You are on page 1of 11

Water Research 102 (2016) 464e474

Contents lists available at ScienceDirect

Water Research
journal homepage: www.elsevier.com/locate/watres

A model for methane production in anaerobic digestion of swine


wastewater
Hongnan Yang a, b, Liangwei Deng a, b, *, Gangjin Liu a, c, Di Yang a, b, Yi Liu a, b, Ziai Chen a, b
a
Biogas Institute of Ministry of Agriculture, Chengdu 610041, PR China
b
Laboratory of Development and Application of Rural Renewable Energy, Ministry of Agriculture, Chengdu 610041, PR China
c
Bioprocess Control AB, Scheeleva €gen 22, 223 63 Lund, Sweden

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

Article history: A study was conducted using a laboratory-scale anaerobic sequencing batch digester to investigate the
Received 15 November 2015 quantitative influence of organic loading rates (OLRs) on the methane production rate during digestion of
Received in revised form swine wastewater at temperatures between 15  C and 35  C. The volumetric production rate of methane
15 May 2016
(Rp) at different OLRs and temperatures was obtained. The maximum volumetric methane production
Accepted 28 June 2016
Available online 1 July 2016
rates (Rpmax) were 0.136, 0.796, 1.294, 1.527 and 1.952 LCH4 L1 d1 at corresponding organic loading rates
of 1.2, 3.6, 5.6, 5.6 and 7.2 g volatile solids L1 d1, respectively, which occurred at 15, 20, 25, 30 and 35  C,
respectively. A new model was developed to describe the quantitative relationship between Rp and OLR.
Keywords:
Anaerobic digestion
In addition to the maximum volumetric methane production rate (Rpmax) and the half-saturation con-
Swine wastewater stant (KLR) commonly used in previous models such as the modified StovereKincannon model and Deng
Model model, the new model introduced a new index (KD) that denoted the speed of volumetric methane
Volumetric methane production rate production rate approaching the maximum as a function of temperature. The new model more satis-
factorily described the influence of OLR on the rate of methane production than other models as
confirmed by higher determination coefficients (R2) (0.9717e0.9900) and lower bias between the
experimental and predicted data in terms of the root mean square error and the Akaike Information
Criterion. Data from other published research also validated the applicability and generality of the new
kinetic model to different types of wastewater.
© 2016 Elsevier Ltd. All rights reserved.

1. Introduction for efficient digester operation and model-based design, kinetic


modeling of the anaerobic digestion process has gained extensive
Associated with the development of an intensive piggery in- attention. The kinetics of biological processes can be addressed
dustry, swine manure management has become an urgent problem using microbial growth models, substrate utilization models and
during recent years. Anaerobic digestion is an economical and product formation models, which are interrelated through the
effective alternative for treating swine manure due to its opera- corresponding yield coefficients (Masse  and Droste, 2000;
tional simplicity and potential for energy recovery (Weiland, 2010; Fernandez-Rodríguez et al., 2013). For such a complex substrate
Deng et al., 2014). containing dissolved and particulate organic matter as swine
Although the use of anaerobic treatment technology is wide- wastewater, the volatile suspended solids, on which an estimate of
spread, optimum process performance seldom is achieved because microorganism concentration typically has been based, is not a
of the high degree of empiricism that prevails in the design and good indicator due to difficulties in differentiating between bac-
operation of anaerobic digesters. As a result of increased demand terial volatile suspended solids and complex biomass volatile solids
(Momoh et al., 2013). In addition, the component of swine waste-
water removed by sedimentation or absorption has often been
neglected because of difficulty in measuring it directly. Using the
* Corresponding author. Biogas Institute of Ministry of Agriculture, Chengdu
microbial growth rate and the substrate removal rate as variables in
610041, PR China.
E-mail address: dengliangwei@caas.cn (L. Deng). models has certain limitations in fitting and evaluating models,

http://dx.doi.org/10.1016/j.watres.2016.06.060
0043-1354/© 2016 Elsevier Ltd. All rights reserved.
H. Yang et al. / Water Research 102 (2016) 464e474 465

particularly for continuous microbial cultures. In contrast, the and at temperatures of 15, 20, 25, 30, and 35  C. The aims were: (1)
amount of biogas (methane) formed during the conversion of to evaluate the methane production performance at gradually
organic matter by microorganisms in anaerobic digestion is the increasing OLR and to obtain the maximum volumetric methane
most common on-line and easily performed measurement production rate at each temperature, (2) to create a rational and
(Donoso-Bravo et al., 2011), and simultaneously reflects the activity suitable model with which to quantify the effect of OLR on the
of microorganisms and the rate and degree of biodegradation volumetric methane production rate, and (3) to evaluate the
because biogas production is directly proportional to substrate quantitative effect of temperature on methane production.
degradation. In many studies, biogas or methane production has
been used as the only measurement by which to estimate model 2. Materials and methods
parameters (Martín et al., 1994; Batstone, 2006; M€ ahnert and Linke,
2009; Ferna ndez-Rodríguez et al., 2013). 2.1. Swine wastewater and inoculum
Numerous types of mathematical models have been developed
to describe the methane production process of animal waste The swine wastewater used in this study was collected from a
digestion. Among them, the anaerobic digestion model 1 (ADM1) is farm located in Jianyang, Sichuan, China, 35 km away from the
the most advanced due to its precise predictability and strong laboratory. Samples were transported back to the laboratory
generality (Girault et al., 2011). The ADM1 model reflects the major immediately after collection and stored at 4  C. The concentration
processes that are involved in the conversion of complex organic of swine manure was adjusted to a volatile solids (VS) content of
substrates into methane and carbon dioxide and inert byproducts 0.80% by adding a certain amount of water before the start of an
(Batstone et al., 2002). However, the model requires a large number experiment.
of constants and coefficients that should be calibrated according to The inoculation sludge for the anaerobic digestion experiments
the characteristics of the substrates; such calibration requires the was obtained from a full-scale digester that treated swine waste-
use of special assays and computing skill, which is difficult for water from the same pig farm that served as the source of the
scientists and engineers dedicated to the plant operation and im- wastewater.
provements (Parker, 2005; Liu et al., 2008). Therefore, simplified
models that consist of only a few variables have been widely 2.2. Anaerobic digestion experiments
studied. The first-order rate equation (Yang et al., 2015) and the
modified Gompertz model (Kafle and Kim, 2013) have been applied As shown in Fig. 1, the methane fermentation experiments were
to batch assays and have satisfactorily predicted methane produc- performed in 1000 mL gas-tight glass reactors with a rubber plug
tion. Unfortunately, the data obtained from batch studies lacks and attached to an influent port, an effluent port and a pipe for
common, universal bases for comparison, and modeling results venting biogas. A 1000 mL wide-mouth glass bottle was used as a
from batch studies are usually provided in terms of the final values biogas gasholder, which was fitted with an influent-effluent port to
of the methane yields from substrates, rather than methane yields allow the entrance of biogas and the discharge of water. The di-
varying with hydraulic retention time (HRT). The volumetric gesters were connected by rubber tubes to the gasholder (Deng
methane production rate, as an important parameter for optimizing et al., 2012). Each digester was inoculated with 500 mL of anaer-
the design of a digester, is difficult to obtain using batch assays obic sludge at the beginning of the experiments. The digester was
(Brule et al., 2014). operated in draw-and-fill mode twice a day. Experiments were
Conversely, simplified models that are calibrated in continuous conducted at 15, 20, 25, 30 and 35  C, and five water baths were
operation can more accurately reflect the actual anaerobic diges- used to maintain the temperature of the digesters. The anaerobic
tion of wastewater and seem to be qualified for design and opti- digesters were mixed manually twice a day.
mization of wastewater treatment plants (Batstone, 2006; Ekama, The OLRs in experiments were increased by reducing the HRT at
2009). Among the kinetic models that predict methane produc- a constant influent concentration until the maximum volumetric
tion based on continuous testing are the modified Sto- methane rate (Rpmax) was achieved (when the Rp stopped rising or
vereKincannon (Yu et al., 1998), CheneHashimoto (Chen, 1983) the deviations of the last two Rps were less than 5%). The difference
and Deng (Deng et al., 2014) models; all are derived assuming that of methane production rate under different temperatures resulted
digesters are operated at steady state conditions. The Chen and in different initial and final OLRs and loading intervals for the five
Hashimoto model was considered to be an appropriate model with temperatures that were studied. The OLRs applied were increased
which to describe the kinetics of methane production from swine by small increments in order to minimize any adverse effects of
wastewater and has been widely used (Pham et al., 2014). Yu et al. sudden increases in loading to obtain the Rpmax. The operating load
(1998) proposed a model to describe the kinetics of methane pro- range in the anaerobic digestion experiments at different temper-
duction based on the StovereKincannon model. The modified atures is listed in Table 1. There were different experimental runs
StovereKincannon model for methane production has been applied for different operation temperatures. A steady-state condition
to soybean wastewater (Yu et al., 1998), synthetic milk wastewater during each run was achieved when the deviations between the
(Ramakant et al., 2002) and synthetic wastewater containing para- observed values of daily methane production were less than 5% and
nitrophenol (Kuşçu and Sponza, 2009), but has been rarely used in each run had a duration of 2e3 times the corresponding HRT (in the
the study of swine wastewater anaerobic treatment. To describe the range of 5e40 days) or of 10 days (in the HRT range 1.1e3.3 days) at
variation in the volumetric methane production rate (Rp) as a steady-state condition. All treatments were conducted in duplicate.
function of the organic loading rate (OLR) in the temperature range The amounts of released biogas and the concentrations of methane
of 15e35  C, Deng et al. (2014) developed a reliable model capable were recorded on a daily basis. The gas produced in each digester
of closely matching observed methane production rates was measured using a water displacement device.
(R2 ¼ 0.989e0.999). However, the lack of widespread verification
and comparison of the fitting results of these models limits their 2.3. Analytical methods
application in biogas engineering.
In this study, the anaerobic digestion of swine wastewater was Analyses of chemical oxygen demand (COD), total solids (TS),
conducted at incremental OLRs by increasing the feed flow-rate and VS were carried out according to standard methods (APHA,
while maintaining a constant influent substrate concentration 1998), The determination of COD was accomplished by digesting
466 H. Yang et al. / Water Research 102 (2016) 464e474

Fig. 1. The scheme of the device for methane fermentation.

Table 1
The operating load range in anaerobic digestion experiment at different Rpmax
temperatures. Rp ¼ (3)
1 þ eðKLR LrÞ
Temperature Organic loading rate (OLR)/g VS L1 d1
In Eqs. (1)e(3), Bo is ultimate methane yield (LCH4 g1 VSadded at

15 C 0.2 0.4 0.6 0.8 1.2 1.6

infinite retention time); Lr is organic loading rate (OLR) (g VS
20 C 0.8 1.2 1.6 2.0 2.4 2.8 3.2 3.6 4.0 4.4
25 
C 1.6 2.4 3.2 4.0 4.8 5.6
L1 d1); KC is a dimensionless parameter; KLR is the half-saturation
30 
C 2.4 3.2 4.0 4.8 5.6 constant (g VS L1 d1); Rp is volumetric methane production rate
35 
C 3.2 4.0 4.8 5.6 6.4 7.2 (LCH4 L1 d1); Rpmax is maximum volumetric methane production
Hydraulic retention time (HRT)/d rate (LCH4 L1 d1); Si is influent substrate concentration (g VS L1);

Q is HRT (days); and mmax is maximum specific microbial growth
15 C 40.0 20.0 13.3 10.0 6.7 5.0
20 
C 10.0 6.7 5.0 4.0 3.3 2.9 2.5 2.2 2.0 1.8 rate (d1).

25 C 5.0 3.3 2.5 2.0 1.7 1.4

30 C 3.3 2.5 2.0 1.7 1.4

35 C 2.5 2.0 1.7 1.4 1.3 1.1

2.5. Model fitting

samples in a microwave digestion system (WMX-Ш-B, China) fol- The parameters for each model were estimated by curve-fitting
lowed by ferrous ammonium sulfate titration. The TS content was using software (Origin version 8.0, OriginLab Corporation, North-
determined by heating samples in an oven (ZXRD-A5110, China) at ampton, MA, USA) that minimized the residual sum of squared
105  C, and VS content was determined by ignition of the residue errors between the experimental data and model predictions.
produced in a muffle furnace (KSL-1100X, China) at 550  C. The pH
of the wastewater was checked using a pH meter (inoLab pH 7200,
Germany). The composition of biogas was measured using a Biogas
Analyzer (ADOS Biogas 401, Germany).
2.6. Model evaluation

In the model fitting stage, three statistical criteria were used to


2.4. Models for methane production evaluate model fit and select the best one. The criteria were the
determination coefficient (R2), the root mean square error (RMSE,
Three simple models (CheneHashimoto model, modified Sto- Eqs. (4) and (5)) and the Akaike Information Criterion (AIC, Eq. (6))
vereKincannon model and Deng model) describing the methane (Pham et al., 2014).
production process were used in the study and are presented here:
pffiffiffiffiffiffiffiffiffiffi
RMSE ¼ MSE (4)
Modified StovereKincannon model (Yu et al., 1998)
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Rpmax Lr Pn 2
Rp ¼ (1) i¼1 ðPi  OiÞ
KLR þ Lr RMSE ¼ (5)
n

Chen-Hashimoto model (Chen, 1983) AIC ¼ n lnðMSEÞ þ 2p (6)


 
BoSi KC In Eqs. (1)e(3), Oi is the experimental value, Pi is the predicted
Rp ¼ 1 (2)
Q Qmmax  1 þ KC value, n is the number of data points, MSE is the mean square error,
and p is the number of fitting parameters employed in the corre-
sponding model. Low values for RMSE and AIC indicate low esti-
Deng model (Deng et al., 2014) mation errors.
H. Yang et al. / Water Research 102 (2016) 464e474 467

3. Results and discussion

3.1. Effect of increasing OLR on methane production at 15e35  C

During the transient condition (the period between the time


when the shock started and when the reactor achieved a steady
state), the volumetric methane production rate gradually increased
over digestion time. When the volumetric methane production rate
reached a stable value and the variation was within 5%, a steady
state condition was obtained. Details are given in Appendix. The
steady-state value of the operational parameters (COD and
methane production) was taken as the average of consecutive
measurements for each parameter. The effect of OLR on the sub-
strate treatment efficiency and methane production performance
of the digesters in the temperature range 15e35  C are summarized
in Figs. 2 and 3, respectively. Increasing the OLR decreased the COD
removal efficiency and methane yield in terms of VS added, and
increased effluent COD concentration, COD removal loading rate at
all temperatures (Figs. 2 and 3). Similar trends have been found in
mesophilic temperature fermentation of swine manure at 25e35  C
(Chae et al., 2008) and psychrophilic temperature fermentation of
dairy manure at 22  C (Ma et al., 2013).
As shown in Fig. 3(A), the methane yields at low organic load-
ings exhibited little change at each temperature. In contrast, the
progressive increase in OLR by reducing HRT during subsequent
stages of the experiments led to a continuous decrease in methane
yield because the short retention time of organic matter prevented
its complete degradation. Methane yields at low temperatures and
low OLRs approached those at high temperatures and high organic
loading rates. For instance, an average methane yield of 0.270
LCH4 g1 VSadded was obtained at 20  C and at OLRs of 0.8e2.0 g VS
L1 d1, corresponding to HRTs of 4e10 days; this yield was equal to
that obtained at 35  C (0.271 LCH4 g1 VSadded) and at OLR of 7.2 g
VS L1 d1, corresponding to HRT of 1.1 days. These results indicated
that anaerobic digestion of swine wastewater at 20  C required
approximately four times as long to achieve the same methane
yield as that achieved at 35  C. The methane yields at 15e35  C
were much lower than the theoretical methane production of 0.516
LCH4 g1 VSadded reported by Panichnumsin et al. (2010), 0.541
LCH4 g1 VSadded reported by Møller et al. (2004) and 0.72 LCH4 g1
VSadded reported by Chae et al. (2008), each of which was calculated
based on the elemental composition of swine wastewater.
The Rp, as a measure of methane production based on the vol-
ume of reactors, is a significant basis of design and investment
calculation of biogas plants. In this study, Rp began to sharply in-
crease with the increase in OLR, and then stabilized at high loading
rates (Fig. 3(B)); the Rp at stabilization was identified as the Rpmax.
As shown in Fig. 3(B), the Rpmax depended on temperature. At OLRs
of 1.2, 3.6, 5.6, 5.6 and 7.2 g VS L1 d1, Rpmax was 0.136, 0.796, 1.294,
1.527 and 1.952 LCH4 L1 d1, respectively, for temperatures of 15,
20, 25, 30 and 35  C, respectively. The HRTs corresponding to those
OLRs were 6.7, 2.2, 1.4, 1.4 and 1.1 days for 15, 20, 25, 30 and 35  C,
respectively, and were considered critical for the proper func-
tioning of the digestion process, because HRTs shorter than these
would lead to washout of sludge and a sharp decrease in pH. At
OLRs in excess of those for Rpmax, a sharp decrease in methane yield
occurred due to organic overloading of the digester. Lo et al. (1994)
obtained the highest Rp value of 0.71 LCH4 L1 d1 in the steady- Fig. 2. Variations of the effluent COD concentration (A), COD removal efficiency (B)
state period of anaerobic treatment of swine wastewater using and COD removal loading rate (C) with the organic loading rate (OLR) and the
temperature.
hybrid upflow anaerobic sludge blanket reactors at an OLR of 3.58 g
COD L1 d1 and ambient temperature (22e28  C); however, this
rate was lower than the 0.796e1.294 LCH4 L1 d1 achieved at
20e25  C in the present study, possibly because of the different
feedstock and digesters. Hill (1982), using a comprehensive dy-
namic model, reported that the maximum organic loading rate
468 H. Yang et al. / Water Research 102 (2016) 464e474

infinity, the value of Bo can be determined graphically by plotting


the methane yield (B) at steady-state against the reciprocal of Q
and extrapolating the curve to where 1/Q ¼ 0 (Chen, 1983). The
values of Bo were determined to be 0.204 LCH4 g1 VSadded at 15  C,
0.310 LCH4 g1 VSadded at 20  C, 0.327 LCH4 g1 VSadded at 25  C, 0.395
LCH4 g1 VSadded at 30  C and 0.400 LCH4 g1 VSadded at 35  C. These
values were well within published values of ultimate methane
yields; for example, Chae et al. (2008) obtained Bo of 0.327, 0.389
and 0.403 LCH4 g1 VSadded at 25, 30 and 35  C, respectively, in batch
experiments. Once the values for Bo were determined, the kinetic
parameters, mmax and KC in the CheneHashimoto model, were
calculated. The Rpmax and KLR in the modified StovereKincannon
model and the Deng model were evaluated directly using the
experimental data.
As shown in Table 2, a wide range of values for the kinetic
constants for each model was obtained depending on the experi-
mental conditions. On the whole, the CheneHashimoto model had
the highest R2 values (average of 0.9840) and the lowest RMSE and
AIC (averages of 0.023 and 41.786, respectively) for each tempera-
ture in comparison to the Deng model (average R2, RMSE and AIC of
0.9387, 0.028 and 38.6610, respectively) and the modified Sto-
vereKincannon model (average R2, RMSE and AIC of 0.9351, 0.034
and 37.242, respectively). The Deng model and the modified
StovereKincannon model demonstrated unsatisfactory fitting re-
sults with R2 of 0.7696 and 0.7816, respectively, at 15  C. Although
both the Deng model and the modified StovereKincannon model
had similar results with high R2 values and low RMSE and AIC
values at 20  C and at 30  C, the R2 values of the Deng model
increased by 2.0% at 25  C and 1.2% at 35  C, and corresponding
RMSE and AIC values decreased by 44.5% and 21.3% at 25  C and
15.6% and 6.8% at 35  C, respectively, compared with those of the
modified StovereKincannon model. These results indicated that
the Deng model was better overall than the modified Sto-
vereKincannon model. The Deng model achieved almost identical
predictions as the CheneHashimoto model with the exception of
the results at 15  C. As described in Section 3.1, the Rp gradually
reached its maximum value as the substrate concentration in the
digester increased, after which production stabilized due to desa-
turation of enzymes. In the anaerobic digestion experiments, Rpmax
Fig. 3. Variations of the methane yield (A) and the volumetric methane production showed the maximum methane production for the specific volume
rate (B) with the organic loading rate (OLR) and the temperature.
of digester. However, the value of Rpmax for the CheneHashimoto
model had to be calculated by taking the derivative of Rp in terms of
prior to process failure at 35  C is 7.5 g VS L1 d1, which is similar Q and equating it to zero (Fongsatitkul et al., 2012). As shown in
to the 7.2 g VS L1 d1 achieved at 35  C in the present study. Table 2, the comparison of experimental Rpmax values and calcu-
lated Rpmax values from the model demonstrated that the Deng
model and the CheneHashimoto model can accurately predict the
3.2. Modeling of methane production peak volumetric methane production rate (the RMSE values were
0.395 and 0.055, respectively), while the predicted values of Rpmax
3.2.1. Fitting of methane production rate versus organic loading rate from the modified StovereKincannon model were 2e3 times
using previously reported models higher than the experimental results (the RMSE was 2.524). On the
The three models described in section 2.4 were used to identify other hand, the half-saturation constant (KLR) in the modified
the one having an appropriate fit for methane production kinetics StovereKincannon model and the Deng model is equated with the
during anaerobic digestion of swine wastewater. Kinetic parame- organic loading rate where Rp is half the value of Rpmax, and can
ters were determined using the nonlinear curve fit procedure of the provide information about the OLR that supports the system for
Origin 8.0 software and the steady-state experimental data at reactor design. The experimental values of KLR, were estimated as
15e35  C. To evaluate model performance, in addition to R2, the 0.4, 1.4, 2.2, 2.4 and < 3.4 g VS L1 d1 for 15, 20, 25, 30 and 35  C,
RMSE and AIC were used to assess the bias and the precision of the respectively, and by definition of KLR, were consistent with those of
models based on experimental and predicted data; low RMSE and Deng model, but much smaller than those of the modified Sto-
AIC indicated low estimation errors. vereKincannon model. These results indicated that the modified
For a given influent substrate concentration (Si) and HRT (Q) in StovereKincannon model has some weaknesses in predicting
the CheneHashimoto model, Rp is a function of ultimate methane maximum volumetric methane production rate. The fact that the
yield (Bo), maximum specific microbial growth rate (mmax) and a prediction ability of the modified StovereKincannon model at high
dimensionless parameter (KC). Using the definition that ultimate OLRs was weak may explain the insensitivity of the model to
methane yield (L g1 VS added) is achieved as the HRT approaches
H. Yang et al. / Water Research 102 (2016) 464e474 469

Table 2
Fitting results using data in this paper.

Temperature Model Rpmaxfrom experiment Rpmax from model KLR mmax KC KD R2 RMSE AIC

15  C Modified Stover-Kincannon 0.136 0.325 ± 0.085 1.591 ± 0.635 0.7816 0.011 41.046
Chen-Hashimoto 0.128 0.510 ± 0.451 1.908 ± 2.275 0.9754 0.005 48.733
Deng 1.000 ± 2.862 3.017 ± 2.225 0.7696 0.016 37.551
Modified Deng 0.137 ± 0.005 0.447 ± 0.023 5.024 ± 0.523 0.9918 0.002 54.248

20  C Modified Stover-Kincannon 0.796 3.084 ± 0.713 9.733 ± 2.883 0.9874 0.021 57.898
Chen-Hashimoto 0.904 1.060 ± 0.461 0.495 ± 0.344 0.9902 0.018 59.893
Deng 0.970 ± 0.031 1.858 ± 0.092 0.9875 0.021 57.967
Modified Deng 0.881 ± 0.036 1.656 ± 0.023 1.242 ± 0.523 0.9923 0.015 61.240

25  C Modified Stover-Kincannon 1.294 3.997 ± 1.099 10.839 ± 4.146 0.9718 0.045 33.105
Chen-Hashimoto 1.306 0.976 ± 0.095 0.159 ± 0.048 0.9921 0.024 40.698
Deng 1.350 ± 0.024 1.858 ± 0.079 0.9913 0.025 40.171
Modified Deng 1.395 ± 0.048 2.383 ± 0.105 0.882 ± 0.092 0.9926 0.020 40.821

30  C Modified Stover-Kincannon 1.527 4.501 ± 0.864 10.601 ± 2.879 0.9737 0.036 29.158
Chen-Hashimoto 1.534 1.246 ± 0.406 0.361 ± 0.231 0.9841 0.028 31.654
Deng 1.557 ± 0.057 2.417 ± 0.177 0.9740 0.034 29.814
Modified Deng 1.839 ± 0.235 2.754 ± 0.435 0.736 ± 0.163 0.9834 0.024 31.456

35  C Modified Stover-Kincannon 1.952 5.182 ± 1.128 11.383 ± 3.693 0.9606 0.055 25.001
Chen-Hashimoto 2.008 1.685 ± 0.367 0.407 ± 0.170 0.9782 0.041 27.953
Deng 1.936 ± 0.038 3.104 ± 0.112 0.9719 0.046 26.692
Modified Deng 2.055 ± 0.066 3.095 ± 0.096 0.715 ± 0.095 0.9900 0.024 31.277

RMSEa Modified Stover-Kincannon 2.524


Chen-Hashimoto 0.055
Deng 0.395
Modified Deng 0.158
a
RMSE is the differences between Rpmax from experiment and that from model under 15e35  C.

mutation of Rp because of overloading. In other words, the Rpmax results for 15  C, producing predictions for Rpmax of 1.000 LCH4
and KLR values obtained from the modified StovereKincannon L1 d1 and the KLR of 3.017 g VS L1 d1 that were approximately
model will not provide an accurate basis for design of reactors and seven times as much as experimental values. This performance
for heating operation. These results demonstrate that the demonstrated that Rpmax and KLR alone were not capable of accu-
CheneHashimoto model and the Deng model matched the exper- rately adjusting the rate of change of Rp with organic loading rate
imental methane production data better than did the modified (Lr) at varying temperature, and that further revision of the Deng
StovereKincannon model. model was needed.
Although the CheneHashimoto model exhibited wellefitting
results as described above in terms of R2, RMSE and AIC, the 3.2.2. The development of a new model
magnitude of the parameters standard error values are much A parameter of KD was introduced to the Deng model as a co-
greater than the fitted values, and the ratio of the parameter esti- efficient of KLR and Lr. The new model, named the “modified Deng
mate to its standard error in was larger than that of the modified model”, is presented as Eq. (7).
StovereKincannon model and Deng model. It indicated that the
poor precision of fitted values was presented in parameters esti- Rpmax
mate in the CheneHashimoto model and parameters should be Rp ¼ (7)
1 þ eKD ðKLR LrÞ
recalibrated in order to gain a better representation of reality by
special efforts. In addition, the evaluation and verification of pa- The kinetic constants calculated from the modified Deng model
rameters such as (mmax) and Bo increased the difficulty of applica- are listed in Table 2. Compared to those of the other three models, a
tion and reduced the practicality of this model. Measuring the higher R2 was obtained using the modified Deng model, ranging
maximum specific microbial growth rate through experimental from 0.9834 to 0.9926, with an average of 0.9900. Likewise, the
methods is impractical due to the requirement of an infinitely short comparatively lower values of RMSE (average of 0.017) and AIC
retention time, which is not feasible for complex biomass (Momoh (average of 43.808) meant that the simulated data from the
et al., 2013). The evaluation of mmax through models based on modified Deng model were in good agreement with the experi-
substrate mass balance can be used as an alternative to experi- mental observations. Although the modified Deng model contains
mental measurement, but the measure of substrate concentration one more fitting parameter than do the other models, the lower AIC
increases the difficulty of application due to problems in differen- values illustrated that this increase did not reduce the simplicity of
tiating between bacterial volatile suspended solids and complex the modified Deng model. On the contrary, the modified Deng
biomass volatile solids. Moreover, the KC in the CheneHashimoto model provided a more satisfactory description of the anaerobic
model, as a dimensionless kinetic parameter, is difficult to deter- digestion process of swine wastewater at 15e35  C compared to the
mine experimentally. other models. Except at 30  C, the ratio of the parameter estimate to
Therefore, only the Deng model came close to meeting the four its standard error did not exceed 10 percent for any parameter in
principles required for adequate model structures: simplicity, the modified Deng model at studied temperatures, and was the
causality, identifiability and predictive capability (Spriet, 1985). smallest than other three models. The fitted values for Rpmax and
However, the Deng model demonstrated unsatisfactory fitting KLR agreed well with the experimental results at different tem-
peratures. As shown in Table 2, the RMSE of Rpmax predictions by
470 H. Yang et al. / Water Research 102 (2016) 464e474

Fig. 4. The relationship between kinetic constants of the maximum volumetric methane production rate (Rpmax), the half-saturation constant (KLR) and kinetic parameter (KD) and
temperature.

the modified Deng model and experimental observations was 0.158 Lr.
compared to a RMSE of 0.055 for the CheneHashimoto model. In Quantification of the effect of temperature on the biokinetic
fact, the Deng model can be viewed as a special form of the coefficients for anaerobic digestion can be accomplished using Eqs.
modified Deng model for which KD is equal to one. The values of KD (8)e(10) (Metcalf, 2003).
were found to be 5.024, 1.242, 0.882, 0.736 and 0.715 for 15, 20, 25,
30 and 35  C, respectively; thus, the KD values for 20e35  C were in RpmaxðT2Þ ¼ RpmaxðT1Þ q
ðT2T1Þ
(8)
close proximity, while the KD value for 15  C was far more than one.
This fact explains why the Deng model produced a poor fit to
ðT2T1Þ
experimental data at 15  C, but good fits for data at 20e35  C. KLRðT2Þ ¼ KLRðT1Þ q (9)

3.2.3. The influence of temperature on kinetic constants of new ðT2T1Þ


KDðT1Þ ¼ KDðT2Þ q (10)
model
The overall effect of temperature changes on methane fermen- In Eqs. (7)e(9), Rpmax(T1) and Rpmax(T2) are the Rpmax at tem-
tation is understood best by considering the variation of the kinetic perature T1 and T2, respectively, L L1 d1; KLR(T1) and KLR(T2) are the
constants Rpmax, KLR and KD with temperature. The three regression KLR at temperature T1 and T2, respectively, g VS L1 d1; KD(T1) and
equations describing the relationships shown in Fig. 4 indicate that KD(T2) are the KD at temperature T1 and T2, respectively; q is the
Rpmax and KLR increase exponentially with temperature, whereas temperatureeactivity coefficient; and T is the temperature ( C).
the KD decreases exponentially with the temperature, but the in- Because KD is negatively correlated with temperature, Eq. (10) is
cremental effects of temperature on Rpmax, KLR and KD gradually expressed differently from Eqs. (8) and (9). The value for q (Table 3)
decrease as temperature increases. The determination coefficients can be obtained using the fitted values for the Rpmax, KLR and KD in
for the three equations (0.9919, 0.9680 and 0.9990 for Rpmax, KLR Table 2 at different temperatures, as determined using the modified
and KD, respectively) confirmed the validity of the equations. Deng model. Table 3 shows that the q values of Rpmax, KLR and KD
Furthermore, it was noted that the observed variation in KD caused between 15 and 20  C are the highest (1.451, 1.299 and 1.322,
by temperature was nearly identical to that caused by increasing respectively), which reflects that the variations of Rpmax, KLR and KD
the maximum OLR from 1.2 to 7.2 g VS L1 d1 as the temperature were the highest when the temperature increased from 15 to 20  C.
ranged from 15  C to 35  C. The KD seems to serve as an index of the When the temperature was greater than 20  C, q values were
processing speed of Rp as it approaches Rpmax due to the change of similar. These results indicate that anaerobic digestion of swine
H. Yang et al. / Water Research 102 (2016) 464e474 471

Table 3
Temperatureeactivity coefficient (q) of kinetic constants at different ranges of temperatures.

Range of temperature Temperatureeactivity coefficient for Rpmax Temperatureeactivity coefficient for KLR Temperatureeactivity coefficient for KD

q(15 Ce20  C) 1.451 1.299 1.322


q(15 Ce25  C) 1.261 1.182 1.190
q(15 Ce30  C) 1.189 1.129 1.137
q(15 Ce35  C) 1.145 1.102 0.907
q(20 Ce25  C) 1.096 1.076 1.071
q(20 Ce30  C) 1.076 1.052 1.054
q(20 Ce35  C) 1.058 1.043 1.037
q(25 Ce30  C) 1.057 1.029 1.037
q(25 Ce35  C) 1.039 1.026 1.021
q(30 Ce35  C) 1.022 1.024 1.006

wastewater was much more sensitive to variation in temperature accomplished using data from an own and single experiment;
from 15 to 20  C than to variation from 20 to 35  C. rather, data from a variety of studies are needed. For that reason, to
validate the modified Deng model, data were collected from pub-
lished studies on several different process typologies involving
3.2.4. The uncertainty of kinetic constants of new model
anaerobic processing of various types of wastewater (Table 4). The
The new model for the anaerobic digestion process was
sources of data were arbitrarily named Study I (Faisal and Unno,
analyzed by the parametric sensitivity method to determine the
2001); Study II (Borja et al., 2002); Study III (Kuşçu and Sponza,
influence of individual model parameters on the selected output
2009); Study IV (Pandian et al., 2011); and Study V (Xiao et al.,
variables. The relative parametric sensitivity value indicated the
2012). The Rpmax values in these studies were obtained by
most influential parameters in the process being the maximum
increasing OLR to match that used in the present study, and were:
volumetric methane production rate.
1.370 L L1 d1 (Study I), 0.560 L L1 d1 (Study II), 0.246 L L1 d1
(Study III), 1.756 L L1 d1 (Study IV) and 5.28 L L1 d1 (Study V). A
3.3. Model verification using data previously reported by other further comparison of all four models (modified Sto-
researchers vereKincannon, CheneHashimoto, Deng and modified Deng) was
conducted using the previously reported experimental data and
Validating a model for applicability and generality cannot be

Table 4
Data from other researchers.

Number Reference Substrate Reactor Temperature Influent COD Hydraulic Organic loading Volumetric methane
concentration/g retention rate/g COD L1 d1 production rate/LCH4
L1 time/d L1 d1

I Faisal and Palm oil mill wastewater ABR e 16 10.0 1.60 0.436
Unno (2001) 8.0 2.00 0.650
7.0 2.28 0.730
6.0 2.67 1.050
5.0 3.20 1.155
3.0 5.33 1.370

II Borja et al., wastewater derived from the A cone-shaped 15e19  C 11.3 20 0.57 0.206
2002 production of protein glass vessel 13.3 0.85 0.280
10 1.13 0.366
8 1.41 0.464
6.7 1.7 0.520
5.7 1.98 0.560
5 2.26 0.425
4.5 2.49 0.460

III Kuşçu and synthetic wastewater AMBR e 3 10.38 0.31 0.075


Sponza 2009 5.19 0.6 0.116
3.4 0.93 0.178
2.4 1.31 0.242
1.5 2.14 0.246
1 3.25 0.026

IV Pandian Pharmaceutical wastewater AHR 30e35  C 1.25 3.2 0.240


et al., 2011 0.75 5.32 0.516
0.5 8.04 0.890
0.33 12.00 1.155
0.25 16.05 1.756
0.125 32.26 1.578

V Xiao et al., municipal biomass waste and CSTR 35  C 100 1.2 0.89
(2012) waste activated sludge 50 2.4 1.72
33 3.6 2.23
25 4.8 3.50
20 6 4.25
15 8 5.28
472 H. Yang et al. / Water Research 102 (2016) 464e474

Table 5
Validation and comparison of models using the data from other researchers.

No. Model Rpmax KLR Bo mmax KC KD R2 RMSE AIC

I Modified Stover-Kincannon 3.804 ± 2.065 8.740 ± 6.607 0.8386 0.115 21.987


Chen-Hashimoto 1.381 0.4 0.481 ± 0.178 0.243 ± 0.272 0.8780 0.100 23.629
Deng 1.488 ± 0.102 2.187 ± 0.206 0.9254 0.078 26.669
Modified Deng 1.375 ± 0.057 2.093 ± 0.079 1.619 ± 0.232 0.9754 0.038 33.360

II Modified Stover-Kincannon 2.056 ± 0.450 5.126 ± 1.456 0.9893 0.105 32.123


Chen-Hashimoto 1.102 0.426 3.776 ± 24.939 10.000 ± 68.611 0.9865 0.097 33.260
Deng 1.112 ± 0.186 1.878 ± 0.285 0.9584 0.126 29.142
Modified Deng 0.633 ± 0.024 0.950 ± 0.045 2.025 ± 0.169 0.9963 0.078 34.721

III Modified Stover-Kincannon 0.424 ± 0.100 1.321 ± 0.608 0.9113 0.114 22.103
Chen-Hashimoto 0.173 0.263 0.902 ± 0.252 1.157 ± 0.417 0.9647 0.073 27.452
Deng 0.368 ± 0.103 1.169 ± 0.607 0.7546 0.126 20.880
Modified Deng 0.255 ± 0.013 0.630 ± 0.059 3.186 ± 0.599 0.9709 0.094 22.403

IV Modified Stover-Kincannon 2.634 ± 0.854 15.506 ± 10.110 0.7981 0.217 14.330


Deng 1.458 ± 0.169 6.846 ± 0.944 0.7507 0.241 13.066
Modified Deng 1.667 ± 0.142 8.022 ± 1.039 0.331 ± 0.097 0.9258 0.114 20.037

V Modified Stover-Kincannon 67.683 ± 86.111 92.880 ± 126.614 0.9866 0.156 18.269


Deng 4.889 ± 0.378 3.546 ± 0.347 0.9141 0.396 7.116
Modified Deng 6.011 ± 0.410 4.330 ± 0.333 0.540 ± 0.063 0.9908 0.112 20.232

Table 6
Fitting results for the maximum volumetric methane production rate (Rpmax) using data from other researchers.

No. Rpmax from experiments reported Rpmax from models

Modified Stover-Kincannon Chen-Hashimoto Deng Modified Deng

I 1.37 3.804 1.381 1.488 1.375


II 0.56 2.056 1.102 1.112 0.633
III 0.246 0.424 0.173 0.368 0.255
RMSEa 1.653 0.316 0.333 0.043
IV 1.756 2.634 1.458 1.667
V 5.28 67.683 4.889 6.011
RMSEb 27.940 0.339 0.331
a
RMSE is for four models using data from I, II and III.
b
RMSE is for three models excepting Chen-Hashimoto using data from IV and V.

results of this comparison are listed in Tables 5 and 6. The fact that methane production rates at various temperatures and (OLRs), as
the influent COD concentration varied at different OLRs in Study IV well as the maximum volumetric methane production rates at each
(Pandian et al., 2011) and Study V (Xiao et al., 2012) increased the temperature. A kinetic study was carried out to improve anaerobic
difficulty of fitting the CheneHashimoto model; therefore, the data process control and design, and compared four methane produc-
from Studies IV and V were not used to verify the CheneHashimoto tion kinetic models: the Modified Stover-Kincannon model,
model. As shown in Table 5, for different types of wastewater, the CheneHashimoto model, Deng model and a new model developed
modified Deng model gave the best simulation of observed data as part of this research. Using data from this study and other pre-
(R2 ¼ 0.9258e0.9963), while the performance of other models viously published research, the new model was shown to yield the
varied with the type of wastewater. For example, the Deng model best simulation of observed methane production in terms of R2,
presented the worst fit (R2 of 0.7546) of all models for synthetic RMSE and AIC. As a result of this study, it can be concluded that the
wastewater (Study III), but better fitting of results for palm oil mill new model accurately describes the quantitative influence of OLRs
wastewater (Study I). Table 6 shows that predictions from the on the production rate of methane, and is superior to other
modified Deng model for Rpmax were better than those of other commonly used kinetic models. Further, the new model is widely
models. For the first three studies of I, II and III, the four models applicable to a variety of digestion process configurations and
were compared. The results indicated that the RMSE of the modi- wastewater, and can be used to improve the design and operation
fied Deng model was 0.043, followed by that of CheneHashimoto of anaerobic digestion facilities.
model (0.316). For all five studies, the three models were compared
excepting the CheneHashimoto model in which the Rpmax could Acknowledgment
not be achieved. The results showed that the RMSE of the modified
Deng model was 0.331, followed by an RMSE of 0.339 for the Deng This study was financially supported by the Natural Science
model. The results of these comparisons support the conclusion Foundation of China (Grant Number: 31572450) and the China
that the modified Deng model is characterized by applicability and Agriculture Research System (Grant Number: CARS-36-10B).
generality for a wide range of conditions.
Appendix
4. Conclusions

This research studied the performance of the anaerobic diges-


tion of swine wastewater at 15e35  C, and obtained the volumetric
H. Yang et al. / Water Research 102 (2016) 464e474 473

0.16 1.5
o
15 C o

-1 -1
1.4 25 C
-1 -1

V o l u m e t ri c m e t h a n e p ro d u c t i o n ra t e / L L d
0.14
V o l u m e t ri c m e t h a n e p ro d u c t i o n ra t e / L L d

1.3
0.12 1.2
OLR:
1.1
0.10 0.2
1.0
0.4
0.6 0.9
0.08
0.8 0.8 OLR:
1.2 0.7
0.06 1.6
1.6
0.6 2.4
0.04 0.5 3.2
4.0
0.4
0.02 4.8
0.3 5.6
0.00 0.2
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 0 5 10 15 20 25 30 35 40 45
Time /d
Time /d

0.90
o 1.6
20 C
-1 -1

o
0.85 30 C

-1 -1
V o l u m e t ri c m e t h a n e p ro d u c t i o n ra t e / L L d

V o l u m e t ri c m e t h a n e p ro d u c t i o n ra t e / L L d
0.80 1.5
0.75 1.4
0.70 1.3
OLR:
0.65
0.8 1.2
0.60
1.2
0.55 1.1
1.6
0.50 1.0
2.0
0.45
2.4 0.9
0.40
2.8
0.35 0.8
3.2
0.30 3.6 0.7
0.25 4.0 0.6
0.20 4.4 OLR: 2.4 3.2 4.0 4.8 5.6
0.15 0.5

0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35
Time /d
Time /d

2.4
-1 -1

o
35 C
V o l u m e t ri c m e t h a n e p ro d u c t i o n ra t e / L L d

2.2
2.0
1.8
1.6

1.4
1.2 OLR:
1.0
3.2
4.0
0.8 4.8
0.6 5.6
6.4
0.4 7.2

0 5 10 15 20 25 30 35 40 45 50
Time /d

Fig. 1. The volumetric methane production rates at different organic loading rates (OLR) and different temperatures. Note: the unit of OLR is g VS L1 d1.

References Batstone, D.J., Keller, J., Angelidaki, I., Kalyuzhnyi, S., Pavlostathis, S., Rozzi, A.,
Sanders, W., Siegrist, H., Vavilin, V., 2002. The IWA anaerobic digestion model
no 1(ADM 1). Water Sci. Technol. 45, 65e73.
APHA, 1998. In: Standard Methods for the Examination of Water and Wastewater,
Borja, R., Gonz
alez, E., Raposo, F., Mill
an, F., Martín, A., 2002. Kinetic analysis of the
twentieth ed. W. American Public Health Association, DC, USA.
psychrophilic anaerobic digestion of wastewater derived from the production of
Batstone, D.J., 2006. Mathematical modelling of anaerobic reactors treating do-
proteins from extracted sunflower flour. J. Agric. Food Chen 50, 4628e4633.
mestic wastewater: rational criteria for model use. Rev. Environ. Sci. Biotechnol.
, M., Oechsner, H., Jungbluth, T., 2014. Exponential model describing methane
Brule
5, 57e71.
474 H. Yang et al. / Water Research 102 (2016) 464e474

production kinetics in batch anaerobic digestion: a tool for evaluation of Technol. 131, 6e12.
biochemical methane potential assays. Bioproc. Biosyst. Eng. 37, 1759e1770. Ma€hnert, P., Linke, B., 2009. Kinetic study of biogas production from energy crops
Chae, K., Jang, A., Yim, S., Kim, I.S., 2008. The effects of digestion temperature and and animal waste slurry: effect of organic loading rate and reactor size. Environ.
temperature shock on the biogas yields from the mesophilic anaerobic diges- Technol. 30, 93e99.
tion of swine manure. Bioresour. Technol. 99, 1e6. Martín, A., Borja, R., Banks, C.J., 1994. Kinetic model for substrate utilization and
Chen, Y., 1983. Kinetic analysis of anaerobic digestion of pig manure and its design methane production during the anaerobic digestion of olive mill wastewater
implications. Agric. Wastes 8, 65e81. and condensation water waste. J. Chem. Technol. Biot. 60, 7e16.
Deng, L., Chen, Z., Yang, H., Zhu, J., Liu, Y., Long, Y., Zheng, D., 2012. Biogas Masse , D.I., Droste, R.L., 2000. Comprehensive model of anaerobic digestion of
fermentation of swine slurry based on the separation of concentrated liquid swine manure slurry in a sequencing batch reactor. Water Res. 34, 3087e3106.
and low content liquid. Biomass Bioenerg. 45, 187e194. Metcalf, E.I., 2003. Wastewater Engineering, Treatment and Reuse. McGraw-Hill,
Deng, L., Yang, H., Liu, G., Zheng, D., Chen, Z., Liu, Y., Pu, X., Song, L., Wang, Z., Lei, Y., New York.
2014. Kinetics of temperature effects and its significance to the heating strategy Møller, H.B., Sommer, S.G., Ahring, B.K., 2004. Methane productivity of manure,
for anaerobic digestion of swine wastewater. Appl. Energ 134, 349e355. straw and solid fractions of manure. Biomass Bioenerg. 26, 485e495.
Donoso-Bravo, A., Mailier, J., Martin, C., Rodríguez, J., Aceves-Lara, C.A., Momoh, O.Y., Anyata, B., Saroj, D., 2013. Development of simplified anaerobic
Wouwer, A.V., 2011. Model selection, identification and validation in anaerobic digestion models (SADM’s) for studying anaerobic biodegradability and kinetics
digestion: a review. Water Res. 45, 5347e5364. of complex biomass. Biochem. Eng. J. 79, 84e93.
Ekama, G.A., 2009. Using bioprocess stoichiometry to build a plant-wide mass Pandian, M., Huu-Hao, N., Pazhaniappan, S., 2011. Substrate removal kinetics of an
balance based steady-state WWTP model. Water Res. 43, 2101e2120. anaerobic hybrid reactor treating pharmaceutical wastewater. J. Water Sust. 1,
Faisal, M., Unno, H., 2001. Kinetic analysis of palm oil mill wastewater treatment by 301e312.
a modified anaerobic baffled reactor. Biochem. Eng. J. 9, 25e31. Panichnumsin, P., Nopharatana, A., Ahring, B., Chaiprasert, P., 2010. Production of
Ferna ndez-Rodríguez, J., Pe rez, M., Romero, L.I., 2013. Comparison of mesophilic methane by co-digestion of cassava pulp with various concentrations of pig
and thermophilic dry anaerobic digestion of OFMSW: kinetic analysis. Chem. manure. Biomass Bioenerg. 34, 1117e1124.
Eng. J. 232, 59e64. Parker, W.J., 2005. Application of the ADM1 model to advanced anaerobic digestion.
Fongsatitkul, P., Elefsiniotis, P., Wareham, D.G., 2012. Two-phase anaerobic diges- Bioresour. Technol. 96, 1832e1842.
tion of the organic fraction of municipal solid waste: estimation of methane Pham, C.H., Triolo, J.M., Sommer, S.G., 2014. Predicting methane production in
production. Waste Manage. Res. 30, 720e726. simple and unheated biogas digesters at low temperatures. Appl. Energ 136,
Girault, R., Rousseau, P., Steyer, J.P., Bernet, N., Beline, F., 2011. Combination of batch 1e6.
experiments with continuous reactor data for ADM1 calibration: application to Ramakant, Satyanarayan, S., Kaul, S., 2002. Kinetics of an anaerobic moving bed
anaerobic digestion of pig slurry. Water Sci. Tech. 63, 2575e2582. reactor system treating synthetic milk wastewater. J. Environ. Sci. Health., Part A
Hill, D., 1982. A comprehensive dynamic model for animal waste methanogenesis. 37, 1737e1755.
T. ASAE 25, 1374e1380. Spriet, J., 1985. Structure characterization: an overview. In: Identification and Sys-
Kafle, G.K., Kim, S.H., 2013. Anaerobic treatment of apple waste with swine manure tem Parameter Estimation (Ed.), Proceedings of the Seventh IFAC/IFORS Sym-
for biogas production: batch and continuous operation. Appl. Energ 103, 61e72. posium. Pergamon Press, Oxford, UK, pp. 749e756.

Kuşçu, O.S., Sponza, D.T., 2009. Kinetics of para-nitrophenol and chemical oxygen Weiland, P., 2010. Biogas production: current state and perspectives. Appl. Micro-
demand removal from synthetic wastewater in an anaerobic migrating blanket biol. Biot. 85, 849e860.
reactor. J. Hazard. Mater 161, 787e799. Xiao, L., Wei, W., Shi, Y., Lei, Z., Gao, X., Wei, Q., Zhou, Y., 2012. Pilot-scale anaerobic
Liu, C.-f., Yuan, X.-z., Zeng, G.-m., Li, W.-w., Li, J., 2008. Prediction of methane yield co-digestion of municipal biomass waste and waste activated sludge in China:
at optimum pH for anaerobic digestion of organic fraction of municipal solid effect of organic loading rate. Waste Manage 32, 2056e2060.
waste. Bioresour. Technol. 99, 882e888. Yang, D., Deng, L.W., Zheng, D., Liu, G.J., Yang, H.N., Wang, L., 2015. Separation of
Lo, K., Liao, P., Gao, Y., 1994. Anaerobic treatment of swine wastewater using hybrid swine wastewater into solid fraction, concentrated slurry and dilute liquid and
UASB reactors. Bioresour. Technol. 47, 153e157. its influence on biogas production. Fuel 144, 237e243.
Ma, J., Yu, L., Frear, C., Zhao, Q., Li, X., Chen, S., 2013. Kinetics of psychrophilic Yu, H., Wilson, F., Tay, J.-H., 1998. Kinetic analysis of an anaerobic filter treating
anaerobic sequencing batch reactor treating flushed dairy manure. Bioresour. soybean wastewater. Water Res. 32, 3341e3352.

You might also like