Journal of Cleaner Production 279 (2021) 123644

Contents lists available at ScienceDirect

Journal of Cleaner Production

journal homepage: www.elsevier.com/locate/jclepro

Improving treatment capacity and process stability via a two-stage

anaerobic digestion of food waste combining solid-state acidogenesis
and leachate methanogenesis/recirculation
Lingkan Ding a, Yan Chen a, b, Yongping Xu b, c, Bo Hu a, *
a
Department of Bioproducts and Biosystems Engineering, University of Minnesota, 1390 Eckles Ave., St. Paul, MN, 55108, USA
b
School of Bioengineering, Dalian University of Technology, Dalian, 116024, China
c
Ministry of Education Center for Food Safety of Animal Origin, Dalian, 116620, China

a r t i c l e i n f o a b s t r a c t

Article history: Food waste (FW) accounts for the largest category of municipal solid wastes in the United States and the
Received 4 May 2020 proper disposal of FW is a unique challenge with urgent environmental concerns. Solid-state anaerobic
Received in revised form digestion (SSAD) is a viable option to treat FW but still hindered by poor mass transfer and severe in-
29 June 2020
hibition caused by the accumulation of acidic intermediates. To improve the treatment capacity and
Accepted 5 August 2020
Available online 13 August 2020
process stability using high-solid FW as the sole feedstock via AD, a bench-scale two-stage system
comprising of a first-stage solid-state digester (Stage 1; 2 L) and a second-stage liquid digester (Stage 2;
Handling Editor: Prof. Jiri Jaromir Klemes 1.4 L) was established and carried out in a semi-batch mode for 99 days. When one-stage SSAD started to
encounter acid accumulation and methanogenesis inhibition at a FW loading of 300 g (15% of working
Keywords: volume), the two-stage system secured robust and stable biogas production when the FW loading
Food waste continued rising to 400 g (20% of working volume), and an increase of 2.4 kJ/gVS (equivalent to 18.2% in
Two-stage system one-stage system) in energy yield was achieved. Stage 1 served as a hydrolysis/acidogenesis reactor with
Solid-state anaerobic digestion a total acids level of 28.5 g/L and a pH of 5.65. A large portion of lactic acid in total acids indicated shifts of
Acid accumulation
metabolic pathways. The transfer of acid-rich leachate from Stage 1 to Stage 2 ensured a high methane
Leachate recirculation
yield of 393.0 mL/gVS (90.2% of the biological methane potential), while the liquid effluent recirculation
provided buffering to Stage 1. Further increase of FW loading triggered a system instability. The enhanced
buffering capacity and increased free ammonia molecules, synergistically induced certain inhibitory
effects on FW acidogenesis in Stage 1. To conclude, the two-stage system outpaced the one-stage SSAD
with higher FW loading and energy yield.
© 2020 Elsevier Ltd. All rights reserved.

1. Introduction The typical treatment methods of FW include anaerobic diges-

tion (AD), incineration, landfill, and composting. AD exhibits the
Around 40% of the food in the United States (US) goes uneaten, highest primary energy production and leaves the least environ-
generating over 280 kg/year per capita of food waste (FW). Ac- mental footprints such as net-negative global warming potential
cording to the US Environmental Protection Agency (EPA), FW was due to the savings on greenhouse gas emissions (Slorach et al.,
the largest single category accounting for 15.1% of the total 2019), thus contributing to a cleaner bioenergy production out of
municipal solid wastes generated in the US in 2015 (EPA, 2018). FW FW. The nature of high solid content in FW makes it problematic for
not only makes adverse impacts on the food security and resource conventional liquid AD process that usually handles total solids (TS)
conservation, but also possesses a heavy burden on the environ- below 15%. The need to dilute FW requires not only water resources,
ment (FAO, 2011). Apart from the actions to reduce FW generation but also larger vessel volumes and higher associated parasitic en-
from the source, significant efforts need to be devoted to its ergy input. Compared with liquid AD, solid-state AD (SSAD) with TS
treatment and disposal (Morone et al., 2019). over 15% holds the superiorities such as smaller reactor volumes
and lower energy requirement for heating and stirring (Li et al.,
2011). Besides, the digestate from dry digesters requires less pro-
* Corresponding author. cessing or can even be directly applied as a biofertilizer or soil
E-mail address: [email protected] (B. Hu).

https://doi.org/10.1016/j.jclepro.2020.123644
0959-6526/© 2020 Elsevier Ltd. All rights reserved.
2 L. Ding et al. / Journal of Cleaner Production 279 (2021) 123644

amendment because of the lower moisture content. However, some performances of one-stage SSAD and two-stage system were
drawbacks of SSAD still limit its industrial application. Firstly, the compared in a semi-batch mode running for 99 days. The im-
poor mass transfer inside dry digesters adversely affects the pro- provements on process stability of this two-stage system were
cess efficiency. The retention time of SSAD needs to be up to three quantified by pushing FW loading into a threshold and character-
times longer than that of liquid AD (Li et al., 2011), thus whittling izing the gaseous and liquid products.
away its advantage of larger treatment capacity. Secondly, the high
degradability of FW can induce severe inhibition on SSAD. The fast 2. Materials and methods
degradation of the components such as carbohydrates and proteins
in FW usually results in the accumulation of volatile fatty acids 2.1. Feedstock and inocula
(VFAs), thus reducing the pH drastically. The pH deviated from the
optimum range for methanogens leads to slower methanogenesis. The FW recipe was based on the annual average of food received
If VFA accumulation occurs in a vast area or the VFA amount ex- by the non-profit food bank Second Harvest Heartland in Minne-
ceeds the buffering capacity, a failure of entire SSAD process is sota. Ten typical foods which are commonly consumed by local
expected (Martin et al., 2003). To tackle these issues associated people were selected as the representatives and the detailed
with SSAD of FW, two countermeasures, including co-digestion and characteristics are shown in Table 1. The foods were manually
two-stage AD, have been studied. Co-digestion typically needs the chopped, ground into pulp, and then stored at
20  C before
addition of bulking agents, such as corn stover (Xu and Li, 2012) and experimentation.
yard waste (Panigrahi et al., 2020), but these agents may not be The inocula, including solid digestate and liquid leachate, were
readily available logistically to some specific institutions such as collected from an industrial dry digester at the University of Wis-
food banks. Two-stage AD processes treating FW have been widely consin Oshkosh fed with a mixture of about 50% FW and 50% other
explored to improve the performances. stuff (yard trimmings, animal bedding wastes, etc.). The solid
By separating AD into two stages undertaking different func- digestate and liquid leachate were firstly mixed, placed at 37 ± 1  C
tions, hydrolysis/acidogenesis of FW in the first stage and aceto- for 7 days to degas, and then used as the inoculum for the biological
genesis/methanogenesis in the second stage could be optimized at methane potential (BMP) tests and the solid digesters in semi-batch
different pH ranges favoring different microbial communities. systems. The TS and volatile solids (VS) of mixed inoculum were
However, most two-stage AD systems treating FW focused on 16.0 wt% and 10.2 wt%, respectively. The liquid leachate degassed at
liquid AD (Srisowmeya et al., 2020), even on a pilot scale recently 37 ± 1  C for 7 days (TS: 2.3 wt%; VS: 0.8 wt%) was employed as the
(Micolucci et al., 2018), whilst investigations involving SSAD are inoculum for the liquid digester in the semi-batch system.
still limited. Browne and Murphy (2014) established a two-stage AD
system comprising of sequential leach bed reactors (LBRs) as the 2.2. Biological methane potential assay
first stage and an upflow anaerobic sludge blanket (UASB) reactor
as the second. FW (TS>30%) was loaded into the LBRs and sprayed The BMP tests on FW were performed in triplicate using 500-mL
by the recirculated effluent from the UASB. The leachate with dis- polyethylene bottles. Three experimental groups at different inoc-
solved/hydrolyzed FW was then transferred to the UASB serving as ulum to feedstock (I/F) VS ratios (i.e., 1:1, 2:1, and 3:1) were set. The
the main methane production reactor and a buffering tank for LBRs. total amount of FW and inoculum was calculated as 250 g in each
Additionally, more strategies, including micro-aeration of LBR (Xu bottle. A group filled with inoculum and no FW was set as a blank
et al., 2014b), reutilization of LBR off-gas in UASB (Yan et al., control. All the filled bottles were purged with N2 for 2 min to
2016), and replacement of leachate with water in recirculation maintain anaerobic and then placed in a warm chamber at 37 ± 1  C
(Xu et al., 2014a), were investigated to improve the FW reduction for AD experiments until no obvious biogas production was
and biogas production using this LBR-UASB system. However, the observed. Due to the high TS in the bottles, the original mechanical
extra requirements in these studies, such as the addition of ancil- stirrers for liquid AD were inappropriate and not employed in the
lary instruments and the dosing of buffering chemicals or water, BMP tests. Alternatively, all the bottles were manually shaken twice
still undermined the superiorities of this two-stage system over
one-stage SSAD, thus impairing its practical applicability. A
different fashion of two-stage SSAD system was established by Li Table 1
Characteristics of raw food waste.
et al. (2018). The first stage was SSAD of mixed FW and chicken
manure, and the second stage was SSAD of the mixture of grass and Parameter Food waste
the digestate from the first stage. By diluting the VFAs from first Category (wet weight%)
stage using new feedstock in the second stage, the methane yield Apple 10
was proved to be higher than that in the one-stage co-digestion of Banana 1
Bread 7
the three substrates. However, the lag phase of the two-stage sys-
Beef 8
tem was still long: no significant increases in methane yields were Cabbage 15
observed during the first 20 days in the batch trial. Cereal 7
To date, there are few studies on the two-stage process incor- Cheese 2
porating SSAD of sole FW feedstock without compromising its Milk 16
Onion 3
practicality. To narrow the knowledge gap in the state of the art, the Potato 31
objective of this study is to investigate the feasibility of a two-stage Proximate analysis
system digesting high-solid FW as the sole feedstock in a long-term TS (wet weight%) 24.9 ± 0.3
run without addition of extra ancillary instruments, dosing of VS (wet weight%) 23.7 ± 0.2
VS/TS (%) 95.2 ± 0.3
buffering chemicals, or mixing with water or bulking agents, thus
Ultimate analysis
providing a more environmentally and economically viable solu- Carbon (VS%) 48.1 ± 0.5
tion for continuous FW-generating organizations such as non-profit Hydrogen (VS%) 7.5 ± 0.1
food banks with limited access to bulking agents as co-substrates. A Nitrogen (VS%) 4.2 ± 0.1
two-stage system comprising of a first-stage solid-state digester Oxygen (VS%) 40.2 ± 0.4
C/N ratio 11.5 ± 0.3
and a second-stage liquid digester was established. The
 L. Ding et al. / Journal of Cleaner Production 279 (2021) 123644 3

every day to increase the mass transfer between FW and inoculum. (CE440, Exeter Analytical, UK), and oxygen was assumed as the
A 1-L tedlar gas bag (SKC, USA) connected to each bottle was used to remaining of VS. The soluble chemical oxygen demand (sCOD), total
collect the biogas. The biogas composition and volume were then ammonia nitrogen (TAN), and total alkalinity of liquid effluents
measured at each time interval. were tested using Hach TNT kits (Hach, USA) and evaluated on a
spectrophotometer (Hach DR5000, USA). Biogas composition (H2,
CO2, and CH4) was determined using a micro gas chromatograph
2.3. Set-up and operation of semi-batch reactors
(Micro-GC, CP-4900, Varian, USA) system equipped with two col-
umns of molecular sieve 5 A and Porapak Q and a thermal con-
Fig. 1 shows the schematic of the semi-batch systems. Two 4-L
ductivity detector. Lactic acid (HLa), acetic acid (HAc), propionic
polypropylene reactors were used as the garage-type solid di-
acid (HPr), iso-butyric acid (Iso-HBu), butyric acid (HBu), and iso-
gesters in both one-stage and two-stage systems, and 150-mm
valeric acid (Iso-HVa) were determined using a high performance
stainless steel mesh screens were used for percolation. A 2-L
liquid chromatography (HPLC, Infinity 1260, Agilent, USA) equip-
polypropylene reactor served as the liquid digester (Stage 2) of
ped with a refractive index detector and an HPX-87H column (Bio-
the two-stage system. The digesters were placed in a warm
Rad, USA).
chamber kept at 37 ± 1  C throughout the experiment. As compared
to continuous solid digesters such as Dranco, Kompogas, and Val-
2.5. Calculations
orga types (Kothari et al., 2014), the garage-type digesters are
technically simple and run without auxiliary instruments for stir-
The theoretical methane yield of FW calculated using the
ring and mixing. Nonetheless, they require big land footprints to
elemental composition as shown in Eq. (1) (Voelklein et al., 2016)
run several batch digesters with different on/off time to mitigate
was 494.0 mL/gVS.
biogas production pulses and reduce storage areas and time for
substrates (Kothari et al., 2014). In this study, the semi-batch
b c 3 a b c 3
feeding strategy shown in Table 2 was employed for the one- Ca Hb Oc Nd þ ða

þ dÞH2 O/ð þ

dÞCH4
4 2 4 2 8 4 8
stage solid digester and Stage 1 of the two-stage system. As
a b c 3
compared to the continuous/semi-continuous feeding strategy, the þ ð
þ þ dÞCO2 þ dNH3 (1)
maintenance requirements such as continuous pumping of highly 2 8 4 8
viscous FW were minimized. As compared to multiple batch re- Acidification yield was defined as the ratio of COD of total acids
actors in line, the semi-batch mode requires less reactor numbers to sCOD of the liquid effluent shown in Eq. (2).
by shortening the feeding time interval. Because of these features,
the organic loading rate (OLR), which is widely adopted for COD of total acids
Acidification yield ¼  100% (2)
assessing continuous/semi-continuous AD systems, was not sCOD
employed to evaluate the garage-type solid digesters in this study.
where total acids stand for the total concentration of acids,
Alternatively, each FW loading was expressed as amount (% of the
including HLa, HAc, HPr, Iso-HBu, HBu, and Iso-HVa.
digester working volume) as shown in Table 2. In the two-stage
Free ammonia nitrogen (FAN) was calculated based on TAN,
system, the transfer of leachate and recirculation of liquid
temperature, and pH according to Eq. (3) (Rajagopal et al., 2013).
effluent between Stage 1 and Stage 2 were synchronized by
adjusting the peristaltic pumps (Cole-Parmer, USA). No water
10
pH
dilution or leachate/effluent discharge was conducted during the FAN ¼ TAN  ð1 þ Þ
1 (3)
10
ð0:09018þ T
2729:92
recirculation, and the liquid samples of leachate and effluent (2 mL
each) were taken for analyses every 1e2 days.
where T stands for the temperature (Kelvin).
The significances of differences between BMP yield, biogas
2.4. Analytical methods production rate (BPR), methane yield, methane content, and energy
yield means were examined through the analysis of variance
TS and VS of FW samples and inocula were determined using the (ANOVA) on Origin 9.0 using Tukey tests (P < 0.05).
Standard Methods 2540 G (APHA, 1999). Carbon, hydrogen, and
nitrogen contents of FW were determined on an elemental analyzer 3. Results and discussion

3.1. BMP performances of FW

Fig. 2 shows the BMP results of FW at I/F ratios of 1:1, 2:1, and
3:1. At the lowest I/F ratio of 1:1, SSAD of FW was totally inhibited.
The highest FW amount led to fast acidification accompanied by the
build-up of VFAs and sharp drop of pH, resulting in the thorough
inhibition of methanogens. Similarly, the lag phase caused by the
fast acidification of FW still lasted for over 30 days when the I/F
ratio increased to 2:1. After that, the methanogens in a higher
inoculum volume got acclimatized to the environment, and the
methane yield reached 321.0 mL/gVS, equivalent to 65.0% of the
theoretical value. When the I/F ratio further increased to 3:1, the
strong inhibitory effect was mitigated, the lag phase was greatly
shortened from about 35 days to 3 days, and the highest methane
yield of 435.9 mL/gVS (88.2% of the theoretical value) was recorded.
With the high I/F ratio of 3:1, the intense accumulation of VFAs and
corresponding sharp pH drop were avoided because of increased
Fig. 1. Schematic of semi-batch one-stage and two-stage systems. microbes and alkalinity of the inoculum with a higher volume.
4 L. Ding et al. / Journal of Cleaner Production 279 (2021) 123644

Table 2
Operation of semi-batch reactors.

Feeding time One-stage reactor Two-stage system

Stage 1 Stage 2

Day 0 In 250 g FW (12.5% of working volume), 1750 g 250 g FW (12.5% of working volume), 1750 g Not connected until Day 13.
inoculum inoculum
Out / /
Day 5 In 300 g FW (15% of working volume), 200 g 300 g FW (15% of working volume), 200 g
inoculum inoculum
Out 500 g solid digestate 500 g solid digestate
Day 10 In 350 g FW (17.5% of working volume), 150 g 350 g FW (17.5% of working volume), 150 g
inoculum inoculum
Out 500 g solid digestate 500 g solid digestate
Day 15 and Day In No feeding after Day 10 350 g FW (17.5% of working volume), 150 g On Days 13e19, 200 mL of liquid leachate from Stage
29 inoculum 1 was transferred to Stage 2, while 200 mL of liquid
Out 500 g solid digestate effluent
Day 43 and Day In 400 g FW (20% of working volume), 100 g from Stage 2 was recirculated to Stage 1 daily.
57 inoculum (HRT ¼ 7 d)
Out 500 g solid digestate After Day 19, the time interval of this liquid effluent
Day 71 and Day In 450 g FW (22.5% of working volume), 50 g exchange
85 inoculum manner was extended to two days. (HRT ¼ 14 d)
Out 500 g solid digestate

Although different optimum I/F ratios were obtained, the trend of hydrogen (8.4%). Thereafter, methanogenesis dominated the pro-
increasing methane production with increasing I/F ratios was cess: the methane content rose to 65.4% along with the increasing
verified in a previous study (Panigrahi et al., 2020). Meanwhile, the specific methane yield (SMY) on Day 3. With the depletion of FW,
high organic contents and poor mass transfer in the bottles the BPR and SMY sharply dropped from Day 3 to Day 5. A similar
required a much longer BMP duration (77 days) than conventional cycle was observed during Days 5e10 with a loading of 300 g FW.
liquid AD of FW (around 30 days) (Zhang et al., 2014) to secure However, after the FW loading increased to 350 g on Day 10, the
enough acclimatization time for the microbial communities. The rebound of biogas production was limited and SSAD inhibition was
further increase of I/F ratio (such as 4:1 and 5:1) may possibly recorded: BPR remained around 1 mL/gVS/d, and carbon dioxide
further shorten the lag phase, but conversely impair the FW was the predominant biogas component. Thereafter, no more FW
treatment capacity. Therefore, the I/F ratio of 3:1 was chosen as the
optimum for starting subsequent semi-batch experiments.

3.2. Semi-batch SSAD of FW

3.2.1. One-stage semi-batch SSAD of FW

The biogas production in the one-stage semi-batch SSAD of FW
is shown in Fig. 3. After the first loading of 250 g FW, biogas pro-
duction was successfully triggered at an initial I/F ratio of 3:1. On
Day 1, the fast hydrolysis and acidogenesis of FW contributed to the
generation of a large amount of carbon dioxide (77.0%) and

Fig. 2. Biogas production through batch BMP test of FW. Different letters (abc) indicate Fig. 3. Biogas production rate, specific methane yield, and biogas content in semi-
significant differences between methane yield means (P < 0.05). batch one-stage reactor (grey bars indicate FW loading points).
 L. Ding et al. / Journal of Cleaner Production 279 (2021) 123644 5

was loaded due to the thorough inhibition or even failure of

methanogenesis in the one-stage system.
The profiles of acids in the liquid effluent of the one-stage
reactor are present in Fig. 4. During the first 5 days, no accumula-
tion of acids was observed with the robust biogas production. After
a higher loading of 300 g FW on Day 5, certain amounts of acids
were accumulated during Days 8e9 but then consumed on Day 10.
After the highest loading of 350 g FW on Day 10, significantly higher
acids were measured. On Day 14, total acids sharply rose to 29.8 g/L
and the corresponding acidification yield was 68.4%. Thereafter, the
acid concentration fluctuated around 20e40 g/L and the acidifica-
tion yield remained high as around 70%. As shown in Fig. 5, the pH
sharply decreased after Day 10 and remained constantly around 6
thereafter, which is beneficial for FW hydrolysis and acidification
but detrimental to methanogenesis (Zhang et al., 2014).
In the composition of total acids, HAc and HBu predominantly
comprised of over 70%, while HPr occupied over 20%. Although HPr
is one of the major metabolites via anaerobic fermentation of
amino acids derived from proteins (Cheng et al., 2015) which are
main FW components, HPr ratio was still higher than that in pre- Fig. 5. Variations of pH values of semi-batch reactors during anaerobic digestion (grey
vious studies on the liquid AD of hydrothermally pretreated FW bars indicate FW loading points of Stage 1).
(Ding et al., 2017) or raw FW (Voelklein et al., 2016). The elevated
HPr level implied the instability or even failure of digestion (Gallert
and Winter 2008). Nonetheless, the predominance of HAc and HBu
and little HLa implied no big shifts from acetic and butyric path-
ways to lactic pathway. These indicators, including the high acid
concentration, the stable acidification yield, and the steady acid
composition, evidently demonstrated the SSAD inhibition.
Fig. 6 shows the variation trend of TAN concentration in the
liquid effluent. As the inoculum originated from a dry digester
mainly treating mixed FW, the degradation of protein-rich com-
ponents contributed to a higher initial TAN (2771 mg/L) than many
other seed inocula in literature (Yenigün and Demirel, 2013). As
such, the inoculum had acclimatized to a relatively high TAN con-
centration. The slightly increasing TAN was recorded during the
first 10 days due to the increasing FW feeding, but biogas produc-
tion was still robust, indicating that a TAN up to 3000 mg/L was not
inhibitory to AD in the one-stage reactor. Moreover, the stable TAN
level (3039e3330 mg/L) during Days 10e15 remained constant
compared with that during Days 5e10, implying its limited rele-
vance to the methanogenesis inhibition. Therefore, the accumula-
tion of acids and sharp pH drop associated with the overfeeding of
FW were to blame for the digester failure. Thereafter, although the Fig. 6. Total ammonia nitrogen variations of semi-batch reactors during anaerobic
digestion (grey bars indicate FW loading points of Stage 1).

methanogenesis was inhibited, the continual hydrolysis and

acidogenesis of proteins in the remaining FW elevated the TAN
concentration to over 6000 mg/L. This high TAN level would
probably induce extra inhibitory effects on methanogenesis in the
one-stage reactor (Rajagopal et al., 2013).

3.2.2. Two-stage semi-batch SSAD of FW with increasing loading

Fig. 7 shows the biogas production in the two-stage system.
During the first 13 days when Stage 2 had not been established, the
feeding strategy of Stage 1 was identical to that of the one-stage
reactor, leading to similar biogas production trends. On Day 15,
350 g FW was added to Stage 1, and an increase of biogas pro-
duction ensued. However, the BPR (66.3 mL/gVS/d) and SMY (182.4
mL/gVS) were markedly lower than that on Days 1, 6, and 11. During
Days 16e29, the BPR remained below 8 mL/gVS/d and carbon di-
oxide dominated the biogas composition, clearly demonstrating
methanogenesis inhibition. During Days 29e57, the biogas pro-
Fig. 4. Profile of acids and acidification yield of one-stage reactor (grey bars indicate duction performance remained stably low. However, the perfor-
FW loading points). mance during Days 57e71 differed: the BPR, SMY, and methane
6 L. Ding et al. / Journal of Cleaner Production 279 (2021) 123644

Fig. 7. Biogas production rate, specific methane yield, and biogas content in semi-batch two-stage system: (a) Stage 1 and (b) Stage 2 (grey bars indicate FW loading points of Stage
1).

content simultaneously increased. This slight rebound of biogas FW into Stage 1. During Days 71e85 with the highest loading of
production was suppressed after a higher loading of 450 g FW on 450 g FW, the BPR and SMY did not rebound as high as that during
Day 71, whereas a stronger rebound ensued thereafter. The SMY Days 57e71. This phenomenon along with the drastic biogas pro-
and methane content reached 404.6 mL/gVS and 79.0%, respec- duction fluctuation in Stage 1 in the decaying phase implied the
tively, illustrating the potential mitigation of inhibition and re- impending system instability. The decreases became worse after
covery of methanogenesis in Stage 1. During Days 85e99 after the the feeding on Day 85: the BPR and SMY averaged 16.9 mL/gVS/
second loading of 450 g FW, the similar “suppression-rebound” d and 186 mL/gVS, respectively, which were low as close to that
trend in biogas production continued. Therefore, the 99-day SSAD during the transition phase.
process in Stage 1 was separated into four periods: (1) starting The pH variations (Fig. 5) and acid profiles (Fig. 8) can explain
phase from Day 0 to Day 13 when Stage 1 functioned as a one-stage the biogas fluctuation. A high total acid concentration of 23.8 g/L in
solid-stage digester; (2) transition phase from Day 13 to Day 29 Stage 1 was recorded on Day 13, but the connection of Stage 2 and
with the methanogenesis inhibition in Stage 1 and the connection subsequent acid-rich leachate transfer led to sharp drops of total
between Stage 2 and Stage 1; (3) stable phase from Day 29 to Day acids and acidification yield, whereas the introduction of liquid
65 when Stage 1 served as a steady hydrolysis and acidogenesis effluent with a pH over 8 from Stage 2 resulted in the direct pH
reactor; and (4) decaying phase from Day 65 to Day 99 when increase in Stage 1. A similar fashion during Days 15e29 was
methanogenesis reoccurred in Stage 1 with drastic biogas observed. The acid accumulation altered the pH in Stage 1, whilst
fluctuation. the pH reinforced the acid production in turn: the range of pH 5e6
Stage 2 was connected to Stage 1 on Day 13, and the biogas was favorable for FW hydrolysis and acidogenesis (Ye et al., 2018).
production performance fluctuation of Stage 2 was closely associ- Meanwhile, the acid-rich leachate from Stage 1 and a certain level
ated to Stage 1 thereafter. During Days 13e19 at a hydraulic of acid accumulation up to 10 g/L led to gradual pH decrease in
retention time (HRT) of 7 days, the anaerobes in Stage 2 were still Stage 2 during Days 13e23. The high total acids to total alkalinity
acclimatizing to the daily loading of 200 mL acid-rich leachate from ratio up to 0.79 on Day 25 also implied that the buffering capacity of
Stage 1, resulting in the limited increases in BPR and SMY and the Stage 2 was challenged. This trend was curbed due to the extension
decrease of methane content. On Day 19, the HRT of Stage 2 was of the HRT from 7 days to 14 days and the acclimatization of mi-
extended to 14 days to relieve the stress. Thereafter, the biogas crobial community to the acid-rich leachate.
production gradually bounced, and the methane content stabilized As the two-stage system stepped into the stable phase from Day
at around 70e80%. Along with the stable phase in Stage 1 from Day 29, the pH of Stage 1 fluctuated between around 5 after FW feeding
29 to Day 57, the BPR and SMY of Stage 2 remained relatively and around 6 before the next until Day 65. Correspondingly, a trend
constant at around 35 mL/gVS/d and 350 mL/gVS, respectively. A of “ascent-plateau-descent” of total acids in Stage 1 between
stronger biogas production performance of Stage 2 affirmed by the feedings could be found in Fig. 8a: (1) the fast hydrolysis and
higher BPR and SMY was observed after the second loading of 400 g acidogenesis of newly fed FW caused fast increase of total acids; (2)
 L. Ding et al. / Journal of Cleaner Production 279 (2021) 123644 7

Fig. 8. Profiles of acids, acidification yield, and total acids to total alkalinity ratio of two-stage system: (a) Stage 1 and (b) Stage 2 (grey bars indicate FW loading points of Stage 1).

the continuous transfer of acid-rich leachate from Stage 1 to Stage 2 fluctuation of total acids and higher upper value of pH in Stage 1
and the recirculation of liquid effluent with low acid concentrations after entering the decaying phase. After the feeding of 450 g FW on
from Stage 2 to Stage 1 increased the buffering capacity and Day 71, the concentration of total acids only increased to
maintained the stable and efficient hydrolysis and acidogenesis 27.6e28.4 g/L, much lower than that before. A drastic drop of total
processes, leading to a stable level of total acids as high as 35e40 g/ acids by 20.5 g/L was recorded on Day 79, only 8 days after the
L in Stage 1; and (3) with the depletion of undegraded FW and the latest feeding. By contrast, such big drops of total acids were
transfer of washed acids, the total acids in Stage 1 gradually observed 12 days after the latest feeding in previous cycles. A faster
decreased to below 10 g/L prior to next FW feeding. During the and stronger pH increase was also recorded. The pH over 7 became
stable phase, the accumulated acids and the total acids to total unfavorable for FW acidogenesis in Stage 1. The lower total acids in
alkalinity ratio in Stage 2 also gradually decreased and then the leachate from Stage 1 led to lower acid accumulation in Stage 2
remained at a low level along with stable biogas production. with stable biogas production during Days 71e85. After the final
Nonetheless, this kind of balance was disturbed with bigger feeding of 450 g FW on Day 85, the acid concentration only bounced
8 L. Ding et al. / Journal of Cleaner Production 279 (2021) 123644

to around 10 g/L. The acidification yield also remained low as 13 when obvious methanogenesis inhibition was observed and the
around 30% and the BPR and SMY decreased to low levels. connection of Stage 2 to Stage 1 was established. After 89 days, the
Combining the low total acids with the limited biogas production, it accumulative methane yield (100.4 mL/gVS) of the one-stage
could be inferred that the acidogenesis process of Stage 1 was reactor was only equivalent to 23.0% of BMP. By comparison, after
partially inhibited. Moreover, the increased fluctuation of total the connection of Stage 2 to Stage 1, the methane yield combined in
acids to total alkalinity ratio also suggested possible methano- the two-stage system at 350 g FW loading was 298.4 mL/gVS,
genesis instability in Stage 2. significantly surpassing the one-stage reactor. When the FW
Beyond the variation of total acid concentration, the acid profile loading increased to 400 g, the methane yield combined reached
of Stage 1 in the two-stage system greatly differed from that of the 393.0 mL/gVS (equivalent to 90.2% of BMP), to which Stage 2
one-stage reactor. HLa was barely detected in the one-stage reactor, contributed 94.3%. The energy yield (15.6 kJ/gVS) was 18.2% higher
whilst it became one of the major acids in addition to HAc and HBu than that in the one-stage reactor at 300 g FW loading. The sCOD
in Stage 1 after Day 15. The highest HLa concentration of 16.3 g/L and total acids levels of Stage 1 almost remained constant, and the
(38.3% of total acids) was recorded on Day 37. A low pH to 5 was highest sCOD removal (70.3%) of Stage 2 was recorded at 400 g FW
proven to facilitate HLa production during FW acidogenesis loading. When the FW loading increased to 450 g, although a higher
(Voelklein et al., 2016). In this study, the HLa concentration closely methane yield of the two-stage system was recorded, the system
coincided with the pH variation: HLa accumulated when the pH instability had been triggered and a downtrend of biogas produc-
was close to 5, but gradually decreased along with the leachate tion was observed (Fig. 7).
transfer and pH increase in the transition phase. Similar patterns Beyond the higher FW treatment capacity and biogas produc-
were recorded between two feedings during subsequent stable tion, this two-stage system holds additional benefits. The methane
phase. Homo-fermentative bacteria (e.g., Lactobacilli, Enterococci, content in the biogas produced in Stage 2 was higher since much
etc.) and hetero-fermentative bacteria (e.g., Leuconostocs, Oeno- carbon dioxide was released via FW hydrolysis and acidogenesis in
cocci, Weissella, etc.) in the microbial communities could account Stage 1. The higher methane content increased the energy content
for the simultaneous HLa and HAc production (Xu et al., 2014a). of biogas from Stage 2, and the lower carbon dioxide content could
Meanwhile, Stage 2 fed with the HLa-rich leachate still exhibited cut down the energy and capital input in subsequent biogas
robust biogas production. The shift from acetic and butyric path- upgrading, thus benefiting a cleaner bioenergy production. More-
ways to lactic pathway during FW acidogenesis in Stage 1 did not over, the stage separation could provide better controllability of the
affect the methanogenesis in Stage 2 when the total acids remained system: Stage 1 could function with a wider pH range suitable for
at reasonable levels (below 5 g/L during the stable phase). This FW hydrolysis and acidogenesis; and stability of Stage 2 could be
finding is in accordance with a previous study: the silage effluents controlled by adjusting the leachate transfer amount/frequency.
from ensiled seaweed biomass which contained high HLa were Concerning about the BPR stability, the two-stage system could
used as feedstocks for AD, and methane yields comparable to or offer an alternative to a set of parallel large batch garage-type di-
even higher than that from fresh seaweeds were obtained gesters running at different loading points that mitigate the
(Herrmann et al., 2015). After entering the decaying phase, the pH instability of BPRs. Additionally, if there is still a need to increase
deviated from the optimum range favoring HLa production, thus FW treatment capacity, more Stage 1 reactors could be connected to
reducing the HLa concentration. one large Stage 2 reactor which not only functions as the main
Fig. 6 presents the TAN variations. In general, the TAN gradually biogas production reactor but also serves buffering agents to all
increased in both reactors along with the increasing FW loading. Stage 1 reactors.
The TAN slightly fluctuated around 4000 mg/L in both reactors Table 4 summarizes the performance comparisons of SSAD of
during the stable phase, indicating that both acidogens and sole FW feedstock between previous research and the present
methanogens were adaptive to such high TAN levels without in- study. The FW origins and characteristics differed in these studies,
hibition. After entering the decaying phase, the TAN increased to leading to different methane potentials. Different digester config-
around 4500e5000 mg/L. Higher ammonium ions strengthened urations (e.g., batch/continuous, one/two/three-stage, etc.) also
the buffering capacity, leading to the gradual pH increase in Stage 1. affected the biogas production from FW. However, the common
Additionally, higher pH value also facilitated the shift of ammo- issue was that the high VS in solid digesters could easily result in
nium to free ammonia molecules, which were suggested to be VFA accumulation and methanogenesis inhibition because of the
much more toxic to microorganisms due to its high cell- fast hydrolysis and acidogenesis. Therefore, strategies including
membrane-permeability causing intracellular proton imbalance maintaining a long HRT of 40e60 d (Cho et al., 2013) and shifting
and potassium deficiency (Rajagopal et al., 2013). As calculated, the mesophilic to thermophilic SSAD (Nguyen et al., 2017) were pro-
higher TAN and pH significantly increased the FAN level from below posed to counteract this problem in the one-stage system. Mean-
1 mg/L on Day 49 to 475 mg/L on Day 99, thus inducing stronger while, different strategies were employed to improve the two-stage
toxicity to the acidogens in Stage 1. Similarly, the increased TAN also LBR-UASB performance, such as intermittent micro-aeration to LBR
contributed to higher FAN in Stage 2 when pH and temperature using an air pump (Xu et al., 2014b), reutilization of LBR off-gas in
remained constant. Although the threshold values of TAN that start UASB (Yan et al., 2016), adjustment of leachate recirculation ratio by
to inhibit AD greatly varied in literature (Rajagopal et al., 2013), the partial water replacement (Luo and Wong, 2019), and change of
increased toxicity of FAN to methanogens might explain the leachate recirculation frequency (Xu et al., 2014a). Although the
instability of methane production in Stage 2. hydrolysis and acidogenesis of FW were improved and the methane
yields were increased by 15e38.6%, the extra equipment require-
3.3. Performance comparison between one-stage reactor and two- ment, water dilution, and pH control using buffering chemicals
stage system whittled the original advantages of SSAD, thus undermining the
practicability of this two-stage system. Zhang et al. (2017) estab-
Table 3 summarizes the AD performances of one-stage reactor lished a semi-continuous three-stage SSAD system in one reactor,
and two-stage system. During the first 13 days, the same experi- improving the VS reduction to 83.5% and obtaining a 24e54%
mental set-up of the two solid digesters led to similar perfor- higher methane yield. However, similar disadvantages included the
mances. The methane yields of 310.2e331.4 mL/gVS were buffering chemical dosing and high parasitic energy for fast stirring
71.2e76.0% of the BMP. The performances started to differ on Day of all three stages. By contrast, aiming at the similar goal of
 L. Ding et al. / Journal of Cleaner Production 279 (2021) 123644 9

Table 3
Comparison of anaerobic digestion performances between one-stage reactor and two-stage system (mean values).

One-stage reactor Two-stage system

Stage 1 Stage 2
X Y
FW loading (g) 250 300 350 250 300 350 400 450 250 300 350 400 450
Biogas
b b a b b a a a a a
Biogas production rate (mL/gVS/d) 123.5 123.9 6.6 135.6 127.5 22.1 10.9 20.5 / / 24.0 35.5 25.0a
Specific methane yield (mL/gVS) 321.7bc 331.4bc 100.4ab 310.2bc 331.2bc 58.3a 22.4a 129.1ab / / 240.1b 370.6c 274.9b
Methane content (%) 55.0bc 58.6bcd 16.9a 52.7bc 58.0bcd 9.9a 9.5a 39.7b / / 60.7c 74.6cd 78.9d
Energy yieldZ (kJ/gVS) 12.8bc 13.2bc 4.0ab 12.3bc 13.2bc 2.3a 0.9a 5.1ab 9.6b 14.7c 10.9b
Liquid effluent
pH 7.12 7.56 6.12 7.22 7.56 5.61 5.65 6.70 / / 7.74 8.06 8.16
Total acids (g/L) 0.0 1.3 29.9 0.2 0.0 25.9 28.5 12.8 / / 5.3 1.2 0.8
Total alkalinity (mg CaCO3/L) / / / / / / / / / / 13933 15330 17188
sCOD (mg/L) 26916 27504 59401 26487 33846 60237 61121 47400 / / 28745 18612 14854
TAN (mg/L) 2871 3097 4366 2846 2886 3496 3926 4610 / / 3565 4000 4463

X: The retention time after 350 g FW loading to the one-stage reactor lasted for 89 days in total.
Y: This first feeding of 350 g FW took place with a retention time of 5 days, while the following two with 14 days each.
Z: Energy yield was calculated based on the methane heating value of 39.8 MJ/m3.
Note: Different letters (abcd) indicate significant differences between biogas production rate, specific methane yield, methane content, and energy yield means (P < 0.05).

Table 4
Comparison of SSAD of sole FW feedstock using different strategies in literature and this study.

FW origin SSAD type Pro Con Reference

FW from a canteen Semi-continuous one-stage A stable volumetric methane production rate of Lower HRT (20 d) triggered SSAD Cho et al. (2013)
(TS: 21.1 wt%; VS: mesophilic SSAD 2.51 m3/m3/d was maintained at the methane instability and continuous
17.4 wt%) content of 66% through HRT control (60 de40 d) stirring increased parasitic
without dosing buffering chemicals. energy.
Source-separated FW Semi-continuous one-stage Thermophilic SSAD of FW achieved higher VS Higher temperature and Nguyen et al. (2017)
from restaurants thermophilic SSAD reduction and biogas production than mesophilic continuous stirring increased
(TS: 23.0 wt%; VS: one by 6.88% and 16.4%, respectively, when parasitic energy requirement.
20.6 wt%) decreasing the HRT to 25 d.
Synthetic FW (TS: Batch two-stage LBR-UASB system Strategies (adequate micro-aeration, off-gas Extra equipment and parasitic Luo and Wong
38.5 wt%; VS: reutilization, leachate recirculation ratio energy were required; water (2019), Xu et al.
37.4 wt%) adjustment) enhanced FW hydrolysis and replacement and pH adjustment (2014a), Xu et al.
acidogenesis, increasing methane yield by 15 in LBR were required. (2014b), Yan et al.
e38.6%. (2016)
FW from a canteen Semi-continuous three-stage SSAD in Separation and optimization of different stages Dosing of buffering chemicals and Zhang et al. (2017)
(TS: 33.8 wt%; VS: one reactor (1st: high-solids improved VS reduction to 83.5% and methane fast stirring in all stages
32.1 wt%) hydrolysis; 2nd: acidification; 3rd: production by 24e54% at high OLRs. undermined the practicability.
wet methane production)
FW based on the Semi-batch two-stage SSAD (1st: Separation and optimization of different stages Ammonia accumulation at higher This study
annual average of a garage-type solid digester; 2nd: increased FW loading by 33.3% and improved FW loading triggered SSAD
food bank (TS: liquid digester) biogas production by 18.2% with a stable AD instability without proper
24.9 wt%; VS: process; no buffering chemical or stirring was stripping or dilution.
23.7 wt%) required.

increasing FW treatment capacity while maintaining efficient 4. Conclusions

biogas production and process stability, the advantages of the two-
stage system in this study fell into the following aspects: easier The findings in this study demonstrated the feasibility of a two-
operation of garage-type solid-state digester in a semi-batch mode, stage system digesting FW as the sole feedstock. By combining a
no extra water/bulking agent addition to reduce the treatment first-stage solid-state digester and a second-stage liquid digester, a
capacity, and no buffering chemicals usage. 33.3% increase in FW loading and an 18.2% increase in energy yield
This study provides a better solution to treat considerable were recorded as compared to the one-stage SSAD of FW. Stage 1
amounts of daily FW generated from food banks that lack easy served as an optimized hydrolysis/acidogenesis reactor with a total
access to bulking agents, thus contributing to a cleaner bioenergy acids level of 28.5 g/L and a pH of 5.65 regardless of the metabolic
production. Nonetheless, some more efforts are required to be flux shift from acetic/butyric pathways to lactic pathway. Stage 2
devoted to the following issues in future studies. First, the microbial served as the main methanogenesis reactor treating the acid-rich
community shifts in both stages with increasing FW loading need leachate from Stage 1, with the highest sCOD removal of 70.3%
to be investigated to provide more insightful and orientational and a contribution of 94.3% in the combined methane yield of 393.0
guidelines for further optimization. Second, the incorporation of mL/gVS. The recirculation of digested liquid effluent from Stage 2
some pretreatments that could promote the production of short- provided buffering and acids-washing to Stage 1. This two-stage
chain VFAs (Liu et al., 2020) is promising to further increase the system provides a better solution to treat considerable amounts
FW loading in the first stage and benefit the second-stage methane of high-solid FW without extra requirement for water or bulking
production. Third, efficient methods other than dilution by adding agents, thus contributing to a cleaner bioenergy production. The
water or fresh inoculum are needed to reduce the high ammonia accumulation of increasing ammonia derived from proteins, which
level in the leachate that might cause inhibitory effects. induced certain inhibitory effects and triggered an instability of the
two-stage system when further increasing the FW loading, requires
10 L. Ding et al. / Journal of Cleaner Production 279 (2021) 123644

future investigations focusing on efficient countermeasures. food waste and horticultural waste in high-solid system. Appl. Energy 209,
400e408.
Li, Y., Park, S.Y., Zhu, J., 2011. Solid-state anaerobic digestion for methane production
CRediT authorship contribution statement from organic waste. Renew. Sustain. Energy Rev. 15 (1), 821e826.
Liu, X., Du, M., Yang, J., Wu, Y., Xu, Q., Wang, D., Yang, Q., Yang, G., Li, X., 2020. Sulfite
Lingkan Ding: Conceptualization, Investigation, Data curation, serving as a pretreatment method for alkaline fermentation to enhance short-
chain fatty acid production from waste activated sludge. Chem. Eng. J. 385,
Writing - original draft. Yan Chen: Investigation, Writing - review & 123991.
editing. Yongping Xu: Writing - review & editing, Supervision. Bo Luo, L., Wong, J.W.C., 2019. Enhanced food waste degradation in integrated two-
Hu: Writing - review & editing, Funding acquisition, Supervision. phase anaerobic digestion: effect of leachate recirculation ratio. Bioresour.
Technol. 291, 121813.
Martin, D.J., Potts, L.G.A., Heslop, V.A., 2003. Reaction mechanisms in solid-state
Declaration of competing interest anaerobic digestion: 1. The reaction front hypothesis. Process Saf. Environ.
Protect. 81 (3), 171e179.
Micolucci, F., Gottardo, M., Pavan, P., Cavinato, C., Bolzonella, D., 2018. Pilot scale
The authors declare that they have no known competing comparison of single and double-stage thermophilic anaerobic digestion of
financial interests or personal relationships that could have food waste. J. Clean. Prod. 171, 1376e1385.
appeared to influence the work reported in this paper. Morone, P., Falcone, P.M., Lopolito, A., 2019. How to promote a new and sustainable
food consumption model: a fuzzy cognitive map study. J. Clean. Prod. 208,
563e574.
Acknowledgements Nguyen, D.D., Chang, S.W., Cha, J.H., Jeong, S.Y., Yoon, Y.S., Lee, S.J., Tran, M.C.,
Ngo, H.H., 2017. Dry semi-continuous anaerobic digestion of food waste in the
mesophilic and thermophilic modes: new aspects of sustainable management
The authors greatly appreciate the funding supports from
and energy recovery in South Korea. Energy Convers. Manag. 135, 445e452.
Minnesota’s Discovery, Research, and InnoVation Economy Panigrahi, S., Sharma, H.B., Dubey, B.K., 2020. Anaerobic co-digestion of food waste
(MnDRIVE), the assistance of Bob Branham and Katharine Reed with pretreated yard waste: a comparative study of methane production, ki-
from Second Harvest Heartland in Minnesota for the annual FW netic modeling and energy balance. J. Clean. Prod. 243, 118480.
Rajagopal, R., Masse, D.I., Singh, G., 2013. A critical review on inhibition of anaerobic
recipe and industrial consultation, and the University of Wisconsin digestion process by excess ammonia. Bioresour. Technol. 143, 632e641.
Oshkosh for the inoculum collection. Slorach, P.C., Jeswani, H.K., Cue llar-Franca, R., Azapagic, A., 2019. Environmental and
economic implications of recovering resources from food waste in a circular
economy. Sci. Total Environ. 693, 133516.
References Srisowmeya, G., Chakravarthy, M., Nandhini Devi, G., 2020. Critical considerations in
two-stage anaerobic digestion of food waste e a review. Renew. Sustain. Energy
APHA, 1999. Standard methods for the examination of water and wastewater. In: Rev. 119, 109587.
American Public Health Association. Voelklein, M.A., Jacob, A., O’ Shea, R., Murphy, J.D., 2016. Assessment of increasing
Browne, J.D., Murphy, J.D., 2014. The impact of increasing organic loading in two loading rate on two-stage digestion of food waste. Bioresour. Technol. 202,
phase digestion of food waste. Renew. Energy 71, 69e76. 172e180.
Cheng, J., Ding, L., Xia, A., Lin, R., Li, Y., Zhou, J., Cen, K., 2015. Hydrogen production Xu, F., Li, Y., 2012. Solid-state co-digestion of expired dog food and corn stover for
using amino acids obtained by protein degradation in waste biomass by com- methane production. Bioresour. Technol. 118, 219e226.
bined dark- and photo-fermentation. Bioresour. Technol. 179, 13e19. Xu, S., Selvam, A., Karthikeyan, O.P., Wong, J.W., 2014a. Responses of microbial
Cho, S.K., Im, W.T., Kim, D.H., Kim, M.H., Shin, H.S., Oh, S.E., 2013. Dry anaerobic community and acidogenic intermediates to different water regimes in a hybrid
digestion of food waste under mesophilic conditions: performance and meth- solid anaerobic digestion system treating food waste. Bioresour. Technol. 168,
anogenic community analysis. Bioresour. Technol. 131, 210e217. 49e58.
Ding, L., Cheng, J., Qiao, D., Yue, L., Li, Y.-Y., Zhou, J., Cen, K., 2017. Investigating Xu, S., Selvam, A., Wong, J.W.C., 2014b. Optimization of micro-aeration intensity in
hydrothermal pretreatment of food waste for two-stage fermentative hydrogen acidogenic reactor of a two-phase anaerobic digester treating food waste. Waste
and methane co-production. Bioresour. Technol. 241, 491e499. Manag. 34 (2), 363e369.
EPA, 2018. Advancing Sustainable Materials Management: Facts and Figures Report. Yan, B.H., Selvam, A., Wong, J.W.C., 2016. Innovative method for increased methane
https://www.epa.gov/facts-and-figures-about-materials-waste-and-recycling/ recovery from two-phase anaerobic digestion of food waste through reutiliza-
advancing-sustainable-materials-management (Accessed June 27 2020). tion of acidogenic off-gas in methanogenic reactor. Bioresour. Technol. 217, 3e9.
FAO, 2011. Global Food Losses and Food Waste. Extent, Causes and Prevention. Ye, M., Liu, J., Ma, C., Li, Y.-Y., Zou, L., Qian, G., Xu, Z.P., 2018. Improving the stability
Gallert, C., Winter, J., 2008. Propionic acid accumulation and degradation during and efficiency of anaerobic digestion of food waste using additives: a critical
restart of a full-scale anaerobic biowaste digester. Bioresour. Technol. 99 (1), review. J. Clean. Prod. 192, 316e326.
170e178. Yenigün, O., Demirel, B., 2013. Ammonia inhibition in anaerobic digestion: a review.
Herrmann, C., FitzGerald, J., O’Shea, R., Xia, A., O’Kiely, P., Murphy, J.D., 2015. Ensiling Process Biochem. 48 (5), 901e911.
of seaweed for a seaweed biofuel industry. Bioresour. Technol. 196, 301e313. Zhang, C., Su, H., Baeyens, J., Tan, T., 2014. Reviewing the anaerobic digestion of food
Kothari, R., Pandey, A.K., Kumar, S., Tyagi, V.V., Tyagi, S.K., 2014. Different aspects of waste for biogas production. Renew. Sustain. Energy Rev. 38, 383e392.
dry anaerobic digestion for bio-energy: an overview. Renew. Sustain. Energy Zhang, J., Loh, K.-C., Li, W., Lim, J.W., Dai, Y., Tong, Y.W., 2017. Three-stage anaerobic
Rev. 39, 174e195. digester for food waste. Appl. Energy 194, 287e295.
Li, W., Loh, K.-C., Zhang, J., Tong, Y.W., Dai, Y., 2018. Two-stage anaerobic digestion of