Modified ADM1 structure for modelling municipal primary sludge hydrolysis

WAT E R R E S E A R C H

42 (2008) 249 – 259

Available at www.sciencedirect.com

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

Modified ADM1 structure for modelling municipal primary sludge hydrolysis H. Yasuia,, R. Goela, Y.Y. Lib, T. Noikec a

Kurita Water Industries, 1-1, Kawada-Gochoyama, Nogi, Tochigi 329-0105, Japan Department of Civil Engineering, Tohoku University, Aza-Aoba, Aramaki, Aoba-ku, Sendai, Miyagi 980-8579, Japan c Advanced Research Institute, Nihon University, 4-2-1, Kudan-kita, Chiyoda-ku, Tokyo 102-0073, Japan b

art i cle info

ab st rac t

Article history:

This study elaborates the rate-limiting steps of particle disintegration/hydrolysis of

Received 21 January 2007

primary sludge using methane production rate (MPR) curves from multiple batch

Received in revised form

experiments. Anaerobic batch degradation of fresh primary sludge showed a complex

7 July 2007

MPR curve marked with two well-defined temporal peaks. The first immediate peak was

Accepted 10 July 2007

associated with the degradation of relatively readily hydrolysable substrates, while the

Available online 12 July 2007

second delayed peak was associated with the degradation of large-sized particles. For

Keywords: ADM1

simulating the second delayed peak, it was necessary to consider a more elaborate particle disintegration/hydrolysis model. Based on the anaerobic respirograms of 17 runs in four datasets and using a substrate characterisation approach similar to activated sludge

ASMs Anaerobic digestion Hydrolysis Primary sludge Sludge composition

models (ASMs), the primary sludge was classified into three biodegradable fractions having different kinetics. These are (1) a hydrolysable substrate (XSettle-I) showing a degradation typical to slowly biodegradable compounds, (2) a substrate fraction (XSettle-II) having a degradation similar to lysis of biomass fraction and (3) a substrate requiring disintegration before hydrolysis (XSettle-III) representing the large-sized particles in primary sludge. Based on these results, modifications in the model structure of anaerobic digestion model no. 1 (ADM1) are proposed to improve the modelling of primary sludge solid degradation in anaerobic digesters. & 2007 Elsevier Ltd. All rights reserved.

1.

Introduction

The generic structure of anaerobic digestion model no.1 (ADM1, Batstone et al., 2002) allows modelling of anaerobic wastewater treatment processes used in industrial wastewater treatment as well as solid degradation processes widely applied in municipal wastewater treatment. The generalised ADM1 structure, however, uses some inevitable simplifications in reactions for solid degradation processes. For example, the solid degradation process in ADM1 has been described using first-order degradation kinetics with respect to a composite state variable of degradable solid concentra-

tions (XC). When ADM1 is applied in combination with activated sludge models (ASMs; Henze et al., 2000), the composite variable of XC is considered to consist of all the degradable solids (XS+XH) originating from primary and secondary sludge. Using a single composite variable of XC in ADM1 to describe the solid degradation reaction is an obvious simplification considering that the nature of primary sludge and secondary sludge is inherently different. While primary sludge is XS rich, secondary sludge is XH rich. Moreover, the physical nature of XS in both sludges is completely different. The XS in primary sludge is mainly composed of settleable particles of variable sizes, while XS in secondary sludge is

Corresponding author. Tel.: +81 280 54 1535; fax: +81 280 57 2633.

E-mail addresses: [email protected] (H. Yasui), [email protected] (R. Goel), [email protected] (Y.Y. Li), [email protected] (T. Noike). 0043-1354/$ - see front matter & 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2007.07.004

250

WA T E R R E S E A R C H

mainly colloidal in nature. These differences in characteristics of degradable solids in primary and secondary sludge are expected to affect their subsequent degradation in anaerobic digestion. For accurately simulating the variable solid mixture arising due to change in primary and secondary sludge ratios at wastewater treatment plants (wwtps), it is necessary to formulate a more comprehensive degradation scheme considering the characteristics of solids in primary and secondary sludge. Considering these needs, Yasui et al. (2006) attempted to experimentally work out the anaerobic degradation kinetics of secondary sludge consisting of degradable organic fractions of XH and XS. Their experimental results suggested that even for anaerobic digestion of activated sludge, it is better to include the degradation reactions for individual components of XH and XS, rather than using a composite variable of XC. For primary sludge degradation, Eastman and Ferguson (1981) have shown through continuous experiments that hydrolysis of solids is the ratelimiting step. They also proposed first-order degradation kinetics with respect to solid concentration. Recent studies describing the degradation of primary sludge, however, point to a much more complex degradation mechanism involving solid disintegration and hydrolysis (Pavlostathis and Giraldo-Gomez, 1991; Vavilin et al., 2001). Sanders et al. (2000) pointed out that surface degradation kinetics was more accurate to express hydrolysis of particulate starch. Vavilin et al. (1996) showed that such kinetic expressions could also adequately reproduce decomposition of heterogeneous solids such as sewage sludge. Although these studies have greatly improved the understanding of the processes involved in primary sludge degradation, the modified kinetics and degradation scheme do not adequately fit into the structured modelling approach of ADM1. Recently there has also been interest in describing the degradation kinetics of organic solids under aerobic conditions (Insel et al., 2002; Ginestet et al., 2002; Dimock and Morgenroth, 2006) mainly for producing soluble substrates in phosphorous and nitrogen removal processes. The results of these aerobic degradation studies also point out to more complex degradation kinetics of primary sludge. Based on the above background, we concentrated on the study of the degradation kinetics of primary sludge under anaerobic conditions with the overall objective of refining the model structure of ADM1 with respect to primary sludge degradation. To achieve these objectives, we focused on (1) characterisation of primary sludge, (2) identification of suitable kinetic expressions for individual component in primary sludge and (3) determination of kinetic coefficients. The outcome of this study will not only help modify the ADM1 structure for primary sludge degradation, but also open up a debate for modifying the ASM structure for including the degradation reactions of settleable fraction that are currently missing.

42 (2008) 249– 259

tests were conducted. To avoid the influence of soluble VFA compounds in anaerobic respirograms, the primary sludge was centrifuged and washed by tap water to remove the soluble organic fractions. Anaerobically digested sludge (seed sludge for the batch tests) was collected from the same plants as that of the samples. Before starting the test, the seed sludge was pre-incubated for about 1–4 days without the addition of substrate at a temperature of 35 1C to remove any remaining readily degradable substrate fraction.

2.2.

Batch respirometric test

2.2.1.

Apparatus for respirometry

Identification of organic fractions and kinetic parameters associated with anaerobic degradation of organic solids was made based on methane production rate (MPR) curves (Guwy, 2004; Yasui et al., 2006). The MPR curves under different experimental conditions were obtained using a batch anaerobic respirometer supplied by Challenging Systems Inc., USA (AER-8). The temperature of the incubation vessel was maintained at 3570.2 1C. A small scrubber consisting of caustic material was set in a gas line to absorb CO2 from the biogas. The data regarding methane gas production were logged at every 2 h interval in the computer.

2.2.2.

2.3.

2.

Material and methods

2.1.

Sludge sample

Fresh primary sludge samples were collected from gravity thickener of two different wwtps during different seasons. The samples were stored at 4 1C for 1–2 days until the batch

Test procedure for anaerobic respirometry

The primary sludge samples were placed in each of the incubation vessels of 1000 mL working volume. Buffer solution (NaHCO3: 872 mg/L; K2HPO4: 80 mg/L; KH2PO4: 80 mg/L; pH ¼ 7.5) having salt concentration comparable to that of the original liquor of anaerobically digested sludge was added to adjust the sludge concentration. After mixing seed sludge, the buffer solution and the primary sludge, the headspace in the incubation vessel was purged with nitrogen gas. The incubation vessels were then sealed and MPRs were measured. Four datasets, containing 17 runs in total, were used for the analysis (Table 1). For collecting respirograms showing a variety of degradation patterns, the tests were carried out under various F/ M ratios ranging from 0.00 to 0.214 (COD/COD). The F/M ratio in the tests was varied by either fixing the substrate concentration while varying the microorganism concentration or vice versa. The former makes it possible to know whether the degradation is affected by seed sludge concentration, while the latter gives the information with respect to the influence of substrate concentration on the degradation pattern. Through preliminary experiments, all the test conditions were adjusted in such a way that no significant VFA accumulation took place during the test period. A blank test was also conducted without addition of primary sludge in order to obtain baseline information about endogenous respiration.

Model calibration procedure

Similar to the procedure of ASMs (Kappeler and Gujer, 1992; Copp et al., 2002), MPR curves were visually analysed to identify regions corresponding to different degradable fractions in primary sludge. The areas of each identified regions were considered to correspond to the concentration of individual degradable fraction. Next, a model structure was developed including the preliminary kinetic expressions for the newly

WAT E R R E S E A R C H

Table 1 – Experimental condition of sludge concentrations and F/M ratios Data sets

Run

Primary sludge (mg COD/L)

Seed sludge (mg COD/L)

F/M ratio (dimensionless)

KM

1 2 3 4 5 B

658 1316 2632 1316 1316 0

12,320 12,320 12,320 9240 6160 12,320

0.053 0.107 0.214 0.142 0.214 0.000

KA

1 2 3 B

558 1116 558 0

28,720 28,720 14,360 28,720

0.019 0.039 0.039 0.000

KD

1 2 3 B

567 1134 567 0

11,200 11,200 5600 11,200

0.051 0.101 0.101 0.000

1 2 B

1959 1313 0

10,400 10,400 10,400

0.188 0.126 0.000

NF

KM: primary sludge ¼ Kitami wwtp on 17/May/2005; digested sludge ¼ pilot-scale tank (HRT ¼ 30 days) fed with secondary sludge only. KA: primary sludge ¼ Kitami wwtp on 23/Aug/2005; digested sludge ¼ full-scale tank (HRT ¼ 30 days) fed with mixture of 60% of primary sludge and 40% of secondary sludge. KD: primary sludge ¼ Kitami wwtp on 5/Dec/2005; digested sludge ¼ full-scale tank (HRT ¼ 30 days) fed with mixture of 60% of primary sludge and 40% of secondary sludge. NF: primary sludge ¼ Nagaoka wwtp in 2/Feb/2002; digested sludge ¼ full-scale tank (HRT ¼ 50 days) fed with mixture of 68% of primary sludge and 32% of secondary sludge.

identified degradable fractions. Finally, the batch experiments were simulated using the developed model structure and rate expression. Calibration of kinetic coefficients was performed until best fits were achieved between the simulated and experimental MPR curves, using a computational process simulator, GPS-Xs (Hydromantis Inc., Canada). To check the validity of the model structure, kinetic rate expressions and kinetic coefficients, the source digester in the wwtp was simulated to determine whether the simulated concentration of the state variables in seed sludge matched with the initial concentration of state variables used in simulating the batch experiments. In case of mismatch, the calibration procedure was repeated until the predicted state of digester matched well with the initial state variable values of the batch tests.

3.

Results

3.1.

Characteristics of MPR curves

Fig. 1 shows all the 17 experimental respirograms belonging to datasets of KM, KA, KD and NF. All the batch experiments with primary sludge resulted in unique double-peak MPR

42 (2008) 249 – 259

251

curves. The first immediate gas production peak was followed by a distinct second peak after 1.5–1.8 days. After the second peak, the MPR showed a gradual decline in a consistent manner and reached the baseline due to depletion of substrate. The first peak was easily attributed to the degradation of readily degradable fraction in primary sludge. As for the second delayed peak, it was necessary to ascertain whether it was due to the large-sized particulate substrate requiring particle disintegration (size reduction), poor sludge acclimation or bacterial inhibition. As discussed later, factors of poor sludge acclimation/bacterial inhibition were ruled out and the peak was attributed to the particulate substrate requiring particle disintegration. Along with the measured methane production rate, the graphs shown in Fig. 1 also contain simulated curves (to be discussed later) in solid line.

3.2.

Fractionation of primary sludge substrate

Based on the experimental results, a typical respirogram (Fig. 2) for primary sludge degradation was prepared. Through visual inspection, the respirogram was broadly divided into three regions representing three distinctive substrate fractions. The substrate fraction (XSettle-I) corresponding to region I was found to degrade within 1 day of incubation. The second fraction (XSettle-II) represented by region II in the respirogram was observed to slowly degrade throughout the experimental period. This fraction was particularly visible in the datasets of KA and NF. The third fraction (XSettle-III) represented by region III indicated delayed degradation kinetics, and as described earlier was mainly attributed to large size particulate substrate.

3.3. Identification of kinetic expression for primary sludge fractions Datasets of KM were used to identify the possible kinetic expressions for the degradation of the individual substrate fractions. Experimentally determined MPR curves for KM are as shown in Fig. 3. For the first fraction (XSettle-I in region I), increase in MPR was observed by either increasing the seed sludge concentration (cf. left graph) or primary sludge concentration (cf. right graph). This suggests that the degradation rate of XSettle-I is a function of both the concentrations of microoragnisms in seed sludge and organics in the primary sludge. Based on this observation, Contois or Monod reaction kinetics was thought to be suitable for this fraction. With respect to the second fraction (XSettle-II in region II), the degradation rate was not influenced by the seed sludge concentration and depended only on the primary sludge concentration. In fact, same boundary line for regions I and II could be drawn when primary sludge was fed in equal amount (‘‘B’’ in left graph). In addition, the intersection on the Y-axis of the line was proportional to the initial primary sludge concentration fed (cf. ‘‘A’’, ‘‘B’’ and ‘‘C’’ in right graph). This suggests that first-order kinetics with respect to XSettle-II concentration is suitable for this fraction. For the third fraction (XSettle-III in region III), almost similar MPR curve patterns were observed when primary sludge was fed in equal amount (cf. left graph). As can be seen from the

252

WA T E R R E S E A R C H

MPR (NL/L/day)

0.6 0.5

F/M = 658/12,320 = 0.053

F/M KM-3 = 2,632/12,320 = 0.214

F/M KM-2 = 1,316/12,320 = 0.107

KM-1

42 (2008) 249– 259

F/M KM-5 = 1,316/6,160 = 0.214

F/M KM-4 = 1,316/9,240 = 0.142

F/M = 0/12,320 = 0.000

KM-B

0.4 0.3 0.2 0.1 0.0 0

MPR (NL/L/day)

0.6

1

2

3

F/M = 558/28,720 = 0.019

0.5

40

1

2

3

40

F/M KA-2 = 1,116/28,720 = 0.039

KA-1

1

2 3 4 0 1 2 Incubation time (day)

F/M = 558/14,360 = 0.039

F/M = 0/28,720 = 0.000

KA-3

3

40

1

2

3

40

1

2

3

4

KA-B

0.4 0.3 0.2 0.1 0.0

MPR (NL/L/day)

0.4 0.3

1

2

3

F/M = 567/11,200 = 0.051

40

1

2 3 40 1 2 Incubation time (day)

F/M KD-2 = 1,134/1,200 = 0.101

KD-1

F/M = 567/5,600 = 0.101

3

4 0

1

2

F/M = 0/11,200 = 0.000

KD-3

3

KD-B

0.2 0.1 0.0 0 0.4

MPR (NL/L/day)

4

1

2

3

40

F/M NF-1 = 1,959/10,400 = 0.188

0.3

1

2 3 40 1 2 Incubation time (day)

F/M NF-2 = 1,313/10,400 = 0.126

F/M = 0/11,200 = 0.000

3

4 0

1

2

3

15 10 5 0 4

Acetate (mg COD/L)

0

NF-B

0.2 0.1 0.0 0

1

2

3

40 1 2 3 40 Incubation time (day)

1

2

3

4

Fig. 1 – Chronological plot of methane gas production rate in the batch tests.  : measured acetate; thin line: simulated acetate.

MPR curves in the left graph, the degradation of this fraction was not governed by seed sludge concentration. Thus, seed sludge concentration-independent kinetics was thought to be suitable for this fraction of primary sludge.

4.

Discussion

4.1.

Model formulation

Based on the experimental results and the theoretical considerations, a modified ADM1 structure was developed

J:

Measured MPR; bold line: simulated MPR;

by replacing the composite state variable of XC with three different solid fractions of XSettle-I, XSettle-II and XSettle-III. While it was straightforward to include the degradation reactions for XSettle-I and XSettle-II solid fractions, modelling the time lag for XSettle-III fraction required further considerations. As the observed time lag in the degradation of large size particle was considered to be due to particle disintegration process, a particle break-up (PBM) model of Fig. 4 was considered for its simulation. The PBM of Fig. 4 was, however, difficult to implement as it requires determination of kinetic coefficients at each individual stage. To overcome this difficulty, a simplified particle break-up model (sPBM) was considered

assuming that the specific disintegration rate at each stage is the same and only particles of the last stage lead to hydrolysed products with specific rate comparable to that of the disintegration. The structure of this sPBM is as shown in Fig. 5. The advantage of this model is that only two parameters of number of disintegration stages (n) and disintegration/hydrolysis rate (k) are required for its application. Based on the above considerations, a modified ADM1

model was formulated which in Peterson matrix form is shown in Table 2. As composition and degradation of soluble component was not the main attention of this study, the ADM1 model structure was simplified with respect to soluble components. Organic soluble substrate fractions were expressed through two composite soluble components of SF ( ¼ Ssu+Saa+Sfa+Sva+Sbu+Spro) and SA ( ¼ Sac+Sh2). The modelling for active biomass in anaerobic digestion was also reduced to two composite state variables of XAcid (acetate and hydrogen producer ¼ Xsu+Xaa+Xfa+XC4+Xpro) and XM (methane producer ¼ Xac+Xh2). This simplification in the description of soluble substrate consumption did not affect the simulation in the present study as degradation was mainly governed by the rate-limiting disintegration/ hydrolysis of solids. In cases, where it is necessary to include soluble substrate degradation, the model structure of ADM1 for soluble substrate can be combined with the modified particulate degradation scheme proposed here without much difficulty. The above model structure was used to simulate the batch experiments conducted under various conditions. The calibrated composition of seed sludge is as shown in Table 3. The variation of the composition was due to the operating condition at each wwtp and pre-incubation time,

0.6

MPR (NL/L/day)

0.5 0.4

Baseline

0.3

Region III

Region I

0.2

(XSettle-I)

0.1

(XSettle-III)

Region II (XSettle-II)

0.0 0

1 2 Incubation time (day)

253

42 (2008) 249 – 259

WAT E R R E S E A R C H

3

Fig. 2 – Classification of three biodegradable fractions in the primary sludge (from KM-3 respirogram).

F/M ratio

0.6

0.5 MPR (NL/L/day)

0.5 MPR (NL/L/day)

0.6

: KM-2 = 1,316/12,320 = 0.107 : KM-4 = 1,316/9,240 = 0.142 : KM-5 = 1,316/6,160 = 0.214

0.4 0.3 0.2 0.1 B 0.0 0

1

2

3

F/M ratio : KM-1 = 658/12,320 = 0.053 : KM-2 = 1,316/12,320 = 0.107 : KM-3 = 2,632/12,320 = 0.214

0.4 0.3 0.2 0.1 A B C 0.0 0

1

Incubation time (day)

2

3

Incubation time (day)

Fig. 3 – MPR pattern in dataset of KM (baseline subtracted) A: boundary line of regions I and II for KM-3, B: for KM2, KM-4 and KM-5, C: for KM-1.

Production of methane RCH4 Production of substrate for methanogens RSA

XSettle-III in Primary sludge SF R1

SF R2

XSettle-III (1)

SF

R3

XSettle-III (2)

r1

SF

r3

…

rj-1

XSettle-III (j)

rj

Rn

XSettle-III (n-1)

rn-2

Production of methanogen Production of acidifier

SF

Rn-1

Rj

XSettle-III (3)

r2

SF

SA

SCH4

⇑ Hydrolysis over time (Production of substrate for acidifier)

XSettle-III (n)

rn-1

⇒Disintegration over time Fig. 4 – Illustration of sequential degradation behaviour of particulate substrate to produce methane.

254

WA T E R R E S E A R C H

42 (2008) 249– 259

Production of methane RCH

SA

Production of substrate for methanogens RSA

XSettle-III in Primary sludge

SCH

4

4

Production of methanogen Production of acidifier

SF Rn = KXSettle-III (n) = kXSettle-III (n) XSettle-III (1)

XSettle-III (2)

XSettle-III (3)

……

XSettle-III (j)

XSettle-III (n-1)

⇑ Hydrolysis over time (Production of substrate for acidifier)

XSettle-III (n)

r2 = k XSettle-III (2) rj-1 = k XSettle-III (j-1) rj = k XSettle-III ( j ) rn-1 = k XSettle-III (n-1) ⇒ Disintegration over time

r1 = k XSettle-III (1)

Fig. 5 – Structure of simplified particle break-up model.

resulting in slight degradation of XH. It appeared that most of the organic solids in seed sludge was inert. The composition of organic fractions in the primary sludge is summarised in Table 4. The composition significantly varied depending on wwtp and seasons. In fact, the XI fraction that was calculated by subtracting sum of biodegradable compounds in regions I, II and III from initial COD concentration of the primary sludge ranged from 5% to 19%. In spite of the variation, individual theoretical ultimate digestion efficiency of the primary sludge as estimated by Eq. (1) was calculated to be an almost comparable value of ca. 80%.

4.2.

Analysis for MPR

All the MPR curves for primary sludge degradation produced two distinct temporal peaks. Even though the primary sludge samples were centrifuged and washed with tap water to remove any soluble substrate, a high initial methane production rate resulted in the first peak. This high methane production rate was most probably the result of degradation of readily hydrolysable colloidal organic fraction (XSettle-I) loosely attached to larger size particles. To model this initial methane production rate, a relatively higher value of specific

Theoretical ultimate digestion efficiency ffi1

ðf I YAcid XSettleI Þ þ ðf I Y Acid ð1  f I ÞXSettleII þ f I XSettleII Þ þ ðf I Y Acid XSettleIII Þ þ XI . ðXSettleI þ XSettleII þ XSettleIII þ XI Þ

The values of kinetic and stoichiometric coefficient obtained from these simulation studies are summarised in Table 5. Comparable values for maximum specific growth rate of XAcid and XM could be used in all of the simulations. The values of maximum specific hydrolysis rate (khXSettle-I) were in the range of 4.0–5.0 day1. The absolute value of khXSettle-I was slightly higher than that that reported in ASM (ca. 3.0 day1 at 20 1C) under aerobic conditions. As compared with the decay rates of acidifier and methanogens, decay rate of XH in seed sludge had significant impact on calibrating the baseline of control runs. Among the specific degradation rates of XH, XAcid and XSettle-II, the degradation rate of XSettle-II was remarkably higher. Although the reason for observed higher degradation rate of XSettle-II is not clear, it may be that the XSettle-II fraction is due to rapid-growing bacteria in sewer networks with short hydraulic retention time. The specific decay rate of these fastgrowing bacteria may be higher than the decay rate of slowgrowing bacteria normally grown at longer retention time in the aeration tank/anaerobic digestion tank (Jo¨nsson et al., 2006). The number of stages (n) and the value of specific first-order disintegration/hydrolysis coefficient k (or K) to simulate the MPR from XSettle-III showed a large variation. The value of n ranged from 6 to 50, while that of k (or K) ranged from 4 to 30 day1. The large variation in the parameter values probably reflects the differences in the compositions of primary sludge at different plants.

ð1Þ

hydrolysis rate of 4.0–5.0 day1 was required. This value of specific hydrolysis rate was obtained considering that only acidifier (XAcid) and heterotrophic (XH) biomass participate in hydrolysis of XSettle-I. However, if the possibility of the presence of active biomass in primary sludge is considered, the absolute value of the specific hydrolysis rate will become lower than that estimated here. Considering the degradation characteristics of XSettle-II, it is quite likely that this fraction is active biomass generated in the sewer network and may contribute towards the hydrolysis of other primary sludge fractions. However, as no direct evidence could be obtained to confirm this, further experimental work is required to verify this hypothesis. As for the second delayed peak, it was necessary to ascertain whether it was due to (1) the time required for bacterial acclimation, (2) bacterial inhibition due to VFA accumulation or (3) due to large-sized particulate substrates requiring particle disintegration (size reduction) before degradation. Time required for bacterial acclimation may produce a delayed peak similar to that observed in this study. The seed sludge used in the experimental datasets of KA, KD and NF was obtained from the digester treating both the primary and secondary sludge. Therefore, it was assumed that the seed sludge microbe population in these datasets was well adapted to the primary sludge. As for the dataset of KM, the seed sludge was acclimated to only activated sludge and thus was most prone to the acclimation effect. As compared with the datasets of KA, KD and NF, the second

Table 2 – Model structure for anaerobic degradation of primary sludge Rate no. (1)

Process

XSettleI

Hydrolysis of XSettleI in primary sludge

1

XSettleII

XSettleIII(1)

XSettleIII(2)

.. XSettleIII(j)..

XSettleIII

(n)

SF

SA

XH

XAcid

XM

XI

SCH4

Rate expression

khX

1

 SettleI

aXSettleI =

KXXSettleI  XAcid þ  aXSettleI XAcid þ Zfe   ð1  aÞ khX SettleI XSettleI =KXXSettleI   XH þ ð1  aÞXSettleI XH

(2) (3) (4) (5) (6)

(7) (8)

(9)

Growth of acidifier Growth of methanogen Decay of acidifier Decay of methanogen Degradation of XSettleII in primary sludge Decay of heterotroph Disintegration of XSettleIII in the 1st stage Disintegration of XSettleIII in the 2nd stage

1/YAcid

1/YAcid1

1 1

1/YM 1fI

1

1fI

1

1fI

1

1fI 1

mAcid ðSF =KSF þ SF ÞXAcid

1

1/YM1

mM ðSA =KSA þ SA ÞXM

fI

bAcidXAcid

fI

bMXM

fI

bXSettleII XSettleII

fI

bHXH

1

kXSettle-III(1)

1

kXSettle-III(2)

Table 2 (continued ) Rate no.

Process

XSettleI

XSettleII

XSettleIII(1)

XSettleIII(2)

(j+6)

Disintegration of XSettleIII in the jth stage (j+7) Disintegration of XSettleIII in the j+1th stage (n+6) Disintegration of XSettleIII in the nth stage (n+7) Hydrolysis of XSettleIII in the nth stage

.. XSettleIII(j)..

XSettleIII

(n)

SF

SA

XH

XAcid

XM

XI

kXSettle-III(j)

y1y

kXSettle-III(j+1)

1

Slowly hydrolysable compounds in primary sludge in the 1st stage

Slowly hydrolysable compounds in primary sludge in the 2nd stage

Rate expression

y1y

1

Slowly Slowly hydrolysable hydrolysable compounds compounds (or XH?) in in primary sludge primary sludge

Slowly hydrolysable compounds in primary sludge in the jth stage

Slowly hydrolysable compounds in primary sludge in the nth stage

kXSettle-III(n)

1

Substrate Substrate Heterotroph Acidifier Methanogen Inert Methane for for (aerobe) solid acidifier methanogen (non growable)

Remarks: XH and XSettleII are assumed to be non-growable in anaerobic digestion processes as this study has no experimental information for their growth possibility. XSettleI is proportioned to the individual biomass of XAcid and XH in rate (1); XSettleI degraded by XAcid ¼ a XSettleI, XSettleI degraded by XH ¼ (1a) XSettleI, where a ¼ XAcid/(XAcid+XH). State variable concentration: mgCOD/L. Rate: mgCOD/L/day.

(1) (2) (3) (4)

SCH4

KXSettle-III(n)

WAT E R R E S E A R C H

Table 3 – Composition of seed sludge State variable in seed sludge

Heterotroph (aerobe) from secondary sludge, XH Acidifier, XAcid ( ¼ Xsu+Xaa+Xfa+Xc4+Xpro) Methanogen, XM ( ¼ Xac+Xh2) Slowly hydrolysable compounds, XSettleI Slowly hydrolysable compounds, XSettleII Slowly hydrolysable compounds, XSettleIII Inert solid, XI

peak appeared to be more prominent in the KM dataset, leading to further speculation about the acclimation effect in this dataset. Further observations revealed that even though the sharpness of the peak was different, the appearance time of the peak was almost similar in all the datasets. If bacterial acclimation had affected the KM dataset, we would expect that the location of peak would shift further down the time scale. As no time shift was observed, bacterial acclimation was ruled out as the reason for the development of second peak. To check the possibility of bacterial inhibition due to VFA accumulation, VFA and soluble COD concentrations were monitored in some of the batch experiments. The measured concentration of soluble components in KD-2 test is as shown in the Fig. 6. The measured data suggested that the only detectable VFAs were acetate and butyrate. The observed maximum concentration in the test was close to the KS values reported in ADM1 and were much below the inhibitory concentration (Batstone et al., 2002). In addition, the measured COD concentration almost remained constant (600–700 mg/L) during the test, suggesting that no significant accumulation of soluble COD took place. The initial high concentration of soluble COD was mainly due to the soluble inert component in the seed sludge. While the measured VFA concentration was low, a distinct second peak could still be observed in the KD-2 experiment. Thus, it was concluded that VFA accumulation leading to bacterial inhibition did not have anything to do with the production of the second peak. Based on these considerations, the development of the second peak in the MPR curve was mainly attributed to largesized particulate substrate degrading through the particle disintegration (size reduction) mechanism and the primary

Dataset KM (%)

KA (%)

KD (%)

NF (%)

6.5

8.2

4.8

10.3

1.3

2.4

2.2

2.3

6.5 0.0

6.0 0.0

5.9 0.0

6.3 0.0

0.0

0.8

0.2

0.3

0.0

0.0

0.0

0.0

85.7

82.6

86.9

80.8

Table 4 – Composition of primary sludge State variable in primary sludge

Slowly hydrolysable compounds, XSettleI Slowly hydrolysable compounds, XSettleII Slowly hydrolysable compounds, XSettleIII Inert solid, XI

Data set KM (%)

KA (%)

KD (%)

NF (%)

21

20

11

8

30

32

49

18

30

43

30

56

19

5

10

18

257

42 (2008) 249 – 259

Table 5 – List of state variables, kinetic and stoichiometric constants Item

Kinetics Maximum specific growth rate of acidifiera Half-saturation coefficient of substrate for acidifiera Maximum specific growth rate of methanogen Half-saturation coefficient of substrate for methanogen Maximum specific hydrolysis rate of XSettleI Half-saturation coefficient for hydrolysis of XSettleI Specific decay rate of XH Specific decay rate of acidifiera Specific decay rate of methanogena Specific decay rate of XSettleII Specific disintegration/hydrolysis rate of XSettleIII Number of disintegration stage of XSettleIII Stoichiometry Production of acidifiera Production of methanogena Production of inert solidb Anaerobic hydrolysis reduction factor of XHc a b c

Batstone et al. (2002). Henze et al. (2000). Yasui et al. (2006).

Symbol

Unit

Dataset KM

KA

KD

NF

macid KSF mM KSA khXSettleI KXSettleI bH bacid bM bXSettleII k n

day1 mg COD/L day1 mg COD/L day1 – day1 day1 day1 day1 day1 stage

4.0 20 0.10 10 5.0 0.06 0.21 0.04 0.001 0.55 31.3 50

’ ’ 0.08 30 4.0 0.08 0.18 ’ ’ 0.60 10.6 18

’ ’ 0.10 10 5.0 0.15 0.18 ’ ’ 0.35 22.2 40

’ ’ 0.10 10 4.0 0.25 0.16 ’ ’ 0.70 3.8 6

YAcid YM fI Zfe

COD/COD COD/COD COD/COD –

0.08 0.04 0.20 0.07

’ ’ ’ ’

’ ’ ’ ’

’ ’ ’ ’

258

WA T E R R E S E A R C H

800

40

42 (2008) 249– 259

RCH (TPeak) 4

iso-Butyrate 400

20

Acetate 200

10

0

0 0

1

2

3

Incubation time (day) Fig. 6 – Soluble compounds in KD-2. }: Lactate; &: formate; J: acetate, m: i-butyrate: n: iso-butyrate;  : propionate; +: soluble COD.

sludge solids were characterised into three different fractions of XSettle-I, XSettle-II and XSettle-III. As shown in Table 4, XSettle-I, XSettle-II and XSettle-III ranged between 8% and21%, 18% and 49%, and 30% and 56%, respectively, in different datasets. While there was some seasonal change in the primary sludge composition for the datasets from same treatment plant (KM, KA and KD), the composition of sludge from Nagaoka treatment plant (NF) was significantly different.

4.3.

Properties of the simplified particle break-up model

The sPBM used for describing the degradation of XSettle-III requires a number of particle break-up stages (n) and disintegration/hydrolysis coefficients (k). The advantage of sPBM is that the model can be analytically solved for the MPR (RCH4). The sPBM equations and analytical solution for the MPR rate is as shown in Eq. (2). The simulated MPR curve based on Eq. (2) is as shown in Fig. 7. The above analysis produces a MPR curve similar to the one observed experimentally (Fig. 1). To obtain the values of n and k, the equation for the MPR rate is manipulated to obtain time (TPeak) of maximum MPR and time of two inflection points (t ¼ TIR, TIL) by equating RCH40 (t) and RCH400 (t) to zero, respectively. These solutions are shown in Eqs. (3) and (4). Both the coefficients of k and n affect the peak height and the delay in the curve. Smaller values of k enhance the delay in appearance of the peak and broaden the curve, making the peak height lower. On the other hand, smaller n makes the peak height lower while accelerating the appearance of the peak. By knowing the time (TPeak) of maximum MPR and the time of one of the inflection point (TIR or TIL), the values of k and n can be uniquely and easily determined. Physically interpreting, the value of n shall be affected by the average particle size of inlet solids, while k shall represent the average degradation rate of

Area = Initial XSettle-III*(1-YAcid)*(1-YM)

=0

4(t)

1 RCH (t) 4

RCH

600

30

Soluble COD concentration (mg/L)

VFA concentration (mgCOD/L)

Sol. COD

TPeak

Time (t)

Fig. 7 – Mathematical properties of sPBM.

particle, which can be limited, by the rate of either disintegration or hydrolysis. 8 dX SettleIIIð1Þ > ffi kXSettleIIIð1Þ > > dt > > > > > dXSettleIIIð2Þ > > ffi kXSettleIIIð1Þ  kXSettleIIIð2Þ > dt > > > > > > >... > > dX > > > SettleIIIðjÞ ffi kXSettleIIIðj1Þ  kXSettleIIIðjÞ > > dt > < ... > > > dXSettleIIðnÞ > ffi kXSettleIIIðn1Þ  KXSettleIIIðnÞ > > dt > > > > > ffi kX SettleIIIðn1Þ  kXSettleIIIðnÞ > > > > > dSF > > ffi kXSettleIIIðnÞ  1Y1 RSA ffi 0 > > dt Acid > > > > > > : dSdtA ffi 1Y1 RSA  ð1Y 1Þð1Y Þ RCH ffi 0 Acid

Acid

M

4

‘RCH4 ðtÞ ¼ ð1  Y Acid Þð1  YM ÞkXSettleIIIðnÞ ðtÞ XSettleIIIðnÞ ðtÞ ¼ XSettleIIIð1Þ

1 ðktÞn1 expðktÞ n!

8 pffiffiffiffiffiffiffiffiffiffi ðn1Þ ðn1Þ > > > TIL ¼ k > < TPeak ¼ ðn1Þ k > pffiffiffiffiffiffiffiffiffiffi > > > : TIR ¼ ðn1Þþ ðn1Þ k

ð2Þ

(3)

RCH4 ðTPeak Þ ¼ ð1  Y Acid Þð1  Y M ÞkXSettleIIIð1Þ

1 ðkTPeak Þn1 expðkTPeak Þ n! ð4Þ

Although application of sPBM to the experimental datasets resulted in good fit to the experimental data, the number of disintegration stages (n) varied from 6 to 50. As the value of n corresponds to the number of intermediate particulate states that are required to set in the model, the large and variable nature of n is seen as the major limitation in the implementation of this model structure in ADM1. The best approach to implement a PBM in ADM1 would probably be to fix the number of particle break-up stages and calibrate the MPR rate response by using the parameters of disintegration rate (k) and hydrolysis rate (K). This approach was tested by fixing the number of disintegration stages to three and conducting simulation studies using the experimental datasets. With this approach, it was not possible to simulate the experimental datasets (data not shown). The major limitation of this approach was that it could reproduce the datasets having

WAT E R R E S E A R C H

delay and a narrow peak. When it is attempted to achieve delay in the second peak with three disintegration stages, the values of disintegration coefficient (k) become very low, making it the rate-limiting step, resulting in a broad peak. Thus, in case of using the PBM, it is necessary that the number of stages be varied depending on the nature of solids in primary sludge. Another approach to produce delay in the biodegradation curve depending on an additional state variable of surface area has been proposed by Dimock and Morgenroth (2006) for aerobic degradation of hard-boiled egg white particles. Extending the modelling approach of Dimock and Morgenroth (2006) to the present datasets of primary sludge degradation is very appealing and is reserved for future study. Besides improving the simulation of municipal digesters under dynamic primary sludge loading, the modified ADM1 can find significant applicability in modelling prefermenters and optimising intermittent solid feeding operations. The new insights into the degradation nature of primary sludge can also be used to assess the strategies to increase the degradation rate of particles by physical/chemical/thermal treatment of primary sludge.

5.

Conclusion

Degradation mechanisms and kinetics of municipal primary sludge were studied using batch anaerobic respirometry experiments producing methane production rate (MPR) curves. The most important and novel conclusion of this study is that the degradable organics in primary sludge contain at least three distinct kinetically differentiable fractions of XSettle-I, XSettle-II and XSettle-III. For simulating the degradation of XSettle-III, which showed delayed degradation kinetics, a simplified break-up model (sPBM) was used. Analytical solutions were obtained to directly determine the parameter values of sPBM using the experimental data of the MPR rate. A modified anaerobic digestion model 1 (ADM1) structure including the degradation processes for the three identified solid fractions was developed and used to simulate all the experimental datasets with reasonable accuracy. The variable number of disintegration stages required in a modified model structure is seen as the major limitation in implementation, necessitating assessment of new modelling approaches in future.

Acknowledgement This study was conducted by Kurita Water Industries and Tohoku University as the research activities in NEDO (New

42 (2008) 249 – 259

259

Energy and Industrial Technology Development Organization, Japan). R E F E R E N C E S

Batstone, D.J., Keller, J., Angelidaki, I., Kalyuzhnyi, S.V., Pavlostathis, S.G., Rozzi, A., Sanders, W.T.M., Siegrist, H., Vavilin, V.A., 2002. Anaerobic Digestion Model No.1. (ADM1). IWA Scientific and Technical Report No. 13, IWA, ISBN:1-900222-78-7. Copp, J.B., Spanjers, H., Vanrolleghem, P.A., 2002. Respirometry in control of the activated sludge process: benchmarking control strategies. IWA Scientific and Technical Report No. 11, IWA, ISBN:1-900222-51-5. Dimock, R., Morgenroth, E., 2006. The influence of particle Size on microbial hydrolysis of protein particles in activated sludge. Water Res. 40, 2064–2074. Eastman, J.A., Ferguson, J.F., 1981. Solubilization of particulate organic carbon during the acid phase of anaerobic digestion. J. WPCF 53, 352–366. Ginestet, P., Maisonnier, A., Spe´randio, M., 2002. Wastewater COD characterization: biodegradability of physico-chemical fractions. Water Sci. Technol. 45 (6), 89–97. Guwy, A.J., 2004. Equipment used for testing anaerobic biodegradability and activity. Rev. Environ. Sci. Biotechnol. 3 (2), 131–139. Henze, M., Gujer, W., Mino, T., van Loosdrecht, M.C.M., 2000. Activated Sludge Models ASM1, ASM2, ASM2D, ASM3. IWA Scientific and Technical Report No. 9, IWA, ISBN:1-900222-24-8. ¨ ., Orhon, D., Vanrolleghem, P.A., Henze, M., Insel, G., Karahan, O 2002. Important limitations in the modelling of activated sludge: biased calibration of hydrolysis process. Water Sci. Technol. 45 (12), 23–36. Jo¨nsson, K., Jansen, J.la C., 2006. Hydrolysis of return sludge for production os easily biodegradable carbon: Effect of pretreatment, sludge age and temperature. Water Sci. Technol. 53 (12), 47–54. Kappeler, J., Gujer, W., 1992. Estimation of kineic parameters of heterotorophic biomass under aerobic conditions and characterization of wastewater for activated sludge modelling. Water Sci. Technol. 25 (6), 125–139. Pavlostathis, S.G., Giraldo-Gomez, E., 1991. Kinetics of anaerobic treatment: a critical review. Critic. Rev. Environ. Control 21 (5/6), 411–490. Sanders, W.T.M., Geerink, M., Zeeman, G., Lettinga, G., 2000. Anaerobic hydrolysis kinetics of particulate substrates. Water Sci. Technol 41 (3), 17–24. Vavilin, V.A., Rytov, S.V., Lokshina, L.Ya., 1996. A description of hydrolysis kinetics in anaerobic degradation of particulate organic matter. Bioresour. Technol. 56, 229–237. Vavilin, V.A., Rytov, S.V., Lokshina, L.Ya., Rintala, J.A., Lyberatos, G., 2001. Simplified hydrolysis models for the optimal design of two-stage anaerobic digestion. Water Res. Yasui, H., Sugimoto, M., Komatsu, K., Goel, R., Li, Y.Y., Noike, T., 2006. An approach for substrate mapping between ASM and ADM1 for sludge digestion. Water Sci. Technol 54 (4), 83–92.