$TXDWLFDQG(QYLURQPHQWDO(QJLQHHULQJ3URJUDPPH 8SSVDOD8QLYHUVLW\6FKRRORI(QJLQHHULQJ
837(&:
DATE OF ISSUEAuthor
6LQJPDQ-HQQ\
Title
(IILFLHQW&RQWURORI:DVWHZDWHU7UHDWPHQW3ODQWV
±D%HQFKPDUN6WXG\
Abstract
Wastewater treatment plants are subject to large disturbances in incoming loads. Numerous control strategies have been suggested, however a fair evaluation is difficult due to different assumptions and scenarios used. In order to improve the evaluation and enhance the use of innovative control strategies the COST 682 Working Group no 2 has developed a benchmark for evaluating control strategies for activated sludge plants. In this study we have used the Simulink/Matlab platform for studying the benchmark. Three controllers were suggested, a carbon flow rate controller, a supervisory DO controller and a recycle rate controller. All the controllers work well locally. The combined control strategies reduce the time when the effluent benchmark criteria are violated compared to the default control strategies. The variances of the effluent concentrations were, however, high.
Keywords:
Activated sludge process; automation; benchmark; control; feedforward control; nutrient removal Supervisor
$VVRFLDWH3URIHVVRU%HQJW&DUOVVRQ
'HSDUWPHQWRI6\VWHPVDQG&RQWURO,QIRUPDWLRQ7HFKQRORJ\8SSVDOD8QLYHUVLW\
Examiner
$VVRFLDWH3URIHVVRU%HQJW&DUOVVRQ
'HSDUWPHQWRI6\VWHPVDQG&RQWURO,QIRUPDWLRQ7HFKQRORJ\8SSVDOD8QLYHUVLW\
Project name Sponsors
Language (QJOLVK Security
ISSN 1401-5765 Pages
Supplementary bibliographical information
'(3$570(172)($57+6&,(1&(68336$/$81,9(56,7<
Address: Villa vägen 16, SE-752 36 Uppsala, Sweden
2%-(&7,9(
,1752'8&7,21
:$67(:$7(575($70(173/$17
3.1 MECHANICAL TREATMENT ...4
3.2 CHEMICAL TREATMENT...5
3.3 BIOLOGICAL TREATMENT ...5
3.4 THE ACTIVATED SLUDGE PROCESS ...5
7+(%(1&+0$5. 4.1 GENERAL INTRODUCTION ...7
4.2 PLANT LAYOUT...7
4.3 THE MODEL ...7
4.4 PERFORMANCE ASSESSMENT...8
&21752//(5()),&,(1&< ())/8(1748$/,7< ,1)/8(1748$/,7< &267)$&7256)2523(5$7,21 4.5 RUNNING THE BENCHMARK ...11
&21752/675$7(*,(6 &21752/2)7+(1,75$7(/(9(/,17+($12;,&=21(686,1* $)(('%$&.)((')25:$5'(;7(51$/&$5%21&21752//(5 '(5,9$7,212)7+(&21752//(5 6,08/$7,216(783 781,1*2)&21752/3$5$0(7(56 &21675$,1762)7+((;7(51$/&$5%21)/2:5$7( &21752//(5+$1'/,1*2)12,6($1''(/$<6 &21752/3(5)250$1&(86,1*67(3',6785%$1&(6,1 ,1)/8(17/2$' 217+(&+2,&(2)1,75$7(6(732,17 6,08/$7,21686,1*:($7+(5'$7$ &21752/2)7+($0021,$/(9(/,17+($(52%,&=21(86,1* $683(59,625<'2&21752//(5 '(5,9$7,212)7+(&21752//(5 781,1*2)&21752/3$5$0(7(56 &21752/3(5)250$1&(86,1*67(3',6785%$1&(6,1 ,1)/8(17/2$' ',6&5(7(0$67(5'2&21752//(5 6,08/$7,21686,1*:($7+(5'$7$ &21752/2)())/8(171,75$7(%<,17(51$/5(&<&/,1* 8.1 ORIGINAL NITRATE CONTROLLER ...28
8.2 SIMPLE CONTROL OF INTERNAL RECYCLING...28
'(5,9$7,212)7+(&21752//(5 &21752//(5+$1'/,1*2)12,6($1''(/$< &21752/3(5)250$1&(86,1*67(3',6785%$1&(6,1 ,1)/8(17/2$' 6,08/$7,21686,1*:($7+(5'$7$ 6,08/$7,212)7+(&203/(7(&21752/675$7(*< 9.1 SIMULATION USING DRY WEATHER CONDITION...34
9.2 SIMULATION USING RAIN WEATHER CONDITION ...38
9.3 SIMULATION USING STORM WEATHER CONDITION ...43
&21&/86,216
$.12:/('*(0(176
5()(5(1&(6
$33(1',;$,1)/8(17'$7$ ERROR! BOOKMARK NOT DEFINED.
$33(1',;%)(('%$&.)((')25:$5'&$5%21&21752//(5
ERROR! BOOKMARK NOT DEFINED.
$33(1',;&683(59,625<'2&21752//(5ERROR! BOOKMARK NOT DEFINED.
$33(1',;',17(51$/5(&<&/(&21752//(5ERROR! BOOKMARK NOT DEFINED.
$33(1',;(7+(:+2/(3/$17 ERROR! BOOKMARK NOT DEFINED.
$33(1',;)25,*,1$/3/$176(783ERROR! BOOKMARK NOT DEFINED.
2%-(&7,9(
Wastewater treatment plants (WWTP) are subject to large disturbances in flow and loads together with uncertainties concerning the composition of the influent wastewater.
Increasingly stringent effluent standard and demand for cost effective treatment call for advanced control and supervision strategies. The objective with this master thesis project is to implement and evaluate a few control strategies for a wastewater treatment plant using a simulation study.
,1752'8&7,21
Increased awareness over the last decades of the effect of wastewater to the recipient water has led to higher and higher standards on effluent wastewater. The original reason for treatment of water was concern for human health, cities that did treat their water was much less sensitive to epidemics. Later came the realization that lakes and rivers that not long ago had been great for recreation smelled badly, the fish died and so forth. At first longer pipes were built, but that only moved the pressure on recipient somewhere else. Today the concerns are not only about human health but also about the welfare of the environment.
As the knowledge on the effects of wastewater to the recipient increases, the demand on the effluent quality from wastewater treatment plants is set higher. To meet the higher requirements on effluent loads the need for advanced control and supervision strategies increases. A good operator probably will be able to auto-tune in accordance to the daily and weekly flow pattern and keep the effluents within given limits as long as nothing exceptional happens. But when margins get smaller and more complex process solutions are used the risk of violating a limit increases.
Wastewater treatment plants are subject to large disturbances in influent flows as well as in influent concentrations. There is also uncertainties concerning the composition of the incoming wastewater. These variations are depending on weather as well as on the time of the day or day of the week. Still the plants need to be operated continuously.
The concentration of pollution in the influent as well as the effluent water is very low, which put high demands on the sensors and also calls for knowledge about what is sensor failure and what is actual fluctuation.
Today many plants have large margins in volumes, that is many plants are built to support larger influents then they receive. Hence they can manage quite well the
disturbances in influent load. But considering the cost to extend a plant there is financial incentives to run the plant efficiently and thereby manage to treat greater amount of water without the need of bigger foundation. The effluent quality is often measured on basis of monthly averages, so a possible failure on one day may well be leveled out during a month. Maybe the requirements of tomorrow will consider spot-checks, then the plant needs to be able to manage to treat the water even in extreme weather.
Good control of a wastewater treatment plant means that only the needed amount of chemicals is used, only the needed amount of oxygen is transferred to the water and no unnecessary water is recycled. In short anything that cost money is used sparingly. The cost to extend a plant is huge, with good control a plant may well manage to treat greater volume of wastewater without the need of bigger foundation.
:$67(:$7(575($70(173/$17
A municipal wastewater treatment plant typically aims to remove suspended substances, organic material and phosphate from the water before releasing it to the recipient. Many plants also have requirements to remove a significant part of the nitrogen from the incoming water. Depending on the location for the plant and the regional tradition, as well as on the sensibility of the recipient and the composition of the wastewater, the exact layout of the plant will differ. The treatment is normally done in a three-step process illustrated in Figure 3.1. In each step there is an accumulation of sludge, which is removed and further treated in a sludge treatment procedure not considered here.
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 Schematic picture of a Wastewater treatment plant. The biological treatment can be situated either before or after the chemical treatment.
The first step is a physical treatment where basically bigger particles are removed by grids, filters and sedimentation units. In the biological treatment suspended organic material is removed with help from microorganisms. Finally in the chemical step, chemicals are added to trap above all phosphorus, which is then separated through sedimentation.
0(&+$1,&$/75($70(17
There are two basic aims with the mechanical treatment. The first is to separate solids that may cause unnecessary wear or jam pumps and such, thereby cause stoppage of the plant. The second is to reduce the load on the following treatment steps by removing some of the contamination.
When the wastewater reaches the treatment plant it consists of many particles of different size. The first step of the physical treatment is a grid where bigger particles, like paper, wood and textiles, are separated. This is followed by sand filters that take care of smaller particles that still are too big to go into the plant. The last part of the physical treatment is presedimentation where smaller but heavier particles are removed.
Although the main aim of the physical step is to remove suspended solids from the wastewater some of the organic material is removed as well since it is aggregated to the
&+(0,&$/75($70(17
The main aim of the chemical treatment is to flocculate and separate phosphorus from the wastewater. The chemical treatment can be performed either before the biological treatment (preprecipitation), in the biological treatment (simultaneous precipitation) or after the biological treatment (postprecipiation). In the process precipitation chemicals are added to the wastewater to precipitate the phosphorus, which then flocculates and can be separated through sedimentation or floatation. Some suspended substances are also removed in the process since substances stick to the formed flocks.
Not all of the removed phosphorus is removed by the chemical treatment. Some of it is removed together with the organic material separated in the physical process and some is built into the microorganisms in the biological treatment. In municipal wastewater treatment plants that lack the chemical step about a third of the influent phosphorus is removed, see Svenska Kommunförbundet (1996).
%,2/2*,&$/75($70(17
The organic material left in the wastewater after the physical step exists mainly in soluble form. To be able to separate the organic material it needs to be transformed into separable particles. In the biological treatment step microorganisms are used to
concentrate and aggregate the organic material through incorporating it into the biomass. Some microorganisms will under certain circumstances transform ammonia and nitrate into nitrogen gas.
The biological treatment is a natural process and the participating microorganisms exist naturally in both wastewater and recipient. In a wastewater treatment plant the high concentration of microorganisms in a confined volume together with abundance of nourishment creates a favorable environment for the processes. The microorganisms need nourishment and the availability to extract energy to propagate and grow. The sludge contains the needed nourishment in the form of organic and inorganic particles as well as living biomass.
The biological treatment can be designed in different ways. In the following only the activated sludge process will be discussed.
7+($&7,9$7('6/8'*(352&(66
The activated sludge process is a biological process in which microorganisms oxidize and mineralize organic material. In the basic activated sludge process, shown in Figure 3.2, microorganisms use oxygen to oxidize organic matter.
Excess sludge
Effluent water water
Influent
Recirculated sludge
Aeration basin Clarifier
)LJXUHA basic activated sludge process with an aerated tank and a settler.
The microorganisms are kept suspended either by blowing air into the tank or by use of agitators. All microorganisms enter the system with the influent wastewater. In the activated sludge process the aim is to keep an ‘active sludge’, i.e. to keep a high concentration of biomass in the reactor. To keep a steady high concentration of sludge in the reactor sludge from the settler is recycled. The growth of the microorganisms and the influent particulate inert material is removed from the process as excess sludge.
After the bioreactor there is a secondary settler. Secondary clarification separates the biomass from the treated wastewater and is a key mechanism in determing the quality of the effluent.
The activated sludge process can be used for biological nitrate removal. In biological nitrate removal microorganisms transform nitrogen and ammonia into nitrogen gas, which is then released to the atmosphere. Nitrogen removal involves two sequential steps. The aerobic growth of autotrophs consumes soluble carbon, ammonia and dissolved oxygen to produce extra biomass and nitrate in solution. This step can be further divided into two, one producing nitrites and the second further oxidizing nitrites to nitrates. The second major step is the anoxic growth of heterotrophs, which use nitrates as oxidizer and produce extra biomass and nitrogen gas. See Newell and Ohlsson () and Svenska Kommunförbundet (1996).
There are two general layout possibilities for the nitrogen removal. In the case of predenitrification the anoxic denitrification processes take place in the beginning of the reactor, see Figure 3.3. An aerated zone, where the nitrification processes take place, follows the anoxic zone. Since the first part of the reactor performs the second step of the process the water needs to be recycled, see Svenska Kommunförbundet (1996).
When the plant has postdenitrification, that is the aerated zone comes before the anoxic one, carbon needs to be added because the denitrification process needs easily
metabolized carbon, see Carlsson and Lindberg (1996).
Effluent water Influent water
Recirculated water
Recirculated sludge
Excess sludge Anoxic Anoxic Aerobic Aerobic Aerobic
)LJXUHA simplified process scheme of an activated sludge process using predenitrification. The first two compartments are anoxic; this is where nitrate is transformed into nitrogen gas. In the following three compartments oxygen is added and ammonia is transformed into nitrate. Since the denitrification process takes place before the nitrification process water needs to be recycled.
The sludge age measures how long time in average the sludge spends in the system. The microorganisms that account for nitrification grow slowly therefore the sludge age needs to be quite long. A high concentration of biomass gives a high sludge age and is favorable, but if the concentration of biomass gets to high there is a risk of sludge
escape due to settler overload. A good rule of thumb is to keep the sludge age at about ten days.
7+(%(1&+0$5.
*(1(5$/,1752'8&7,21
The evaluation and comparison of different control strategies either practical or by simulation are difficult. This is due to the variability of the influent as well as on the complexity of the processes in the plant and on the differences in time constants. Time constants in a wastewater treatment plant may vary from a few minutes to several days or weeks. There is also a lack of standard evaluation criteria. Different regions have different effluent requirements as well as different cost levels.
Most often a situation is suboptimal, therefore it is difficult to judge the particular influence of an applied control strategy. As the systems have such high level of
complexity the effort needed to develop alternative control strategies is big. Because of this different approaches are seldom compared in a fair way. Even when they are it is hard to conclude to what extent the reached solution is process or location specific.
A benchmark, for evaluation of control strategies by simulation, has been developed by the COST 682 Working Group No.2. The evaluation is based on a rigorous
methodology including a simulation model, plant layout, controllers, performance criteria and test procedures, see http://www.ensic.u-nancy.fr/COSTWWTP for the full plant setup. Pons et al (1999) and Alex et al (1999) provides further information of the benchmark.
The notation used in this report will if nothing else is said follow the notation used in the benchmark description. Index number on a variable is to define which of the five zones that is described.
3/$17/$<287
The benchmark defines a plant layout, influent loads, test procedures and evaluation criteria. The benchmark tries to combine plainness with realism.
The benchmark has a fairly simple plant layout; it combines nitrification with predenitrification. The plant has a five-compartment reactor where the first two
compartments are anoxic and the following three are aerobic. The reactor is followed by a secondary settler.
The influent files, one with constant influent loads called FRQVWLQIOXHQW, and three
different weather files with varying influent loads are described in detail in Appendix A.
The file GU\LQIOXHQW contains two weeks of dry weather. The file UDLQLQIOXHQW contains one week of dry weather and a long rain event during the second week. The file
VWRUPLQIOXHQW contains one week of dry weather and two storms during the second week.
Each weather file takes into account the weekend effect on influent flow and composition.
7+(02'(/
The platform used for simulation is an implementation of the benchmark in
Simulink/Matlab. The model was implemented in Simulink by Dr Ulf Jeppsson, LTH, Lund, see Jeppson (1996).
Simulink is a general all-purpose software for mathematical modelling and simulation.
The user has immediate access to a large number of predefined building blocks,
numerical solvers, etc. The combination of the graphic interface in Simulink and the full access to Matlab makes the model very easy to handle and also very illustrative
The full benchmark model includes approximately 150 non-linear differential equations, the complete model can be found on a website (http://www.ensic.u-
nancy.fr/COSTWWTP). In order to enhance the computional performance the actual bioreactor and settler model have been implemented in C and integrated into Simulink.
The biological process is modeled by the IAWQ Activated Sludge Model no 1 (Henze et al 1987), abbreviated ASM1. It takes into account the aerobic and anoxic growth of heterotrophs, the aerobic growth of autotrophs as well as the decay of both types of bacteria. The model also considers the ammonification of soluble organic nitrogen and the hydrolysis of entrapped organics and organic nitrogen.
The behavior of the secondary settler is modeled by a double-exponential settling velocity model, seethe web site cited above.
3(5)250$1&($66(660(17
The performance assessment is made at two levels where the first level concerns the local control loops and the second provides measure for the effect of the proposed control strategy on plant performance. When looking at the performance the three different influent files (dry, rain and storm weather) are used. Each of the weather files describes a two-week period, but the tests are conducted on the last seven days.
&RQWUROOHU(IILFLHQF\
The local control loops are assessed by four criteria: IAE (Integral of the Absolute Error), ISE (Integral of the Square Error), the maximal deviation from setpoint and error variance. The error, e, is defined as the difference between the actual value and the reference value at any given time:
; UHI
; 6
6
H= , −
The IAE is defined as:
=
∫
=
= 14
7 W
W
GW H
,$(
The ISE is defined as:
=
∫
=
= 14
7 2 W
W
GW H
,6(
The maximal deviation from setpoint is defined as:
H
'HYmax =max
The error variance is defined as:
2
)
(
−
= 7
,$(
7 H ,6(
9DU
(IIOXHQW4XDOLW\
The constraints with respect to the effluent quality are defined as that the average
effluent concentration should not violate certain limits. The number of times these limits are violated should be reported as well as the percentage of time these constrains are not met. The constrains for the effluent are defined as follow:
• Total N (= SNO + SNKj) < 18 g/m3
• CODt < 100 g/m3
• SNH < 4 g/m3
• TSS < 30 g/m3
• BOD5 < 10 g/m3
Often fines have to be paid due to the discharge of pollution to the recipient. The Effluent Quality (E.Q.), [kg/d] is averaged over the last seven days for each weather file. It is based on weighting of the effluent loads of compounds that have a major influence on the quality of the receiving water. The compounds included in the E.Q. are usually included in regional legislation. It is defined as:
∫
== ⋅
⋅ +
⋅ +
⋅ +
⋅ +
⋅
= ⋅ 14
7
, 5 5
,
, ( )
) (
) ( )
(
) ( )
( 1000
. 1
. W
W H
H
%2'
H 12 12 H
1.M 1.M
H
&2' H
66
GW W 4 W
%2'
%
W 6
% W 6
%
W
&2'
% W 66
% 4 7
(
where
) (
)
( , , , , , ,
, ,
,
,H 1+H 12H 12H ;% %+H %$H ;3 3H LH
1NM 6 6 ; L ; ; L ; ;
6 = + + + + + +
) (
75 .
0 6,H ,,H %,+,H %,$,H 3,H
H ; ; ; ; ;
66 = ⋅ + + + +
)) (
) 1 ( (
25 .
0 , , , , , ,
5 66H ;6H IS ;%+H ;%$H
%2' = ⋅ + + − ⋅ +
H 3 H
$
% H +
% H , H 6 H , H
6 6 ; ; ; ; ;
6
&2'5 = , + , + , + , + , , + , , + .
The Bi are weighting factors to convert the different types of pollution into pollution units according to their influence on the reciepient.
• BSS = 2
• BCOD = 1
• BNKj = 20
• BNO = 20
• BBOD5 = 2
,QIOXHQW4XDOLW\
As have been mentioned before different plants will have different influent loads. To be able to compare wastewater with different composition there is a need for a defined influent quality. The Influent Quality (I.Q.) is defined as:
∫
== ⋅
⋅ +
⋅ +
⋅
+
⋅ +
⋅
= ⋅ 14
7 0
0 , 5 5
0 , 0
,
0 0
) ( )
(
) ( )
(
) ( )
( 1000
. 1
. W
W
%2'
12 12 1.M
1.M
&2' 66
GW W 4 W
%2'
%
W 6
% W 6
%
W
&2'
% W 66
% 4 7
,
) (
)
( , ,0 , ,0 ,0 ,0
0 , 0
, 0 , 0
, 1+ 12 12 ;% %+ %$ ;3 3 L
1NM 6 6 ; L ; ; L ; ;
6 = + + + + + +
) (
75 .
0 ,0 ,0 , ,0 , ,0 ,0
0 ;6 ;, ;%+ ;%$ ;3
66 = ⋅ + + + +
)) (
) 1 ( (
25 .
0 ,0 ,0 , ,0 , ,0
0 ,
5 66 ;6 I ;%+ ;%$
%2' = ⋅ + + − ⋅ +
0 . 0 , , 0 , , 0 , 0 , 0 , 0 , 0 ,
5 66 6, ;6 ;, ;%+ ;%$ ;3
&2' = + + + + + +
&RVW)DFWRUVIRU2SHUDWLRQ
This take into account the sludge production, the aeration and pumping energy and is calculated over the last seven days of each weather file. It also deals with controllers output variation since this gives an indication on peak loads and the wear of the pumps and aeration devices due to usage.
The amount of solids in the system at time t: TSS(t) = TSSa(t) + TSSs(t), where TSSa(t) is the amount of solids in the reactor and TSSs(t) is the amount of solids in the settler.
L 6 L 3 L
$
% L +
% L , Q
L
L 6L
D W ; ; ; ; ; 9
766 , , , , , , ,
1
, )
( 75 . 0 )
( = ⋅
∑
= + + + + ⋅=
M 6 M 3 M
$
% M +
% M , P
M
M 6 M
V W ; ; ; ; ; 9
766 , , , , , , ,
1
, )
( 75 . 0 )
( = ⋅
∑
= + + + + ⋅=
⋅
+ +
+
⋅ + + +
= =
∫
= 14
7 , , ,
, , ,
, ( )
75 . 0 ) ( )
1 ( W
W %$ 3
+
% , 6 V
D
VOXGJH 4 W GW
;
;
;
; W ;
766 W 7 766
3
The sludge production to be disposed, Psludge, is calculated from the total solid flow from wastage and the solids accumulated in the system during the period considered, i.e.
seven days for each weather file.
The total sludge production, Ptotal sludge, takes into account both the sludge production to be disposed and the sludge lost at the weir:
⋅
+ +
+ +
+ +
+
= =
∫
= 14
7 , , ,
, , , _ ,
) ( 75
. 0
) ( )
(
1 W
W
H H 3 H
$
%
H +
% H , H 6
V D
VOXGJH VOXGJH
WRWDO 4 W GW
;
;
;
;
; W 766 W
766 3 7
3
The pumping energy (PE) measures the energy use [kWh/d] for both internal and external recycle pumps. The pumping energy is calculated as:
=
∫
=
+ +
= 14
7
)) ( ) ( ) ( 04 (
.
0 W
W
Z
U W 4 W GW
4 W 7 4
3(
The aeration energy (AE) is calculated from the KLa in the three aerated tanks using the following relation, which is valid for Degrémont DP230 porous disks at an immersion depth of four meters:
{ }
=
∫ ∑
=
=
=
+
=
14
7 5
3
2 7.8408( )
) ( 4032 . 24W 0
W L
L N/D L N/D L GW
$( 7
5811,1*7+(%(1&+0$5.
The simulations are run in two sequential steps. First a quasi steady state are reached by simulating the FRQVWLQIOXHQWfile over a 150 day period without noise and delays on the sensors. For these simulations a stiff ODE 15 solver is used. The quasi steady state values are saved and the dry weather file is simulated as beginning at steady state. The other two weather files (UDLQLQIOXHQW and VWRUPLQIOXHQW) are simulated as beginning after a dry weather period. The weather files are simulated using an ODE 45 solver and noise and delays are added to the sensors.
&21752/675$7(*,(6
Three control strategies are suggested and evaluated. The results are first presented locally, i.e. with respect to the control signal and the output signal. The results are presented sequentially, that is, at each step a new controller is added to the plant. Finally all controllers are implemented and the objective is to get a stable well behaving plant performance.
To get a more realistic situation, the incoming measurements that are needed in the different controllers are simulated with noise and time delay. For flow and oxygen concentration though the measurements are assumed ideal, but with a delay of one
sample. The simulated sensors for nitrate, ammonia and readily biodegradable carbon have a delay of two thirds of a sample, or ten minutes, and there is white noise added.
The noise is Gaussian distributed with zero mean and variance 0.01.
In the case of delays it is important to get good values to use for the initial values. In the cases where a reference value exists that value has been used. Elsewhere the quasi steady-state value has been used.
&21752/2)7+(1,75$7(/(9(/,17+($12;,&
=21(686,1*$)(('%$&.)((')25:$5'(;7(51$/
&$5%21&21752//(5
Biological nitrogen removal in the activated sludge process is dependent on sufficient supplies of easily metabolized carbon compounds for the denitrifying bacterial
population. An external carbon source can increase denitrification rates and compensate for deficiencies in the influent carbon/nitrogen ratio. The flow rate of the external carbon needs to be controlled. If the dosage is too low the denitrification process will be constrained and the full capacity for nitrogen removal is not used. If the carbon dosage is too high it may cause carbon spill and increase the sludge production. A too high carbon addition is unnecessary expensive. A natural control strategy is to control the carbon flow rate so that the nitrate level in the last anoxic zone is kept at a desired level.
For a further discussion see Lindberg and Carlsson (1996).
'HULYDWLRQRIWKH&RQWUROOHU
The feedforward controller is based on a simple static mass-balance. The main idea is to derive a simple controller that will react on changes in the incoming water, such as changes in the influent flow rate and influent load of nitrate and carbon. See also Carlsson and Samuelsson (1999) and Samuelsson and Carlsson (1999).
In the derivation we assume a completely mixed anoxic reactor with inflow = outflow = Q [m3/day] and volume V [m3]. The dilution rate is defined as D = Q/V [day-1] and the flow rate of external carbon is denoted u [g COD/day]. The flow rate of external carbon is expressed as a mass flow, i.e. as the carbon concentration times the flow rate. If the hydrolysis and decay processes in the ASM1 model are ignored the nitrate dynamics simplifies to:
) (
)
( , , , ,
,
+
% LQ +
% +
% +
% ; ' ; ;
GW
G; =µ θ ⋅ + ⋅ −
9X 6 6 '
< ; GW G6
6 LQ 6 +
% +
6 1
) (
) 1 (
,
, + ⋅ − +
⋅
−
= µ θ
) (
)
( %,+ 12,LQ 12
+
12 ; ' 6 6
<
GW
G6 =− β µ θ ⋅ + ⋅ −
with
67 . 0 86 ,
. 2
1− =
= <+ <+
β
Considering steady state, (6.2) and (6.3) become
) (
) 1 ( 1
,
,+ 6LQ 6
% +
6 6 '
< ;
9 X= µ θ ⋅ − ⋅ −
) )(
( ,
, + 12LQ 12
+
% ' < 6 6
; −
⋅
= ⋅
θ µ
β
Inserting (6.5) in (6.6) and replacing SNO with a reference value SNO,ref gives:
− − −
⋅
= 1( ) ( )
, ,
,LQ 12UHI 6LQ 6
12 6 6 6
6 4
X β
Equation (6.7) shows that the control signal will react instantly on changes in the influent load of substrate and nitrate. This control strategy is now used to govern the nitrate level in zone 2 (SNO,2). Since a simplified model was used for the design there is no guarantee the stationary level of nitrate will equal the setpoint SNO,ref. To get the right static gain, the feedforward controller is completed with a feedback PI-controller. The complete controller with the benchmark notation is given by:
( )
∫
= −+
−
⋅ +
− − −
+ +
=
W W
W 12 12UHI
, UHI 12 12
3
6 6
UHI 12 12
V U
G 6
6 . 6
W 6 .
W 6 W 6 W 6 W 6 W
4 W 4 W 4 W X
0
) )
( ( )
) ( (
)) ( ) ( ( )) ( )
( 1 (
) ( ) ( ) ( ) (
, 2
, ,
2 ,
2 , 0
, ,
0 ,
τ τ
β
Where u(t) is a flow of mass and is expressed in g COD/day. The index zero denotes inflow to zone 1 and other index number denotes in what zone the concentration is valid. When the external carbon is added to the flow, the concentration of SS going into the first anoxic basin will be:
( )
(
U V)
6 V QHZ U
6 4 4 4
X 6 4 4 6 4
+ +
+
⋅ +
= + ,0
,
where the increase in flow originating from the flow of external carbon is considered to be negligible.
6LPXODWLRQ6HWXS
The recirculation rate is held constant at 5*Qin0. To avoid too high effluent of suspended solids but to still get a sludge age that is high enough Qw = 300 m3/day. That is about 20 – 25% lower than the original Qw = 385 m3/day.
In all the simulations run with the carbon controller the original controller for dissolved oxygen is used. This means that KLa in reactor 3 and 4 is held constant at 240 day-1 and the DO controller in reactor 5 has its setpoint to 2 mg O2/l. The proportional gain KP in equation 6.8 is taken as KP = 0.
7XQLQJRI&RQWURO3DUDPHWHUV
As mentioned above the P part of the PI controller is turned off by letting KP = 0.
KI has to be chosen so that the feedforward and feedback part of equation (6.8) is of the same magnitude. The concept was to design a fairly slow PI controller to reduce the risk of instability. Most of the disturbances are being attenuated by the feedforward; the main reason with the integration part is to get rid of biases concerning the setpoint.
After a lot of trial and error KI ended up at 7*106. Two different values of KI can be seen in Figure 6.1.
0 5 10 15
0 0.5 1 1.5 2 2.5
SNO reactor 2 [g/m3]
KI = 7e6
0 5 10 15
0 0.5 1 1.5 2 2.5x 106
External carbon [g/d]
time [days]
0 5 10 15
0 0.5 1 1.5 2 2.5
KI = 3.5e6
0 5 10 15
0 0.5 1 1.5 2 2.5x 106
time [days]
)LJXUH Illustration of different KI values, all sensors are ideal. Upper left:
Setpoint and nitrate level in zone 2 [g/m3] for KI = 7*106. Lower left: Dosage of external carbon [g COD/d]. Upper right: Setpoint and nitrate level in zone 2 [g/m3] for KI = 3.5*106. Lower right: Dosage of external carbon [g COD/d].
&RQVWUDLQWVRIWKH([WHUQDO&DUERQ)ORZ5DWH
A typical choice of external carbon is ethanol, which has a carbon content of: CODethanol
= 1,200,000 [g/m3]. Then how to choose a limit to the carbon addition? One possibility is of course to have no limit to the carbon dosage. Running different simulations give that in the worst possible case, there is at the most an addition of external carbon equal
to about 3 m3 of ethanol per day. Considering the cost of adding external carbon this may be considered as too much, but as seen from Figure 6.2, a too low limit on the carbon flow deteriorates the control performance significantly.
0 5 10 15
0 0.5 1 1.5 2
SNO rector 2 [g/m3]
umax=inf
0 5 10 15
0 0.5 1 1.5 2 2.5 3 3.5x 106
External carbon [g/day]
Time [days]
0 5 10 15
0 0.5 1 1.5 2
umax=2.4*10e6
0 5 10 15
0 0.5 1 1.5 2 2.5x 106
Time [days]
0 5 10 15
0 0.5 1 1.5 2 2.5 3
umax=1.2*10e6
0 5 10 15
0 2 4 6 8 10 12x 105
Time [days]
)LJXUH Illustration of different saturation values for u. The influent file GU\LQIOXHQW is simulated. Top left: SNO in zone 2 [g/m3] when there is no limit for the carbon dosage. Lower left: Added mass flow of external carbon [g COD/d], no limits to the flow. Upper middle: SNO in zone 2 [g/m3] when the limit for the carbon dosage is 2.4 * 106 g COD/day. Lower middle: Added mass flow of external carbon [g COD/d], umax = 2.4 * 106 g COD/day. Top right: SNO in zone 2 [g/m3] the limit for the carbon dosage is 1.2 * 106 g COD/day. Lower right:
Added mass flow of external carbon [g COD/d], umax = 1.2 * 106 g COD/day.
The limit on the carbon flow was decided to umax = 2,400,000 g COD/day, this
corresponds to 2 m3/day of ethanol. This limit is illustrated in the middle plots of Figure 6.2. The benchmark plant treats around 20,000 m3 wastewater/day, then to have a carbon flow of, at the most, 2 m3/day may give a quite high cost but is chosen here.
&RQWUROOHU+DQGOLQJRI1RLVHDQG'HOD\V
Naturally, the controller act better when there is no sensor delay or noise. Still without further tuning of the parameters the controller works more than acceptable when measurement noise and delay are added to the sensors, as can be seen in Figure 6.3.
0 2 4 6 8 10 12 14 0
0.5 1 1.5 2 2.5
SNO reactor 2 [g/m3]
0 2 4 6 8 10 12 14
0 0.5 1 1.5 2 2.5x 106
External carbon [g/d]
time [days]
)LJXUHStep response to changes in setpoint for the non-ideal case when delays as well as noise is added to the sensors. Upper plot: The controlled nitrate level in zone 2 together with setpoint [g/m3]. Lower plot: The amount of added external carbon [g COD/d].
As can be seen in Figure 6.3 the controller is fast when the reference value of SNO,2 is changed and the overshoot is fairly small. The controller is slightly slower when the setpoint is increased than when it is decreased.
&RQWURO3HUIRUPDQFH8VLQJ6WHS'LVWXUEDQFHVLQ,QIOXHQW/RDG A good controller should level out disturbances from the influent load. To test the control performance concerning this, step changes were added to the influent load on parameters directly affecting the controller.
As can be seen in equation (6.8), the controller reacts instantly on changes in influent SS. In order to illustrate how the controller can attenuate variations in influent SS, the influent file FRQVWLQIOXHQW, see Appendix A.1, was used but step changes were added to the influent SS concentration. The control performance is shown in Figure 6.4. (ideal sensor). The control performance is very good. Also when a sensor with delay was used, the control performance was good (this is not illustrated).
0 2 4 6 8 10 12 14 20
40 60 80 100
Ssin [g/m3]
0 2 4 6 8 10 12 14
0 0.5 1 1.5
Sno, reactor 2 [g/m3]
0 2 4 6 8 10 12 14
0 0.5 1 1.5
2x 106
time [days]
External Carbon [g/d]
)LJXUHControl performance when influent SS makes step changes. An ideal sensor is used for measuring SNO in zone 2. Upper plot: Influent SS [g/m3].
Middle plot: The controlled nitrate level in zone 2 [g/m3]. The setpoint is equal to 1 g/m3. Lower plot: Amount of added external carbon [g COD/d].
Next the effect of the variations in SNH,in concentration is illustrated. This was achieved by using the influent file FRQVWLQIOXHQW but step changes were added to SNH,in. The results are shown in Figure 6.5. The control performance is excellent both with and without sensor delay although the later is not illustrated.
0 2 4 6 8 10 12 14
20 30 40 50
Snhin [g/m3]
0 2 4 6 8 10 12 14
0 1 2 3
Sno, reactor 2 [g/m3]
0 2 4 6 8 10 12 14
0 1 2 3x 106
External Carbon [g/d]
time [days]
)LJXUH Control performance when influent SNH makes step changes using ideal sensor for measuring SNO in zone 2. Upper plot: Influent SNH [g/m3].
Middle plot: The controlled nitrate level in zone 2 [g/m3]. The setpoint is equal to 2 g/m3. Lower plot: Amount of added external carbon [g COD/d].
The last disturbance to be illustrated is step changes of the influent flow rate. As before the step changes are added using the influent file FRQVWLQIOXHQW. The step response to a change in Qin is more sensitive to the use of delays (not illustrated) than the disturbances shown above. Still as Figure 6.6. shows, when the sensors are ideal, it only takes a few hours before the SNO,ref is reached.
0 2 4 6 8 10 12 14
0 1 2 3x 104
Qin [m3/d]
0 2 4 6 8 10 12 14
0.5 1 1.5 2 2.5
Sno, reactor 2 [g/m3]
0 2 4 6 8 10 12 14
0 0.5 1 1.5
2x 106
time [days]
External Carbon [g/d]
)LJXUH Control performance when influent flow rate makes step changes.
Ideal sensors are used for all measurements. Upper plot: Influent Q [m3/d].
Middle plot: The controlled nitrate level in zone 2 [g/m3]. The setpoint is equal to 2 g/m3. Lower plot: Amount of added external carbon [g COD/d].
The influent concentration of nitrate is zero in all influent files connected to the benchmark. Therefore it has not been tested how well the controller attenuates disturbances in the influent flow of nitrate.
These step responses give an indication on how effective the controller is. For all disturbances the controller can keep the nitrate level close to the setpoint. The response to changes in Qin is slower than in the other cases but not excessively slow. Using a lower rate of recycling would probably improve the performance for disturbances in Qin, but considering that a controller for internal recycling will be added later there is little need for that.
2QWKH&KRLFHRI1LWUDWH6HWSRLQW
According to Yuan et al (1996) a good choice for the setpoint of SNO in the anoxic part
3
g/m3 is almost as good for the outcome and the need for external carbon is lower, see Figure 6.7. The final choice of the setpoint to 1 g/m3 comes from that the fluctuations in SNO in zone 2, when Q is high, are lower for a low setpoint. Further, the control strategy for internal recycling, which will be added later, assumes a complete denitrification.
0 5 10 15
0 1 2 3
SNO reactor 2 [g/m3]
0 5 10 15
0 1 2 3x 106
External carbon [g/d]
0 5 10 15
4 5 6 7 8
time [days]
SNO effluent [g/m3]
0 5 10 15
0 0.5 1 1.5 2
0 5 10 15
0 1 2 3x 106
0 5 10 15
3 4 5 6 7
time [days]
)LJXUH A comparison between different setpoints for SNO,2. The weather file GU\LQIOXHQW is used as influent. Upper left: SNO in zone 2 [g/m3], the setpoint equals 2 g/m3. Middle left: Amount of added external carbon [g COD/d]. Lower left: Effluent concentration of nitrate [g/m3] when the setpoint in zone 2 equals 2 g/m3. Upper right: Nitrate level in zone 2 [g/m3] together with the setpoint SNO,ref = 1 g/m3. Middle right: Corresponding amount of added external carbon [g COD/d]. Bottom right: Effluent concentration of nitrate [g/m3] when the setpoint in zone 2 is 1 g/m3.
6LPXODWLRQV8VLQJ:HDWKHU'DWD
It is of interest to also see how well the controller responds to the three different weather files. All three weather files are run using noise and delay on the sensors as described in section 5. Except for the influent files everything is kept the same as in the simulations above.
The first weather file to be illustrated is the influent file GU\LQIOXHQW. Figure 6.8 shows control performance for the carbon controller. The overall plant performance is not really interesting in the case of only one control strategy. The only benchmark criterion that is violated for the influent file GU\LQIOXHQW is the maximum effluent ammonia limit.
For a full review of the plant and control performance see Appendix B.1.
0 2 4 6 8 10 12 14 0
0.5 1 1.5 2
SNO reactor 2 [g/m3]
0 2 4 6 8 10 12 14
0 1 2 3x 106
External carbon [g/d]
0 2 4 6 8 10 12 14
3 4 5 6 7
SNO effluent [g/m3]
time [days]
)LJXUH Control performance for the influent file GU\LQIOXHQW. Top plot:
Setpoint and nitrate level in zone 2 [g/m3]. Middle plot: Added amount of external carbon [g COD/d]. Lower plot: Effluent nitrate [g/m3].
Next, the input file UDLQLQIOXHQW was simulated. The long period of rain makes it harder for the controller to stay close to the setpoint. Still nothing exceptional happens, the only limit that is violated is the maximum effluent ammonia. As expected the controller has to work harder and the variables to describe the control efficiency are slightly higher.
For a full review on plant and control performance see Appendix B.2.
0 2 4 6 8 10 12 14
0 0.5 1 1.5 2
SNO reactor 2 [g/m3]
0 2 4 6 8 10 12 14
0 1 2 3x 106
External carbon [g/d]
0 2 4 6 8 10 12 14
3 4 5 6 7
SNO effluent [g/m3]
time [days]
)LJXUH Controller performance using the influent file UDLQLQIOXHQW. Top plot:
Nitrate level and setpoint in zone 2 [g/m3]. Middle plot: External carbon [g COD/d]. Lower plot: Effluent concentration of nitrate [g/m3].
Finally, the input file VWRUPLQIOXHQW was used. The two storm events can clearly be noticed in the top plot of Figure 6.10. During more occasions than during the other weather files, the controller uses the maximum level of carbon. As in the other cases the maximum level of effluent ammonia is violated, but now the maximum level of total suspended solids is also violated.
For a full review on plant performance see Appendix B.3.
0 2 4 6 8 10 12 14
0 0.5 1 1.5 2
SNO reactor 2 [g/m3]
0 2 4 6 8 10 12 14
0 1 2 3x 106
External carbon [g/d]
0 2 4 6 8 10 12 14
3 4 5 6 7
time [days]
SNO effluent [g/m3]
)LJXUHControl performance using the influent file VWRUPLQIOXHQW. Upper plot: Setpoint and nitrate level in zone 2 [g/m3]. Middle plot: Added external carbon [g COD/d]. Lower plot: Effluent concentration of nitrate [g/m3].
&21752/2)7+($0021,$/(9(/,17+($(52%,&
=21(86,1*$683(59,625<'2&21752//(5
The nitrification process, that is biological transformation of ammonia to nitrate, is dependent on the availability of oxygen. To run the plant efficiently the amount of dissolved oxygen (DO) in the aerobic reactor needs to be controlled. Too high dissolved oxygen concentration will lead to unnecessary cost due to high aeration and may also effect the anoxic processes when water is recycled. With too low dissolved oxygen concentration in the reactor the nitrification process will not be at its optimum. A good trade-off can be achieved by adjusting the dissolved oxygen setpoint so that the effluent ammonia is kept at a low and constant value, see also Lindberg and Carlsson (1996).
