterization of selected

24 municipal solid waste components to estimate their biodegradability. J. Environ. Manage.

ACCEPTED MANUSCRIPT

19

1 216, 4–12. https://doi.org/10.1016/J.JENVMAN.2017.04.087

2 Bayard, R., Liu, X., Benbelkacem, H., Buffiere, P., Gourdon, R., 2016. Can Biomethane

3 Potential (BMP) Be Predicted from Other Variables Such As Biochemical Composition in

4 Lignocellulosic Biomass and Related Organic Residues? BioEnergy Res. 9, 610–623.

5 https://doi.org/10.1007/s12155-015-9701-3

6 Bekiaris, G., Triolo, J.M., Peltre, C., Pedersen, L., Jensen, L.S., Bruun, S., 2015. Rapid

7 estimation of the biochemical methane potential of plant biomasses using Fourier

8 transform mid-infrared photoacoustic spectroscopy. Bioresour. Technol. 197, 475–481.

9 https://doi.org/10.1016/j.biortech.2015.08.050

10 Bruni, E., Jensen, A.P., Angelidaki, I., 2010. Comparative study of mechanical, hydrothermal,

11 chemical and enzymatic treatments of digested biofibers to improve biogas production.

12 Bioresour. Technol. 101, 8713–8717. https://doi.org/10.1016/j.biortech.2010.06.108

13 Chen, Y., Cheng, J.J., Creamer, K.S., 2008. Inhibition of anaerobic digestion process: A review.


15 Cu, T.T.T., Nguyen, T.X., Triolo, J.M., Pedersen, L., Le, V.D., Le, P.D., Sommer, S.G., 2015.

16 Biogas production from vietnamese animal manure, plant residues and organic waste:

17 influence of biomass composition on methane yield. Asian-Australasian J. Anim. Sci. 28,

18 280–9. https://doi.org/10.5713/ajas.14.0312

19 Dandikas, V., Heuwinkel, H., Lichti, F., Eckl, T., Drewes, J.E., Koch, K., 2018. Correlation

20 between hydrolysis rate constant and chemical composition of energy crops. Renew.

21 Energy 118, 34–42. https://doi.org/10.1016/J.RENENE.2017.10.100

22 Doublet, J., Boulanger, A., Ponthieux, A., Laroche, C., Poitrenaud, M., Cacho Rivero, J.A.,

23 2013. Predicting the biochemical methane potential of wide range of organic substrates by

24 near infrared spectroscopy. Bioresour. Technol. 128, 252–258.

ACCEPTED MANUSCRIPT

20


2 Ebrahimi-Nik, M., Heidari, A., Ramezani Azghandi, S., Asadi Mohammadi, F., Younesi, H.,

3 2018. Drinking water treatment sludge as an effective additive for biogas production from

4 food waste; kinetic evaluation and biomethane potential test. Bioresour. Technol. 260,

5 421–426. https://doi.org/10.1016/J.BIORTECH.2018.03.112

6 Fortela, D.L., Hernandez, R., French, W.T., Zappi, M., Revellame, E., Holmes, W., Mondala,

7 A., 2016. Extent of inhibition and utilization of volatile fatty acids as carbon sources for

8 activated sludge microbial consortia dedicated for biodiesel production. Renew. Energy

9 96, 11–19. https://doi.org/10.1016/j.renene.2016.04.068

10 Hashimoto, A.G., 1989. Effect of inoculum/substrate ratio on methane yield and production rate

11 from straw. Biol. Wastes 28, 247–255. https://doi.org/10.1016/0269-7483(89)90108-0

12 Holliger, C., Alves, M., Andrade, D., Angelidaki, I., Astals, S., Baier, U., Bougrier, C.,

13 Buffiere, P., Carballa, M., de Wilde, V., Ebertseder, F., Fernandez, B., Ficara, E., Fotidis,

14 I., Frigon, J.-C., de Laclos, H.F., Ghasimi, D.S.M., Hack, G., Hartel, M., Heerenklage, J.,

15 Horvath, I.S., Jenicek, P., Koch, K., Krautwald, J., Lizasoain, J., Liu, J., Mosberger, L.,

16 Nistor, M., Oechsner, H., Oliveira, J. V., Paterson, M., Pauss, A., Pommier, S., Porqueddu,

17 I., Raposo, F., Ribeiro, T., Rusch Pfund, F., Stromberg, S., Torrijos, M., van Eekert, M.,

18 van Lier, J., Wedwitschka, H., Wierinck, I., 2016. Towards a standardization of

19 biomethane potential tests. Water Sci. Technol. 74, 2515–2522.

20 https://doi.org/10.2166/wst.2016.336

21 Ibrahim, M.A.-R., 2014. Kinetics of anaerobic digestion of labaneh whey in a batch reactor.

22 African J. Biotechnol. 13, 1745–1755. https://doi.org/10.5897/AJB2013.13310

23 ISO, 1995. Evaluation of the ultimate anaerobic biodegradability of organic compounds in

24 digested sludge method by measurements of the biogas production. International Standard,

25 ISO/DIS 11734.

ACCEPTED MANUSCRIPT

21

1 Kafle, G.K., Chen, L., 2016. Comparison on batch anaerobic digestion of five different

2 livestock manures and prediction of biochemical methane potential (BMP) using different

3 statistical models. Waste Manag. 48, 492–502.

4 https://doi.org/10.1016/j.wasman.2015.10.021

5 Larsen, S.U., Hjort-Gregersen, K., Vazifehkhoran, A.H., Triolo, J.M., 2017. Co-ensiling of

6 straw with sugar beet leaves increases the methane yield from straw. Bioresour. Technol.

7 245, 106–115. https://doi.org/10.1016/j.biortech.2017.08.117

8 Llaneza Coalla, H., Blanco Fernández, J.M., Morís Morán, M. a., López Bobo, M.R., 2009.

9 Biogas generation apple pulp. Bioresour. Technol. 100, 3843–3847.


11 Mata-Alvarez, J., Dosta, J., Romero-Güiza, M.S., Fonoll, X., Peces, M., Astals, S., 2014. A

12 critical review on anaerobic co-digestion achievements between 2010 and 2013. Renew.

13 Sustain. Energy Rev. 36, 412–427. https://doi.org/10.1016/j.rser.2014.04.039

14 Møller, H.B., Sommer, S.G., Ahring, B.K., 2004. Methane productivity of manure, straw and

15 solid fractions of manure. Biomass and Bioenergy 26, 485–495.

16 https://doi.org/10.1016/j.biombioe.2003.08.008

17 Oslaj, M., Mursec, B., Vindis, P., 2010. Biogas production from maize hybrids. Biomass and

18 Bioenergy 34, 1538–1545. https://doi.org/10.1016/j.biombioe.2010.04.016

19 Pham, C.H., Triolo, J.M., Cu, T.T.T., Pedersen, L., Sommer, S.G., 2013. Validation and

20 Recommendation of Methods to Measure Biogas Production Potential of Animal Manure.

21 Asian-Australasian J. Anim. Sci. 26, 864–873. https://doi.org/10.5713/ajas.2012.12623

22 Raju, C.S., Ward, A.J., Nielsen, L., Møller, H.B., 2011. Comparison of near infra-red

23 spectroscopy, neutral detergent fibre assay and in-vitro organic matter digestibility assay

24 for rapid determination of the biochemical methane potential of meadow grasses.

ACCEPTED MANUSCRIPT

22


2 Raposo, F., De la Rubia, M.A., Fernández-Cegrí, V., Borja, R., 2012. Anaerobic digestion of

3 solid organic substrates in batch mode: An overview relating to methane yields and

4 experimental procedures. Renew. Sustain. Energy Rev. 16, 861–877.

5 https://doi.org/10.1016/j.rser.2011.09.008

6 Saidu, M., Yuzir, A., Salim, M.R., Salmiati, Azman, S., Abdullah, N., 2013. Influence of palm

7 oil mill effluent as inoculum on anaerobic digestion of cattle manure for biogas

8 production. Bioresour. Technol. 141, 174–6.


10 Strömberg, S., Nistor, M., Liu, J., 2015. Early prediction of Biochemical Methane Potential

11 through statistical and kinetic modelling of initial gas production. Bioresour. Technol. 176,

12 233–41. https://doi.org/10.1016/j.biortech.2014.11.033

13 Syaichurrozi, I., Budiyono, B., Sumardiono, S., 2015. Triple Batch Digesters in Series Method

14 to Analyze Biogas Potential from Bioethanol Vinasse. World Chem. Eng. J. 1.

15 Symons, G.E., Buswell, A.M., 1933. The methane fermentation of carbohydrates. J. Am. Chem.

16 Soc. 55, 2028–2036. https://doi.org/10.1021/ja01332a039

17 Thomsen, S.T., Spliid, H., Østergård, H., 2014. Statistical prediction of biomethane potentials

18 based on the composition of lignocellulosic biomass. Bioresour. Technol. 154, 80–86.


20 Triolo, J.M., Pedersen, L., Qu, H., Sommer, S.G., 2012. Biochemical methane potential and

21 anaerobic biodegradability of non-herbaceous and herbaceous phytomass in biogas

22 production. Bioresour. Technol. 125, 226–232.


24 Triolo, J.M., Sommer, S.G., Møller, H.B., Weisbjerg, M.R., Jiang, X.Y., 2011. A new algorithm

ACCEPTED MANUSCRIPT

23

1 to characterize biodegradability of biomass during anaerobic digestion: Influence of lignin

2 concentration on methane production potential. Bioresour. Technol. 102, 9395–9402.


4 Triolo, J.M., Ward, A.J., Pedersen, L., Løkke, M.M., Qu, H., Sommer, S.G., 2014. Near

5 Infrared Reflectance Spectroscopy (NIRS) for rapid determination of biochemical methane

6 potential of plant biomass. Appl. Energy 116, 52–57.

7 https://doi.org/10.1016/j.apenergy.2013.11.006

8 Triolo, J.M., Ward, A.J., Pedersen, L., Sommer, S.G., 2013. Characteristics of Animal Slurry as

9 a Key Biomass for Biogas Production in Denmark, in: Biomass Now - Sustainable Growth

10 and Use. InTech. https://doi.org/10.5772/54424

11 Tuesorn, S., Wongwilaiwalin, S., Champreda, V., Leethochawalit, M., Nopharatana, A.,

12 Techkarnjanaruk, S., Chaiprasert, P., 2013. Enhancement of biogas production from swine

13 manure by a lignocellulolytic microbial consortium. Bioresour. Technol. 144, 579–586.


15 Van Soest, P.J., 1963. Use of Detergents in the Analysis of Fibrous Feeds. II. A Rapid Method

16 for the Determination of Fiber and Lignin. J. Assoc. Off. Anal. Chem. 46, 829–835.

17 VDI 4630, 2006. Fermentation of organic materials: characterization of the substrate, sampling,

18 collection of material data, fermentation tests. In: Verein Deutscher Ingenieure (Ed.), VDI-

19 Handbuch Energietechnik, Beuth Verlag GmbH, 10772 Berlin, Germany.

20 Weissbach, F., Strubelt, C., 2008. Correction of Dry Matter Content in Sugar Beet Silages as a

21 Substrate for Biogas Production. Landtechnik 63, 354–356.

22 Wilson, D.B., 2011. Microbial diversity of cellulose hydrolysis. Curr. Opin. Microbiol. 14, 259–

23 263. https://doi.org/10.1016/j.mib.2011.04.004

24 Xu, F., Wang, Z.W., Li, Y., 2014. Predicting the methane yield of lignocellulosic biomass in

ACCEPTED MANUSCRIPT

24

1 mesophilic solid-state anaerobic digestion based on feedstock characteristics and process

2 parameters. Bioresour. Technol. 173, 168–176.


4 Yenigün, O., Demirel, B., 2013. Ammonia inhibition in anaerobic digestion: A review. Process

5 Biochem. 48, 901–911. https://doi.org/10.1016/j.procbio.2013.04.012

6

ACCEPTED MANUSCRIPT

25

Figure Captions

Figure 1. Specific CH4 rate and VFA concentration plotted versus time course for 4 AD reactors.

The average TAN for HRT of 15, 20. 30 and 45 days are 4.02, 3.88, 3.92 and 3.77 g kg ww-1

Figure 2. Degradation of crude lipids, crude proteins, hemicelluloses and cellulose. St: content of

organic compound residues; S0: content of initial organic compound

Figure 3: Remaining fraction of hemicellulose and cellulose in prediction versus measured one (Above)

and its simulation results of 200 days extended retention time.

Figure 4. model validation of the suggested BMP predicting model: Predicted versus measured BMP. The solid line indicates the best linear relationship (1:1). R2 of grass and crop residue including winter harvested sample 0.62; R2 of grass and crop residue excluding winter harvested samples 0.47; R2 of woody biomass 0.33; R2 of maize 0.00.

ACCEPTED MANUSCRIPT

26

Figure 1.

ACCEPTED MANUSCRIPT

27

Figure 2.

ACCEPTED MANUSCRIPT

28

Figure 3.

ACCEPTED MANUSCRIPT

29

Figure 4.

ACCEPTED MANUSCRIPT

30

Table 1. Concentration of major organic compounds, methane potentials measured (BMP), stoichiometry methane potential(TBMP), pig manure, processed meat waste, brewery wastewater, slaughterhouse waste and feed mixture.

Pig manure Processed meat waste

Brewery wastewater

Slaughterhouse waste

Feed mixture

DM (g kg ww-1) 39.55 (±0.96) 278.71 (±3.76) 7.74 (±0.97) 324.29 (±4.88) 70.0VS (g kg ww-1) 30.41 (±0.45) 239.08 (±4.57) 6.23 (±0.10) 314.20 (±5.02) 58.26TKN (g kg ww-1) 3.14 (±0.10) 19.09 (±0.61) 0.09 (±0.00) 7.66 (±0.41) 4.70TAN (g kg ww-1) 2.15 (±0.17) 0.00 (±0.00) 0.08 (±0.00) 1.66 (±0.12) 1.66VFAs (g kg ww-1) 7.12 (±0.85) 0.00 (±0.00) 0.29 (±0.09) 2.53 (±0.11) 5.44

(% of VS) 23.41 (±2.43) 0.00 (±0.00) 0.81 (±0.15) 4.65 (±0.79) 9.33Ethanol (g kg ww-1) 0.00 (±0.00) 0.00 (±0.00) 3.74 (±0.36) 0.00 (±0.00) 0.42

(% of VS) 0.00 (±0.00) 0.00 (±0.00) 60.03 (±1.75) 0.00 (±0.00) 0.72XP (g kg ww-1) 6.19 (±0.46) 120.00 (±2.35) 0.00 (±0.00) 37.45 (±1.13) 19.08

(% of VS) 20.35 (±0.55) 50.19 (±1.82) 0.00 (±0.00) 11.92 (±0.66) 32.74XL (g kg ww-1) 0.00 (±0.00) 100.00 (±2.15) 0.00 (±0.00) 230.00 (±4.34) 17.00

(% of VS) 0.00 (±0.00) 41.83 (±1.05) 0.00 (±0.00) 73.20 (±2.92) 29.18Hemicellulose (g kg ww-1) 4.72 (±0.54) 0.00 (±0.00) 0.00 (±0.00) 0.00 (±0.00) 3.54

(% of VS) 15.52 (±0.95) 0.00 (±0.00) 0.00 (±0.00) 0.00 (±0.00) 6.08Cellulose (g kg ww-1) 7.32 (±0.46) 0.00 (±0.00) 0.00 (±0.00) 0.00 (±0.00) 5.49

(% of VS) 24.07 (±1.96) 0.00 (±0.00) 0.00 (±0.00) 0.00 (±0.00) 9.42Lignin (g kg ww-1) 5.03 (±0.78) 0.00 (±0.00) 0.00 (±0.00) 0.00 (±0.00) 3.77

(% of VS) 16.54 (±1.10) 0.00 (±0.00) 0.00 (±0.00) 0.00 (±0.00) 6.48Carbohydrate (g kg ww-1) 0.03 19.08 2.20 44.22 3.52

(% of VS) 0.11 7.98 14.07 35.31 6.05

TBMP (NL CH4 kg VS–1) 473 706 602 862 635BMP (NL CH4 kg VS–1) 297±8 605±15 577±20 821±14 513TBMP (NL CH4 kg ww-1) 14.39 168.74 3.75 270.97 36.97SMP (NL CH4 kg ww-1) 9.03 (±0.26) 144.70 (±5.35) 3.60 (±0.35) 258.03 (±5.34) 29.91BMP/TBMP (%) 62.8 85.8 95.9 95.2 80.9

Figures in parentheses are standard deviations.

XP: Crude protein; XL: crude lipid;

SMP: Specific methane potential in terms of wet weight (NL CH4 kg ww-1)

NL: normal liter at 273 K, 1.013 bar

ACCEPTED MANUSCRIPT

31

Table 2. The concentration of each compound is indicated as TBMP in terms of wet weight (NL CH4 kg ww -1)

Substrates VFA Ethanol XP XL Hemicellulose Cellulose Lignin Carbohydrate

Pig manure 2.66 0.00 3.07 0.00 1.96 3.04 3.66 0.01

Processed meat 0.00 0.00 59.47 101.36 0.00 0.00 0.00 7.91

Brewery wastewater 0.11 2.73 0.00 0.00 0.00 0.00 0.00 0.91

Slaughterhouse waste 0.94 0.00 18.56 233.12 0.00 0.00 0.00 18.34

Feed for reactor 2.03 0.31 9.45 17.23 1.47 2.28 2.74 1.46

ACCEPTED MANUSCRIPT

32

1 Table 3. Hydrolysis rate, ultimate biodegradable and inert fraction of hemicellulose, cellulose and 2 lignocellulose from the first-order kinetic regression test and its model calibration results

ki df∞ (100-df∞) 𝑅𝑀𝑆𝐸𝐶 𝑟𝑅𝑀𝑆𝐸𝐶 R2

Hemicellulose(VSHE) 0.0298 59.3 40.7 1.46 4.60 0.97Cellulose (VSCE) 0.0699 36.3 63.7 0.48 1.65 0.99Lignocellulose 0.0602 27.4 72.6 0.77 3.68 0.99

3 dfi∞ is the ultimate biodegradable fraction (%); (100-dfi∞): inert fraction (%); ki: the decay rate constant for a

4 fraction (day-1); θ: HRT (days).

5 : Root Mean Square error in model calibration (NL CH4 kg VS-1)𝑅𝑀𝑆𝐸𝐶

6 : Relative Root Mean Square error in model calibration (%)𝑟𝑅𝑀𝑆𝐸𝐶

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

ACCEPTED MANUSCRIPT

33

1 Table 4. Model validation results of suggested BMP predicting model using the 4 datasets of samples used in

2 this study.

𝐵𝑀𝑃𝑚

(NL CH4 kg VS-1)𝐵𝑀𝑃𝑃

(NL CH4 kg VS-1) SDPig manure 305 297 5.5Processed meat 639 605 24.0Slaugterhouse waste 847 821 18.3

Tested substrates Brewery wastewater 602 577 17.8

(NL CH4 kg VS-1)𝑅𝑀𝑆𝐸𝑃 25.1 (%)𝑟𝑅𝑀𝑆𝐸𝑃 4.2

Statistical parameters

R2 0.993 : measured BMP (NL CH4 kg VS-1)𝐵𝑀𝑃𝑚

4 : predicted BMP (NL CH4 kg VS-1)𝐵𝑀𝑃𝑃

5 : Root Mean Square error in prediction (NL CH4 kg VS-1)𝑅𝑀𝑆𝐸𝑃

6 : Relative Root Mean Square error in prediction (%)𝑟𝑅𝑀𝑆𝐸𝑃

7 SD: Standard deviation between measured and predicted BMP

8

9

10

11

12

13

14

15

16

17

18

19

ACCEPTED MANUSCRIPT

34

1 Table 5. Model errors in prediction using 64 external dataset

Oslaj et al. (2010) Triolo et al. (2012)Biomass type Maize Grass and residues woody samplesna 10 31 27 23

(mean) (NL CH4 kg VS-1)𝐵𝑀𝑃𝑚 309.1 254.8 276.3 267.0 (min) (NL CH4 kg VS-1)𝐵𝑀𝑃𝑚 251.4 104.0 177.2 133.7 (max) (NL CH4 kg VS-1)𝐵𝑀𝑃𝑚 294.7 391.3 391.3 249.5

SDa (NL CH4 kg VS-1) 26.0 77.3 59.9 31.8Validation results

(NL CH4 kg VS-1)𝑅𝑀𝑆𝐸𝑃 37.4 49.9 43.7 86.1 (%)𝑟𝑅𝑀𝑆𝐸𝑃 12.1 19.6 15.8 47.8

2 a n : number of dataset used for validation.

3 b SD: Standard deviation of BMP of attended dataset for validation (n).

4

ACCEPTED MANUSCRIPT

Highlights

We explored a new approach to determine biochemical methane potentials. The model was validated using 65 internal and external datasets. 45% of hemicelluloses and 34% of cellulose decomposed despite long

HRT. The relative error for precision of the best model was 12.1%. The model has better performance for gently lignified biomasses

A novel mathematical modelling of waste biomass ...

Documents

Transcript of A novel mathematical modelling of waste biomass ...