Professional Documents
Culture Documents
a r t i c l e i n f o a b s t r a c t
Article history: An optimization study on the mesophilic anaerobic treatment of palm oil mill effluent (POME) in an up-
Received 29 August 2014 flow anaerobic sludge blanket–hollow centered packed bed (UASB–HCPB) reactor was conducted for the
Received in revised form 20 January 2015 first time by using kinetics modeling. Monod model was opted for the optimization due to its capability of
Accepted 26 January 2015
describing the anaerobic digestion and the involvement of adequate amount of parameters in the kinetics
model. The best-fit kinetic constants obtained from previous work were incorporated in the optimization
Keywords:
model. Fuzzy optimization was adopted and it yielded the optimized chemical oxygen demand (COD)
Multi-objective optimisation
removal of 86.7% and volumetric methane production rate of 0.448 LCH4 /L.day at hydraulic retention time
Anaerobic
Palm oil mill effluent (POME)
(HRT) of 4.33 days, organic loading rate (OLR) of 14.75 gCOD/L.day and mixed liquor volatile suspended
Fuzzy optimisation solid (MLVSS) concentration of 19,130 mg/L. A confirmative experiment was conducted and the results
Kinetics modeling validate the legitimacy of the fuzzy optimization model. Further studies are required to improve the
reliability of this newly introduced optimization technique in the wastewater treatment industry.
© 2015 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jwpe.2015.01.005
2214-7144/© 2015 Elsevier Ltd. All rights reserved.
Y.J. Chan et al. / Journal of Water Process Engineering 5 (2015) 112–117 113
Table 1
Nomenclature Summary of biochemical analytical methods.
Parameter Procedure
Symbols
B specific methane yield (L CH4 /g CODremov ) COD Sample was diluted and COD was measured using HACH
spectrophotometer (DR2800, Loveland, CO).
Bo maximum specific methane yield (L CH4 /g
BOD Sample was diluted with de-ionized water and incubated
CODremov ) for 5 days at 30◦ C and BOD5 was evaluated with a DO
KS half velocity coefficient (g/L) meter.
Kd endogenous decay coefficient (/d) TSS MLSS Diluted sample was analysed by using calorimeter.
k first order kinetics constant (s−1 ) TS ML VSS VSS Follow the standard Methods for the Examination of Water
and Wastewater where the samples was filtered through a
M volumetric methane production rate (L CH4 /L.d)
glass fiber filter (Whatman grade GF/A, 1.6 m, UK) and the
Q flow rate (L/d) residue retained on the filter was dried in the oven
So influent COD (mg/L) (Memmert, Germany) at 105 ◦ C for 24 h where MLSS and
Se effluent COD (mg/L) VSS was determined by ash-ing for 3 h in muffled furnace
(Carbolite, UK) at 550 ◦ C.
V effective working volume of reactor (L)
Biogas Biogas production was measured by using water
X MLVSS concentration in the reactor (mg/L) displacement method with a 1 L inverted water-filled
Y growth yield coefficient (g/g) graduated cylinder for a collection period of 1 h, four times
a day and average value for the particular day was
Greek letters estimated.
TA 20 ml of sample was being tritrated with 0.1 N of sulphuric
m maximum specific microbial growth rate (/d)
acid. The total volume of titrant required to acidify the
uf upflow velocity (m/h) sample to pH of 4.3 and pH of 4.0 was used to calculate TA.
H hydraulic retention time, HRT (d) VFA Measurement of VFA concentrations was based on the
C solid retention time, SRT(d) HACH Method 8196 esterification for volatile acid.
fuzzy degree of satisfaction‘
Table 2
Operating conditions of UASB–HCPB.
Parameter Values
to maximize the treatment performance at lower treatment cost. HRT (d) 4.33
Optimization of the UASB–HCPB reactor has only been conducted MLVSS (mg/L) 19,310
SRT (d) 136
by using one-at-a-time strategy [17]. Thus far, optimization of
OLR (gCOD/L.d) 14.75
bioreactor for POME treatment by using kinetics modeling has Influent Flow rate, Q (L/d) 1.15
not been performed. The conventional one-at-a-time optimization
strategy is time consuming, fails to generate insightful results and
it does not guarantee a global optimum solution. Fuzzy optimiza-
2.2. Bacteria source
tion resembles RSM, where optimization of multiple objectives can
be achieved with several predetermined constraints. However, it
The seed granules collected from Pan Century Oleo Chemi-
is more straightforward and produces more insightful results than
cals, Pasir Gudang in Johor, Malaysia were used to inoculate the
that of RSM.
UASB–HCPB bioreactor. The characteristic of the seed granule is
Besides, certain parameters are contradicting in nature. For
presented in Table 2.
instance, hydraulic retention time, (HRT) of a bioreactor is required
to be as high as possible in order to yield high COD removal. How-
ever, high HRT is not desired as it would lead to high treatment 2.3. Reactor configuration
and maintenance costs. Hence, it is worthy to mention that fuzzy
optimization has been adopted in the modeling works of [9,14] in The basic configuration of UASB–HCPB is described in Fig. 1 and
the integrated biorefinery industry where multiobjective optimiza- the operating procedure can be found elsewhere [17]. As mentioned
tions from a set of contradicting parameters in a design problem above, UASB–HCPB is a hybrid reactor which highly resembles the
have been achieved successfully. However, the application of fuzzy UASFF. The effective working volume of the reactor is 5 L with an
optimization on wastewater treatment is still in its infancy. There- internal diameter of 12 cm, height of 53 cm and the packed bed
fore, the aim of this study is to optimize the performance of the section was filled with ceramic packing materials. The reactor was
UASB–HCPB in order to strike a balance between the contradict- operated at mesophilic condition (28 ◦ C) as no external heating was
ing parameters. Fuzzy optimization coupled with kinetics modeling applied.
is presented for the first time in the optimization of reactor for
POME treatment. Fuzzy Optimization is able to identify the opti- 2.4. Biochemical analytical methods
mum operating conditions such that the satisfactory level () for
the optimization problem is maximized. The parameters that were evaluated during the entire experi-
mental period are COD, BOD, Total suspended solid (TSS), volatile
suspended solids (VSS), total alkalinity (TA), and volatile fatty acids
2. Methodology (VFA) concentrations of the effluent, as well as pH, mixed liquor
suspended solid (MLSS), and mixed liquor volatile suspended solid
2.1. Wastewater preparation (MLVSS) of the reactor together with biogas production rate.
The raw POME was obtained from Seri Ulu Langat Palm Oil Mill, 2.5. Operating conditions
Dengkil. Several characteristics of the POME were analysed so that
they can be used to design the experiments. The characteristic of The UASB–HCPB reactor was operated under the operating con-
the POME is presented in Table 1. ditions obtained from the optimization model in order to verify
114 Y.J. Chan et al. / Journal of Water Process Engineering 5 (2015) 112–117
the legitimacy of the model. The operating conditions are given as Bo was obtained from plotting against HRT based on Eq. (5),
Table 2. given by:
So H 1
= + (5)
2.6. Development of optimization model M Bo Bo k
Eq. (4) and non linear form of Monod equation (Eq. (6)) are used
Volumetric methane production rate, M is defined as unit vol- to form the optimization model,
ume of methane produce per unit volume of reactor per day. It is
another important parameter that has to be optimized. The devel- Q (So − Se ) Xm Se
= (6)
opment of the kinetic model was revised from Ma et al. [11] and V Y (ks + Se )
reported as below:
2.7. Fuzzy optimisation
Bo
M= (So − Se ) (1) The idea of fuzzy optimisation is to integrate the multiple objec-
H
tives into a continuous interdependence variable, within the
From Monod Eq. (2), Se can be rearranged in terms of the SRT model. The is a measure of the satisfactory level of the solution.
and other kinetic constants to give Eq. (3), The optimization objective of is maximized subject to the prede-
fined upper and lower bounds of the objectives. A value of unity
1 m Se signifies all the objective functions have attained the maximum or
= − kd (2)
c Ks + S e minimum values in the bounds. On the contrary, value of 0 indi-
cates that the objectives that are to be maximize have attained a
ks (1 + kd × c ) minimum in the bounds and vice versa. LINGO 14.0 was used for the
Se = (3)
c × (m − kd ) − 1 optimization problem and the steps for building the optimization
model are as follow:
Substituting (3) into (1), the volumetric methane production
I. Eqs. (4) and (6) were used to form part of the optimiza-
rate equation was developed to give (4):
tion model. Then, the objective functions were defined. For Monod
model, the four objective functions set were: (i) Maximize COD
Bo Ks (1 + kd × c )
M= So − (4) removal; (ii) Minimize HRT required; (iii) Maximize volumetric
H c × (m − kd ) − 1
methane production; (iv) Maximize MLVSS required
Y.J. Chan et al. / Journal of Water Process Engineering 5 (2015) 112–117 115
(obtained from Step IV). The fuzzy equation was defined as follow Objective Function Lower bound Upper bound max
to complete the optimization model:
Se (mg/L) 8320 8753.6 8,503
HRT (d) 4.28 4.42 4.33
P − PL
For maximizingavariable : U ≥ (7) MLVSS (mg/L) 19129.63 19475.34 19,310
P − PL M (LCH4/L.d) 0.27 0.28 0.28
COD Removal Efficiency (%) – – 86.7
PU − P SRT (d) – – 136
For minimizing a variable : ≥ (8)
PU − PL OLR (gCOD/L.d) – – 14.75
Influent Flow rate, Q (L/d) – – 1.15
Where P represents the variables, U indicates the upper bound – – 0.52
value while L indicates the lower bound variables.
VII. The single objective function which was defined in the ini-
tial step was removed and the objective function was altered to
treatment unit [3]. The optimum OLR should fall between 12.8 and
maximize the degree of satisfactory, . The complete optimiza-
15.0 gCOD/L.d. From the authors’ experience, the HRT should be
tion problem created was solved to identify the optimum operating
greater than 3 days to avoid washout of biomass.
condition for the reactor.
SRT should be sufficient to allow sufficient growth of anaerobic
Objective function (i) is set as maximizing COD removal will
bacteria in order to compensate the biomass loss in effluent. The
lead to high treatment efficiency and thus reducing the complex-
SRT for anaerobic process must be > 20 days [13].
ity of the subsequent treatment and operating cost. On top of that,
Up-flow velocity <0.1 m/h [6], generally attainable with
treated effluent with high quality can be achieved. Besides, in objec-
HRT > 3days
tive function (ii), minimizing HRT can increase the productivity of
the reactor because larger amount of effluent can be treated for the
same amount of timeframe which reduces the operating cost. Also, 3. Results and discussion
minimum HRT reduces the reactor volume which leads to lower
capital cost. Thirdly, volumetric methane production rate is maxi- 3.1. Fuzzy optimization
mized in objective function (iii). This is owing to the facts that larger
amount of methane will increase the revenue of the treatment Monod model (Eq. (6)), and volumetric methane production
plant as the methane generated can be used by the plant itself for rate equation (Eq. (4)) had been incorporated into the optimization
power/heat generation or to be sold. On top of that, higher methane model. By using Lingo 14.0, with the use of optimum set of kinetic
production signifies high COD removal. Lastly, MLVSS is maximized constants as shown in Table 3 and appropriate fuzzy equations, the
in objective function (iv). It is because based on Monod model, the optimized value for each objective function is tabulated in Table 4.
operating parameters that can affect the COD removal efficiency are Note that the influent COD used in the model was assumed to
HRT and MLVSS. MLVSS is used as measures of the microorganism be 64,000 mg/L [2,3]. Nevertheless, by assuming different influent
concentration in the wastewater treatment system. High MLVSS COD concentration in the model, another solution can be obtained
concentrations are required to ensure the bacterial population is Table 5.
adequate to anaerobically degrade the organic matter efficiently. The value of Bo determined from Fig. 2 was 0.019 LCH4 /gCOD.
From the literatures, the values of Bo for daily manure ranges from
2.8. Constraints 0.21 to 0.27 LCH4 /gCOD [4,7]. All the constraints set in the opti-
mization model are satisfied such that the HRT is 4.33 days, MLVSS
The effluent from UASB–HCPB will be sent to aerobic treatment is 19,310 mg/L, SRT is 136 days, OLR is 14.75 gCOD/L.d and up-flow
unit to polish the treatment efficiency. Hence, it is crucial to ensure velocity is 0.004 m/h. A low value of was obtained from the opti-
adequate but not excessive removal of organic matters in anaerobic mization model due to the highly inter-correlated variables [10].
treatment unit for effective functioning of aerobic unit. Chan, et al. The value also signifies that each variable considered in the model
[3] recommended that the COD removal in anaerobic treatment was settled in the mid of their bounds. It is observed that the bound
should be maintained in the range of 77–87%. for each variable of interest was highly constricted. As such, the sat-
MLVSS concentration in the reactor is another important fac- isfaction value is highly sensitive and a small change in the value
tor that needs to be selected as a controlled operating condition.
Deficient in MLVSS concentration will lead to failure to sustain Table 5
high bioactivity in the reactor while excessive MLVSS concentra- Comparison between response surface methodology (RSM) and fuzzy optimisation.
tion will lead to a lower methane yield [17]. Therefore, the MLVSS Variable Response surface Fuzzy
concentration should be maintained between 15,000 mg/L [17] and methodology optimization
42,000 mg/L [3]. (RSM)
A low value of OLR has to be set in order to avoid unnecessary OLR (gCOD/L.day) 12.8 14.75
investment of treatment facilities. In other words, the maximum MLVSS (mg/L) 40200 19310
HRT is settled by this constraint. Conversely, high OLR will lead to COD removal efficiency (%) 85 86.7
Reference Chan et al. [3] This study
low capital cost but poorer treatment efficiency of the anaerobic
116 Y.J. Chan et al. / Journal of Water Process Engineering 5 (2015) 112–117
[14] T.L. Ng, M.H. Hassim, K.S. Ng, Process synthesis and optimization of a [17] P.E. Poh, Treatment of palm oil mill effluent (POME) under thermophilic
sustainable integrated biorefinery via fuzzy optimization, Am. Inst. Chem. condition using upflow anaerobic sludge blanket–hollow centered packed
Eng. J. 59 (2013) 4212–4217. bed (UASB–HCPB) reactor. Ph.D Thesis, University of Nottingham Malaysia
[16] P. Poh, M. Chong, Upflow anaerobic sludge blanket-hollow centered packed Campus, Selangor, 2012.
bed (UASB–HCPB) reactor for thermophilic palm oil mill effluent (POME)
treatment, Biomass Energy 67 (2014) 231–242.