Modeling of two-phase anaerobic process treating traditional Chinese medicine wastewater with the IWA Anaerobic Digestion Model No. 1

Bioresource Technology 100 (2009) 4623–4631

Contents lists available at ScienceDirect

Bioresource Technology journal homepage: www.elsevier.com/locate/biortech

Modeling of two-phase anaerobic process treating traditional Chinese medicine wastewater with the IWA Anaerobic Digestion Model No. 1 Zhaobo Chen a,b,3, Dongxue Hu b,1, Zhenpeng Zhang c,2, Nanqi Ren a,*, Haibo Zhu a,3 a

State Key Laboratory of Urban Water Resource and Environment, Research Center of Environmental Biotechnology 2614#, School of Municipal and Environmental Engineering, Harbin Institute of Technology, 202 Haihe Road, Harbin 150090, China b School of Materials Science and Chemical Engineering, Harbin Engineering University, 145 Nantong Street, Harbin 150001, PR China c Beijing Enterprises Water Group, Beijing 100195, China

a r t i c l e

i n f o

Article history: Received 21 January 2009 Received in revised form 28 April 2009 Accepted 28 April 2009 Available online 23 May 2009 Keywords: Traditional Chinese medicine wastewater Two-phase anaerobic process ADM1 Dynamic modeling Simulation

a b s t r a c t The aim of the study was to implement a mathematical model to simulate two-phase anaerobic digestion (TPAD) process which consisted of an anaerobic continuous stirred tank reactor (CSTR) and an upflow anaerobic sludge blanket (UASB) reactor in series treating traditional Chinese medicine (TCM) wastewater. A model was built on the basis of Anaerobic Digestion Model No. 1 (ADM1) while considering complete mixing model for the CSTR, and axial direction discrete model and mixed series connection model for the UASB. The mathematical model was implemented with the simulation software package MATLABTM/Simulinks. System performance, in terms of COD removal, volatile fatty acids (VFA) accumulation and pH fluctuation, was simulated and compared with the measured values. The simulation results indicated that the model built was able to well predict the COD removal rate (4.8–5.0%) and pH variation (2.9–1.4%) of the UASB reactor, while failed to simulate the CSTR performance. Comparing to the measured results, the simulated acetic acid concentration of the CSTR effluent was underpredicted with a deviation ratios of 13.8–23.2%, resulting in an underprediction of total VFA and COD concentrations despite good estimation of propionic acid, butyric acid and valeric acid. It is presumed that ethanol present in the raw wastewater was converted into acetic acid during the acidification process, which was not considered by the model. Additionally, due to the underprediction of acetic acid the pH of CSTR effluent was overestimated. Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction Traditional Chinese medicine (TCM), a splendid culture of Chinese nation for thousands of years, occupies the very important status in the medicine domain due to its unique effectiveness. In recent years, TCM industry rapidly develops, which results in a correspondingly increase in TCM wastewater production. It is estimated that approximately half of the TCM wastewater was discarded directly without specific treatment (Shi et al., 2005). The composition of TCM wastewater was determined by raw materials and production process, mainly coming from the washing water of raw material, the residual liquid of original medicine and floor washing water (Jiang et al., 2006; Zhu, 2007; Zhao

* Corresponding author. Fax: +86 451 8628 2008. E-mail addresses: [email protected] (Z. Chen), [email protected] (D. Hu), [email protected] (Z. Zhang), [email protected] (N. Ren), zhuhaibo1616@163. com (H. Zhu). 1 Fax: +86 451 8628 2195. 2 Fax: +65 6792 1291. 3 Fax: +86 451 8628 2008. 0960-8524/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.biortech.2009.04.066

et al., 2007). TCM wastewater contains a lot of various natural organic pollutants, such as carbohydrates (mainly polysaccharides), alkaloids, phenols,alcohols, amino acid, lignin, proteins, pigments and their hydrolysates, in the form of water-soluble, colloidal, suspended particles (Li et al., 2006; Zhu, 2007). Therefore, TCM wastewater is a kind of high concentration organic wastewater with complicated composition, which will cause the serious pollution to the environment if emitted directly. Due to the high COD concentration in pharmaceutical wastewaters, attempts have already been made to work with anaerobic processes (Yang et al., 2003; Göblös et al., 2008). It has been noted that the acid inhibition usually occurs in the one-phase anaerobic digester because of the differences in the rates of acidogenesis and methanogenesis, which depresses the methane yield and the stability, resulting thereby in low treatment efficiency. The different growth rates and pH optima for acidogenic and methanogeic organisms, and thus different requirements regarding reactors, have led to the development of two-phase anaerobic digestion (TPAD) processes (Liu, 1998). Currently, research on TPAD processes was concentrated on the evaluation of operating conditions, biological characteristics, and process performance, but not

4624

Z. Chen et al. / Bioresource Technology 100 (2009) 4623–4631

the model development. Model study might be favorable to comprehensively evaluate the impacts of all process variables on the performance of TPAD processes. Many models had been developed for anaerobic digestion processes, but with increasing complexity of the advanced digestion technologies, more complex models that can represent the impacts of changing environments on chemical and microbial species were required (Parker, 2005). Recently, there had been a move by the International Water Associations (IWA) Task Group for mathematical modelling of anaerobic digestion processes to develop a common model (Batstone et al., 2002a). ADM1 model has been used as a common basis for further development or as a common platform for dynamic simulation of different anaerobic processes. Blumensaat and Keller (2005) used ADM1 model to implement the anaerobic two-stage digestion of sewage sludge, a thermophilic pre-treatment followed by a mesophilic main treatment stage. The accuracy of the optimized parameter sets has been assessed against experimental data from pilot-scale experiments. Under these conditions, the model predicted reasonably well the dynamic behaviour of a two-stage digestion process in pilot-scale. Considering microbial storage, Toshio et al. (2007) demonstrated that the modified model adequately described anaerobic sequencing batch reactor ASBR dynamics in the degradation of trehalose during a 24h cycle. There were also many implementations of this powerful tool that have been tested and proved their success in simulating the anaerobic digestion of several organic wastes (Fuentes and Nicolás, 2008; Fezzani and Ridha, 2008). However, application of ADM1 model to the two-phase anaerobic process, especially for the treatment of drugs manufacturing with two-phase anaerobic process wastewater remained limited. The objective of this study was to make ADM1 model modified based on the process configurations and then to simulate experimental results of two-phase anaerobic process treating TCM wastewater with the modified ADM1 model. 2. Methods 2.1. Pilot-scale setup A schematic diagram of pilot-scale anaerobic process is shown in Fig. 1. The system consisted of two anaerobic reactors, that is,

a continuous stirred tank reactor (CSTR) and an upflow anaerobic sludge blanket anaerobic reactor (UASB), working as acidogenic conditions and mehtanogenic conditions, respectively. The CSTR had a working volume of approximately 12 m3 with an internal diameter of 2.2 m and a height of 3.2 m. A blender was equipped for culture mixing. The UASB was built cuboidally with a square section of 9 m2 and a height of 6.5 m, providing a working volume of approximately 55 m3. The top half of the reactor was filled with wave-shaped epoxy glass cells (cell thickness, 1 mm; crest height of wave, 45 mm; crest distance of wave 100 mm; slant distance of wave, 130 mm; and specific surface area, 360 m2 m3). A three-phase separator was installed upside the reactor to prevent biomass washout. The HRTs of CSTR and UASB were maintained consistently at 12 h and 55 h, respectively. The reactor temperatures were maintained through auto-controlled heat exchangers to be 35 °C. 2.2. Wastewater The wastewater, obtained from a local TCM company (Harbin, China) was used as the feedstock of the CSTR-UASB system. The wastewater was generated from the processes of washing, distillation, separation, concentration, purification washing, refining, torrefaction, and moulding. Its biochemical oxygen demand (BOD5) was estimated to be around 4200 mg/L and COD around 17000 mg/L, indicating a low biodegradability (BOD5/COD  0.25). Volatile fatty acids (VFA), including acetic, propionic and butyric acids, and ethanol accounted for about 3.65% of total wastewater COD, with the averages as follows: acetic acid (241 mg/L), propionic acid (63 mg/L), butyric acid (68 mg/L) and ethanol (358 mg/L). 2.3. Analytical methods The TCM wastewater and effluent of CSTR and UASB were collected daily for the analysis of COD. Measurement of COD was performed directly on the sampled slurry according to Standard Methods (APHA, 1995). VFA and ethanol were determined by gas chromatography (GC) as described in the previous work (Ren et al., 2006). The GC (GC-122, Shanghai Analytical Apparatus Corporation, China) was equipped with an injector (220 °C), a hydrogen flame ionization detector (220 °C) and a 2 m

Methane

Elevated water tank

Solenoid valve Solenoid valve Blender

Pump

Temperature controller

Effluent

Fermented gas

Heating

Temperature controller ORP/pH determinator Inflnent

Pump Solenoid valve

CSTR Solenoid valve Pump

Pump

Electric cabinet UASB Buffer tank Fig. 1. Scheme of the TPAD system.

4625

Z. Chen et al. / Bioresource Technology 100 (2009) 4623–4631

(length)  5 mm (ID) stainless steel column packed with supporter of GDX-103 (60–80 meshes). The operation of the stainless steel column was amenable to a temperature programming process within 100–200 °C (N2 carrier at a flow rate of 50 mL/min, and combustion gas H2 at 50 mL/min, combustion-supporting gas O2 at 500 mL/min). The retention times were of 0.474 min for ethanol, 0.690 min for acetic acid, 1.412 min for propionic acid, 3.020 min for butyric acid and 6.727 min valeric acid. Before measurement, the sample was filtrated through a 0.45 lm filter and acidified with 6 M HCl. Two lL acidified sample was injected manually for analysis. ORP and pH were monitored daily by ORP determinator (ORP412, Cany Precision Instruments Co., Ltd, China). 3. Mathematical model 3.1. Substrate degradation and main hypotheses of modeling It is recognized that a number of the conversion processes that are active in anaerobic digestion can be inhibited by the accumulation of intermediate products such as molecular hydrogen, ammonia or by extremes of pH. In this study, it is assumed that all microbially mediated substrate conversion processes are only inhibited by extremes of pH. Fluid-gas exchange process is neglected. The microorganism content is neglected in influent and effluent. These simplifying assumptions are introduced to make the model workable, although they do not completely reflect reality. 3.2. Main mathematical model equations In ADM1 model, cellular kinetics is described by three expressions: uptake, growth and decay. This model includes three overall biochemical (cellular) steps, this is, acidogenesis (fermentation), acetogenesis (anaerobic oxidation of organic acids) and methanogenesis. In a two phase anaerobic process, acidogenic-phase digestion process includes acidogenesis and acetogenesis steps, while methanogenic-phase process involves methanogenesis. Matrix form is adopted in ADM1 model to describe the biokinetics model (Batstone et al., 2002a). An important advantage of matrix form is able to easily understand dynamic process of the various components used in the model. The matrix includes biochemical rate coefficients (v i;j ) and kinetic rate equations (qj ). BeTable 1 Model component. Component

Name

Value

Unit

Ssu Saa Sfa Sva Sbu Spro Sac Sh2 Sch4 SI Xc X ch X pr X li X su X aa X fa X c4 X pro X ac X h2 Xi

Monosaccharides Amino acid Long chain fatty acid Valerate acid Butyric acid Propionic acid Acetic acid Dissolved H2 Dissolved CH4 Soluble inerts Composites Carbohydrates Proteins Lipids Sugar degraders Amino acid degraders LCFA degraders Valerate and butyrate degraders Propionate degraders Acetate degraders Hydrogen degraders Particulate inerts

Simulation Simulation Simulation Simulation Simulation Simulation Simulation Simulation Simulation Simulation Simulation Simulation Simulation Simulation Simulation Simulation Simulation Simulation Simulation Simulation Simulation Simulation

kgCOD/m3 kgCOD/m3 kgCOD/m3 kgCOD/m3 kgCOD/m3 kgCOD/m3 kgCOD/m3 kgCOD/m3 kgCOD/m3 kgCOD/m3 kgCOD/m3 kgCOD/m3 kgCOD/m3 kgCOD/m3 kgCOD/m3 kgCOD/m3 kgCOD/m3 kgCOD/m3 kgCOD/m3 kgCOD/m3 kgCOD/m3 kgCOD/m3

cause the microorganism content is neglected in influent and effluent flows, the calculation of the state variables is shown in Eqs. (1) and (2).

V

X dSi qj v i;j ¼ qin Si;in  qout Si þ dt j¼119

ð1Þ

V

X dX i ¼ qw X i þ qj mi;j dt j¼119

ð2Þ

P Where j¼119 qj v i;j is the sum of the biochemical rate coefficients ðv i;j Þ multiplies kinetic rate equations ðqj Þ for process; V is the liquid reactor volume; qin is flow rate of influent; qout is flow rate of effluent; qw is flow rate of sludge discharge; Si;in is the input concentration of the soluble components; Si is the concentration of soluble components and Xi is the concentration of the particulate in the reactor. The components are shown in Table 1.

"  2 # pH  pHUL I1 ¼ I2 ¼ I3 ¼ exp 3 pHUL  pHLL

ð3Þ

To reduce the complexity of the model, only pH inhibition mechanisms are considered, as shown in Eq. (3). 3.2.1. The balance equations of CSTR In the present study, the CSTR is considered to be completely mixed through continuous stirring, and the frequency and intensity of mixing are sufficient to prevent the formation of large substrate gradients in the reactor. Based on the rate equations matrix of ADM1, the following balance equations of CSTR can be derived, as shown in Eqs. (4–20) (Appendix A). 3.2.2. The balance equations of UASB According to the distribution of sludge concentration along the reactor height, UASB reactor was divided into three internal sections, i.e., bed section, blanket section and settler section. In order to simple the flow model, the axial direction discrete model and mixed series connection model were introduced here. Discrete model, one kind of revised ideal model, mainly describes non-ideal reactors, particularly applicable to the non- back-mixing systems. This model has assumptions as follows: (1) each section vertical to flow direction has homogeneous radial concentration; (2) axial direction discrete coefficient DL is constant, regardless of operating time and reactor position; (3) substrate concentration is continuous functions of flow position. Two equations [Eqs. (21) and (22)] representing inflow and outflow states can de derived.

   d dS Sþ A inflow : qS þ DL dl dl

ð21Þ

    dS dS outflow : q S þ þ DL A dl dl

ð22Þ

Where A is the section area of reactor, l is the reactor height and V is the reactor volume.As high order differential items are neglected, a balance equation [Eq. (23)] can be derived. 2

dS d S dS ¼ DL 2  q dt dl dl

ð23Þ

Boundary conditions for Eq. (23) are described as follows:

l ¼ 0;

qSin ¼ qSþin  DL

l ¼ L;

  dS ¼0 dl L

  dS dl þin

ð24Þ

ð25Þ

4626

Z. Chen et al. / Bioresource Technology 100 (2009) 4623–4631

According to the mixed series connection model, the actual reactor is divided into several equal volume ideal-mixing sectors. The mass balance of those sectors is expressed by Eqs. (26) and (27).

X V dSi qj mi;j ¼ qSi;in þ n dt j¼119

ð26Þ

X V dX i qj mi;j ¼ qw X i þ n dt j¼119

ð27Þ

By combining the rate equations matrix of ADM1 and the flow model, the balance equations of UASB are derived, as shown in Eqs. (28)–(38) (Appendix B). Boundary conditions for the above material balance equations are given in following equations [Eqs. (24) and (25)]: 3.3. Model parameters estimation In ADM1, there are three main parameters: stoichiometric coefficients, equilibrium coefficients, and kinetic parameters. Many of the stoichiometric coefficients, equilibrium coefficients as well as kinetic parameters are assumed to be fixed due to their low variability and sensitivity and taken from the scientific and technical work reported by the IWA task group and Chynoweth et al. (Chynoweth et al., 1998a; Batstone et al., 2002b; Masse et al., 1996, 1997; Masse and Droste, 1997, 2000; Stumm and Morgan, 1996). The reactor parameters were taken from the experimental setups for both datasets and applied to the model. The operating temperatures were 35 °C for CSTR and UASB. The HRTs of CSTR and UASB were maintained consistently at 12 and 55 h, respectively. All simulations were run to steady-state using a constant input including a constant flow rate that was given through the initial conditions of the dynamic experiments, to avoid numerical probTable 2 Model parameters with low sensitivity and low variability from the literature. Parameter A. Hydrolysis khyd-CH khyd-PR khyd-Li

Name parameter Hydrocarbon hydrolysis coefficient Proteins hydrolysis coefficient Lipid hydrolysis coefficient

B. Kinetic parameters Monod largest absorption rate of sugar km-su Half saturated coefficient of sugar KS-su Y su Yield of sugar Monod largest absorption rate of amino km-aa acid K S-aa Half saturated coefficient of amino acid Y aa Yield of amino acid Monod largest absorption rate of long km-fa chain fatty acid K S-fa Half saturated coefficient of long chain fatty acid Y fa Yield of long chain fatty acid Monod largest absorption rate of butyric km-Cþ 4 acid K S-Cþ Half saturated coefficient of butyric acid 4 Y Cþ Yield of butyric acid 4 Yield of propionic acid Y pro Yield of acetic acid Y ac Monod largest absorption rate of H2 km-H2 Yield of H2 Y H2 C. Stoichiometric coefficients Butyric acid generated by sugar fbu; su Acetic acid generated by sugar fac; su Propionic acid generated by sugar fpro; su Acetic acid generated by amino acid fac; aa Propionic acid generated by amino acid fpro; aa Butyric acid generated by amino acid fbu; aa fva; aa Valeric acid generated by amino acid

Value

Unit

8 8 8

d1 d1 d1

30 0.5 0.1 50

COD/(CODd) kgCOD/m3 COD/COD COD/(CODd)

0.3 0.08 6

kgCOD/m3 COD/COD COD/(CODd)

0.4

kgCOD/m3

0.06 20

COD/COD COD/(CODd)

0.3 0.06 0.04 0.05 35 0.06

kgCOD/m3 COD/COD COD/COD COD/COD COD/(CODd) COD/COD

0.59 0.45 0.20 0.6 0.02 0.2 0.3

– – – – – – –

lems that could occur when starting a simulation with dynamic input (Pauss et al., 1990). The prior simulations were run over an integration period of 120 days to find the effective steady state. Carbon contents were recalculated by implementing the balance terms in the rate equation matrix. The mass balance has been checked for COD and carbon contents using the Excel spreadsheet developed by Batstone et al. (2002a). These specific biochemical process rates and coefficients are tabulated in Table 2. Parameters with high sensitivity and high variability are estimated using experimental data of Masse et al., 1996, 1997; Masse and Droste, 2000; Siegrist et al., 2002; and the selected optimization methods (Masse et al., 1996, 1997; Masse and Droste, 1997, 2000; Siegrist et al., 2002). The nonlinear constrained optimization method was implemented using MATLAB 6.5 optimization toolbox. The two-parameter optimization around optimum using the secant method is implemented using Aquasim 2.1d version. Estimation procedures were applied for the following parameters: the dissociation coefficient, kdis; Monod largest absorption rate of acetic acid and propionic acid, km,ac, km,pro; Half saturated coefficient of acetic acid, propionic acid and hydrogen, Ks,ac, Ks, pro, Ks, H2. Estimation results of the six parameters were shown in 4.1 sections. 3.4. Ordinary differential equation solutions Anaerobic digestion processes was formed the work of which led to the Anaerobic Digestion Model No. 1 (Batstone et al., 2002b). ADM1 is a highly complex model, characterized by 19 biochemical conversion processes and 24 dynamic state variables. For this paper calculations were executed with the software MATLAB version 6.5 (release 13) and Simulink 5. MATLAB/Simulink system provides ordinary dynamic equation solver systems (Copp et al., 2002; Rosen and Jeppson, 2002. Solver ode23, a one-step solver, is an implementation of an explicit Runge–Kutta (2,3) pair of Bogacki and Shampine. It could be more efficient than ode45 in cases of crude tolerances and moderate stiffness. Solver ode45, a one-step solver, is based on an explicit Runge–Kutta (4,5) formula, the Dormand-Prince pair. It could be considered as the best function to apply as a ‘‘first try” for most problems. Solver ode15s, a multi-step solver, is a variable order solver based on the numerical differentiation formulas. Optionally, it uses the backward differentiation formulas (also known as Gear’s method) that are usually less efficient. It is considered to be good for stiff problems while solving a differential-algebraic problem. Solver ode23s, a one-step solver is based on an extended Rosenbrock formula of order 2. It is could be more efficient than ode15s at crude tolerances and is used for the solution of specific type of stiff problems where ode15s are not efficient.

4. Results and discussion 4.1. Parameter estimation results The results of the parameter estimation indicated a good fit between the model and the measured data. Parameters with low sensitivity and low variability were not further optimized after carrying out sensitivity analysis using Aquasim 2.1d software and comparing with previous results of similar cases. Previous study indicated that the effect of these parameters on the model outcomes is quite limited (Batstone et al., 2002a). Optimization of the low sensitivity parameters was not carried out since further tuning of these parameters requires highly accurate experimental data, and changes in the model outputs would be relatively small as compared to errors in the data. Those typical parameters were selected for the optimization in this work (i.e., similar to the ones generally selected for the optimization).

4627

Z. Chen et al. / Bioresource Technology 100 (2009) 4623–4631 Table 3 Model parameter estimation results with high sensitivity and high variability.

4.2. The simulation result of COD

Parameter

The raw wastewater COD had a concentration of 15000– 20000 mg/L, corresponding to an OLR of 30–40 kgCOD/m3 d. An average COD removal rate of 41.7% was achieved by the CSTR, while simulated COD removal rates averaged at around 30.0%, showing a deviation ratio of 10.1–24.6% between the measured and predicted values (Fig. 2). Comparing to the COD removal rates (20–30%) in acidogenic phase reported by Li et al. (2006), the present CSTR system exhibited a superior performance in COD removal. The inconsistency is likely owing to the difference of wastewater composition as TCM wastewater was used in the present study. The deviation ratios between the measured and predicted values indicate that the coefficients and parameters obtained by previous work might not be suitable for the acidogenic phase simulation of TCM with ADM1 model. In the two-phase anaerobic process, wastewater COD is mainly removed in the methanogenic phase. The effluent of UASB stabilized at around 1400 mg/L, corresponding to a COD removal rate of 86.7%. Only a COD removal efficiency of 65% was achieved in a similar, but one-phase UASB reactor while treating chemical synthesis based pharmaceutical wastewater (Ince et al., 2002). Results

K S-ac K S-H2

Dissociation coefficient Monod largest absorption rate of propionic acid Half saturated coefficient of propionic acid Monod largest absorption rate of acetic acid Half saturated coefficient of acetic acid Half saturated coefficient of H2

Unit 1

0.75 13

d COD/(CODd)

0.2 18

kgCOD/m3 COD/(CODd)

0.2 5  105

kgCOD/m3 kgCOD/m3

In the estimation procedure, the disintegration constant was first estimated by matching the model outputs with the measured outputs. Then the model outputs for concentrations of acetic acid and propionic acid were changed by changing the half saturation constants and maximum uptake rates. Parameters km, Ks for acetic acid and propionic acid were optimized together since they exhibit the lowest correlation and the highest relevance (Batstone et al., 2003). The optimum values for the parameters are given in Table 3.

COD simulated

COD meatured

deviation ratio

12000

50

11000

45 40

COD (mg/L)

10000

35

9000

30 8000 25 7000

20

6000

deviation ratio

K S-pro km-ac

Value

15

5000

10 5 120

4000 0

10

20

30

40

50

60

80

70

90

100

110

Time (days) Fig. 2. Measured COD value of CSTR comparison with model simulated value.

COD simulated

COD meatured

deviation ratio

10

1800

5 1700

COD (mg/L)

0 -5

1600

-10 1500

-15 -20

1400 -25 1300 0

10

20

30

40

50

60

70

80

90

100

110

-30 120

Time (days) Fig. 3. Measured COD value of UASB comparison with model simulated value.

deviation ratio

kdis km-pro

Name

4628

Z. Chen et al. / Bioresource Technology 100 (2009) 4623–4631

of this study show that the TPAD system is much superior compared to one-phase processes. It is probably owing to the fact that complicated organics were first broken down and converted to easily biodegradable components in the CSTR. The experimental results did indicate a substantially increase in ethanol concentration of CSTR effluent. As shown in Fig. 3, the model was able to predict the COD of UASB effluent with a considerable accuracy. It was found that the deviation ratio of the measured and simulated values fluctuated between 4.8% and 5.0%.

acetic acid ethanol was much underpredicted. Compared to the measured values, there was a deviation ratio of 13.8–23.2% in the simulated results. As a consequence, the VFA content was underpredicted in this model. It is likely that ethanol present in the raw wastewater was converted into acetic acid during the acidification process, which was not considered by the model. It was noted that Parker (2005) showed that the concentrations of VFAs were consistently over-predicted in digesters at short solid retention times. These results seems to confirm the previous observation that it was difficult to accurately predict the VFA concentration with ADM1 model due to complicated behavior of acidification process (Blumensaat and Keller, 2005).

4.3. The simulation result of VFA In the two-phase anaerobic process, the function of acidogenic phase reactor is mainly to enhance wastewater biodegradability through decomposing large molecular compounds into small molecular products. In the present study, VFA and ethanol concentrations of CSTR effluent increased substantially, suggesting the acidification occurring in the CSTR. It was found an average increase in acetic acid from 241 ± 1.6 mg/L to 873 ± 2.8 mg/L, in propionic acid from 63 ± 1.2 mg/L to 174 ± 1.7 mg/L, in butyric acid from 68 ± 0.9 mg/L to 340 ± 2.5 mg/L, valeric acid from null to 150 ± 2.06 mg/L, and in ethanol from 358 ± 2.1 mg/L to 767 ± 3.2 mg/L. As a result, the proportions of wastewater VFA and ethanol increased from 3.65% to 23.09% after the TCM wastewater reacted in the CSTR. Fig. 4 shows the measured and simulated results of VFA and ethanol in the CSTR. It can be seen that propionic acid, butyric acid and valeric acid were well predicted with respective deviation ratios of 6.5–7.1%, 2.1–5.7%, and 4.4–4.3%. However,

4.4. The simulation result of pH Previous studies indicated that the major products of carbohydrates acidogenesis are acetate, butyrate, propionate, and ethanol, whose proportion is strongly dependent on culture pH (Fang and Liu, 2002). The effluent pH of CSTR was in a range of 4.8–5.2, and the simulated pH was higher than the measured values (deviation ratio, 5.8–14.5%), as shown in Fig. 5. It is probably due to the underpredicted VFA concentrations with the ADM1 model. The effluent pH of UASB fluctuated in a narrow range of 6.5–7.0 since the acidified intermediates were fermented in the UASB (Fig. 6). The model well predicted the effluent pH of UASB with a deviation ratio of 2.9–1.4%, indicating that methanogenic bacteria grew well. 35 deviation ratio

acetic acid measured

30

800 700

25

600

20

500 15

400

propionic acid (mg/L)

300

deviation ratio

acetic acid simulated

900

10 40

225

propionic acid meatured

propionic acid simulated

deviation ratio

20

210 0 195

-20

180

-40

165

-60

150

-80

deviation ratio

aceticacid (mg/L)

1000

butyric acid simulated

360

10

deviation ratio

butyric acid meatured

0

340

-10

320 300

-20

280

-30

deviation ratio

butyric acid (mg/L)

380

190

deviation ratio

valeric acid meatrued

valeric acid simulated

180

20 0

170 -20

160 150

-40

140 130

-60 0

10

20

30

40

50

60

70

80

90

100

110

Time (days) Fig. 4. Measured VAF value of CSTR comparison with model simulated value.

120

deviation ratio

valeric acid (mg/L)

200

4629

Z. Chen et al. / Bioresource Technology 100 (2009) 4623–4631

deviation ratio

pH meatured

5

5

0

pH

5.5

4.5

-5

4

deviation ratio

pH simulated

-10

3.5 0

10

20

30

40

50

60

70

80

90

100

110

-15 120

Time (days) Fig. 5. Measured pH of CSTR comparison with model simulated value.

pH simulated

pH meatured

deviation ratio

7.2 9 7.1 7 5

pH

6.9

3

6.8

1

6.7 6.6

-1

6.5

-3

6.4 0

10

20

30

40

50

60

70

80

90

100

110

deviation ratio

7

-5 120

Time (days) Fig. 6. Measured pH of UASB comparison with model simulated value.

5. Conclusions ADM1 model was applied to the TCM wastewater treatment with TPAD process while considering complete mixing model for the CSTR, and axial direction discrete model and mixed series connection model for the UASB. The simulation results indicated that the model built was able to well predict the COD removal rate (4.8–5.0%) and pH variation (2.9–1.4%) of the UASB reactor, while failed to simulate the CSTR performance. Comparing to the measured results, the simulated acetic acid concentration of the CSTR effluent was underpredicted with a deviation ratios of 13.8–23.2%, resulting in an underprediction of total VFA and COD concentrations despite good estimation of propionic acid (deviation ratio, 6.5–7.1%), butyric acid (deviation ratio, 2.1–5.7%) and valeric acid (deviation ratio, 4.4–4.3%). It is presumed that ethanol present in the raw wastewater was converted into acetic acid during the acidification process, which was not considered by the model. Additionally, due to the underprediction of acetic acid the pH of CSTR effluent was overestimated. Acknowledgements The authors are grateful to Research Center of Environmental Biotechnology in Harbin Institute of Technology for their technical

and logistical assistance during this work which was supported by State Key Laboratory of Urban Water Resource and Environment (HIT-QAK200808) and China National ‘‘863” Hi-Tech R & D Program (Grant No. 2007AA06Z348).

Appendix A

V

V

V

dSsu ¼ qSsu;in  qSsu þ khyd;ch X ch þ ð1  ffa;li Þkhyd;li X li dt "  2 # Ssu pH  5:5  km;su X su exp 3 1:5 K Ssu þ Ssu dSaa ¼ qSaa;in  qSaa þ khyd;pr X pr dt "  2 # Saa pH  5:5  km;aa X aa exp 3 1:5 K Saa þ Saa dSfa ¼ qSfa;in  qSfa þ ffa;li khyd;li X li dt "  2 # Sfa pH  5:5  km;fa X fa exp 3 1:5 K Sfa þ Sfa

ð4Þ

ð5Þ

ð6Þ

4630

V

V

Z. Chen et al. / Bioresource Technology 100 (2009) 4623–4631

dSv a Saa X aa ¼ qSv a;in  qSv a þ ð1  Y aa Þfv a;aa km;aa dt K Saa þ Saa "  2 # pH  5:5  exp 3 1:5

V

ð7Þ

dSbu Ssu X su ¼ qSbu;in  qSbu þ ð1  Y su Þfbu;su km;su dt K Ssu þ Ssu "  2 # pH  5:5  exp 3 1:5

V

"  2 # Saa pH  5:5 X aa exp 3  þð1  Y aa Þfbu;aa km;aa 1:5 K Saa þ Saa

V

ð8Þ V

V

dSpro Ssu X su ¼ qSpro;in  qSpro þ ð1  Y su Þfpro;su km;su dt K Ssu þ Ssu "  2 # pH  5:5 þ ð1  Y aa Þfpro;aa km;aa  exp 3 1:5 "  2 # Saa pH  5:5  X aa exp 3 1:5 K Saa þ Saa

V

ð9Þ

dX fa Sfa X fa ¼ qw X fa þ Y fa km;fa dt K Sfa þ Sfa "  # 2 pH  5:5  kdec;Xfa X fa  exp 3 1:5 dX i ¼ qw X i þ fxl;xc kdis X c dt

ð18Þ

ð19Þ

ð20Þ

2

DL

d Sv a dl

2

dSv a q dl

!

Sv a 1 X c4 1 þ Sbu =Sv a K SC þ þ Sv a 4 "  2 # pH  5:5  exp 3 1:5

¼ qSv a;in  qSv a  km;c4

"  2 # Saa pH  5:5 X aa exp 3 þ ð1  Y aa Þfac;aa km;aa 1:5 K Saa þ Saa "

 2 # Sfa pH  5:5 X fa exp 3 þ 0:7ð1  Yfa Þkm;fa 1:5 K Sfa þ Sfa ð10Þ V

dX aa Saa X aa ¼ qw X aa þ Y aa km;aa dt K Saa þ Saa   pH  5:5 2  exp 3ð Þ  kdec;X aaa X aa 1:5

ð17Þ

Appendix B

V n

dSac Ssu X su ¼ qin Sac;in  qout Sac þ ð1  Y su Þfac;su km;su dt K Ssu þ Ssu "  2 # pH  5:5  exp 3 1:5

dX su Ssu X su ¼ qw X su þ Y su km;su dt K Ssu þ Ssu "  # 2 pH  5:5  kdec;Xsu X su  exp 3 1:5

dSh2 Ssu X su ¼ ð1  Y su Þfh2;su km;su dt K Ssu þ Ssu "  2 # pH  5:5 Saa ð1  Y aa Þfh2 ;aa km;aa  exp 3 X aa 1:5 K Saa þ Saa   Sfa pH  5:5 2  exp 3ð X fa Þ þ 0:3ð1  Y fa Þkm;fa 1:5 K Sfa þ Sfa   pH  5:5 2 ð11Þ  exp 3ð Þ 1:5

V n

2

DL

d Sbu dl

2

q

dSbu dl

ð28Þ

!

Sbu 1 X c4 1 þ Sv a =Sbu K SCþ þ Sbu 4 "  2 # pH  5:5  exp 3 1:5

¼ qSbu;in  qSbu  km;c4

2

V d Spro dSpro DL q 2 n dl dl

ð29Þ

!

"  2 # Spro pH  5:5 X pro exp 3 1:5 K Spro þ Spro "  2 # Sv a 1 pH  5:5 þ 0:54ð1  Y c4 Þkm;c4 X c4 exp 3 1 þ Sbu =Sv a 1:5 K SC þ þ Sv a ¼ qSpro;in  qSpro  km;pr

4

V

dSI ¼ qw SI þ fsI;xc kdis X c dt

dX c V ¼ qw X c  kdis X c þ kdec;X su X su þ kdec;X aa X aa þ kdec;Xfa X fa dt

ð30Þ

ð12Þ 2

ð13Þ

V d Sac dSac Þ ðDL 2  q dl n dl ¼ qSac;in  qSac  km;ac

V

dX ch ¼ qw X ch þ fch;xc kdis;ch X c  khyd;ch X ch dt

ð14Þ

V

dX pr ¼ qw X pr þ fpr;xc kdis X c  khyd;pr X pr dt

ð15Þ

þ 0:31ð1  Y c4 Þkm;c4

Sac X ac exp½3ðpH  7Þ2  K Sac þ Sac Sv a 1 pH  5:5 2 X c4 exp½3ð Þ  K SC þ þ Sv a 1 þ Sbu =Sv a 1:5 4

Sbu 1 pH  5:5 2 X c4 exp½3ð þ 0:8ð1  Y c4 Þkm;c4 Þ  K SC þ þ Sbu 1 þ Sv a =Sbu 1:5 4

dX li V ¼ qw X li þ fli;xc kdis X c  khkd;li X li dt

ð16Þ

þ 0:57ð1  Y pro Þkm;pr

Spro pH  5:5 2 X pro exp½3ð Þ  K Spro þ Spro 1:5

ð31Þ

Z. Chen et al. / Bioresource Technology 100 (2009) 4623–4631 2

V d Sh2 dSh2 q DL 2 n dl dl

!

¼ 0:15ð1  Y c4 Þkm;c4

Sv a 1 pH  5:5 2 X c4 exp½3ð Þ  1 þ Sbu =Sv a 1:5 K SCþ þ Sv a 4

Sbu 1 pH  5:5 2 X c4 exp½3ð þ 0:2ð1  Y c4 Þkm;c4 Þ  1 þ Sv a =Sbu 1:5 K SC þ þ Sbu 4

þ 0:43ð1  Y pro Þkm;pr  km;h2

Spro pH  5:5 2 X pro exp½3ð Þ  1:5 K Spro þ Spro

Sh2 X h2 exp½3ðpH  6Þ2  K S þ Sh2

2

V d Sch4 dSch4 q DL 2 n dl dl

! ¼ ð1  Y ac Þkm;ac

ð32Þ

Sac X ac exp½3ðpH  6Þ2  K S þ Sac

Sh2 X h2 K S þ Sh2 pH  5:5 2  exp½3ð Þ  1:5 þ ð1  Y h2 Þkm;h2

V n

2

DL

d Xc dl

2

q

dX c dl

ð33Þ

! ¼ qw X c þ kdec;X c4 X c4 þ kdec;X pro X pro þ kdec;Xac X ac þ kdec;X h2 X h2

2

V d X c4 dX c4 q DL 2 n dl dl

ð34Þ

!

¼ qw X c4 þ Y c4 km;c4

Sv a 1 pH  5:5 2 X c4 exp½3ð Þ  1 þ Sbu =Sv a 1:5 K SCþ þ Sv a 4

Sbu 1 pH  5:5 2 X c4 exp½3ð þ Y c4 km;c4 Þ  1 þ Sv a =Sbu 1:5 K SC þ þ Sbu 4

 kdec;Xc4 X c4 2

ð35Þ

V d X pro dX pro q DL 2 n dl dl

!

¼ qw X pro  kdec;X ac X ac þ Y pro km;pr pH  5:5 2 Þ  1:5 ! 2 V d X ac dX ac q DL 2 n dl dl  exp½3ð

¼ qw X ac þ Y ac km;ac

2

V d X h2 dX h2 q DL 2 n dl dl

Spro X pro K Spro þ Spro ð36Þ

Sac X ac exp½3ðpH  7Þ2   kdec;Xac X ac K Sac þ Sac ð37Þ

!

¼ qw X h2 þ Y h2 km;h2

Sh2 X h2 exp½3ðpH  6Þ2   kdec;X h2 X h2 K SH2 þ Sh2 ð38Þ

References APHA, 1995. Standard Methods for the Examination of Water and Wastewater, 20th ed. American Public Health Association, Washington, DC.

4631

Batstone, D.J., Keller, J., Angelidaki, I., Kalyuzhnyi, S.V., Pavlostathis, S.G., Rozzi, A., Sanders, W.T.M., Siegrist, H., Vavilin, V.A., 2002a. The IWA Anaerobic Digestion Model No. 1 (ADM1). Water Science and Technology 45 (10), 65–73. Batstone, D.J., Keller, J., Angelidaki, I., Kalyuzhnyi, S.V., Pavlostathis, S.G., Rozzi, A., Sanders, W.T.M., Siegrist, H., Vavilin, V.A., 2002b. Anaerobic Digestion Model No. 1 (ADM1), International Water Association Scientific and Technical Report No. 13. IWA Publishing, London, U.K. Batstone, D.J., Pind, P.F., Angelidaki, 2003. Kinetics of thermophilic, anaerobic oxidation of straight and branched chain butyrate and valerate. Biotechnology and Bioengineering 84 (2), 195–204. Blumensaat, F., Keller, J., 2005. Modelling of two-stage anaerobic digestion using the IWA Anaerobic Digestion Model No. 1 (ADM1). Water Research 39, 171–183. Chynoweth, D.P., Wilke, A.C., Owens, J.M., 1998. Anaerobic Processing of Piggery Wastes: A Review. Paper presented at the ASAE Meeting, Orlando, FL. Copp, J.B., Alex, J., Beteau, J.F., Dudley, J., Dupont, R., Gillot, S., Jeppsson, U., Lelann, J.M., Pons, M.N., Vanrolleghem, P.A., 2002. The Cost Simulation Benchmark. Office for Official Publications of the European Communities, Luxemburg. Fang, H.H.P., Liu, H., 2002. Effect of pH on hydrogen production from glucose by a mixed culture. Bioresource Technology 82, 87–93. Fezzani Boubaker, Ridha Ben Cheikh, 2008. Implementation of IWA anaerobic digestion model No. 1 (ADM1) for simulating the thermophilic anaerobic codigestion of olive mill wastewater with olive mill solid waste in a semicontinuous tubular digester.Chemical Engineering Journal. Fuentes, M., Nicolás, J., 2008. Application of two anaerobic digestion models to biofilm systems. Biochemical Engineering Journal 38, 259–269. Göblös, S., Portörõ, P., Bordás, D., Kálmán, M., Kiss, I., 2008. Comparison of the effectivities of two-phase and single-phase anaerobic sequencing batch reactors during dairy wastewater treatment. Renewable Energy 33, 960–965. Ince, K.B., Selcuk, A., Ince, O., 2002. Effect of a chemical synthesis-based pharmaceutical wastewater performance, acetoclastic methanogenic activity and microbial population in an upflow anaerobic filter. J. Chem. Technol. Biotechnol. 77, 711–719. Jiang, Bin, Lu, Xi-wu, Liu, Zhuang, Ge, Jun, Gui, Yi-jun, 2006. Application of reactor theory to technical reformation works for traditional Chinese medicine wastewater treatment. Technology of Water Treatment 32, 86–88. Li, Dong-wei, Li, Dou, Wang, Ke-hao, Yang, Jian, Yuan, Xue, 2006. Wastewater treatment of traditional Chinese medicine by two phase anaerobic technology. Technology of Water Treatment 32, 81–83. Liu, Tuanchi, 1998. Anaerobic digestion of solid substrates in an innovative twophase plug-flow reactor (tppfr) and a conventional single-phase continuously stirred-tank reactor. Water Science and Technology 38, 453–461. Masse, D., Droste, R.L., 1997. Microbial interaction during the anaerobic treatment of swine manure slurry in a sequencing batch reactor. Canadian Journal of Agricultural Engineering 39 (1), 35–41. Masse, D., Droste, R.L., 2000. Comprehensive model of anaerobic digestion of swine manure slurry in a sequencing batch reactor. Water Research 34 (12), 3087– 3106. Masse, D., Patni, N.K., Droste, R.L., Kennedy, K.J., 1996. Operation strategies for psychrophilic anaerobic digestion of swine manure slurry in sequencing batch reactors. Canadian Journal of Civil Engineering 23, 1285–1294. Masse, D., Patni, N.K., Droste, R.L., Kennedy, K.J., Munroe, J.A., 1997. Potential for the psychrophilic anaerobic treatment of swine manure using a sequencing batch reactor. Canadian Journal of Agricultural Engineering 39 (1), 25–33. Parker, W.J., 2005. Application of the ADM1 model to advanced anaerobic digestion. Bioresource Technology 96, 1832–1842. Pauss, A., Andre, G., Perrier, M., Guiot, S.R., 1990. Liquid-to-gas mass transfer in anaerobic processed: inevitable transfer limitations of methane and hydrogen in the biomethanation process. Applied and Environmental Microbiology 56 (6), 1636–1644. Rosen, C., Jeppson, U., 2002. Anaerobic COST Benchmark Model Description (Report). San Sebastian, Spain. Shi, Yue, Ren, Nanqi, Gao, Yu, Wang, Aijie, Chen, Zhaobo, 2005. Fuli-scale two-phase anaerobic process treating traditional Chinese medicine wastewater. Environmental Science 26, 106–110. Siegrist, H., Renggli, D., Gujer, W., 2002. Mathematical model for mesophilic thermophilic anaerobic sewage sludge digestion. Environmental Science and Technology 36 (5), 1113–1123. Stumm, W., Morgan, J.J., 1996. Anaerobic chemistry: chemical equilibria and rates in natural waters, third ed. John Wiley and Sons, New York, NY. Toshio, Shimada, Julie, Zilles, Lutgarde, Raskin, Eberhard, Morgenroth, 2007. Carbohydrate storage in anaerobic sequencing batch reactors. Water Research 41, 4721–4729. Yang, K.Y., Yu, Y., Hwang, S., 2003. Selective optimization in thermophilic acidogenesis ofcheese-whey wastewater to acetic and butyric acids: partial acidification and methanation .Water Research 37, 2467–2477. Zhao, Zheng, Tai, Mingqing, Niu, Qikai, et al., 2007. The Practice on Engineering Treatment of Chinese Traditional Medicine Wastewater. China Resources Comprehensive Utilization 25, 19–21. Zhu, Jian, 2007. Progress in study of the characterstics and treatments of traditional Chinese medicine wastewater. Energy Environmental Protection 21, 15–17.