Dynamic evaluation of integrated control strategies for enhanced nitrogen removal in activated sludge processes

Dynamic evaluation of integrated control strategies for enhanced nitrogen removal in activated sludge processes

ARTICLE IN PRESS Control Engineering Practice 14 (2006) 1269–1278 www.elsevier.com/locate/conengprac Dynamic evaluation of integrated control strate...

357KB Sizes 0 Downloads 47 Views

ARTICLE IN PRESS

Control Engineering Practice 14 (2006) 1269–1278 www.elsevier.com/locate/conengprac

Dynamic evaluation of integrated control strategies for enhanced nitrogen removal in activated sludge processes M. Yonga, P. Yongzhena,b,, U. Jeppssonc a

School of Municipal and Environmental Engineering, Harbin Institute of Technology, PO Box 2607, Haihe Road 202, Nangang District, Harbin 150090, China b Key Laboratory of Beijing for Water Environmental Recovery Engineering, Beijing University of Technology, Pingleyuan 100, Chaoyang District, Beijing 100022, China c Department of Industrial Electrical Engineering and Automation (IEA), Lund University, PO Box 118, SE-221 00 Lund, Sweden Received 25 January 2005; accepted 20 June 2005 Available online 25 August 2005

Abstract In this paper, six different strategies for integrated control of nitrate recirculation flow rate and external carbon addition in a predenitrifying biological wastewater treatment system are proposed and evaluated using the COST simulation benchmark. The proposed control strategies consist of two or three feedback control loops, which manipulate the nitrate recirculation and the carbon dosage flow rates in an integrated manner, such that the consumption of external carbon is minimized while the nitrate discharge limits are met. The control systems require measurements of nitrate concentrations at the end of both the anoxic and the aerobic zones. A number of simulations are performed and the results show that the control strategies can maintain the designated effluent quality in the presence of external disturbances, while obtaining excellent results in terms of consumption of external carbon, as well as meeting the general plant performance and controller criteria as defined by the benchmark group. r 2005 Elsevier Ltd. All rights reserved. Keywords: Benchmark simulation model; Control; External carbon addition; Nitrate recirculation; Pre-denitrification process; Wastewater

1. Introduction Today, eutrophication in receiving waters due to the presence of nutrients including nitrogen is a wellrecognized environmental problem worldwide. As a result, stringent standards have been imposed on total nitrogen (TN) concentration levels in effluent waters from wastewater treatment systems and the need for better operational control of wastewater treatment plants (WWTPs) has been emphasized. In an activated Corresponding author. School of Municipal and Environmental Engineering, Harbin Institute of Technology, PO Box 2607, Haihe Road 202, Nangang District, Harbin 150090, China. Tel.: +0861067392627; fax: +0861067391827. E-mail addresses: [email protected] (M. Yong), [email protected] (P. Yongzhen), [email protected] (U. Jeppsson).

0967-0661/$ - see front matter r 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.conengprac.2005.06.018

sludge process with pre-denitrification, as in the COST Benchmark Simulation Model (Copp, 2002), denitrification is a key process for removing nitrate nitrogen. In this case, the strict standards on TN discharge limits have posed a particular challenge for biological denitrification, especially for treatment plants with an unfavourable COD to N ratio. The most effective solutions known to date are to supplement an external carbon source into the anoxic part of the process to enhance denitrification and to control the nitrate recirculation flow rate to provide an appropriate amount of nitrate to the anoxic zones. These methodologies are being used at many WWTPs already (e.g. Nyberg, Andersson, & Aspegren, 1996; Purtschert, Siegrist, & Gujer, 1996), and are expected to be implemented even more within the wastewater treatment industry in the years to come, as the demand on effluent wastewater quality continues to increase.

ARTICLE IN PRESS 1270

M. Yong et al. / Control Engineering Practice 14 (2006) 1269–1278

Nitrate recirculation flow rate has long been identified as a manipulative variable in pre-denitrification systems. The on-line control possibilities of this variable to improve nitrate removal have been studied by several researchers (e.g. Balslev, Lynggaard-Jensen, & Nickelsen, 1996; Londong, 1992; Yuan, Oehmen, & Ingildsen, 2002). A common strategy is to control the nitrate concentration at the end of the anoxic zone at a level of 1–3 mg N l1. Such a strategy maximizes the usage of available influent COD for denitrification but the effectiveness for maintaining the effluent nitrate level is limited, as the amount of nitrate that can be removed is predominantly determined by the ratio of influent COD to N (assuming that the nitrification process is working properly). When the influent wastewater to a WWTP has a low COD to N ratio, it should be supplemented with an external carbon source to the anoxic zone to increase the denitrification rate. However, while the lack of readily available COD leads to incomplete denitrification, excess dosing of external carbon may significantly increase operating costs or reduce the effluent quality. Consequently, many researchers have studied the problem of determining the appropriate amount of carbon source to be added. For a pre-denitrification system, it has been found that controlling the nitrate nitrogen concentration at the end of the anoxic zone at a low set-point (1–2 mg N l1) minimizes the amount of external carbon required, while maintaining the long-term average effluent nitrate nitrogen concentration at a pre-specified level (e.g. Lindberg & Carlsson, 1996; Samuelsson & Carlsson, 2001; Yuan, Bogaert, & Vanrolleghem, 1997). The above two control handles cannot simply be added together to form an integrated controller, as both use the anoxic nitrate concentration as the controlled variable. Until now, limited work has been reported on integrated control of nitrate recirculation and external carbon dosage, see for example Yuan and Keller (2003). The main study of this work comprises the development and evaluation of different integrated control strategies of nitrate recirculation flow rate and external carbon addition. The goal is to improve the effluent quality with a limited increase of operational costs by controlling the process using these control handles. The paper is organized as follows. In the next section, the benchmark simulation model is briefly recapitulated. Then, six suggested integrated control strategies are described in Section 3 and in the following section their performances are evaluated using the criteria defined by the benchmark group and the results are discussed. Finally, Section 5 provides some general conclusions.

2. Material and methods Many control strategies have been proposed in the literature, but it is often difficult to evaluate and compare

them to each other. This is partly due to the variability of the influent, use of different mathematical models, different plant configurations, etc. Also complicating the evaluation is the lack of standard evaluation criteria. To enhance the development and acceptance of innovative control strategies, the evaluation procedure must be made easier and more objective. Therefore, a simulation benchmark of the activated sludge process has been developed by a working group of COST Actions 624 and 682, together with the IWA Task Group on Respirometry. The overall goal of the work has been to develop a general benchmarking protocol and a software tool for benchmarking, i.e. performance assessment and evaluation of control strategies for wastewater treatment systems. The benchmark simulation model no. 1 (BSM1) defines a plant layout, influent wastewater characteristics, models and sets of model parameters, test procedures and evaluation criteria (Copp, 2002). In Fig. 1, the schematic outline of the benchmark plant is given, which represents a standard activated sludge process including a bioreactor and a secondary settler. The bioreactor is composed of five completely mixed compartments, the first two are anoxic and the last three are aerated. The total reactor volume is 6000 m3 with an anoxic fraction of 33%. The average dry weather flow is 18446 m3 d–1, with a total COD/N ratio of 5.6. The IWA activated sludge model no 1 (ASM1) is used to model the biological processes in the bioreactor (Henze, Grady, Gujer, Marais, & Matsuo, 1987). The secondary settler is modelled as a non-reactive, onedimensional ten-layer system and the double exponential settling velocity model proposed by Taka`cs, Patry, and Nolasco (1991) is chosen to capture the settling properties of the sludge. The full set of equations and all parameter values are available on the benchmark website (http://www.benchmarkWWTP.org). In this paper, the controllers are implemented in the benchmark using Matlab/Simulink. It is assumed, as a physical constraint, that the maximum nitrate recirculation flow rate is 5 times the average dry weather influent flow rate. The maximum allowed dosage of external carbon is limited to 3 m3 d1, with a COD concentration of 1,200,000 mg l1 (the added COD is assumed to be readily biodegradable). The performance of the plant is always evaluated for the last 7 days of dynamic simulations (from a total of 28 days) using the dry weather influent wastewater according to the benchmark definition. Settler Influent

Effluent 1

Zone:

2

3

4

5

Recirculated water Recirculated sludge Fig. 1. Benchmark plant layout.

Excess sludge

ARTICLE IN PRESS M. Yong et al. / Control Engineering Practice 14 (2006) 1269–1278

Perfect and continuous measurements of the nitrate concentration (without delay time or noise) in the second anoxic and the last aerobic tank are assumed. This assumption is made to simplify the evaluation and comparison procedure. In order to make an objective evaluation of the different control structures, the nitrate set-point at the end of second anoxic zone is 2 mg N l1 in all cases (SNO2, sp), regardless of the chosen manipulated variable; and the nitrate set-point at the end of the last aerobic zone is also identical: 7 mg N l1 in all cases (SNO5, sp). The relay function used in the controllers (see Fig. 2: cases c–f) is designed so that the high-load controller is activated when eo0 and the low-load controller is activated when e40.5 mg N l–1 (e is defined in Fig. 2).

3. Proposed integrated control strategies for nitrate recirculation and external carbon addition Control strategy no. 1: Comprises of two feedback control loops as that of Singman (1999) (Fig. 2a), one to

SNO2,sp=2mg/L +

determine the flow rate of external carbon source to the first anoxic zone (Qcarbon) to keep the nitrate concentration at the end of the second anoxic zone (SNO2) at a pre-specified level (SNO2, sp), and the other to adjust the flow rate of the nitrate recirculation (QR) to maintain the nitrate concentration at the end of the last aerobic zone (SNO5) at a pre-specified level (SNO5, sp). Control strategy no. 2: As shown in Fig. 2b, one feedback loop is to determine the flow rate of external carbon source to the first anoxic zone, keeping the nitrate concentration at the end of the aerobic zone at a prespecified level, and the other is to adjust the nitrate recirculation flow rate to maintain the nitrate concentration at the end of second anoxic zone at a pre-specified level. Control strategy no. 3: The proposed control system comprises of the same main components as that of Yuan and Keller (2003) (Fig. 2c). PID A manipulates the nitrate recirculation flow rate, controlling the nitrate concentration at the end of the second anoxic zone at a pre-specified level. This loop is coupled with zero addition of external carbon and is switched on/off when

SNO2,sp=2mg/L +

Qcarbon PID A

WWTP

1271

QR PID A

WWTP

SNO2 SNO5,sp=7mg/L

+

QR PID B

SNO2 SNO5,sp=7mg/L

SNO5

system

Qcarbon PID B

+

-

(a)

system

SNO5

-

(b) Low-load controller

Qcarbon=0

SNO5,sp=7mg/L

PID A

flag

Relay

e

QR

QR

+

PID B

-

SNO2,sp=2mg/L

Qcarbon

PID C

Decision controller

+

SNO2,sp=2mg/L

Qcarbon=0

+

QR Qcarbon

plant

SNO2 SNO5

QR

PID A

flag

Relay

SNO5,sp=7mg/L e

+

Qcarbon

PID B

-

Decision controller

Low-load controller

-

SNO2,sp=2mg/L

SNO2,sp=2mg/L

QR

PID A

QR Qcarbon

plant

+

+

-

-

High-load controller

High-load controller

(d)

(c) Low-load controller

QR=2*Qin,avg PID A Relay

e

flag

SNO5,sp=7mg/L

+

PID B

-

SNO2,sp=2mg/L PID A

Low-load controller

Qcarbon

QR Qcarbon

+

PID A

QR Qcarbon

plant

SNO2 SNO5

Relay

e

+

Q

flag

SNO5,sp=7mg/L

Q PID A

-

SNO2,sp=2mg/L PID B

+

(e)

Qcarbon=0

-

SNO5,sp=7mg/L

Decision controller

+

Decision controller

-

SNO2,sp=2mg/L

Q

+

-

-

High-load controller

High-load controller

(f) Fig. 2. Six integrated control strategies of nitrate recirculation and external carbon addition.

SNO5

Q Q

plant SNO2

SNO2 SNO5

ARTICLE IN PRESS 1272

M. Yong et al. / Control Engineering Practice 14 (2006) 1269–1278

4. Results and discussion

resulting from control strategies numbers 1, 2, 3, 4, 5 and 6 are shown during seven days of dynamic simulations. Also, the settler effluent total nitrogen (TNe) and nitrate concentrations (SNOe) are presented for comparisons. Apart from the impact of the proposed control strategies, the plant is simulated under identical conditions for all cases (influent wastewater, sludge recycle flow rate, sludge wastage flow rate, oxygen control, etc.). From these figures, it is clear that the nitrate concentration in the last aerobic reactor (SNO5) was successfully controlled around its set-point of 7 mg N l1 applying control strategies 1, 3, 4, 5 and 6. For strategy no. 6, the standard deviation is higher but the average value is still maintained around 7 mg N l1. The major variations appear for strategy no. 2, where the standard deviation and mean value are 1.22 and 6.66 mg N l1, respectively. Strategy no. 5 demonstrates the smallest standard deviation (0.20) although the mean value of SNO5 (7.21 mg N l1) is somewhat high. The nitrate concentrations in the last anoxic tank (SNO2) are also successfully controlled around the selected set-point of 2 mg N l1 using strategies 1, 3 and 4. There are more significant variations for strategies 2 and 5, where the average values of SNO2 are 1.66 and 1.0 mg N l1, respectively. Strategy no. 6 demonstrates the biggest standard deviation (0.90) although the mean value of SNO2 (2.14 mg N l1) is close to 2 mg N l1. The SNO2 controller of strategy no. 1 has the smallest standard deviation (0.34). During low-load periods, using a fixed nitrate recirculation flow rate, not enough nitrate for denitrification is recycled, which explains why the average effluent nitrate nitrogen concentration of strategy no. 5 is somewhat higher. The resulting effluent TN concentrations from applying all the tested control strategies are below the defined limit of 18 mg N l1 during the entire period. A few incidents also appear where SNO2 and SNO5 are considerably lower than their set-points. This is caused by low nitrate loading, because the nitrate recirculation flow rate during these periods reaches its physical maximum (5 times the average dry weather flow rate) and the nitrate concentration in the last aerobic tank is low at the same time. Generally, the nitrate recirculation flow rate should be increased at night (low-load periods) to recycle more nitrate and maintain anoxic conditions, thus maximizing the use of influent and intracellular COD for denitrification. During daytime (high-load periods), the nitrate recirculation flow rate is lower and just sufficient to keep the nitrate concentration around its set-point. The carbon dosing is low at night and high during daytime, reflecting the general strategy of ‘adding only when necessary’.

4.1. Simulation of proposed control strategies

4.2. Evaluation and discussion of proposed strategies

In Figs. 3–8, the concentration profiles of the controlled variables as well as the manipulated variables (flow rates)

Table 1 summarizes the results obtained from implementing and simulating the six integrated control

PID B is switched off/on. It ensures the maximum use of the influent COD for denitrification during low-load periods. PID B manipulates the nitrate recirculation flow rate, controlling the nitrate concentration at the end of the last aerobic zone at a pre-specified level, which is based on the effluent standard required. This loop is activated only when SNO5 exceeds the set-point, ensuring (together with controller PID C) that the aerobic effluent nitrate request is met. PID C manipulates the external carbon dosage to the first anoxic zone, controlling the nitrate concentration at the end of the anoxic zone at a pre-specified set-point. This loop is switched on/off when PID A is switched on/off. Control strategy no. 4: As shown in Fig. 2d, this control strategy is similar to strategy no. 2, except that in this case the second controller (PID B) is active only during high-load periods and is switched off during lowload periods to reduce consumption of external carbon. Control strategy no. 5: This control system also comprises of two feedback control loops (Fig. 1e). During low-load periods, PID A manipulates the external carbon dosage to the first anoxic zone, controlling the nitrate concentration at the end of the anoxic zone at a prespecified set-point, with a fixed nitrate recirculation flow rate of 2  18446 m3 d1 (200% of the average influent flow rate). During high-load periods, PID B manipulates the nitrate recirculation flow rate, controlling the nitrate concentration at the end of the aerobic zone at a requested set-point. This loop is activated only when SNO5 exceeds the set-point SNO5, sp (together with the controller PID A), ensuring that the aerobic effluent request is met. Control strategy no. 6: As shown in Fig. 2f, this control system also comprises of two feedback control loops. During low-load periods, PID A manipulates the nitrate recirculation flow rate, controlling the nitrate concentration at the end of the aerobic zone at a pre-specified level. This loop is coupled with zero addition of external carbon. During high-load periods, together with controller PID A, PID B manipulates the external carbon dosage to the first anoxic zone, controlling the nitrate concentration at the end of the anoxic zone at a pre-specified set-point. This loop is activated only when SNO5 exceeds the set-point SNO5, sp. PID A, PID B and PID C are all proportional–integral–derivative controllers. A relay function to switch between PID A (low load) and PID B/C (or A/B, high load) is provided to avoid oscillations. The decision controller acts upon the signal from the relay to determine which controller pathways are currently active.

ARTICLE IN PRESS M. Yong et al. / Control Engineering Practice 14 (2006) 1269–1278

20

8

concentration (mg N/L)

concentration (mg N/l)

10

1273

6 SNO2

4

SNO5

2

TN e

18

SN O e

high-load

16 14 12 10 low-load

8 6

0 7

8

9

10

11

12

13

4

14

7

9

10

11

12

13

14

100000

3

80000 2

QR (m3/d)

Qcarbon (m3/d)

8

1

60000 40000

0

20000 7

8

9

10

11

12

13

7

14

8

9

10

time (d)

11

12

13

14

13

14

time (d) Fig. 3. Simulation results of control strategy no. 1.

10

20 concentration (mg N/l)

concentration (mg N/l)

18 8 6 4

SNO2

SNO5

2

TNe

16

SNOe

14 12 10 8 6

0

4 7

8

9

10

11

12

13

14

7

8

9

10

11

12

100000

3

QR (m3/d)

Qcarbon (m3/d)

80000 2

1

60000

40000 0

20000 7

8

9

10

11

12

13

14

7

8

time (d)

9

10

11

time (d) Fig. 4. Simulation results of control strategy no. 2.

12

13

14

ARTICLE IN PRESS M. Yong et al. / Control Engineering Practice 14 (2006) 1269–1278

1274

20

10

TNe

18 concentration (mg N/l)

concentration (mg N/l)

8 6 SNO2

SNO5

4 2

16 14 12 10 8 6 4

0 7

8

9

10

11

12

13

14

7

3

8

9

10

11

12

13

14

13

14

13

14

13

14

100000 80000

2 QR (m3/d)

Qcarbon (m3/d)

SNOe

1

60000 40000

0

20000 7

8

9

10

11

12

13

14

7

8

9

10

time (d)

11

12

time (d) Fig. 5. Simulation results of control strategy no. 3.

22

10

20 concentration (mg N/L)

concentration (mg N/l)

8 6 SNO2

4

SNO5

2 0

TN e

SNOe

18 16 14 12 10 8 6 4

7

8

9

10

11

12

13

14

7

3

8

9

10

11

12

100000

QR (m3/d)

Qcarbon (m3/d)

80000 2

1

60000 40000 20000

0

0 7

8

9

10 11 time (d)

12

13

14

7

8

Fig. 6. Simulation results of control strategy no. 4.

9

10 11 time (d)

12

ARTICLE IN PRESS M. Yong et al. / Control Engineering Practice 14 (2006) 1269–1278

20

10

TNe

18 concentration (mg N/l)

concentration (mg N/l)

8 6 SNO2

SNO5

4 2

SNOe

16 14 12 10 8 6 4

0 7

8

9

10

11

12

13

14

7

3

8

9

10

11

12

13

14

13

14

13

14

13

14

100000 80000

2 QR (m3/d)

Qcarbon (m3/d)

1275

1

60000 40000

0

20000 7

8

9

10

11

12

13

14

7

8

9

10

11

12

time (d)

time (d) Fig. 7. Simulation results of control strategy no. 5.

20

9 SNO2

8

NO5 concentration (mg N/l)

concentration (mg N/l)

7 6 5 4 3 2 1

SNOe

16 14 12 10 8 6 4

0 7

8

9

10

11

12

13

14

7

3

8

9

10

11

12

100000 80000

2 QR (m3/d)

Qcarbon (m3/d)

TNe

18

1

60000 40000

0

20000 7

8

9

10

11

12

13

14

7

8

9

10

11

time (d)

time (d) Fig. 8. Simulation results of control strategy no. 6.

12

ARTICLE IN PRESS 1276

M. Yong et al. / Control Engineering Practice 14 (2006) 1269–1278

Table 1 Simulation results of integrated control strategies of nitrate recirculation and external carbon addition Evaluation criteria

Reference case

Strategy no. 1

Strategy no. 2

Strategy no. 3

Strategy no. 4

Strategy no. 5

Strategy no. 6

EQ (kg d1) PE (kWh d1) AE (kWh d1) PSdisp (kg d1) PStotal (kg d1) Effluent average nitrate (gN m3) Effluent average ammonia (gN m3) Effluent average TN (gN m3) Effluent average TCOD (g m3) Average external carbon (m3 d1)

7556.5 1488.1 7241 2440.6 2675.5 12.44 2.53 16.93 48.22 0

5716.8 3955.8 7415.6 2677.6 2924.7 7.0 2.65 11.70 49.15 0.689

5640.8 3965.6 7442.3 2715.8 2962.5 6.65 2.79 11.45 49.16 0.774

5793.3 3929.1 7410.1 2673.6 2918.9 7.04 2.87 11.96 49.03 0.656

5798.0 3805.4 7410.5 2677.9 2923.4 7.06 2.84 11.94 49.04 0.668

5809.4 3078.7 7458.6 2706.8 2955.7 7.22 2.67 11.95 49.26 0.768

5753.9 4139.6 7413.0 2688.9 2934.8 7.0 2.76 11.82 49.07 0.712

18.30 17.26

18.46 0

19.19 0

21.13 0

20.68 0

18.51 0

19.64 0

1.0 1.54 0.11 0.29 0.09

1.95 2.36 1.09 0.34 0.11

1.66 3.13 0.4 0.74 0.55

1.99 3.29 0.72 0.57 0.32

1.88 3.05 0.32 0.64 0.40

1.00 2.44 0.05 0.56 0.31

2.14 4.79 0.45 0.90 0.81

7.0 7.91 6.00 0.39 0.15

6.66 9.42 3.48 1.22 1.49

7.01 8.06 5.61 0.47 0.22

7.05 8.26 5.45 0.56 0.31

7.21 7.50 6.56 0.20 0.04

7.00 8.40 5.22 0.60 0.36

% time of effluent violation SNH TN SNO2 controller Average nitrate in zone 2 (gN m3) Maximum value (gN m3) Minimum value (gN m3) Standard deviation of error Variance of error SNO5 controller Average nitrate in zone 5 (gN m3) Maximum value (gN m3) Minimum value (gN m3) Standard deviation of error Variance of error

— — — — —

EQ: effluent quality index (kg pollution units per day), PE: pumping energy, AE: aeration energy, PSdisp: average sludge for disposal per day, PStotal: total average sludge production per day.

strategies. For comparison, the default closed-loop benchmark reference case is also presented. The default benchmark system uses two controllers. The dissolved oxygen level in zone five is controlled at 2 mg l1 by manipulation of the oxygen mass transfer coefficient (KLa) and the nitrate concentration in the last anoxic zone is controlled at 1 mg N l1 by manipulation of the internal recycle flow rate. The air flow rates into zones three and four are assumed constant (KLa ¼ 240 d1) and the excess sludge flow rate is constant at 385 m3 d1. The reference case only represents a pre-defined simple control example and it is clear that all the integrated control strategies can significantly improve the effluent quality. Through the addition of an external carbon source, the average concentration of nitrate and TN in the effluent are significantly reduced by 42–47% and 29%–33%, respectively. However, for the proposed strategies, the effluent ammonia concentration is increased by 3.5–13%. The main reason for this is, because the addition of carbon source to the anoxic zone inevitably results in a small part of extra COD being spilled into the aerobic tanks. The limited supply of oxygen will then to a higher extent be used for removal of organic substrate rather than for the

nitrification process. Comparing the different strategies, control strategies 1 and 5 produce a slightly lower increase in effluent ammonia, while control strategies 3 and 4 increase the effluent ammonia concentration more. As shown in Table 1, control strategies no. 3 and 4 are using a somewhat lower carbon dosage (control strategy no. 3 are using the smallest dosage). The central idea of the control scheme is to get the effluent nitrate concentration as low as possible during low-load periods by making maximum use of influent and intracellular COD for denitrification and to maintain the effluent nitrate concentration at the highest allowable level (SNO5, sp) during high-load periods. Among all the control strategies, numbers 2, 5 and 6 give a slightly higher external carbon consumption (control strategy no. 2 are using the largest dosage). For strategy no. 2, the reason for the high external carbon dosing is that external carbon is dosed all the time, and the carbon addition is used to control SNO5 at the pre-specified setpoint of 7 mg N l1, which results in a delayed response. This inevitably results in more external COD being spilled into the aerobic tanks. For strategy no. 5, the reason for the high COD addition is that the external

ARTICLE IN PRESS M. Yong et al. / Control Engineering Practice 14 (2006) 1269–1278

carbon source is dosed also during low-load periods, which does not ensure the maximum use of influent COD for denitrification. Therefore, it increases the consumption of carbon source, aeration energy and leads to a higher sludge production. For strategy no. 6, the reason is that in order to keep the nitrate concentration SNO5 at the pre-specified set-point of 7 mg N l1, the nitrate recirculation flow rate must be increased to provide more nitrate to the anoxic zones, which results in higher external carbon consumption and more carbon source spilling into the aerobic tanks. Among all the tested control strategies, strategy no. 5 consumes the smallest amount of pumping energy (3078.7 kWh d1), which is a result of the low fixed nitrate recirculation flow rate during low-load periods. Strategy no. 6 consumes most energy for pumping (4139.6 kWh d1). The aeration energy for all control strategies is approximately the same with a slightly higher consumption for control strategy no. 5 (7458.6 kWh d1). Strategy no. 5 also produces the highest effluent quality index (5809.4 kg d1). The reason is that, during low-load periods with a fixed nitrate recirculation flow rate, not enough nitrate is recycled into the anoxic zone. This results in very low nitrate concentrations in the second anoxic zone, which limits the amount of denitrification taking place and, consequently, leads to a higher effluent nitrate concentration. The simulation results clearly show that the nitrate recirculation flow rate needs to be controlled dynamically, along with dynamic control of the external carbon addition. Strategy no. 2 produces the best effluent quality index (5640.8 kg d1) but requires the highest dosing of external carbon (average of 0.774 m3 d1). Control strategies 3, 4 and 5 give slightly higher average effluent TN concentrations compared with strategies no. 1 and 2, but note that the carbon dosage of control strategy no. 2 is increased by 12% compared to that of control strategy no. 1. The control strategies have no clear influence on the effluent total COD. The SNO2 controller of strategy no. 1 and the SNO5 controller of strategy no. 5 are better than that of other strategies in terms of overall control goals (maximum and minimum values, standard deviation and variance of error). Based on the above analysis (including more criteria not given in this text), it can be concluded that strategy no. 1 is the best integrated control strategy for nitrate recirculation flow rate and external carbon dosage in terms of external carbon consumption, plant performance criteria defined by the benchmark and the evaluation of the controllers. The strategy makes better use of the plant denitrification capacity and maximizes the use of the influent COD for denitrification during low-load periods. Moreover, the dosage of external carbon can be controlled to very precisely meet the effluent criteria for nitrate during high-load periods.

1277

5. Conclusions Six different integrated control strategies for nitrate recirculation and external carbon addition were evaluated using the COST simulation benchmark. The investigation focuses on biological treatment systems with a low influent COD to N ratio and high demands on nitrogen removal. The following conclusions can be drawn: The integrated dynamic control of the nitrate recirculation flow rate and the external carbon addition is essential for minimizing the consumption of external carbon source in a pre-denitrifying activated sludge process and meeting high-quality effluent standards. Controlling the nitrate recirculation flow rate without any or with only constant external carbon dosage will not meet effluent standards or result in excessive use of external carbon source. Controlling the external carbon dosage alone with a constant nitrate recirculation flow rate does not yield the best use of influent COD for the denitrification process or may not recycle enough nitrate into the anoxic zones, thus resulting in anaerobic conditions. The cost for nitrate recirculation in terms of energy consumption is normally limited compared to the cost of external carbon source in a well-designed plant. Therefore, only the nitrate recirculation flow rate is controlled without any external carbon addition during low-load periods. If the system does not meet the effluent nitrate criteria after applying the nitrate recirculation controller, then external carbon is added. Consequently, the dosage of external carbon should be controlled so as to just meet the effluent criteria for nitrate. Strategy no. 1 represents the best integrated control strategy for nitrate recirculation and external carbon dosage in terms of plant performance criteria as defined by the benchmark group and the evaluation of the controller performance. It comprises of two feedback control loops: one is to determine the dosing of external carbon source, keeping the nitrate concentration at the end of the anoxic zone at a pre-specified level, and the other is to adjust the flow rate of the nitrate recirculation to maintain the nitrate concentration at the end of the aerobic zone at a pre-specified level. Current research is focused on practical experiments in a pilot-scale plant to validate the behaviour of the proposed control strategies. The COST benchmark simulation model (BSM1) is a suitable and useful platform for objective evaluation of control strategies for the activated sludge system. A future extension will be to integrate the aeration control within the same framework as discussed in this paper and thereby be able to control the total nitrogen and also the ratio of nitrate and ammonia in the effluent in the best possible way. Such integration will be especially important for plants where also the

ARTICLE IN PRESS 1278

M. Yong et al. / Control Engineering Practice 14 (2006) 1269–1278

nitrification capacity is limited during high-load periods. Purposeful by-passing, step-feeding and aeration tank settling (ATS) represent further essential control elements to manage excessive hydraulic disturbances of the activated sludge process.

Acknowledgements This work was supported by the ‘‘863’’ Programme of China (contract no. 2004AA601020 and 2003AA601110) and by the National Natural Science Foundation of China (NSFC) (contract no. 20377003). References Balslev, P., Lynggaard-Jensen, A., & Nickelsen, C. (1996). Nutrient sensor based real-time on-line process control of a wastewater treatment plant using recirculation. Water Science Technology, 33(1), 183–192. Copp, J. B. (Ed.). (2002). The COST Simulation Benchmark— Description and Simulator Manual. Luxembourg: Office for Official Publications of the European Communities. Henze, H., Grady, C.P.L., Jr., Gujer, W., Marais, G.V.R., & Matsuo, T., (1987). Activated sludge model no. 1. IWA Scientific and Technical report no. 1, IWA Publishing, London, UK

Lindberg, C., & Carlsson, B. (1996). Adaptive control of external carbon flow rate in an activated sludge process. Water Science Technology, 34(3/4), 173–180. Londong, J. (1992). Strategies for optimized nitrate reduction with primary denitrification. Water Science Technology, 26(5/6), 1087–1096. Nyberg, U., Andersson, B., & Aspegren, H. (1996). Long-term experiences with external carbon sources for nitrogen removal. Water Science Technology, 33(12), 109–116. Purtschert, I., Siegrist, H., & Gujer, W. (1996). Enhanced denitrification with methanol at WWTP Zurich-Werdholzli. Water Science Technology, 33(12), 117–126. Samuelsson, P., & Carlsson, B. (2001). Feed-forward control of the external carbon flow rate in an activated sludge process. Water Science Technology, 43(1), 115–122. Singman, J., (1999). Efficient control of wastewater treatment plant—a benchmark study. MSc thesis, ISSN 1401-5765, Department of Earth Sciences, Uppsala University, Sweden. Taka`cs, I., Patry, G. G., & Nolasco, D. (1991). A dynamic model of the clarification thickening process. Water Research, 25(10), 1263–1271. Yuan, Z., Bogaert, H., & Vanrolleghem, P. (1997). Control of external carbon addition to predenitrifying systems. Journal of Environmental Engineering, 123(11), 1080–1086. Yuan, Z., & Keller, J. (2003). Integrated control of nitrate recirculation and external carbon addition in a predenitrification system. Water Science Technology, 48(11/12), 345–354. Yuan, Z., Oehmen, A., & Ingildsen, P. (2002). Control of nitrate recirculation flow in predenitrification systems. Water Science Technology, 45(4/5), 29–35.