Automatic Control of an Activated Sludge Process in a Wastewater Treatment Plant - a Benchmark Study.

(QJLQHHULQJ3K\VLFV3URJUDPPH

8SSVDOD8QLYHUVLW\6FKRRORI(QJLQHHULQJ 837(&)

Anders Rehnström

Title

Automatic Control of an Activated Sludge Process in a Wastewater Treatment Plant - a Benchmark Study.

Abstract

In this master thesis work some control strategies for the activated sludge process in a wastewater treatment plant are evaluated. A benchmark developed within the COST 682 Working Group No.2, implemented on a Matlab/Simulink platform, was used to implement the control strategies and also to evaluate them by simulation. Five different controllers were considered; an external carbon flow rate controller, a supervisory dissolved oxygen set point controller, an aeration volume controller, an internal recycling flow rate controller and an excess sludge flow rate controller. The results from this study were compared to those obtained when simulating the two basic control strategies included in the benchmark model.

The effluent water quality, assessed by the effluent quality index in the benchmark, was significantly improved when applying the control strategies developed in this study.

Keywords

Supervisor(s)

Professor Bengt Carlsson, Dept of Systems and Control, Information Technology, Uppsala University Examiner

Professor Bengt Carlsson, Dept of Systems and Control, Information Technology, Uppsala University

Project name Sponsors

Language

English

Security

ISSN 1401-5757

Pages

150

Supplementary bibliographical information

6FKRRORI(QJLQHHULQJ6WXGLHV2IILFH 3KRQH +46-(0)18-4713003

9LVLWLQJDGGUHVV: Lägerhyddsvägen 2, bldg 4, Uppsala )D[ +46-(0)18-4713000 3RVWDODGGUHVV Box 823, SE-751 08 Uppsala, Sweden (PDLO kansli@uth.uu.se

 ,1752'8&7,21  

1.1 BACKGROUND... 2

1.2 PURPOSE... 3

 $*(1(5$/:$67(:$7(575($70(173/$17   2.1 MECHANICAL TREATMENT... 4

2.2 BIOLOGICAL TREATMENT... 4

2.3 CHEMICAL TREATMENT... 7

 7+(%(1&+0$5.  3.1 PLANT LAYOUT... 8

3.2 PROCESS MODEL... 8

3.3 INFLUENT LOAD... 8

3.4 EFFLUENT CONSTRAINTS... 8

3.5 PERFORMANCE ASSESSMENT... 9

 $)(('%$&.)((')25:$5'(;7(51$/&$5%21)/2:5$7(&21752//(5  4.1 SIMULATION CONDITIONS... 13

4.2 DERIVATION OF THE CONTROLLER... 13

4.3 TUNING OF CONTROL PARAMETERS IN THE FEED BACK PART... 16

4.4 UPPER BOUND ON EXTERNAL CARBON FLOWRATE... 16

4.5 CHOICE OF SET POINT... 18

4.6 STEP DISTURBANCES IN INFLUENT LOAD... 19

4.7 SIMULATIONS USING WEATHER DATA... 21

 683(59,625<',662/9('2;<*(16(732,17&21752//(5  5.1 DERIVATION OF THE CONTROLLER... 23

5.2 IDENTIFICATION OF THE PROCESS AND POLE PLACEMENT... 25

5.3 TUNING OF CONTROL PARAMETERS IN THE MASTER CONTROLLER... 27

5.4 GAIN SCHEDULING... 29

5.5 STEP DISTURBANCES IN INFLUENT LOAD... 31

5.6 SIMULATIONS USING WEATHER FILES... 32

 $(5$7,2192/80(&21752//(5  6.1 DERIVATION OF THE CONTROLLER... 38

6.2 TUNING OF CONTROL PARAMETERS... 41

6.3 STEP DISTURBANCES IN INFLUENT LOAD... 44

6.4 SIMULATIONS USING WEATHER DATA... 46

 ,17(51$/5(&<&/,1*)/2:5$7(&21752//(5  7.1 DERIVATION OF THE CONTROLLER... 48

7.2 TUNING OF CONTROL PARAMETERS... 51

7.3 STEP DISTURBANCES IN INFLUENT LOAD... 53

7.4 SIMULATIONS USING WEATHER FILES... 55

 (;&(666/8'*()/2:5$7(&21752//(5  8.1 DERIVATION OF THE CONTROLLER... 58

8.2 SIMULATIONS USING WEATHER FILES... 59

 6,08/$7,216:,7+7+(),1$/&21752/675$7(*<  9.1 DRY WEATHER CONDITIONS... 62

9.2 RAIN WEATHER CONDITIONS... 65

9.3 STORM WEATHER CONDITIONS... 68

 237,0,=$7,21  

 &21&/86,216 

 $&.12:/('*(0(176  

5()(5(1&(6  

$33(1',&(6 

A INFLUENT FILES... 76

B A FEEDBACK - FEEDFORWARD EXTERNAL CARBON FLOW RATE CONTROLLER... 80

C SUPERVISORY DISSOLVED OXYGEN SET POINT CONTROLLER... 87

D AERATION VOLUME CONTROLLER... 96

E INTERNAL RECYCLING FLOW RATE CONTROLLER... 105

F EXCESS SLUDGE FLOW RATE CONTROLLER... 115

G FINAL PLANT CONFIGURATION... 130

H ORIGINAL PLANT CONFIGURATION... 140

I OPTIMIZATION... 148

 ,QWURGXFWLRQ

 %DFNJURXQG

Water is one of the most fundamental things for the survival of mankind. Without drinking water the population of earth cannot survive. To provide the population of earth with drinking water might seem to be a minor problem since more than 70 percent of the earth’s surface is covered by water. Unfortunately it is not that simple. Only a fraction of the water on earth, about 0.5 percent, can serve as drinking water or be used in

agriculture.

The water in our environment has always followed a natural cycle, and in that natural cycle polluted water can be purified by nature itself. However, as the population grew bigger in the nineteenth century, more and more people moved into cities and the fast urbanisation led to sanitary problems and epidemics. In order to improve the health of the citizens, cloak systems were built and the water closet was introduced. A new forced water cycle was now created by man. Here polluted sewage water directly was led to the surrounding recipients without any treatment at all. As the condition of the recipients grew worse it was realized that the wastewater had to be treated in some way before it was released back into the recipients.

The first construction for wastewater treatment in Sweden was built in 1911, but not until the 60’s in connection to a new environmental protective law the real expansion of wastewater treatment plants took place. The demands on the effluent water quality from the wastewater treatment plants have become stricter with time, and this trend will probably continue. At the same time the loads to existing plants are predicted to increase due to the growth of urban areas.

One way to meet the higher demands of efficiency is to over-dimension the plants, but

this is neither cost effective nor a long-term-solution. Instead it takes more elaborate

control and supervision strategies to achieve higher operational efficiency, but

unfortunately many plants use very simple control strategies or no automatic control at all today.

 3XUSRVH

In order to meet stricter effluent quality standards at a minimum cost, additional research within the area is necessary to find new control strategies for wastewater treatment plants.

In this master thesis work some control strategies for wastewater treatment plants are implemented and evaluated. A benchmark developed within the COST 682 Working Group No.2, implemented on a matlab/simulink platform, is used to implement the control strategies and also to evaluated them by simulation.

 $JHQHUDOZDVWHZDWHUWUHDWPHQWSODQW

The main object of municipal wastewater treatment plants is to separate pollutions such as solid particles, floating material, organic material and nutritive salts (such as

phosphorus and nitrogen) from the wastewater before it is returned to the recipients.

Depending on the structure of the surrounding society the composition and amount of the

wastewater varies. Thus the layout and size of the wastewater treatment plants must vary

from location to location. Another factor that is important on the choice of type of plant

for a certain location is the character of the recipients in connection to the plant. Most

plants along the coast in the southern part of Sweden, for example, have more advanced

wastewater treatment than plants have in the north of Sweden, since the recipients are

more sensitive in the south ( ,QWURGXNWLRQWLOODYORSSVWHNQLQNHQ1996). There are national

regulations concerning treatment of wastewater, but as a consequence of the different

conditions from location to location the requirements on wastewater treatment plants

differs. In a general municipal wastewater treatment plant the wastewater passes three

different steps in the plant where the pollutions mentioned above are removed, see Figure

2.1.

Chemical treatment

Sludge treatment

Primary Sedimentation

Dewatered sludge

water

Sludge

thickening Stabilization

Dewatering

Biological treatment

Sand filter Grid

Activated sludge

Supernatants + Backwashing

Effluent

Mechanical treatment 2 3

1

4

Chemicals

Preciptation

)LJXUH



The three steps are: mechanical treatment, biological treatment and chemical treatment.

 0HFKDQLFDOWUHDWPHQW

When the wastewater enters the plant it first passes the mechanical treatment step where bigger objects such as rags are removed in a grating. After the grating sand and other heavier particles are removed in a sand trap. Somewhat lighter particles are finally removed in a primary settler where much of the suspended solids settles. Phosphorus, nitrogen and some organic material are also to some extent removed here as the removed suspended solids contain these contaminations. The sludge being produced in the primary settler is led to the sludge treatment where it is further processed, while the material that are trapped in the two proceeding steps are transported to a dump.

 %LRORJLFDOWUHDWPHQW

After the mechanical step objects and larger particles have been removed from the wastewater, and the water now enters the biological treatment step. A great deal of the organic material that is left in the wastewater occurs in soluble form, and must in some way be transformed into removable particles. Microorganisms, among which bacteria are most important, carry this out. To grow and propagate the bacteria need energy and nourishment, and most bacteria use organic material for growth, but other sources for growth are used by other bacteria as well. The newly created bacteria gather into flocks that are heavy enough to be removed in sedimentation basins. In this way the

microorganisms transform and concentrate soluble organic material. All bacteria

originate from the influent wastewater and form a so-called mixed culture of bacteria,

which contains many different types of bacteria. The bacteria composition is never

constant since the character of the wastewater never is the same; temperature, pH value

and other factors are constantly changing. The biological processes in the biological

treatment step are also present in natural watercourses, but in the wastewater treatment

plant these processes are more effective. The high concentration of microorganisms and

the good supply of nourishment make the conditions more favourable. There are many

different techniques for biological treatment of wastewater. One of the most common is

the activated sludge process, shown in Figure 2.2, which will be described in the following.

Inluent water

Recirculated sludge

Excess sludge Effluent water

Aeration basin Settler

)LJXUH

 7KHDFWLYDWHGVOXGJHSURFHVV

The activated sludge process consists of an aerated basin and a settler. In the aerated basin bacteria live and propagate. By blowing air into the water oxygen is added, which is used by the bacteria to oxidize the organic material from the wastewater. When the bacteria oxidize the organic material they get energy to maintain their living functions and to propagate. Some of the organic material is then transformed to carbon dioxide and some is incorporated into new cell mass. The new cell mass forms sludge that is

separated in the secondary settler and transported to the sludge treatment. The sludge contains both living and death microorganisms and thus contains organic material, but also some phosphorous and nitrogen.

In order to transform the contaminations at a desired pace the amount of sludge, or the concentration of microorganisms, in the aerated basin must be controlled. The

microorganisms grow fairly slowly and to maintain a certain population size in the aerated tank, sludge containing microorganisms is recirculated back to the aerated basin from the secondary settler. The concentration of microorganisms in the aerated basin should in principle be kept constant. To achieve this some sludge must be removed as excess sludge from the settler, since new sludge continuously is formed in the aeration basin.

Nitrogen is present in wastewater in several forms such as: ammonia ( NH ), ammonium

( NH ), nitrate (

NH

), nitrite ( NO ) and as organic compounds. High concentrations of

nitrogen in the effluent water may cause different problems. Nitrogen stimulates the growth of aquatic plants and algae, which in due course die and are degraded by bacteria.

In the degrading process large quantities of oxygen are consumed by the bacteria, which may result in a severe lack of oxygen. High concentrations of ammonium in the effluent water will also reduce the oxygen store in the recipient, since the ammonia is oxidized to nitrate causing heavy oxygen consumption. To avoid these and other problems caused by too high nitrogen concentrations in the effluent wastewater the amount of nitrogen in the incoming wastewater must be reduced. At an ordinary biological treatment of the

wastewater approximately 10-30% of the nitrogen is removed with the excess sludge. To

accomplish a higher degree of nitrogen removal in the biological treatment a special kind

of technique is required. The most common one is called biological nitrogen removal and is based on transformation of the ammonia to nitrogen gas by the aid of bacteria. This process can be described by a two step procedure. During the first step, the nitrification process, oxygen must be present when ammonia is oxidized to nitrate (aerobic

conditions). The nitrification process actually consists of two sub processes, which can be described by the following simplified chemical reaction formulas:

−

−

+

− +

⇒ +

+ +

⇒ +

3 2

2

2 2 2

4

5 . 0

2 5

. 1

12 2

12

+ 2 + 12 2

1+ (2.1)

The two types of bacteria that are active in these two sub processes grow very slowly and hence a long retention time (sludge age) is required. The second step in the biological nitrogen reduction is called denitrification, see equation (2.2). The denitrification process takes place during anoxic conditions where no dissolved oxygen is present. To respire the bacteria make use of the oxygen bound to the nitrate, instead of using dissolved oxygen from the water. In this way the bacteria can oxidize organic material at which nitrate is transformed to nitrogen gas.

2 2

2

3

2 2 . 5

2 12

+ +

⇒ 1 + + 2 + 2 (2.2)

The denitrification process does not demand as long retention time (sludge age) as the nitrification process does. On the other hand plenty of organic material is needed for the denitrification process to work and the dissolved oxygen concentration must be low.

How is the biological nitrogen removal practically accomplished then? The right

conditions of life for the nitrification and denitrification processes must be created. This means that anoxic zones must be incorporated into the aerated basin in the activated sludge process.

The anoxic zones can be placed either before the aerobic zones (pre-denitrification), as in Figure 2.3, or after the aerobic zones (post-denitrification).

Zone: 1 2 3 4 5

Inluent water

Recirculated water

Recirculated sludge

Effluent water

Excess sludge Settler

)LJXUH

If the post-denitrification system is used oxidation of organic material and nitrification

take place in the aerated zones at the same time. Since a great deal of the organic material

from the incoming wastewater then is degraded in the aerated zones, carbon must be

added to the anoxic zones to keep the denitrification process going. Methanol, for example, can serve as an external carbon source. When pre-denitrification is applied the denitrification bacteria better utilise the organic material in the incoming wastewater.

Since the nitrification takes place in the aerated zones that are placed after the anoxic ones, water with high concentrations of nitrate from the end of aerobic zones must be recirculated back to the beginning of the anoxic zones. The flow rate of the recirculated water is typically three to five times larger than the flow rate of the incoming water. The recirculated water contains oxygen that makes the denitrification process less efficient and hence an external carbon source may be required to reduce the oxygen concentration.

If an external carbon source is necessary or not to cover the carbon need of the

denitrifying bacteria depends of course on the composition of the influent waste water.

 &KHPLFDOWUHDWPHQW

The chemical treatment first and foremost aims to remove phosphorus from the wastewater. In a wastewater plant without chemical treatment about one third of the phosphorus is removed, some in the mechanical treatment and some in the biological treatment. If chemical treatment is applied around 80-95% of the phosphorus can be removed, depending on the chemical being used. The added chemical precipitates the phosphorus, and when the precipitated phosphorus has gathered it can be separated in the following sedimentation basin. Se for example ( ,QWURGXNWLRQWLOODYORSSVWHNQLNHQ1996) for more details.

 7KHEHQFKPDUN

Wastewater treatment plants are non-linear systems consisting of a great number of complex processes. The influent water, or the system input, varies both in amount and composition and the variations are difficult to predict. Still the wastewater plants have to be operated uninterruptedly while meeting stricter and stricter regulations.

Many different control strategies have been proposed for wastewater treatment plants, but it has been troublesome or even impossible to evaluate and compare the strategies, either practically or by simulation. This is mainly due to the large variations in the influent, to the complexity of the chemical and biochemical processes and to the great span of time constants (from a few minutes to several days) present in the activated sludge process. An additional factor complicating the evaluation is the lack of standard evaluation criteria.

This follows from the fact that regulations regarding effluent water quality and labour costs often are location specific. This makes it difficult to judge the impact of a control strategy when a performance increase is reported, since the conditions often is less than optimal. If the conditions had been better the same control strategy could have given better performance results. The complexity of the systems makes it difficult to put new control strategies into practise, which in turn means that different strategies rarely are compared in a fair way and some ideas are never realised at all.

To enhance the development and acceptance of new control strategies the evaluation

procedure must be made easier and in some way uniform. Therefore a so-called

“benchmark” has been developed within the COST 682 Working Group No.2. The benchmark is a simulation environment defining a plant layout, a process model, influent loads, test procedures and evaluation criteria. Demand for realism, simplicity and

accepted standards were taken into consideration when developing these different parts.

For more information see: http://www.ensic.unancy.fr/COSTWWTP/ and Alex et al.

1999.

 3ODQWOD\RXW

The layout of the plant is relatively simple and common, and consists of a bioreactor and a secondary settler. The bioreactor consists of five compartments and predenitrification is applied. The first two compartments are anoxic while the last three are aerated. All five compartments are considered to be fully mixed. The secondary settler is modeled as a series of ten layers and the double exponential settling velocity model proposed by Takás et al. (1991) is chosen to resemble the behavior of the settler.

 3URFHVVPRGHO

The IAWQ Activated Sludge Model No 1 (ASM1) was selected to model the biological processes in the bioreactor. The ASM1 model is probably the most widely used model for the activated sludge process and can be considered as a ‘state of the art’ model. The ASM1 model consists of thirteen state variables and eight processes. The processes are:

anoxic growth of heterotrophs, aerobic growth of heterotrophs and autotrophs, decay of heterotrophs and autotrophs, ammonification of soluble organic nitrogen, hydrolysis of entrapped organics and finally hydrolysis of entrapped organic nitrogen.

 ,QIOXHQWORDG

There are three influent files, representing three different weather conditions during 14 days. The data files aim to mimic real operating conditions. The first file is constructed to resemble a dry weather period with decrease in flow and load during weekends. The other two files are designed with the dry weather file as a starting point with an added rain event during the second week. The first of the two rain files simulates a period of steady downpour during the second week, which results in a constant increase in influent for two days. Compared to the dry weather file this file has a constant hydraulic load increase without any increase in carbon oxygen demand (COD) or nitrogen. The second rain file has two storm events during the second week that are shorter in time compared to the rain events, but more intense. The storm events give rise not only to an increase in the hydraulic load, but also an increase in particulate load. The increase in particulate load illustrates a first flush event in the sewer system. Plots of the weather files can be found in appendix A.

 (IIOXHQWFRQVWUDLQWV

The assessment is based on data generated during the last seven days when the weather

files are used as input. There are constraints with respect to the effluent water quality that

should not be violated, see table 3.1. The flow-weighted average effluent concentrations

of the following variables must meet their corresponding limitations.

e

N

<18 [ g/m ]

COD

<100 [ g/m ]

e

S

<4 [ g/m ]

SS

<30 [ g/m ]

e

BOD

<10 [ g/m ]

7DEOH

The limiting variables are calculated according to the following expressions:

Total effluent nitrogen concentration:

H 12 H 1.M H

727

6 6

1

=

+

(3.1)

Effluent Kjeldahl nitrogen concentration:

(

)

(

)

;%

H 1' H 1' H 1+

H

1.M

6 6 ; L ; ; L ; ;

6

=

+

+

+ ⋅

+

+ ⋅

+

(3.2)

Effluent carbon oxygen demand:

H , H 3 H

%$

H

%+

H 6 H , H 6

H

6 6 ; ; ; ; ;

&2' =

+

+

+

+

+

+

(3.3) Effluent suspended solids concentration:

(

)

H

; ; ; ; ;

66 = 0 . 75 ⋅

+

+

+

+

(3.4)

Effluent biological oxygen demand – 5 days:

( ) ( )

(

)

H

6 ; I ; ;

%2'

= 0 . 25 ⋅

+

+ 1 − ⋅

+

(3.5)

The number of violations as well as the percentage of time the limitations are not met should be reported.

 3HUIRUPDQFHDVVHVVPHQW

The performance assessment is made at two levels. At the first level the local control

loops are assessed and at the second level the effect of the control strategy on plant

performance is studied. This level can be further divided into two sub levels; one

concerning effluent and influent water quality and one concerning cost factors for

operation.

 &RQWUROOHUSHUIRUPDQFH

The local control loops are assessed by means of the IAE (Integral of the Absolute Error) and the ISE (Integral of the Squared Error) criteria, by registering maximum deviation from set point, and by measuring the variance and the standard deviation of the error:

GW H ,$(

W L L ⁼

∫

=

=

7

(3.6)

=

∫

=

=

14

7 2 W

W L

L

H GW

,6( (3.7)

L

L

H

'HY

= max (3.8)

( ) 



 



− 

= 7

,$(

7 H ,6(

9DU

(3.9)

( ) H

9DU ( ) H

6WG = (3.10)

 (IIOXHQWDQGLQIOXHQWZDWHUTXDOLW\

To measure the effluent water quality an effluent quality index (E.Q.) is used, see

equation (3.11). Compounds in the effluent that have a major negative effect on receiving water bodies are included in the index. Some compounds are also interesting in this context since they are included in legislation; taxes have to be paid due to discharge of pollution to receiving water bodies, and violation of threshold values might result in fines. These kinds of compounds are also included in the index. All compounds are weighted and the index value is an average over the period of observation. The E.Q. is defined as follows:

( ) ( )

( ) ( )

( )

( ) W GW 4 H W

W % %2' %2' H W

H W 6 12

% 12 H W

6 1.M

% 1.M

H W

&2'

&2'

% H W 66 66

%

(4 7 = ∫ ⋅

=     





 

 







⋅ +

⋅ +

⋅ +

⋅ +

⋅

= ⋅ 14

7

, 5 5

, , 1000

1 (3.11)

The different variables are already defined above and B are weighting factors to convert

the different types of pollution to pollution units.

2 20

20 1

5

2

=

=

=

=

=

%2' 12 1.M

&2'

%

%

%

%

%66

An index for the influent water quality has also been defined since the composition of the incoming water often varies from situation to situation. The influent quality (I.Q.) is defined as:

( ) ( )

( ) ( )

( ) ⁴ ( ) ^{W GW}

W

%2'

%

W 6

% W 6

%

W

&2'

% W 66

%

,4 7

W

LQ

%2'

LQ 12 12 LQ

1.M 1.M

LQ

&2' LQ

66

⋅



 







 





⋅ +

⋅ +

⋅ +

⋅ +

⋅

= ⋅

∫

= 14

7

, 5 5

,

1000

1 (3.12)

The weighting factors are the same as in the expression for the E.Q. The different variables are defined above.

 &RVWIDFWRUVIRURSHUDWLRQ

There are three costs factors taken into consideration for operation of the wastewater plant, namely: the variation of the controller output, the sludge production and the aeration and pumping energy being consumed. The variation of the controller output is interesting since it gives an indication of the wear of the pumps and aeration devices. The total sludge production consists of the sludge lost at the weir and the sludge to be

disposed. The sludge to be disposed is calculated from the total solid flow from wastage and the solids accumulated in the system during the last seven days. Amount of solids in the system at time t:

( ) W 766 ( ) W 766 ( ) W

766 =

+

(3.13)

( ) ^t

TSS

is the amount of solids in the reactor at time t:

( ) ( )

1

L 6L %+L %$L 3L ,L

D

W ; ; ; ; ; 9

766 = ⋅ ∑

+ + + + ⋅

=1 , , , , ,

75 .

0 (3.14)

where 5 N

= is the number of compartments. ^TSS

( ) ^t is the amount of solids in the settler at time t:

( ) ( )

1

M 6 M %+ M %$M 3 M , M

V

W ; ; ; ; ; 9

766 = ⋅ ∑

+ + + + ⋅

=1 , , , , ,

75 .

0 (3.15)

where N

= 10 is the number of layers in the settler. Thus the sludge to be disposed at time t will be:

( ) ( ) 4 GW

;

;

;

; 766 ;

7 766

3

W 3Z ,Z

Z

%$

Z

%+

Z 6

VOXGJH

  ⋅





 













+ +

+

⋅ + +

−

⋅

=

∫

= 14

7 , ,

, ,

75

. 0 7 1 14

(3.16)

And finally the total sludge production at time t is:

∫ ( )

=

=

 ⋅







+ +

+

⋅ + +

=

7 , ,

, ,

, _

75 .

0

W

H H

, H 3

H

%$

H

%+

H 6 VOXGJH

VOXGJH

WRWDO

4 W GW

;

;

;

;

; 3 7

3 (3.17)

Plant peculiarities like bubble size, depth of submersion, e t c should be taken into consideration when calculating the aeration energy (AE). The AE is calculated from the

a

K

-values in the aerated compartments according to the following formula:

( ) ( )

( )

=

∫ ∑

= =

⋅ +

⋅

⋅

=

14

7 5

3

2

7 . 8408 4032

. 24

0

W L

N

D

N

D

GW

$( 7 (3.18)

Index i denotes the compartment number.

The pumping energy is a measure of the energy consumed by the internal and external recycle pumps and is calculated as:

( ) ( ) ( )

( 4 W 4 W 4 W ) GW

3( 7

W

Z U

∫

=

=

+ +

⋅

=

7

04 .

0 (3.19)

 $IHHGEDFNIHHGIRUZDUGH[WHUQDOFDUERQIORZUDWH FRQWUROOHU

As mentioned before biological nitrogen removal in an activated sludge process is carried out by two biological processes, namely nitrification and denitrification. The

denitrification bacteria is in need of sufficient amounts of readily metabolised carbon in order to function optimal. In a post denitrification system an external carbon source is normally necessary and in a pre denitrification system, which is studied here, additional carbon might be needed if the carbon/nitrogen ratio is too low in the influent water.

Since the load to a typical plant may vary a factor six or more a constant flow rate of external carbon will give a pore result. It may therefore be better to control the dosage of external carbon automatically in some way, so that changes in the influent can be

compensated for. One should strive to add just enough carbon to cover the need of the denitrification bacteria. A too high dosage of external carbon leads to increased sludge production, unnecessary high operational costs and may lead to carbon spill in the effluent.

The aim of the external carbon controller is to keep the nitrate concentration in the last

anoxic zone at a low pre specified level by adjusting the external carbon flow rate. The

goal is also to design a controller that quickly attenuates disturbances from the influent

load. See also Samuelsson & Carlsson 1999.

 6LPXODWLRQFRQGLWLRQV

The excess sludge flow rate has been reduced from 385 to 300 [ m

/d ] in order to get a sufficiently high sludge age, and the internal recirculation rate is set to a constant value of

Q

5 ⋅ [ m

/d ]. The airflow rate is kept constant at 240 [ m

/h ] in zones three and four and the original controller regulates the dissolved oxygen concentration in zone five by manipulation of the airflow rate. Its set point is 2 [ g O/m

] and the upper bound of the control signal is 240 [ m

/h ].

 'HULYDWLRQRIWKHFRQWUROOHU

The control strategy is based on a steady state analysis of a simplified IAWQ Activated Sludge Model No 1 (ASM1). The full ASM1 is rather complex and may therefore be difficult to use for controller design. To describe the denitrification process, or the anoxic growth of heterotrophs, at least three components must be taken into consideration;

heterotrophic biomass X

[g COD/ m ], readily biodegradable substrate

S [g

COD/ m ] and soluble nitrate

S

[g COD/ m ].

Q Q

V Ss,in

Sno,in Xbh,in

Ss Sno Xbh

)LJXUH

If the mass balance is studied and the notation in Figure 4.1 is used the following set of equations is obtained.

( ^; _{% + LQ} ^; _{% +} )

+ '

; % GW 5

+ G; %

, ,

, ,

, = + − (4.1)

( ⁶ _{6 LQ} ⁶ ₆ )

6 ' 5 6 GW 6

G6 = + , − (4.2)

( ⁶ _{12 LQ} ⁶ ₁₂ )

12 ' 5 6 GW 12

G6 = + , − (4.3)

Here V

D = Q is the dilution rate and R denotes the reaction rate. Before the reaction rates are inserted into the equations some simplifications of the model are made in order to reduce the complexity of the reaction rates.

A.1 It is assumed that the dissolved oxygen is S

≈ 0 . A.2 The hydrolysis and decay processes are ignored.

A.3 The external carbon source is assumed to be readily biodegradable.

If the reaction rates are simplified according to the assumptions A.1 and A.2 and assumption A.3 is considered the following model is obtained.

( ) ^; _{% +} ^' ( ^; _{% + LQ} ^; _{% +} )

GW + G; %

, ,

, ,

, = µ θ + − (4.4)

( ) ^; _{% +} ^' ( ⁶ _{6 LQ} ⁶ ₆ )

< + GW 6

G; = − , + , −

1 µ θ (4.5)

( ) ^; _{% +} ^' ( ⁶ _{12 LQ} ⁶ ₁₂ ) ₉ ^X

< + GW 12

G6 1

,

, + − +

−

= β µ θ

(4.6)

Where

( )

86 . 2 1 − < +

β = , <

= 0 . 67 (4.7)

and u is the flow rate of external carbon defined as:

FDU FDU

&2' 4

X = (4.8)

where Q

[ m

/d ] is the flow rate of the external carbon source and COD

[ g COD/m

] is its COD content. Observe that u is expressed as a mass flow and has the unit [g COD/d]. At steady state equation (4.4) and equation (4.6) become:

( )

(

)

+

%

'< 6 6

;

=

−

θ

βµ (4.9)

( )

(

)

+

6 6

'

< ;

9 1 X = βµ θ

−

−

(4.10)

Inserting equation (4.9) in equation (4.10), replacing S

with a reference value S

and solving for u gives:

( ) ( ) _



 



 − − −

= 4 6

6

6

6

X

1

β (4.11)

Note that Q actually consists of three different flows; the influent flow, the flow

recirculated from zone five and the flow recirculated from the settler. To compensate for the model simplifications and possible unmeasurable disturbances the feedforward strategy is combined with a feedback PI-part including anti windup. With the benchmark notation the final control law becomes:

( ) ( ( ) ( ) ( ) ) ( ( ) ( ) ) ( ( ) ( ) )

( ) ( )

( ) ( ( ) ( ) )

( ) ( )

( )

∫

∫

− +

− +

− +

 

 



 − − −

+ +

=

W W VDW

$

W

W 12 12UHI

, UHI

12 12

3

6 6

UHI 12 12

U D

LQ

G X X

.

G 6

6 . W 6 W 6 .

W 6 W 6 W 6 W 6 W

4 W 4 W 4 W X

0

0

, 2

, ,

2 ,

0 , ,

0 ,

1

τ τ τ

τ τ τ

β

(4.12)

Since the feed forward part attenuates disturbances relatively fast the feedback part can be made slow. In Figure 4.2 a schematic picture of the control law is shown.

Q Sno,ref Carbon controller

External carbon

Qr

Zone: 1 2 3 4 5

Qa Sno,2 Ss,2

Ss,0 Sno,0

)LJXUH

This control law is derived for an ASP with one anoxic compartment, but often the ASP

consists of several compartments. It can be shown that the same control law also is

applicable for the multi-compartment case. See Samuelsson and Carlsson (1999).

 7XQLQJRIFRQWUROSDUDPHWHUVLQWKHIHHGEDFNSDUW

According to Singman (1999) the parameters should be chosen as K

= 7 ⋅ 10

and 0

K

= . The fact that K is superfluous is not rather surprising since the feed forward

part speeds up the controller. As several simplifications are made in the derivation of the controller K is utmost necessary in order to prevent bias in relation to the set point. The

integral part should be chosen in the same region as the size of the feed forward part, and its final value was 7 ⋅ 10

. The controller is quite fast and keeps the reference value without any difficulties. By multiplying the flow in the feed forward part with different constants the speed of the controller can be manipulated. This is illustrated in Figure 4.3.

0 2 4 6

0.8 1 1.2 1.4 1.6 1.8 2 2.2

SNO,2 [g/m3]

0 2 4 6

0 0.5 1 1.5 2 2.5

x 10⁶

External carbon [g/d]

Time [days]

0 2 4 6

0.8 1 1.2 1.4 1.6 1.8 2 2.2

0 2 4 6

0 0.5 1 1.5 2 2.5

x 10⁶

Time [days]

0 2 4 6

0.8 1 1.2 1.4 1.6 1.8 2 2.2

0 2 4 6

0 0.5 1 1.5 2 2.5

x 10⁶

Time [days]

)LJXUH

10 I 7

K = ⋅ .

The flow was kept unscaled since this gave the desired speed. When measurement noise and delay was added to the sensors the controller still worked satisfactory and any retuning was not necessary.

 8SSHUERXQGRQH[WHUQDOFDUERQIORZUDWH

Ethanol was chosen as an external carbon source with COD

= 1,2 ⋅ 10

[ g COD/m

].

To find out how to choose the upper bound the dry weather file was simulated, which

represents normal operational conditions. First the carbon flow from the external carbon

source was unbounded and reached at most about 2 , 6 ⋅ 10

[ g COD/d ], which

corresponds to just over 2 m

ethanol/d . Then some simulations were run with different values on the upper bound, and the performance of the controller was evaluated by studying the nitrate concentration in zone two. If too little carbon was added due to a low value of the upper bound the control performance deteriorated significantly.

0 5 10

0 0.5 1 1.5 2 2.5 3

SNO,2 [g/m3]

0 5 10

0 0.5 1 1.5 2 2.5

x 10⁶

External carbon [g/day]

Time [days]

0 5 10

0 0.5 1 1.5 2 2.5 3

0 5 10

0 0.5 1 1.5 2

x 10⁶

Time [days]

0 5 10

0 2 4 6 8 10 12

0 5 10

−1

−0.5 0 0.5 1

Time [days]

)LJXUH

10 1.8 6, 10 max 2.4

u = ⋅ ⋅ ^[gCOD/d]. The

file dryinfluent is simulated without noise and delay on the sensors.

After some consideration and guidance from Singman 1999 the upper bound on the external carbon flow rate was chosen to 2 , 6 ⋅ 10

[ g COD/d ], which exactly corresponds to ethanol/d 2 m

. As seen in Figure 4.4 the chosen limit of the flow rate is only

exceeded at a few peaks and is thus fairly high. But taking into consideration that two

additional controllers will be added later, one governing the dissolved oxygen level and

one governing the recirculated flow rate, the high level is reasonable. A more efficient

oxygen controller improves the nitrification process in the aerobic zones and this will

lead to higher concentrations of nitrogen there, and this water will be recirculated to the

anoxic zones. Such conditions are of course more demanding for the external carbon

controller. To show the necessity of an external carbon source it is also shown in Figure

4.4 what happens when u

= 0 .

 &KRLFHRIVHWSRLQW

According to Yuan et al (1996) the set point of the nitrate level in the anoxic part may be chosen as two times the value of the half saturation coefficient for nitrate, K

. That would give a setpoint equal to one, since K

= 0.5 . In broad outline the half saturation coefficient determines at how low nitrate concentrations the denitrification process can be active. The smaller value of the constant, the lower concentrations of nitrate are possible without stopping the denitrification process and hence the set point can be chosen

smaller. Three different values of the set point have been simulated using the dryinfluent weather file; S

= 0.5 , 1.0, 2.0 [ g/m ], see Figure 4.5.

0 5 10

0 1 2 3

SNO,2 [g/m3]

0 5 10

0 1 2 3

x 10⁶

External carbon [g/day]

0 5 10

3 4 5 6 7 8

Time [days]

SNO,e [g/m3]

0 5 10

0 1 2 3

0 5 10

0 1 2 3

x 10⁶

0 5 10

3 4 5 6 7 8

Time [days]

0 5 10

0 1 2 3

0 5 10

0 1 2 3

x 10⁶

0 5 10

3 4 5 6 7 8

Time [days]

)LJXUH

0.5 1, ref 2,

SNO, = ^[ 3

g/m ].

The controller manages to keep all three set points fairly well. The deviation from the set point is actually smallest for the lowest set point, but the control signal is then quite high in relation to the its upper bound and reaches it at a few peaks. The effluent

concentrations of nitrogen are of course lowest for S

= 0.5 [ g/m ]. The set point

1.0

S

= [ g/m ] is finally chosen since the deviations from the set point are

acceptable and the control signal is not to high in relation to its upper bound. Again it

must be taken into consideration that the conditions will be more demanding later when

the other two controllers are added, and hence the set point can not be chosen to low.

 6WHSGLVWXUEDQFHVLQLQIOXHQWORDG

A well-designed controller should level out disturbances in the influent load. Therefore it is of interest to evaluate the control performance when applying steps to the influent components that are directly affecting the controller. The components of interest are:

readily biodegradable substrate S

, nitrate S

and flow Q . The concentrations of

incoming nitrate are normally very low and hence it is more realistic to apply steps in incoming ammonia than in incoming nitrate. The effect will be similar since the ammonia is converted to nitrate in the aerated zones and recirculated back to the anoxic zones. The steps were added to the file constinfluent. To get realistic steps the mean, maximum and minimum values of S

, S

and Q were taken from the file dryinfluent. The sizes of

the steps were then chosen 10 percent lower than the maximum values and 10 percent higher than the minimum values and the steps were centred around the mean value.

0 2 4 6 8 10 12 14

0.6 0.8 1 1.2

SNO,2 [g/m3]

0 2 4 6 8 10 12 14

0 0.5 1 1.5 2

x 10⁶

External carbon [g/day]

0 2 4 6 8 10 12 14

40 60 80 100 120

SS,in [g/m3]

Time [days]

)LJXUH

in

SS, . The file constinfluent is simulated without noise and delay on the sensors.

As seen in Figure 4.6 the effects of the steps in S

are effectively suppressed by the controller. The control signal is far from reaching its upper bound of

10

2,4 ⋅ [ g COD/d ]. When steps are added to Q the peaks in

S

are roughly of the

same size, but the set point tracking is slower, see Figure 4.7.

0 2 4 6 8 10 12 14 0.6

0.8 1 1.2

SNO,2 [g/m3]

0 2 4 6 8 10 12 14

0 0.5 1 1.5 2

x 10⁶

External carbon [g/day]

0 2 4 6 8 10 12 14

0 1 2 3

x 10⁴

Qin [m3/day]

Time [days]

)LJXUH

Finally steps are added to S

and as seen in Figure 4.8 the controller output saturates during the positive step time, and the nitrate concentration deviates somewhat from the reference value. The maximum supply of external carbon is quite simply not enough to keep the nitrate set point when the nitrate concentration is growing this high. It is important to keep in mind that it may not be realistic to have such a high constant concentration of influent ammonia during a three-day-period.

0 2 4 6 8 10 12 14

0.6 0.8 1 1.2

SNO,2 [g/m3]

0 2 4 6 8 10 12 14

0 0.5 1 1.5 2

x 10⁶

External carbon [g/day]

0 2 4 6 8 10 12 14

20 30 40 50

SNH,in[g/m3]

Time [days]

)LJXUH

in

SNH, . The file constinfluent is simulated without noise and delay on the sensors.

 6LPXODWLRQVXVLQJZHDWKHUGDWD

Now when suitable controller parameters, a suitable set point and a suitable constraint on the control signal have been found it is time to evaluate the performance of the controller using weather data. All simulations with the weather files are run with measurement noise and delays on the sensors, but except from these changes everything is identical to the former simulations. Since additional controllers will be added later the overall plant performance is not of any greater interest right now, but it might be interesting to look back later and compare the results. For a complete review of the plant performance see appendix B.

When the influent file dryinfluent was simulated the controller manages well to keep the nitrate concentration around the reference value as seen in the upper plot in Figure 4.9.

The lower plot shows the control signal and it just touches its upper bound at a few occasions. The threshold value for effluent ammonia was exceeded a number of times, but none of the other threshold values were violated.

0 2 4 6 8 10 12 14

0 0.5 1 1.5 2

SNO,2 [g/m3]

0 2 4 6 8 10 12 14

0 0.5 1 1.5 2 2.5

x 10⁶

External carbon [g/day]

Time [days]

)LJXUH

The weather file dryinfluent is simulated with noise and delay on the sensors.

The heavy rain period of the raininfluent file (around day eight to ten) can clearly be

noticed in Figure 4.10; the control signal decreases significantly and the effluent

concentrations of ammonia are lower during this period. The effluent constraint for

ammonia is not met.

0 2 4 6 8 10 12 14 0

0.5 1 1.5 2

SNO,2 [g/m3]

0 2 4 6 8 10 12 14

0 0.5 1 1.5 2 2.5

x 10⁶

External carbon [g/day]

Time [days]

)LJXUH

The weather file raininfluent is simulated with noise and delay on the sensors.

The two flow peaks in the storminfluent file symbolising storm events can clearly be seen in the upper plot in Figure 4.11. The threshold value for effluent ammonia is violated several times and the constraint for total effluent suspended solids is violated at two occasions.

0 2 4 6 8 10 12 14

0 0.5 1 1.5 2

SNO,2 [g/m3]

0 2 4 6 8 10 12 14

0 0.5 1 1.5 2 2.5

x 10⁶

External carbon [g/day]

Time [days]

)LJXUH

The weather file storminfluent is simulated with noise and delay on the sensors.