Scholarly article on topic 'Optimization of fermentative hydrogen production from palm oil mill effluent in an up-flow anaerobic sludge blanket fixed film bioreactor'

Optimization of fermentative hydrogen production from palm oil mill effluent in an up-flow anaerobic sludge blanket fixed film bioreactor Academic research paper on "Chemical sciences"

Share paper
Academic journal
Sustainable Environment Research
OECD Field of science
{"Fermentative hydrogen production" / "Palm oil mill effluent" / "Up-flow anaerobic sludge blanket fixed film"}

Abstract of research paper on Chemical sciences, author of scientific article — Parviz Mohammadi, Shaliza Ibrahim, Mohamad Suffian Mohamad Annuar, Maryam Khashij, Seyyed Alireza Mousavi, et al.

Abstract Response surface methodology with a central composite design was applied to optimize fermentative hydrogen production from palm oil mill effluent (POME) in an upflow anaerobic sludge blanket fixed film reactor. In this study, the concurrent effects of up-flow velocity (Vup) and feed flow rate (QF) as independent operating variables on biological hydrogen production were investigated. A broad range of organic loading rate between 10 and 60 g COD L−1 d−1 was used as the operating variables. The dependent parameters as multiple responses were evaluated. Experimental results showed the highest value of yield at 0.31 L H2 g−1 COD was obtained at Vup and QF of 0.5 m h−1 and 1.7 L d−1, respectively. The optimum conditions for the fermentative hydrogen production using pre-settled POME were QF = 2.0–3.7 L d−1 and Vup = 1.5–2.3 m h−1. The experimental results agreed very well with the model prediction.

Academic research paper on topic "Optimization of fermentative hydrogen production from palm oil mill effluent in an up-flow anaerobic sludge blanket fixed film bioreactor"

Accepted Manuscript

Optimization of fermentative hydrogen production from palm oil mill effluent in an up-flow anaerobic sludge blanket fixed film bioreactor

Parviz Mohammadi, Shaliza Ibrahim, Mohamad Suffian Mohamad Annuar, Maryam Khashij, Seyyed Alireza Mousavi, Aliakbar Zinatizadeh

PII: S2468-2039(16)30049-8

DOI: 10.1016/j.serj.2016.04.015

Reference: SERJ 37

To appear in: Sustainable Environment Research

Received Date: 11 January 2016

Revised Date: 12 April 2016

Accepted Date: 27 April 2016

Please cite this article as: Mohammadi P, Ibrahim S, Annuar MSM, Khashij M, Mousavi SA, Zinatizadeh A, Optimization of fermentative hydrogen production from palm oil mill effluent in an up-flow anaerobic sludge blanket fixed film bioreactor, Sustainable Environment Research (2016), doi: 10.1016/ j.serj.2016.04.015.

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

Received 11 January 2016 Received in revised form 12 April 2016 Accepted 27 April 2016

Optimization of fermentative hydrogen production from palm oil mill effluent in an up-flow anaerobic sludge blanket fixed film bioreactor

12 2 3

Parviz Mohammadi ' '* Shaliza Ibrahim , Mohamad Suffian Mohamad Annuar , Maryam Khashij1, Seyyed Alireza Mousavi1, Aliakbar Zinatizadeh4

department of Environmental Health Engineering Kermanshah University of Medical Science Kermanshah 6719851351, Iran 2Department of Civil Engineering University of Malaya Kuala Lumpur 6719851351, Malaysia 3Institute of Biological Sciences University of Malaya Kuala Lumpur 6719851351, Malaysia "Department of Applied Chemistry Razi University Kermanshah 6719851351, Iran

Key Words: Fermentative hydrogen production, palm oil mill effluent, up-flow anaerobic sludge blanket fixed film

Corresponding author: E-mail:


Fermentative hydrogen production from palm oil mill effluent (POME) in an upflow anaerobic sludge blanket fixed film bioreactor was optimized using response surface methodology with a central composite design. The simultaneous effects of two independent operating variables, i.e. feed flow rate (QF) and up-flow velocity (Vup) on biological hydrogen production was investigated. The operating variables were varied to cover a wide range of organic loading rates from 10 to 60 g COD L-1 d-1. The dependent parameters as multiple responses were evaluated. Experimental results showed the highest value of yield at 0.31 L H2 g-1 COD was obtained at QF and Vup of 1.7 L d-1 and 0.5 m h-1, respectively. The optimum conditions for the fermentative hydrogen production using pre-settled POME were QF = 2.03.7 L d-1 and Vup = 1.5-2.3 m h-1. The experimental results agreed very well with the model prediction. 1. Introduction

Bio-based energy is a sustainable and promising alternative to fossil fuel-based energy which can defend against a crisis in the energy supply and can protect the world from the approaching environmental calamity. Since the last couple decades, due to the limited availability of global fossil fuel reserves and concerns about global climate change from greenhouse gases emissions, microbial hydrogen production has received wide attention from many researchers due to the fact that hydrogen is a renewable energy source known for its non-polluting, sustainability, and environmentally friendly nature [1,2].

Biological H2 production as high H2 production rate (HPR) is a less energy intensive alternative to processes like water electrolysis, where processes can be operated at ambient temperature and pressure [3-5]. The biological technique necessitates a certain condition under which acidogens (H2 producing bacteria) and methanogens (H2 consuming bacteria) exhibit an imbalance in their activities resulting in the rapid accumulation of H2 [6,7]. H2

producing bacteria can utilize various forms of substrates. Published studies have shown that glucose [8,9], sucrose [10,11] and starch [12,13] are used as substrates for fermentative H2 production. However, numerous works also make use of wastewater as substrates such as rice winery wastewater [14,15], palm oil mill effluent (POME) [2,17], food waste [17,18] and dairy wastewater [19,20] due to their readily availability, low cost, high carbohydrate content and biodegradability in the biological processes of H2 production [17,21].

The POME is rich in organic carbon with a biochemical oxygen demand (BOD) value higher than 20 g L-1 and nitrogen content around 0.2 and 0.5 g L-1 as ammonia nitrogen and total nitrogen, respectiely [20]. Three main sources of the POME are sterilization runoffs (36%), clarification (60%) and hydrocyclone (4%) units. Raw POME as a colloidal suspension contains 95-96% water, 0.6-0.7% oil and 4-5% total solids [15,23]. It is estimated that 5-7.5 t of water are required for each tonne of crude palm oil production; which accounts for more than 50% of the used water converted to POME [24]. Biological H2 production from POME has been studied before [16,21,23].

Several studies of H2 production from POME have been accomplished using different reactors such as upflow anaerobic sludge blanket (UASB) that uses an anaerobic process whilst forming a blanket of granular sludge in the tank [24], anaerobic sequencing batch reactor which is a type of activated sludge process for the treatment of wastewater, and batch reactors (i.e. fermentors and serum bottles) [24,25]. The UASB reactor is capable of retaining high microorganism concentration and high rate of waste stabilization and could be an alternative reactor to generate biological H2. Nevertheless, the long start-up period (2-4 months), extreme variation in the up-flow velocities, and granules washout due to hydraulic stresses are the major problems associated with UASB reactors. Therefore, modification of the UASB process is needed to eliminate the problems in order to generate high performance H2 production from POME. In this study, a combination of UASB and up flow fixed film

bioreactor in a single reactor was used as modified UASB-fixed film (UASB-FF) bioreactor. The UASB-FF bioreactor has been examined in the treatment of different industrial wastewaters, such as slaughterhouse, distillery spent wash, starchy, fiberboard manufacturing, whey wastewater, and POME [25,26].

Vijayaraghavan and Ahmad [23] studied the effects of pH, chemical oxygen demand (COD) and hydrualic retention time (HRT) on biological H2 production from POME in an up-flow anaerobic contact filter. The highest H2 production and COD removal were achieved at the pH, COD concentration, and HRT of 5.0, 20 g L-1, and 7 d, respectively. Atif et al. studied the effect of H2 production from POME using microflora isolated from the sludge of an anaerobic pond treating POME. The batch experiments showed a total yield of 4708 mL H2 L-1 of POME with a maximum evolution rate of 454 mL H2 L-1 POME h-1 [21].

In this study, application of a modified UASB-FF bioreactor on biological H2 production from POME was investigated. The simultaneous effects of two independent operating variables, i.e., feed flow rate (QF) and up-flow velocity (Vup) on biological H2 production from POME in the UASB-FF bioreactor were studied, along with the optimization of the operating conditions using response surface methodology (RSM). 2. Materials and methods 2.1. Wastewater preparation

The reactor was fed with a pre-settled POME. POME samples were taken from SIME Darby Plantation Palm Oil Industry Sdn. Bhd., Nilai, Malaysia. The samples were transferred to the laboratory immediately and stored in a cold room (4 °C) before use. Various dilutions of POME were made using tap water. The pH and alkalinity of the feed was adjusted to 5.5 and 1400-1600 mg CaCOs L-1 using NaOH solution (3 N) and NaHCOs, respectively. Supplementary nutrients such as nitrogen (NH4Cl) and phosphorous (KH2PO4) were added to

give a ratio of COD: N: P of 550:7.4:1 [27]. The characteristics of the raw and pre- settled POME are summarized in Table 1.

2.2. Seed sludge preparation

The inoculum for the reactor was a digested sludge from the same palm oil mill industry. The sludge was initially passed through a screen to remove debris and solid particles and then heated at 100 °C for 1 h to enrich hydrogenic microbes. The initial volatile suspended solids (VSS) concentration of the seed was measured at 4600 mg L-1.

2.3. Experimental set-up

The laboratory-scale UASB-FF bioreactor (total volume 3.5 L, working volume 2.55 L, liquid height 80 cm) rig set-up was used in this study (Fig. 1). The glass bioreactor column was fabricated with an internal diameter of 55 mm at the bottom and middle parts and 75 mm at top part. The bioreactor comprised of three sections; the lowest section of the bioreactor was UASB reactor with the height of 60 cm (granular sludge portion) accommodated 67.8% of the working volume, the middle section of the bioreactor was fixed film reactor with the height of 15 cm accommodated 14.5% of working volume and the top section of the bioreactor accommodated 17.7% of the working volume consisting of gas-solid separator and outlet zone for fermented POME. The middle part of the column was packed with 70 Pall

rings (diameter and height 16 mm; specific surface area 341 m m- ). The void space ratio of the packed-bed reactor was 90.9%. To ensure isothermal operation of the UASB-FF reactor at the pre-determined temperature, water was circulated through the bioreactor jacket from a thermostated water bath equipped with a centrifugal pump (Lab. Companion, model: CW-05G, Korea).

2.4. Experimental design and mathematical model

Dark fermentative H2 production in up-flow bioreactors depends on a multitude of variables. The main factors that affect the process are QF, Vup, COD concentration, alkalinity,

pH and biomass concentration. In this study, two factors, i.e., QF and Vup were chosen as the independent variables for investigation.

The range of QF and Vup investigated for biological H2 production from POME were 1.7-10.2 L d-1 and 0.5-3.0 m h-1, respectively. In this study, the influent COD concentration of pre-settled POME was maintained at 15000 mg L-1 in all experiments. Therefore, QF was determined to be in the range of 1.7 to 10.2 L d-1 (corresponding to HRT of 36 to 6 h) in order to find the optimum conditions for improved effluent quality and process stability. This would cover an organic loading rate (OLR) range of 10 to 60 g COD L-1 d-1. The attainment of steady state condition for each experiment is expected after several turnovers.

The statistical method of design of experiment eliminates systematic errors with an estimate of the experimental error and minimizes the number of experiments [26,28]. In the present study, H2 production from POME in an UASB-FF bioreactor was evaluated and optimized employing RSM with a central composite design (CCD). The main and interactive effects of QF and Vup on the process responses were studied. Table 2 shows the experimental conditions for fermentative H2 production from POME based on the CCD. The design includes of 2k factorial points augmented by 2k (the number of variables) axial points and a center point. The levels of each variable vary from a low to high value, either numerically expressed in their actual values or coded as -1 (low) and 1 (high) with intermediate value of 0 (middle).

In order to perform a comprehensive analysis of the fermentative H2 production process, eight dependent parameters viz. H2 percentage in biogas, H2 yield, hydrogen production rate, specific hydrogen production rate, COD removal efficiency, effluent pH, effluent bicarbonate alkalinity (BA), and effluent total volatile fatty acid (TVFA) as simultaneous responses were evaluated. The coefficients of the polynomial model were obtained using polynomial as shown below [29]:

7 = bo + biXi + bjXj + bn x i2 + bjjXj2 + bjXlx] +...

where 7 is the predicted response, Д is the ith linear coefficient, fin is the ith quadratic coefficient and fa is the ijth interaction coefficient, x;x/ are input variables which influence the response variable 7. The terms of model may be selected or rejected based on P value with 95% confidence level. The results were analyzed by applying analysis of variance (ANOVA) in Design Expert software. The effects and simultaneous interaction of the variables on the responses were visualized using three dimensional plots and their respective contour plots. The optimum region is also recognized based on the main responses in the overlay plot. 2.5. Analytical techniques

The parameters viz. BOD, COD, total suspended solids, VSS, alkalinity, total Kjeldahl nitrogen, oil and grease, and pH were analyzed using procedures outlined in the APHA Standard Methods [30]. The biogas composition was determined using a gas chromatograph (Perkin Elmer, Auto system GC), equipped with thermal conductivity detector (TCD) and data acquisition system with Total Chrom® software. H2 content wasalso analyzed by GC-TCD fitted with a 1.5 m stainless steel column (SS350 A) packed with a molecular sieve (80/100 mesh). The temperature of the injection port, oven and detector were 80, 200, and 200 °C, respectively. Argon was used as a carrier gas at a flow rate of 30 mL min-1. The liquid samples for VFA determination were analyzed by a gas chromatograph (Perkin Elmer, Auto system GC) equipped with a flame ionization detector. The oven temperature was maintained at 175 °C while the injector and detector temperatures were kept at 200 and 220 °C, respectively. The carrier gas (helium) flow rate was set at 40 mL min-1.

3. Results and discussion 3.1. Effects of Qf and Vup on biogas composition, H2 yield, HPR, and Specific HPR (SHPR)

To evaluate the bioreactor performance, H2 production was monitored in terms of H2 content in the biogas, H2 yield, HPR and SHPR as responses. The results obtained based on

the experimental conditions are shown in Table 2. The three dimensional plots (Fig. 2) show the variations in H2 content in the biogas, H2 yield, HPR and SHPR as functions of QF and Vup. According to Table 3, a reduced quadratic model described the variation in these responses as a function of variables being studied. The response surface plot for H2 content in the biogas is shown in the Fig. 2a. According to the model presented in Table 3 and the Fig. 2a, positive effect of increase in Vup and interactive effect of QF and Vup on H2 content at Vup values less than 1.75 m h-1 was attributed to the increase in pH caused by recycled alkalinity; while the negative effect of the increase in Vup with second order on H2 percentage dominated at Vup values greater than 1.75 m h-1. The H2 percentage was found to increase with an increase in Vup at QF higher than 8 L d-1, which indicated the influencing effect of Vup on the improvement of the H2 content in the biogas from the system. The contour plot of H2 yield (Fig. 2b) showed that an increase in the response was given by a decrease in the variables. The highest value of yield was 0.31 L H2 g-1 COD at Qf and Vup of 1.7 L d-1 (i.e., HRT = 1.5 d) and 0.5 m h-1 (i.e., recycle flow rate, QR = 15.8 L d-1), respectively. The model predicted 0.30 L H2 g-1 COD under these conditions.

The contour plots of of HPR and SHPR (Fig. 2c-2d) show the HPR and SHPR as a function of Vup and QF. In the lower range of Vup, HPR and SHPR decrease with increasing Qf, while in the higher range of Vup, increasing QF did not affect significantly HPR and SHPR. This was interpreted as an indication of the commencement of an unstable condition at lower range of Vup due to very high QF/QR ratio and inappropriate food to microorganism (F/M) ratio. It was also found that HPR and SHPR values under reactor destabilization condition (the highest QF) was high because of higher F/M ratio as SHPR is calculated by the value of produced H2 and the value of biomass concentration within the bioreactor (Table 2). The interactive effects of the studied variables on the responses related to H2 production were well explained by the models (probability of lack of fit at P > 0.05).

3.2. Effects of Qf and Vup on COD removal

The COD removal efficiency was also monitored to evaluate the reactor performance. Figure 3 shows three-dimensional plots for the variations of COD removal efficiency as a function of QF and Vup. The COD concentration of the feed flow was constant throughout the experiments. The different influence of Vup on COD removal efficiency was observed at constant feed flow rates QF. The COD removal efficiency at feed flow rates lower than 3.6 L d-1 decreased slightly as Vup increased while at feed flow rates higher than 3.6 L d-1 the COD removal efficiency decreased significantly with declining Vup that was attributed to high OLR with short HRT. The lowest COD removal efficiency was 44% at the OLR of 60 g COD L d-1 (the highest feed flow rate). It was attributed to an incomplete fermentation process and significant production of acids (the system pH was lowest at this point). The COD removal efficiency decreased considerably as the feed flow rate increased. The negative effect of feed flow rate on COD removal efficiency was deminishing with increasing the Vup. The change in COD removal at the lower limit of Vup (0.5 m h-1) was 22%, while the corresponding value at the upper limit of Vup (3 m h-1) was 5%. The improvement of substrate diffusion rate into the granules due to higher flow velocity may determine the trend of COD removal efficiency at low Qf [27,31]. Meanwhile, trend at high OLR demonstrated that at the effect of substrate concentration on the diffusion rate was stronger than that of Vup. It is concluded that to eliminate the negative effects of organic shock caused by high QF, appropriate value of Vup for each OLR is required. The trend of COD variationsas a function of the studied variables could be described well by a first order polynomialmodel with a good correlation coefficient of R2 = 0.93 (Table 3).

3.3. Effects of Qf and Vup on pH, bicarbonate alkalinity and total volatile fatty acids

Figure 4 shows the three-dimensional contour plots for pH, BA and TVFA obtained from the models listed in Table 3. There was a relatively strong interaction between the

variables (P values < 0.0005) as demonstrated by the graphs' curvatures. In Fig. 4a, increasing Vup at various feed flow rates resulted in dynamic changes in the pH. An increase in Vup caused a decrease in pH at low QF which could be due to low rate of alkalinity production and higher CO2 dissolution while at high feed flow rates, the increase in pH was attributed to alkalinity build-up following intensified metabolic activities. Similar observation was made during effluent recycle.

The significant effects of the variables interaction on BA are shown in Fig. 4b. From the figure, the highest value of BA was obtained at the middle level of the variables and was reduced as QF and Vup moved away from the central region simultaneously. It is suggested that the BA was generated from dark fermentation H2 production reactions from POME, allowing some extent of pH control, except for overloading conditions (QF 10.2 L d-1, Vup 0.5 m h-1).

The values of the effluent VFA measured at steady state condition of each run are summarized in Table 4. The main aqueous products in this study were butyrate, acetate and ethanol. Propionate concentration was relatively low during most runs. The effect of QF and Vup on the yield of VFAs and ethanol was evaluated using RSM with a CCD. To fit the data, a logarithmic function with base 10 was applied due to the high ratio between the maximum and minimum effluent TVFA value (1.61). The high regression

coefficient R2 (0.955)

suggests that the regression equation of TVFA yield was appropriate for modeling the experimental results. Figure 4c shows the three-dimensional contour plots for the TVFA as a function of the studied variables. Both QF and Vup had a significant individual influence on the yield of TVFA. The minimum TVFA yield was estimated from model to be 2454 mg L-1 when Vup and QF were 1.13 m h-1 and 10.2 L d-1, respectively. Figure 3c illustrates that the TVFA increased significantly as the Vup increased at fixed QFs. At Vup higher than 1.45 m h-1, the TVFA yield increased logarithmically with increasing Vup. In addition, the model was

significant (P < 0.0001), which indicated a satisfactory explanation of the dynamics of the effluent TVFA by the model. The minimum and maximum levels of butyrate were 471 and 1500 mg L-1 corresponding with the lowest and highest H2 yields 0.01 and 0.31 L H2 g-1 COD, respectively. At overloading (QF 10.2 L d-1, HRT 6 h) where reactor destabilization was observed, the effluent VFA composed of 1210, 471, 630, and 510 mg L-1 of acetic, butyric, propionic acids and ethanol, respectively. 3.4. Process optimization

To determine the optimum region for the studied variables, the responses that were used to check the bioreactor performance in this study were modeled. H2 percentage, H2 yield, SHPR and COD removal as responses were each optimized as a function of the studied variables viz. QF and Vup. Figure 5 illustrates the identified optimum region (shaded in white) based on the four responses where their values are greater than those shown in the overlay plot. These parameters were chosen due to their considerable importance for a reliable representation and optimization of the fermentative H2 production process.

To confirm the reliability of the models' predictions, two points within the optimum region were chosen for verification experiments. They were as follows: one point at the maximum feed flow rate (QF 3.71 L d-1 with the corresponding Vup 1.48 m h-1) and the other point at maximum Vup (2.31 m h-1 with the corresponding QF 2.03 L d-1) within the optimum region. Experiments were carried out to verify the experimental results against the model predicted values. Table 5 presents the experimental conditions and results. It is shown that the verification experimental results agree reasonably well to those corresponding values predicted by the model under the optimum conditions. This strongly demonstrated that the optimization of the H2 production in the UASB-FF bioreactor using RSM with a CCD analysis was successful. 4. Conclusions

RSM was used to optimize simultaneous effects of two independent operating variables (QF and Vup) on biohydrogen production from POME in a UASB-FF bioreactor in this study. The operating variables were varied to cover a wide range of OLR from 10 to 60 g COD L d-1. Experimental results show that the highest H2 yield was 0.31 L H2 g-1 COD at QF and Vup of 1.7 L d-1 and 0.5 m h-1. The optimum ranges for the fermentative hydrogen production of the pre-settled POME were Qf = 2.1-3.7 L d-1 and Vup = 1.5-2.3 m h-1. References

[1] Modarresi A, Wukovits W, Foglia D, Friedl A. Effect of process integration on the exergy

balance of a two-stage process for fermentative hydrogen production. J Clean Prod 2010;18:63-71.

[2] Mohammadi P, Ibrahim S, Annuar MSM, Law S. Effects of different pretreatment methods on anaerobic mixed microflora for hydrogen production and COD reduction from palm oil mill effluent. J Clean Prod 2011;19:1654-8.

[3] Younesi H, Najafpour G, Ku Ismail KS, Mohamed AR, Kamaruddin AH. Biohydrogen

production in a continuous stirred tank bioreactor from synthesis gas by anaerobic photosynthetic bacterium: Rhodopirillum rubrum. Bioresour Technol 2008;99:2612-9.

[4] Peintner C, Zeidan AA, Schnitzhofer W. Bioreactor systems for thermophilic fermentative hydrogen production: evaluation and comparison of appropriate systems. J Clean Prod 2010;18:15-22.

[5] Chen CC, Sen B, Chuang YS, Tsai CJ, Lay CH. Effect of effluent recycle ratio in a continuous anaerobic biohydrogen production system. J Clean Prod 2012;32:236-43.

[6] Mohammadi P, Ibrahim S, Annuar MSM. Comparative study on the effect of various pretreatment methods on the enrichment of hydrogen producing bacteria in anaerobic granulated sludge from brewery wastewater. Korean J Chem Eng 2012;29:1347-51.

[7] Jamil Z, Annuar MSM, Ibrahim S, Vikineswary S. Optimization of phototrophic hydrogen production by Rhodopseudomonas palustris PBUM001 via statistical experimental design. Int J Hydrogen Energy 2009;34:7502-12.

[8] Hallenbeck PC, Benemann JR. Biological hydrogen production; fundamentals and limiting processes. Int J Hydrogen Energ 2002;27:1185-93.

[9] Monmoto M, Atsuka M, Atif AAY, Ngan MA, Fakhru'l-Razi A, Iyuke SE, et al. Biological production of hydrogen from glucose by natural anaerobic microflora. Int J Hydrogen Energ 2004;29:709-13.

[10] Chen CC, Lin CY, Chang JS. Kinetics of hydrogen production with continuous anaerobic cultures utilizing sucrose as the limiting substrate. Appl Microbiol Biotechnol 2001;57:56-64.

[11] Lee KS, Hsu YF, Lo YC, Lin PJ, Lin CY, Chang JS. Exploring optimal environmental factors for fermentative hydrogen production from starch using mixed anaerobic microflora. Int J Hydrogen Energ 2008;33:1565-72.

[12] Akutsu Y, Li YY, Harada H, Yu HQ. Effects of temperature and substrate concentration on biological hydrogen production from starch. Int J Hydrogen Energ 2009;34:2558-66.

[13] Su HB, Cheng J, Zhou JH, Song WL, Cen KF. Improving hydrogen production from cassava starch by combination of dark and photo fermentation. Int J Hydrogen Energ 2009;34:1780-6.

[14] Yu HQ, Zhu ZH, Hu WR, Zhang HS. Hydrogen production from rice winery wastewater in an upflow anaerobic reactor by using mixed anaerobic cultures. Int J Hydrogen Energ 2002;27:1359-65.

[15] Mohammadi P, Ibrahim S, Annuar MSM. Effects of biomass, COD and bicarbonate concentrations on fermentative hydrogen production from POME by granulated sludge in a batch culture. Int J Hydrogen Energ 2012;37:17801-8.

[16] Chong ML, Rahim RA, Shirai Y, Hassan MA. Biohydrogen production by Clostridium butyricum EB6 from palm oil mill effluent. Int J Hydrogen Energ 2009;34:764-71.

[17] Van Ginkel SW, Oh SE, Logan BE. Biohydrogen gas production from food processing and domestic wastewaters. Int J Hydrogen Energ 2005;30:1535-42.

[18] Kim SH, Han SK, Shin HS. Optimization of continuous hydrogen fermentation of food waste as a function of solids retention time independent of hydraulic retention time. Process Biochem 2008;43:213-8.

[19] Mohan SV, Babu VL, Sarma PN. Anaerobic biohydrogen production from dairy wastewater treatment in sequencing batch reactor (AnSBR): effect of organic loading rate. Enzyme Microb Technol 2007;41:506-15.

[20] Yang PF, Zhang RH, MeGarvey JA, Benernann JR. Biohydrogen production from cheese processing wastewater by anaerobic fermentation using mixed microbial communities. Int J Hydrogen Energ 2007;32:4761-71.

[21] Atif AAY, Fakhru'l-Razi A, Ngan MA, Morimoto M, Iyukec SE, Veziroglu NT. Fed batch production of hydrogen from palm oil mill effluent using anaerobic microflora. Int J Hydrogen Energ 2005;30:1393-7.

[23] Vijayaraghavan K, Ahmad D. Biohydrogen generation from palm oil mill effluent using anaerobic contact filter. Int J Hydrogen Energ 2006;31: 1284-91.

[24] Pakarinen O, Lehtomaki A, Rintala J. Batch dark fermentative hydrogen production from grass silage: The effect of inoculum, pH, temperature and VS ratio. Int J Hydrogen Energ 2008;33:594-601.

[25] Ismail I, Hassan MA, Rahman NAA, Soon CS. Thermophilic biohydrogen production from palm oil mill effluent (POME) using suspended mixed culture. Biomass Bioenergy 2010;34:42-7.

[26] McHugh S, Collins G, O'Flaherty V. Long-term, high-rate anaerobic biological treatment of whey wastewaters at psychrophilic temperatures. Bioresour Technol 2006;97:1669-78.

[27] Zinatizadeh AAL, Mohamed AR, Mashitah MD, Abdullah AZ, Isa MH. Optimization of pre-treated palm oil mill effluent digestion in an up-flow anaerobic sludge fixed film bioreactor: a comparative study. Biochem Eng J 2007;35:226-37.

[28] O-Thong S, Prasertsan P, Karakashev D, Angelidaki I. Thermophilic fermentative hydrogen production by the newly isolated Thermoanaerobacterium thermosaccharolyticum PSU-2. Int J Hydrogen Energ 2008;33:1204-14.

[29] Khuri AI, Cornell JA. Response Surfaces: Designs and Analyses. 2nd ed. New York: CRC Press; 1996.

[30] APHA. Standard Methods for the Examination of Water and Wastewater. 20th ed. Washington, DC: American Public Health Association; 1999.

[31] Beyenal H, Lewandowski Z. Combined effect of substrate concentration and flow velocity on effective diffusivity in biofilms. Water Res 2000;34:528-38.

Table 1. Characteristics of raw and per-settled POME for 3 samples (All in mg L-1 except pH)

Parameters Raw POME Pre-Settled POME

Amount Standard deviation Amount Standard deviation

BOD 22500 ± 2150 19400 ± 235

TCOD 49800 ± 4250 28800 ± 260

SCOD 21950 ± 2320 17140 ± 165

TSS 18800 ± 1240 850 ± 21

TKN 430 ± 32 382 ± 4

TP 84 ± 7.5 84 ± 2

Oil & Grease 4250 ± 62 1480 ± 14

pH 4.2 - 4.2 -

Table 2. Experimental conditions and results of central composite design

Variables Response

Run no.

Qf (l d-1)

Vup (m h-1)

COD removal


H2 percentage (%)

H2 yield (L H2 g-1 COD)

(L H2 g-1

CODrem d-1)

SHPR (L H2 g-1

VSS d-1)

TVFA (mg acetic

acid L-1)

(mg CaCOs L-1)

1 1.7 0.5 68 53 0.31 0.31 0.15 5.70 3650 523

2 5.95 0.5 59 43 0.06 0.24 0.10 5.38 2860 1930

3 5.95 1.75 60 52 0.07 0.26 0.12 5.47 2860 2210

4 5.95 1.75 59 53 0.07 0.26 0.12 5.45 2965 1990

5 1.7 1.75 66 70 0.25 0.25 0.12 5.63 2920 1280

6 5.95 1.75 60 54 0.07 0.25 0.11 5.50 2770 2310

7 10.2 0.5 44 21 0.01 0.10 0.03 4.78 2705 125

8 1.7 3.0 65 60 0.20 0.21 0.10 5.42 3990 890

9 10.2 1.75 56 53 0.03 0.22 0.09 5.50 2475 1540

10 5.95 1.75 61 53 0.07 0.25 0.11 5.57 2850 2210

11 5.95 1.75 59 53 0.06 0.23 0.10 5.47 2880 2090

12 10.2 3.0 59 58 0.03 0.20 0.09 5.39 3780 245

13 5.95 3.0 61 47 ^0.05 0.21 0.09 5.27 3970 1410

3 Table 3. ANOVA results for the equations of the Design Expert 6.0.8 for studied responses as predictor variables

4 (A: Qf; B: Vup; SD: standard deviation; CV: coefficient of variation; PRESS: prediction error sum of squares)

Response Transformation Modified Equations with Significant Terms Probability R2 SD CV PRESS Probability for lack of fit

H2 fraction - 53.27 - 8.5A + 7.63B + 6.26A2 - 10.24B2 + 7.13AB 0.003 0.890 4.79 9.3 1299 0.059

H2 yield - 0.062 - 0.11A - 0.016B +0.076A2 + 0.032AB < 0.0001 0.991 0.01 11.1 4.7*10-3 0.16

HPR - 0.25 - 0.039A - 0.037B2 + 0.051AB 0.0004 0.859 0.02 9.1 0.01 0.093

SHPR - 0.11 - 0.026A - 0.015B2 + 0.27AB < 0.0001 0.919 0.01 8.5 1.9*10-3 0.14

COD removal - 59.66 - 6.78A + 2.4B + 4.4AB < 0.0001 0.929 1.82 3.1 111 0.11

Effluent pH TVFA - 5.53 - 0.18A - 0.21B2 + 0.22AB 0.0005 0.848 0.10 1.9 0.33 0.16

Base 10 log 3.45 -0.027A + 0.054B + 0.09B2 + 0.027AB < 0.0001 0.9551 0.018 0.52 0.01 0.0723

Effluent - 2201.76 - 113.67A - 955.67A2 - 694.66B2 < 0.0001 0.9061 267.66 18.59 1.668*10+6 0.1071

6 Table 4. The amounts of the effluent VFAs at various trials

Run Acetate Butyrate Propionate Ethanol TVFA

no. mg L-1 mg L-1 mg L-1 -1 mg L (mg Acetic acid L-1)

1 1370 1500 570 615 3654

2 1080 1340 160 570 2864

3 1215 1155 210 525 2855

4 1260 1175 230 552 2965

5 653 1468 340 765 2924

6 1151 1149 208 516 2774

7 1210 471 630 510 2705

8 1320 1650 435 915 3988

9 850 1085 145 590 2475

10 1218 1153 214 517 2850

11 1226 1174 215 520 2877

12 1720 865 570 774 3779

13 1154 1376 835 925 3972

8 Table 5. Verification experiments at optimum conditions (Eff. = Effluent)

Responses COD removal efficiency (%)

Run Conditions H2 Percentage (%) H2 yield (L H2 g-1 COD) HPR (L H2 g-1 CODrem d-1 SHPR ^ (L H2 g-1 VSS d- 1) Eff. pH Eff. BA (mg CaCOs L-1) Eff. TVFA (mg acetic acid L-1)

1 Qf = 3.71 L d-1 Vup = 1.48 m h-1 Experimental values 53.3 0.14 0.26 0.12 61 5.55 1820 3145

Predicted values 52.8 0.15 0.27 0.13 63 5.64 1964 2910

2 Qf = 2.03 l d-1 Vup = 2.31 m h-1 Experimental values 62.2 0.21 0.25 0.11 61 5.53 1315 3380

Predicted values 64.9 0.21 0.26 0.12 65 5.57 1355 3265

9 10 11 12

Feed Tank

Gas SampSrtg port

Gas i-cfcd separator ~~


CraerSBJc graiular Siudge portion

---- "

S&aQ . - f-

N Flow distributer

Peri&tdtk Pump [Feed}

Setting tank

Perietiftii: Pump {Recyde Effluent)

Fig. 1. Schematic diagram of the experimental set-up.

2.38 ss 1.75

Up-Flow Velocity (m h-1) i 13

0.50 N 1.70

Feed Flow Rate (L d-1)

) 0.301

8 0.227 £ 0.154 | 0.081 % 0.008

2.38 s„. 1.75

Up-Flow Velocity (m h-1) 1.13

0.50 N 1.70

Feed Flow Rate (L d-1)

0.30 0.25 0.21 0.16 0.12

Up-Flow Velocity (m h-1) 1.13

Feed Flow Rate (L d-1)

Up-Flow Velocity (m h-1) 1.13

Feed Flow Rate (L d-1)

0.50 1.70

0.50 1.70

Fig. 2. Three-dimensional contour plots for (a) H2 percentage, (b) H2 yield, (c) HPR, and (d) SHPR.

16 Fig. 3. Three-dimensional contour plot of the model for COD removal.

20 Fig. 4. Three-dimensional contour plots of the two-factor interaction models for (a)

21 TVFA, (b) bicarbonate alkalinity, (c) effluent pH.

22 Fig. 5. Overlay plot showing optimal region (white color).