Application of fuzzy logic control for Benchmark simulation ... - SER

235

Sustain. Environ. Res., 24(4), 235-243 (2014)

Application of fuzzy logic control for Benchmark simulation model.1 Mahmoud Nasr,1,* Medhat Moustafa,1 Hamdy Seif1 and Galal El-Kobrosy2 1

Department of Sanitary Engineering Alexandria University Alexandria 21544, Egypt 2 Department of Mathematical Engineering Alexandria University Alexandria 21544, Egypt

Key Words: Activated sludge, Benchmark, biological process, fuzzy logic, MATLAB

ABSTRACT Wastewater treatment processes are difficult to control because of their complex and nonlinear bio-chemical reactions. This study compared a fuzzy logic control (FLC) to classical (on/off and Proportional-Integral) methods in order to maintain the effluent quality within specified limits, as well as acceptable aeration energy (AE) consumption. Data were collected from the COST Benchmark simulation model.1 that comprises anoxic/aerobic modules for a combined biological carbon and nitrogen removal. Fuzzy logic toolbox in MATLAB was used to develop the fuzzy logic rule based system. The data of variables were implemented into the fuzzy inference system with Mamdani's method. Results showed that, good performance was achieved under dynamic influent characteristics, especially concerning the nitrogen-related species. In the anoxic section (denitrification process), nitrate was utilized by the heterotrophic organisms, and decreased from 4.8 ± 1.2 to 2.8 ± 0.9 mg L-1. In the subsequent aerobic section, ammonium was oxidized by the autotrophs and dropped from influent value of 30 ± 7 to 5 ± 4 mg L-1 (nitrification process). Degradation of the readily biodegradable substrate (98% removal) was associated with the utilization of nitrate in the anoxic tanks and oxygen in the aerated reactors. Moreover, fuzzy controller was able to handle variations in the system, and showed lower AE consumption, by 18.5 and 8.3%, as compared to uncontrolled and Proportional-Integral controlled systems, respectively. Additionally, FLC was able to self-adapt the aeration supply to handle different influent wastewater characteristics, i.e., rain and storm weather. The results showed that, FLC could be effectively used to control wastewater treatment process with good effluent quality and adequate AE consumption. . INTRODUCTION During the last few decades, numerous possible control actions and strategies have been proposed for wastewater treatment processes (WWTP) [1]. Such strategies have been conceived as a result of logical reasoning, modelling and simulation studies, lab-scale experiments, pilot-scale experiments, full-scale applications, or different combinations of those [2-4]. Obviously, one relevant control objective of WWTP operation would be to achieve sufficiently low concentrations of biodegradable matter and nutrients in the effluent with minimal sludge production and minimum costs. To enhance the acceptance of inno*Corresponding author Email: [email protected]

vating control strategies, the performance evaluation should be based on a rigorous methodology including a simulation model, plant layout, controllers, performance criteria, and test procedures [5]. Moreover, to increase safety and improve operational performance of WWTP, it is important to develop computer operational decision support systems that are able to play a similar role to the expert in daily operation [6]. Operational control of a biological WWTP is often complicated because of variations in raw wastewater compositions, strengths and flow rates owing to the changing and complex nature of the treatment process. Most theoretical models are mathematically complex, computationally expensive and they ideally require a

236

Nasr et al., Sustain. Environ. Res., 24(4), 235-243 (2014)

very detailed knowledge of the WWTP as well as characterization of the influent. Therefore, there is a need to find an alternative means for predicting process performance by exploiting available process data . and extending them to unavailable data. Intelligence control, such as fuzzy systems, have been proven to model nonlinear systems and successfully applied for modelling highly complex processes [7]. The most significant advantage of intelligence control is that no precise mathematical model is needed. Fuzzy inference systems have found wide range of industrial and commercial control applications that require analysis of uncertain and imprecise information [8]. Fuzzy logic, first proposed by Lotfi A. Zadeh in 1965 [9], deals with the concept of a fuzzy set and measures the degree to which the event may occur while probably predicts unknown outcomes based on known parameters [10]. Moreover, fuzzy logic analyzes analog input values that admit continuous variation from 0 to 1, in contrast to classical or digital logic, which operates on discrete values of either 0 or 1 (true or false). According to Kuo and Lin [11], there are two difficulties in designing any fuzzy logic systems: the shape of the membership functions and the choice of the fuzzy rules which can define the variables to be used and the ways in which the rules have to be combined. Membership functions are combined interchangeably with a logical “and” or “or” a conjunction [11], whereas the rules are combined with the logical “or” the conjunction. The shape of membership functions of fuzzy sets can be triangular, trapezoidal, bell-shaped, sigmoidal, or another appropriate form, depending on the nature of the system . being studied [12-14]. Previous researchers have investigated the application of fuzzy inference system (FIS) in the field of environmental engineering for wastewater treatment. Zhu and Peng [15] reviewed the application and development of fuzzy control in urban wastewater treatment. Moreover, in a study by Erdirencelebi and Yalpir [16], selection and prediction of anaerobic effluent quality were modelled using adaptive network FIS (ANFIS). It was found that the ANFIS models developed were successful in predicting the effluent parameters of pH and chemical oxygen demand (COD). Furthermore, Pai et al. [17] employed three types of ANFIS and Artificial Neural Network to predict suspended solids (SS) and COD in the effluent from a hospital wastewater treatment plant. According to the results, ANFIS could predict the effluent variation with maximum R values for SS-eff and COD-eff of 0.75 and 0.92, respectively. Turkdogan-Aydinol and Yetilmezsoy [18] developed a fuzzy-logic-based model to predict biogas and methane production rates in an up-flow anaerobic sludge blanket reactor treating molasses wastewater. The proposed model produced smaller deviations and exhibited a superior predictive performance on forecasting of both biogas and

methane production rates with satisfactory determina. tion coefficients over 0.98. The benchmark is a simulation environment defining plant layout, simulation model, influent loads, test procedures, and evaluation criteria. From 1998 to 2004, the development of benchmark tools for simulation-based evaluation of control strategies has been undertaken in Europe by working groups of COST Action 682 and 624 [19]. The COST benchmark was originally defined as ''a protocol to obtain a measure of performance of control strategies for activated sludge plants based on numerical, realistic simulations of the controlled plant''. It was quickly decided that the official plant layout would be aimed at removing carbon and nitrogen in a series of activated sludge reactors followed by a secondary settling tank. More details about the progress and development of the benchmark are found in previous studies [20-23]. . The overall goal of this study is to develop a fuzzy control strategy for benchmarking to maintain the effluent quality within regulations-specified limits, as well as acceptable aeration energy (AE) consumption. . MATERIALS AND METHODS 1. Benchmark Model.1 (BSM.1) Layout

.

The BSM.1 combines nitrification with pre-denitrification in a configuration that is commonly used for achieving biological carbon and nitrogen removal in full-scale plants (Fig. 1 and Table 1). The plant is designed to treat an average flow rate of 20,000 m3 d-1, and an average biodegradable COD in the influent of 300 mg L-1. The plant is composed of five biological tanks with a total volume of 5,999 m3; tanks 1 and 2 are anoxic, while 3, 4 and 5 are aerobic, and a settler with a volume of 6,000 m3. Hydraulic retention time based on average flow rate and total tank volume (i.e., biological reactor + settler of 12,000 m3) is 14.4 h. Waste activated sludge is pumped continuously from the secondary settler underflow at a default rate of 385 m3 d-1, which corresponds approximately to a sludge . age of 10 d. In BSM.1, two recycle flows are: (1) External recycle flow: A return activated sludge from the underflow of the settler to the front of the plant at a default flow rate of 18,500 m3 d-1, and (2) Internal recycle flow rate: A flow rate from tank-5 to tank-1 at a . default value of 55,300 m3 d-1. Activated sludge model no. 1 (ASM1) [24] was recognized to describe the biological processes, whereas the chosen settler model was a one-dimensional model together with a double-exponential settling velocity model proposed by Takács et al. [25]. The settler was modelled as a 10 layers non-reactive unit, and the 6th layer (counting from bottom to top) is the feed layer. Furthermore, the simulation procedure was established by creating M-file + SIMULINK/

237


Table 1. Physical configuration of the biological reactors and settling tank, and selection of system variables Physical configuration

Value

Default system flow rates

Value

Volume of tank-1

1000 m

Influent flow rate

18450 m3 d-1

Volume of tank-2

1000 m3

External recycle flow (Qr)

18450 m3 d-1

Volume of tank-3

1333 m3

Internal recycle flow (Qa)

55340 m3 d-1

Volume of tank-4

1333 m3

Wastage flow rate (Qw)

385 m3 d-1

Volume of tank-5

1333 m3

kLa1 & kLa2

0 d-1

Depth of settler

4m

kLa3

240 d-1

Area of settler

1500 m2

kLa4

240 d-1

Volume of settler

6000 m3

kLa5

84 d-1

3

kLa5

Qe

PI SOe

Qo

Tank-1

Tank-2

Tank-3

Tank-4

Tank-5

settler

Qa Qr

Qw

Fig. 1. Benchmark model.1 (BSM.1) layout.

MATLAB software. The M-file was used to program every detail of the activated sludge process, as it defines the number of inputs, outputs, parameters, initial states, etc. From the SIMULINK model point of view, the S-function block is like a black box, the output can be generated with the given input and . parameters. .

2. DO Control in the Fifth Tank The conversion rate of DO is given by Eq. 1:

.

(1) where the process ñ1 is the aerobic growth of heterotrophs modelled as Eq. 2: (2) The process ñ3 is the aerobic growth of autotrophs modelled as Eq. 3: (3) where YH and YA are the stoichiometric parameters, and µH, KS, KO,H, µA, KNH and KO,A are the kinetic

parameters. SS is the soluble inert organic matter, XB,H is the active heterotrophic biomass, SNH is the NH4+ + NH3 nitrogen and XB,A is the active autotrophic biomass . [24]. A basic control strategy is proposed to test the benchmark via Proportional-Integral (PI)-controller system; its aim is to control the DO level in the 5th tank (SOe) at a set point (yset) of 2 g m-3 by manipulating the oxygen transfer coefficient (kLa5) [26]. Moreover, the control approach has been selected to maintain the quality of the plant effluent (total nitrogen 18 mg L-1, ammonium 4 mg L-1, BOD 10 mg L-1 and COD 100 mg L-1), regardless of the variations of the incoming wastewater, in terms of flow rate and composition. To incorporate the PI controller into the simulated system, two ordinary differential equations will be needed. The first equation represents the derivative of the integral part of the error due to difference between set point and actual measurement (Eq. 4). The second equation represents the derivative of the integral part of the error due to anti-windup correction (Eq. 5). . (4) (5)

238


Accordingly the control action of the PI-controller will be based on the formula: u(t) equals proportional gain plus integration part plus anti-windup part (Eq. 6). .

Table 2. The chosen shapes and parameters of the used membership functions Membership function

Shape

Parameters

VL

gaussmf

[0.05 0.9]

L

trimf

[1 1.2 1.5]

M

gaussmf

[0.09 1.55]

H

trimf

[1.6 1.9 2.1]

VH

gaussmf

[0.05 2.2]

close-fast

trimf

[50 65 85]

close-slow

trimf

[75 95 115]

no-change

trimf

[105 125 145]

open-slow

trimf

[135 155 175]

open-fast

trimf

[165 185 200]

SOe

(6) Since DO probes are proven to be robust and reliable with relatively fast response times, DO measurement will be assumed to be ideal with no delay and noise. Moreover, the controller tuning parameters K, Ti, Tt, yset, umin and umax will be set to the values provided in the COST simulation benchmark's specifications [3]: 500 d-1 g-1 m3, 0.001 d, 0.0002 d, 2 g m-3, 0 d-1 and . 240 d-1, respectively. .

SOe: Effluent soluble oxygen, kLa5 : Oxygen transfer coefficient in the fifth tank

(a) Degree of membership

In this study, a FLC was examined instead of PIcontroller to manipulate the oxygen level in the fifth tank. Fuzzy logic toolbox/MATLAB, as a graphical user interface, was designed to work in SIMULINK environment. As similarly conducted by several researchers [27-29], a Mamdani's fuzzy inference method [30] was chosen to implement IF-THEN rules. There are basically two kinds of the inference operators: minimization (min), which truncates the output fuzzy set and product (prod) which scales the output fuzzy set. In this study, we employed the prod method due to its better performance [31]. The model was built on five rules, and each of the rules depends on resolving the inputs into a number of different fuzzy linguistic sets: soluble oxygen (SOe) is very low (VL), low (L), medium (M), high (H), or very high (VH). However, kLa5 is close-fast, close-slow, no-change, open-slow, or open-fast. Those inputs and outputs were fuzzified according to membership functions, as . described in Figs. 2a and 2b and Table 2. The rule editor was used for editing the list of five rules that defined the behaviour of the system. The fuzzy rule base system was constructed to connect the input variables to the output by means of IF - THEN . rules. The five rules were identified as: . Rule-1: If SOe is “VL” then kLa5 is “open-fast” . Rule-2: If SOe is “L” then kLa5 is “open-slow” . Rule-3: If SOe is “M” then kLa5 is “no-change” . Rule-4: If SOe is “H” then kLa5 is “close-slow” Rule-5: If SOe is “VH” then kLa5 is “close-fast” . The surface viewer screen obtained from Fuzzy Logic Toolbox, as shown in Fig. 3, was used to display the dependency of the output (kLa5) on the input (SOe), and then plot an output surface map for the system. . After creating the fuzzy rule base system, the rules must be aggregated in some manner in order to make a decision. Aggregation is the process by which the

VL 1

L

M

H

VH

0.5

0 1.0

(b) Degree of membership

3. Application of Fuzzy Logic Control (FLC)

kLa5

close-fast

1.2

1.4 1.6 SOe (mg L-1)

close-slow

no-change

1.8

open-slow

2.0

2.2

open-fast

1

0.5

0 50

100

150

200

kLa5 (d-1)

Fig. 2. Membership function editor for input and output variables. a) Effluent soluble oxygen, b) Oxygen transfer coefficient in the fifth tank.

fuzzy sets that represent the outputs of each rule are combined into a single fuzzy set. Three built-in methods are supported; max (maximum), probor (probabilistic OR), and sum (simply the sum of each rule's output set) [31,32]. In this study, the max method was applied for the aggregation step. The last step in the fuzzy inference process is to converts the resulting fuzzy outputs to a number. This operation is known as defuzzification. The defuzzification method used was a centre of gravity where the crisp output value is the abscissa under the centre of gravity of the fuzzy set. . The influent file “Inf_dry_2006.txt” can be downloaded from the website (http://www.ensic.inpl-nancy. fr/COSTWWTP). AE (kWh d-1) is an index about the


operational issue. It is calculated by integrations performed on the final 7 days of weather simulations (Eq. 7): .

(7) where, kLai(t) = the mass transfer coefficient in ith aerated reactor at time t (h-1). . RESULTS AND DISCUSSION 1. Performance Assessment of the Benchmark . Model-1 (BSM-1) In anoxic section, i.e., in the absence of oxygen, the heterotrophic organisms are capable of using nitrate as the terminal electron acceptor with (SS) as a substrate resulting in biomass growth and nitrogen gas (not included in the ASM1 model). Nitrate was decreased in the first two tanks from 4.8 ± 1.2 to 2.8 ± 0.9 mg L-1 (NO3NO2NO N 2O N2), then increased again in the following aerobic reactors to a value of 9.7 ± 1.6 mg L-1 (Fig. 4a). This was mainly due to the aerobic growth of autotrophs (nitrification), where ammonium concentration (SNH) was oxidized to nitrate. Moreover, SNH was incorporated into cell mass as the nitrogen source for synthesis. Accordingly, ammonium was decreased from 30 ± 7 mg L-1 (influent) to 5 ± 4 mg L-1 in the subsequent aerobic units (Fig. 4b). During the nitrification process, it is essential to maintain the DO in a biological reactor at the right level. Both SNH and SO are rate limiting to the process. Piotrowski et al. [33] stated that at least 0.2 mg L-1 is required for the nitrification processes to be carried out, whereas, Fan et al. [34] found that in many cases, the DO concentration needs to be increased to 4

239

mg L-1 to achieve the best nitrification performance. Moreover, O'Brien et al. [35] claimed that a typical average value of 0.8 mg DO L-1 is sufficient for aerobic removal of carbonaceous material. Efficient nutrient removal is linked to the availability of a sufficient carbon source for heterotrophs in the anoxic section of the biological reactor [36]. Zhu et al. [37] showed that for conventional pre-denitrification process, 1850% internal recycle ratio is needed to maintain carbon source for the denitrification process, as well as 50% sludge recycle ratio. Another study by Lee et al. [38] found that C/N ratio of 13:1 was an effective organic carbon source for promoting denitrification process for total nitrogen removal in saturated zone with vermicompost. Moreover, previous studies investigated the optimization of anoxic/aerobic step feeding for improving nitrogen removal [39,40]. A fraction of readily biodegradable substrate (SS) was degraded from the influent value of 65 ± 19 mg L-1 to 1.3 ± 0.3 mg L-1 due to its consumption for the biomass growth (Fig. 4c). This process was generally the main contributor to the production of new biomass and removal of COD. It was also associated by the utilization of nitrate in the anoxic tanks and oxygen in the third aerated reactor, where the SO was decreased from a set point of 8 to 1.8 ± 0.5 mg L-1 (Fig. 4d). Similarly, Holenda et al. [41] stated that the DO level in the aerobic reactors has significant influence on the behaviour and activity of the heterotrophic and autotrophic microorganisms. In the fourth aerated unit, DO levels started to increase, and reached up to 2.7 ± 1.0 mg L-1. This indicated that there was no significant increase in SS removal, as it decreased to 1.1 ± 0.2 mg L-1 (only 16% deterioration). An excessive DO concentration does not accelerate the nitrification but only increases the energy cost due to increased airflow. Thus providing the same amount of aeration was not . economically feasible.

kLa5 (d-1)

2. Results of Manipulating the DO in the Fifth . Aerated Tank

SOe (mg L-1)

Fig. 3. The surface viewer screen: the input (SOe) on Xaxis, and the output (kLa5) on Y-axis.

In the uncontrolled system, the kLa5 in the fifth tank was kept constant at a certain value. Results revealed that, at kLa5 = 84 d-1: effluent average concentration of COD = 48 mg L-1 and BOD = 3 mg L-1 obeys to the constraints, COD = 100 mg L-1 and BOD = 10 mg L-1, while the ammonium concentration (SNH) = 4.8 mg L-1 is higher than the effluent criteria of 4 mg L-1. However, AE has recorded the minimum value of 6480 kWh d-1 to achieve the objective of saving consumption energy. On the other hand, at kLa5 = 240 d-1: effluent average concentration of COD = 48 mg L-1 and BOD = 3 mg L-1, as well as SNH = 2.5 mg L-1 obeying the constraints. Conversely, AE has recorded the maximum value of 8530 kWh d-1, indicating the . maximum energy consumption.

240


c)

8

12 10 8 6 4

7 6 5 4 3 2

2

1

0

0

25

6

b) Soluble oxygen (mg L-1)

Ammonium concentration (mg L-1)

9

a)

14

Soluble substrate (mg L-1)

Nitrate concentration (mg L-1)

16

20 15 10 5

d)

5 4 3 2 1 0

0 0

2

4

Tank-1

6 8 Time (d) Tank-2

Tank-3

10

12

14

0

2

4

6

8

10

12

14

Time (d) Tank-1

Tank-4

Tank-2

Tank-3

Tank-4

Fig. 4. Concentration of the estimated parameters in the anoxic (1st and 2nd) and aerobic (3rd and 4th) tanks using influent dry weather data; a) Nitrate, b) Ammonium, c) Substrate, d) Oxygen.

For the controlled system via PI-controller: effluent average concentration of COD = 48 mg L-1, BOD = 3 mg L-1, in addition to SNH = 2.5 mg L-1 obeying the constraints with acceptable AE of 7250 kWh d-1. On the other hand, for the controlled system via fuzzy logic controller: effluent average concentration of COD, BOD and SNH of 48, 3 and 2.9 mg L-1, respectively follow the limitations with lower AE of 6650 kWh d-1. Table 3 presents the simulation results of BSM.1 under the action of FLC as compared to both uncontrolled and PI-controlled systems. In a similar study by Belchior et al. [42], DO levels of the activated sludge wastewater treatment process were controlled using stable adaptive fuzzy control. The paper described the development of an adaptive fuzzy control strategy for tracking the DO reference trajectory applied to the Benchmark Simulation Model n.1. Moreover, Piotrowski et al. [33] proposed a hierarchi-

cal DO control for activated sludge processes by considering both nutrient and phosphorous removal. The study has led to the least energy cost and robust DO tracking. Another study by Traoré et al. [12] compared between different types of DO control (on/off; PID and FLC) during a nitrification and denitrification process in the sequential batch reactor pilot plant. The study found that, fuzzy controller was effective and superior as compared to the traditional . regulation techniques. In order to investigate the applicability of the model at different influent conditions, FLC was examined under rain and storm weather (http://www.ensic. inpl-nancy.fr/COSTWWTP). The FLC showed a stability performance in terms of effluent SNO (Fig. 5a), SNH (Fig. 5b) and SS (Fig. 5c): 11.7 ± 1.5, 2.7 ± 2.8, and 0.9 ± 0.2 mg L-1 respectively. However, results of PI controller showed fluctuation in character as shown in

Table 3. BSM.1 simulation results for uncontrolled, PI, and fuzzy logic control systems

SNH TKN Ntot COD AE

Units

Influent

mg L-1 mg L-1 mg L-1 mg L-1 kWh d-1

30 55 55 300

Uncontrolled system (kLa5 = 84 d-1) 4.8 6.8 15.5 48 6480

Uncontrolled system (kLa5 = 240 d-1) 2.5 4.5 17.5 48 8530

PI controller

Fuzzy logic controller

2.5 4.5 16.7 48 7250

2.9 4.9 16.4 48 6650

SNH: ammonia nitrogen; TKN: Total Kjeldahl Nitrogen; Ntot: total nitrogen; COD: chemical oxygen demand; BOD: biochemical oxygen demand; AE: aeration energy

241


CONCLUSIONS FLC was applied to the COST BSM.1 to maintain the effluent quality within regulations-specified limits, Soluble substrate (mg L-1)

b)

further investigations should be examined to test the stability of the FLC in terms of different configuration of reactors, and that will be the primary focus of our future work. .

c)

d) Soluble oxygen (mg L-1)

Nitrate concentration (mg L-1)

a)

Ammonium concentration (mg L-1)

Fig. 6a (SNO: 11 ± 3 and 11 ± 3 mg L-1), Fig. 6b (SNH: 2.4 ± 2.6 and 2.7 ± 2.9 mg L-1), and Fig. 6c (SS: 0.9 ± 0.2 and 0.9 ± 0.3 mg L-1), for rain and storm weather, respectively. This can be attributed to the fact that FLC is able to self-adapt the aeration supply to handle variations in the system. As a result, SO was varied under the action of FLC at 1.6 ± 0.5 mg L-1 (Fig. 5d), whereas as depicted in Fig. 6d PI controller showed constancy at a value of 2.0 ± 0.1 mg-DO L-1. Yet,

Time (d) Dry weather

Rain weather

Dry weather

Storm weather

Time (d) Rain weather

Storm weather

Soluble substrate (mg L-1)

a)

b)

c)

d) Soluble oxygen (mg L-1)

Ammonium concentration (mg L-1) Nitrate concentration (mg L-1)

Fig. 5. Concentration of the estimated parameters in the fifth aerated tank under dry, rain and storm weather using fuzzy logic control; a) Nitrate, b) Ammonium, c) Substrate, d) Oxygen.

Time (d) Dry weather

Rain weather

Storm weather

Dry weather

Time (d) Rain weather

Storm weather

Fig. 6. Concentration of the estimated parameters in the fifth aerated tank under dry, rain and storm weather using PI control; a) Nitrate, b) Ammonium, c) Substrate, d) Oxygen.

242


as well as minimum AE consumption. The simulated model compared between uncontrolled and controlled systems in terms of environmental and economic points of view. Good performance was achieved under dynamic influent characteristics, especially concerning the nitrogen-related species. The effect of FLC was visible on decreasing COD, BOD and ammonium concentration which were considered as being of major importance. It was found that simulation of the uncontrolled system with constant kLa5 value of 84 d-1 did not achieve the effluent limitation of SNH, but AE consumption was the minimum. On the other hand, simulating the with kLa5 value of 240 d-1 achieved the effluent requirements but AE of 8530 kWh d-1 was the maximum of all. Control strategy via PI-controller achieved the effluent limitation requirements with acceptable AE. However, FLC recorded minimum AE consumption (lower than uncontrolled and PIcontroller by 18.5 and 8.3%, respectively), as well as achieving the effluent environmental aspects in terms of BOD, COD, SNH and Total Kjeldahl Nitrogen. Additionally, at different influent feed of rain and storm weather, FLC was able to self-adapt the aeration . supply rather than the PI controller. In conclusion, the controlled system achieved the environmental goals; however, FLC recorded more stability and lower AE consumption as compared to . the PI controller.

7.

8. 9. 10.

11.

12.

13.

14.

REFERENCES 1. Nasr, M.S., M.A.E. Moustafa, H.A.E. Seif and G. El Kobrosy, Application of Artificial Neural Network (ANN) for the prediction of EL-AGAMY wastewater treatment plant performance-EGYPT. . Alexandria Eng. J., 51(1), 37-43 (2012). 2. Rieger, L., J. Alex, S. Winkler, M. Boehler, M. Thomann and H. Siegrist, Progress in sensor technology - Progress in process control? Part I: Sensor property investigation and classification. . Water Sci. Technol., 47(2), 103-112 (2003). 3. Alex, J., L. Rieger, S. Winkler and H. Siegrist, Progress in sensor technology - Progress in process control? Part II: Results from a simulation benchmark study. Water Sci. Technol., 47(2), . 113-120 (2003). 4. Hayet, C., T. Youssef, S. Hédil and H. Moktar, Optimization of Tunisian wastewater treatment plant - The Kelibia case. Sustain. Environ. Res., . 21(1), 65-72 (2011). 5. Nasr, M.S., M.A.E. Moustafa, H.A.E. Seif and G.El Kobrosy, Modelling and simulation of German BIOGEST/EL-AGAMY wastewater treatment plants - Egypt using GPS-X simulator. . Alexandria Eng. J., 50(4), 351-357 (2011). 6. Wan, J.Q., M.Z. Huang, Y.W. Ma, W.J. Guo, Y. Wang, H.P. Zhang, W.J. Li and X.F. Sun, Prediction of effluent quality of a paper mill

15.

16.

17.

18.

19.

20.

wastewater treatment using an adaptive networkbased fuzzy inference system. Appl. Soft Comput., . 11(3), 3238-3246 (2011). Bhat, N. and T.J. Mcavoy, Use of neural nets for dynamic modeling and control of chemical process systems. Comput. Chem. Eng., 14(4-5), . 573-583 (1990). Akcayol, M.A., Application of adaptive neurofuzzy controller for SRM. Adv. Eng. Softw., . 35(3-4), 129-137 (2004). Zadeh, L.A., Fuzzy sets. Inform. Control., 8(3), . 338-353 (1965). Rihani, R., A. Bensmaili and J. Legrand, Fuzzy logic modelling tracer response in milli torus reactor under aerated and non-aerated conditions. . Chem. Eng. J., 152(2-3), 566-574 (2009). Kuo, K.Y. and J. Lin, Application of fuzzy set theory to the change intervals at a signalized intersection. Fuzzy Set. Syst., 129(3), 387-403 . (2002). Traoré, A., S. Grieu, S. Puig, L. Corominas, F. Thiery, M. Polit and J. Colprim, Fuzzy control of dissolved oxygen in a sequencing batch reactor pilot plant. Chem. Eng. J., 111(1), 13-19 (2005). . Metternicht, G. and S. Gonzalez, FUERO: Foundations of a fuzzy exploratory model for soil erosion hazard prediction. Environ. Modell. . Softw., 20(6), 715-728 (2005). Sadiq, R., M.A. Al-Zahrani, A.K. Sheikh, T. Husain and S. Farooq, Performance evaluation of slow sand filters using fuzzy rule-based modelling. Environ. Modell. Softw., 19(5), 507-515 (2004). . Zhu, G.B. and Y.Z. Peng, Application and development of intelligent control technology in urban wastewater treatment system. J. Harbin Inst. . Technol., 36, 328-332 (2004). Erdirencelebi, D. and S. Yalpir, Adaptive network fuzzy inference system modeling for the input selection and prediction of anaerobic digestion effluent quality. Appl. Math. Model., 35(8), 3821. 3832 (2011). Pai, T.Y., T.J. Wan, S.T. Hsu, T.C. Chang, Y.P. Tsai, C.Y. Lin, H.C. Su and L.F. Yu, Using fuzzy inference system to improve neural network for predicting hospital wastewater treatment plant effluent. Comput. Chem. Eng., 33(7), 1272-1278 . (2009). Turkdogan-Aydinol, F.I. and K. Yetilmezsoy, A fuzzy-logic-based model to predict biogas and methane production rates in a pilot-scale mesophilic UASB reactor treating molasses wastewater. J. Hazard. Mater., 182(1-3), 460-471 . (2010). Copp, J., The COST Simulation Benchmark: Description and Simulator Manual. European Cooperation in Science and Technology, Brussels, . Belgium (2001). Spanjers, H., P. Vanrolleghem, K. Nguyen, H. Vanhooren and G.G. Patry, Towards a simulation-


21.

22.

23. 24.

25. 26. 27.

28.

29.

30. 31.

32. 33.

benchmark for evaluating respirometry-based control strategies. Water Sci. Technol., 37(12), . 219-226 (1998). Alex, J., J.F. Beteau, J. Copp, C. Hellinga, U. Jeppsson, S.M. Libelli, M.N. Pons, H. Spanjers and H. Vanhooren, Benchmark for evaluating control strategies in wastewater treatment plants. The ECC'99 Conference. Karlsruhe, Germany, . Aug. 31-Sep. 3 (1999). Pons, M.N., H. Spanjers and U. Jeppsson, Towards a benchmark for evaluating control strategies in wastewater treatment plants by simulation. . Comput. Chem. Eng., 23, 403-406 (1999). Copp, J.B., Defining a simulation benchmark for . control strategies. Water, 21, 44-49 (2000). Henze, M., C.P.L. Jr. Grady, V. Gujer, G.R. Marais and T. Matsuo, Activated Sludge Model No. 1. IAWPRC Scientific and Technical Reports No. 1. . IWA Publishing, London, UK (1987). Takács, I., G.G. Patry and D. Nolasco, A dynamic model of the clarification thickening process. . Water Res., 25(10), 1263-1271 (1991). Pons, M.N. and J.P. Corriou, Implementation of storage tanks on the COST 624 benchmark. Water . Sci. Technol., 45(4-5), 159-166 (2002). Akkurt, S., G. Tayfur and S. Can, Fuzzy logic model for the prediction of cement compressive strength. Cement Concrete Res., 34(8), 1429-1433 . (2004). Acaroglu, O., L. Ozdemir and B. Asbury, A fuzzy logic model to predict specific energy requirement for TBM performance prediction. Tunn. Undergr. . Sp. Tech., 23(5), 600-608 (2008). Adriaenssens, V., P.L.M. Goethals and N. De Pauw, Fuzzy knowledge-based models for prediction of Asellus and Gammarus in watercourses in Flanders (Belgium). Ecol. Model., . 195(1-2), 3-10 (2006). Mamdani, E.H., Application of fuzzy logic to approximate reasoning using linguistic synthesis. . IEEE T. Comput., 26(12), 1182-1191 (1977). Jantzen, J., Design of Fuzzy Controllers (Lecture Notes). Department of Automation, Technical University of Denmark, Kongens Lyngby, . Denmark (1998). Kosko, B., Fuzzy Thinking: The New Science of . Fuzzy Logic. Hyperion, New York (1994). Piotrowski, R., M.A. Brdys, K. Konarczak, K. Duzinkiewicz and W. Chotkowski, Hierarchical dissolved oxygen control for activated sludge processes. Control Eng. Pract., 16(1), 114-131 . (2008).

243

34. Fan, C., W.C. Wang and S.T. Cheng, Economical analysis and performance evaluation for municipal wastewater treatment by an aerobic biofilm reactor using recycle materials as the contact media. . Sustain. Environ. Res., 22(4), 261-269 (2012). 35. O'Brien, M., J. Mack, B. Lennox, D. Lovett and A. Wall, Model predictive control of an activated sludge process: A case study. Control Eng. Pract., . 19(1), 54-61 (2011). 36. Shen, W.H., X.Q. Chen and J.P. Corriou, Application of model predictive control to the BSM1 benchmark of wastewater treatment process. Comput. Chem. Eng., 32(12), 2849-2856 . (2008). 37. Zhu, G.B., Y.Z. Peng, S.Y. Wang, S.Y. Wu and B. Ma, Effect of influent flow rate distribution on the performance of step-feed biological nitrogen removal process. Chem. Eng. J., 131(1-3), 319. 328 (2007). 38. Lee, L.Y., L. Tan, W. Wu, S.K.Q. Yeo and S.L. Ong, Nitrogen removal in saturated zone with vermicompost as organic carbon source. Sustain. . Environ. Res., 23(2), 85-92 (2013). 39. Zhu, G.B., Y.Z. Peng, B. Ma, Y. Wang and C.Q. Yin, Optimization of anoxic/oxic step feeding activated sludge process with fuzzy control model for improving nitrogen removal. Chem. Eng. J., . 151(1-3), 195-201 (2009). 40. Guo, W., H.H. Ngo, Z. Wu, A.Y.J. Hu and A. Listowski, Application of bioflocculant and nonwoven supporting media for better biological nutrient removal and fouling control in a submerged MBR. Sustain. Environ. Res., 21(1), . 53-58 (2011). 41. Holenda, B., E. Domokos, Á. Rédey and J. Fazakas, Dissolved oxygen control of the activated sludge wastewater treatment process using model predictive control. Comput. Chem. Eng., 32(6), . 1270-1278 (2008). 42. Belchior, C.A.C., R.A.M. Araújo and J.A.C. Landeck, Dissolved oxygen control of the activated sludge wastewater treatment process using stable adaptive fuzzy control. Comput. . Chem. Eng., 37, 152-162 (2012). Discussions of this paper may appear in the discussion section of a future issue. All discussions should be submitted to the Editor-in-Chief within six months of publication. . Manuscript Received: September 24, 2013 Revision Received: December 09, 2013 and Accepted: January 04, 2014