Both the nitrification process and the denitrification process are complete

This means that S

_NO,a

= 0 and S

_NH,5

= 0 , and the control law can be simplified to:

( ) ( ) ( ) ( 4 ( ) W 4 ( ) W )

6

W 6 W W 4

4

_{LQ U}

UHI 12

LQ 1+

D

=

LQ

− +

,

,

(7.9)

An integral part is also included in the control laws in order to avoid bias in relation to the set point. The integral part is especially important in the second control law where

a

S

NO,

is assumed to be zero, since we know that S

_NO,a

≈ 1 [g/ m ] due to the chosen set

³

point of the carbon controller. The integral part looks like:

( ) ^{N .} ( ⁶ ( ) ^{N 6} ( ) ^N )

, =

_{, 12},5

−

₁₂,_UHI

∆ (7.10)

( ) ( ) ( )

( ) ( ) ( ) ( )

( ) ( )

 

 



<

∆ +

−

≤

∆ +

−

≤

∆ +

−

>

∆ +

−

=

min min

max min

max max

1

1 1

1

, N , N

, LI ,

, N , N

, , LI N

, N

,

, N , N

, LI ,

N

, (7.11)

The complete control law is given by:

( ) ( ) ( ( ) ( ) )

( ) ( 4 ( ) W 4 ( ) W ) ( ) , W W

6 6

W 6 W 6 W W 4

4

_{LQ U}

D 12 UHI 12

1+

LQ 1+

LQ

D

− + +

−

= −

, ,

5 ,

,

(7.12)

A schematic picture of the simplified control law (7.9) with the integral part (7.11) is

shown in Figure 7.2.

Zone: 1 2 3 4 5

Qr Qa

Sno,ref Qr

Sno,5

Sno,5

Snh,in

Qin Snh,in

Qin

)LJXUH

Schematic picture of the simplified control law.

 7XQLQJRIFRQWUROSDUDPHWHUV

The upper limit of the recirculation rate is chosen to 5 ⋅ Q

_in,stab

= 92230 [ m /d], as in the

³

original benchmark setup. If the recirculation rate is too high the performance of the two other controllers may deteriorate. A set point of 6 [g/ m ] for the nitrate concentration in

³

zone five turned out to be a realistic value with the chosen constraint on the control signal. First a controller based on equation (7.8) is tuned. The over shoot can not be reduced by decreasing K since a smaller value on the integrating constant leads to an

_I

unacceptable set point tracking, see Figure 7.3.

0 1 2 3 4 5 6 7

5.5 6 6.5 7 7.5

SNO,5 [g/m3]

0 1 2 3 4 5 6 7

5.5 6 6.5 7 7.5 8 8.5 9 9.5

x 10⁴

Time [days]

Qa [m3/d]

)LJXUH

Upper plot shows nitrate concentration in zone five and reference value, and lower plot shows the internal recirculation rate. A controller based on equation (7.8) is used with 4

10 I 6

K = ⋅ ^and the file constinfluent is simulated with ideal sensors.

Instead a controller based on equation (7.9) is tested. With this simplified controller design the over shoot is much smaller without any significant decrease in speed, see Figure 7.4. The simplified controller will be used in all remaining simulations.

0 1 2 3 4 5 6 7

5.5 6 6.5 7 7.5

SNO,5 [g/m3]

0 1 2 3 4 5 6 7

5.5 6 6.5 7 7.5 8 8.5 9 9.5

x 10⁴

Time [days]

Qa [m3/d]

)LJXUH

Upper plot shows nitrate concentration in zone five and reference value, and lower plot shows the internal recirculation rate. A simplified controller based on equation (7.9) is used with

104 I 6

K = ⋅ and the file constinfluent is simulated with ideal sensors.

When noise and delay is added to the sensors the controller performance is still satisfying

and no retuning is necessary, see Figure 7.5.

0 1 2 3 4 5 6 7 5.5

6 6.5 7 7.5

SNO,5 [g/m3]

0 1 2 3 4 5 6 7

5.5 6 6.5 7 7.5 8 8.5 9 9.5

x 10⁴

Time [days]

Qa [m3/d]

)LJXUH

Upper plot shows nitrate concentration in zone five and reference value, and lower plot shows the internal recirculation rate. A simplified controller based on equation (7.9) is used with

104 I 6

K = ⋅ and the file constinfluent is simulated with noise and delay on the sensors.

 6WHSGLVWXUEDQFHVLQLQIOXHQWORDG

The controller performance will be evaluated when steps are added to the influent

parameters Q and

_in

S

_NH,in

that directly are affecting the controller. The file constinfluent is simulated with all sensors ideal. The maximum step size of 45 [g/ m ] in the influent

³

ammonia concentration that was used in section 4 is too high. As seen in Figure 4.8 the external carbon controller could not keep the reference value during the positive three day step. To decrease the nitrate concentration in zone two, and indirectly in zone five, the upper bound on the external carbon flow must be increased. The internal recirculation controller does not itself contribute in decreasing the nitrogen concentration directly, but

“shuffles” the problem further to the external carbon controller. Since the carbon controller can not manage such high concentrations of influent ammonia the step sizes are decreased to get fairly reasonable results. In the following simulations the step sizes have been reduced by 25 percent instead of 10 percent as in section 4.

First steps are added to the influent ammonia concentration and the result can be seen in

Figure 7.6. The internal recirculation controller can still not keep the reference value

when .

0 2 4 6 8 10 12 14 5

6 7 8 9

SNO,5 [g/m3]

0 2 4 6 8 10 12 14

2 4 6 8

x 10⁴

Qa [m3/d]

0 2 4 6 8 10 12 14

20 25 30 35 40

Time [days]

SNH,in [g/m3]

)LJXUH Nitrate concentration in zone five, internal recirculation flow rate and influent ammonia concentration. The steps in influent ammonia concentration are added to the file constinfluent and all sensors are ideal. The step sizes are chosen as 25 percent higher and 25 percent lower than the minimum and maximum values in the influent data file dryinfluent.

When the original step sizes of influent ammonia were used both the carbon controller and the internal recycling controller worked at maximum capacity, but now when the step sizes have been reduced the carbon controller has unused capacity. By increasing the upper bound of the internal recycling flowrate the idle capacity of the carbon controller can be utilized. The upper bound of the internal recycling flowrate must be set in relation to the capacity of the external carbon controller. If the upper bound is set too low the capacity of the carbon controller is not fully utilized. If it on the other hand is set too high one tries to utilize a nonexistent capacity without any use, and the high recirculation flows may deteriorate the overall performance. The controller is not fast enough when the influent ammonia concentration increases at day twelve, and the nitrate concentration deviates from the set point, see Figure 7.6. The control signal increases very fast in the beginning due to the contribution from the first term in equation (7.9) when S

_NH,in

increases. When the step in S

_NH,in

has reached its final value the control signal increases slower since the dominating contribution now comes from the integral part.

The steps in the influent flow also had to be reduced with 25 percent instead of 10 in

order to get reasonable results. As seen in Figure 7.7 the internal recycle controller does

not manage to keep the reference value during the positive step in the influent flow rate.

0 2 4 6 8 10 12 14 5

6 7 8 9

SNO,5 [g/m3]

0 2 4 6 8 10 12 14

2 4 6 8

x 10⁴

Qa [m3/d]

0 2 4 6 8 10 12 14

1 1.5 2 2.5

x 10⁴

Time [days]

Qin [m3/d]

)LJXUH Nitrate concentration in zone five, internal recirculation flow rate and influent flow rate. The steps in the influent flow rate are added to the file constinfluent and all sensors are ideal. The step sizes are chosen as 25 percent higher and 25 percent lower than the minimum and maximum values in the influent data file dryinfluent.

When the water flow through the system increases the recirculation flow naturally also must increase, but the upper bound of 5 ⋅ Q

_in,stab

[ m /d] is obviously not high enough.

³

The capacity of the external carbon controller was not fully utilized during day three to six. The increase in the nitrate concentration in zone five day twelve can be explained in the same way as above. Q is present in both terms in equation (7.9), but the first term

_in

that is positive is dominating, and the positive step in Q results in an increase in the

_in

control signal.

 6LPXODWLRQVXVLQJZHDWKHUILOHV

All necessary preparations have now been made and the control performance can be evaluated by simulation of the weather files. All simulations are run with noise and delay on the sensors.

The first weather file to be run is constinfluent. The nirate concentration in zone five centers quite well around the reference value, but the peaks in influent ammonia can not be fully surpressed, see Figure 7.8. During these peaks both the carbon controller and the internal recycling controller work at maximum capacity. The limit for the effluent

ammonia concentration was violated at four different occasions.

0 2 4 6 8 10 12 14 4

6 8

SNO,5 [g/m3]

0 2 4 6 8 10 12 14

0 5 10

x 10⁴

Time [days]

Qa [m3/d]

0 2 4 6 8 10 12 14

4 5 6 7 8 9

SNO,e [g/m3]

)LJXUH Nitrate concentration in zone five, effluent nitrate concentration and internal recirculation flow rate. The weather file dryinfluent is simulated with noise and delay on the sensors.

Next the weather file raininfluent is simulated. Around day eight to ten when there is a constant increase in the influent flow the peaks in the nitrate concentration are gone, see Figure 7.9. The limit for the effluent ammonia concentration was violated at five different occasions when the file raininfluent was simulated.

0 2 4 6 8 10 12 14

4 6 8

SNO,5 [g/m3]

0 2 4 6 8 10 12 14

0 5 10

x 10⁴

Time [days]

Qa [m3/d]

0 2 4 6 8 10 12 14

4 5 6 7 8 9

SNO,e [g/m3]

)LJXUH Nitrate concentration in zone five, effluent nitrate concentration and internal recirculation flow rate. The weather file raininfluent is simulated with noise and delay on the sensors.

Finally the file storminfluent is simulated. At the two major flow peaks around day nine and day eleven the nitrate concentration drops as seen in Figure 7.10.

0 2 4 6 8 10 12 14

4 6 8

SNO,5 [g/m3]

0 2 4 6 8 10 12 14

0 5 10

x 10⁴

Time [days]

Qa [m3/d]

0 2 4 6 8 10 12 14

4 5 6 7 8 9

SNO,e [g/m3]

)LJXUH Nitrate concentration in zone five, effluent nitrate concentration and internal recirculation flow rate. The weather file storminfluent is simulated with noise and delay on the sensors.

The limit for the total effluent suspended solids was violated at two different occasions and the limit concerning the effluent ammonia concentration was violated six times.

 ([FHVVVOXGJHIORZUDWHFRQWUROOHU

By applying the control strategies presented in former chapters almost all effluent

constraints have been met when simulating the weather files. However, the threshold

value for the effluent ammonia concentration has not been met during any simulations. In

sections 5 and 6 different strategies to control the dissolved oxygen concentration have

been presented and those methods have improved the conditions of the nitrification

process. To lower the effluent ammonia concentration enough it has turned out to be

insufficient only to control the dissolved oxygen concentration. If the amount of sludge in

the system is increased there are more bacteria that can contribute in the nitrification

process and the ammonia concentration can be lowered. To increase the amount of sludge

in the system the excess sludge rate Q must be reduced, but this might cause sludge

_w

escape. A lower value on Q than the original of 385 [

_w

m

³

/day ] will lead to too high

sludge concentrations in the effluent when the weather file storminfluent is simulated. To solve the problem Q must be variable in time. In this chapter the sludge blanket height

_w

in the settler will be controlled with Q as control variable.

_w

 'HULYDWLRQRIWKHFRQWUROOHU

A simple PI-regulator on incremental form is chosen:

( ) N . ( H ( ) N H ( N ) ) . H ( ) N

4

_Z

=

_{3 /}

−

_/

− +

_{, /}

∆ 1 (8.1)

( ) ( ) ( )

( ) ( ) ( ) ( )

( ) ( )

 

 



<

∆ +

−

≤

∆ +

−

≤

∆ +

−

>

∆ +

−

=

min min

max min

max max

1

1 1

1

Z Z

Z Z

Z Z

Z Z

Z Z

Z Z

Z Z

Z

4 N 4 N

4 LI 4

4 N 4 N

4 4

LI N

4 N

4

4 N 4 N

4 LI 4

N

4 (8.2)

In document Automatic Control of an Activated Sludge Process in a Wastewater Treatment Plant - a Benchmark Study. (Page 51-59)