Benchmark 1 (BSM1)

This page described the first benchmark (BSM1) and has been prepared initially by the different Working Groups of COST Actions 632 and 624. A pdf revised version is under development.

Simulation model

• Influent (General characteristics)
  • Flow: 20,000 m³d^-1
  • Average biodegradable COD: 300 mg l^-1
  • Hydraulic retention time (based on total volume, i.e. aeration tank + settler = 12 000 m³): 14.4 hours
    • Reactor volume: 6 000 m³
    • Settler volume  : 6 000 m³
  • Wastage flowrate: 385 m³/day, which corresponds approximatively to a sludge age of 10 days
  • Dynamics:
    • Dry weather
    • Rain 1: Dry weather + 2 storm events
    • Rain 2: Dry weather + long rain period



• Process model

The Activated Sludge Model 1 is selected (Henze et al., 1986).



Processes

Conversion rates

Parameter values

List of variables


• Soluble inert organic matter S_I
• readily biodegradable substrate S_S
• Particulate inert organic matter X_I
• Slowly biodegradable substrate X_S
• Active heterotrophic biomass X_B,H
• Active autotrophic biomass X_B,A
• Particulate products arising from biomass decay X_P
• Oxygen S_O
• Nitrate and nitrite nitrogen S_NO
• NH₄⁺ + NH₃ nitrogen S_NH
• Soluble biodegradable organic nitrogen S_ND
• Particulate biodegradable organic nitrogen X_ND
• Alkalinity S_ALK
  

Variables

Conversion rates

Parameter values

List of processes
Process j:

• j = 1: Aerobic growth of heterotrophs

  
  

• j = 2: Anoxic growth of heterotrophs

  
  

• j = 3: Aerobic growth of autotrophs

  
  

• j = 4: Decay of heterotrophs

  
  

• j = 5: Decay of autotrophs

  
  

• j = 6: Ammonification of soluble organic nitrogen

  
  

• j = 7: Hydrolysis of entrapped organics

  
  

• j = 8: Hydrolysis of entrapped organic nitrogen

  
  

Variables

Processes

Parameter values

Observed conversion rates:

• S_I (i = 1)

  

• S_S (i = 2)

  

• X_I (i = 3)

  

• X_S (i = 4)

  

• X_B,H (i = 5)

  

• X_B,A (i = 6)

  

• X_P (i = 7)

  

• S_O (i = 8)

  

• S_NO (i = 9)

  

• S_NH (i = 10)

  

• S_ND (i = 11)

  

• X_ND (i = 12)

  

• S_ALK (i = 13)

  
  

Variables

Processes

Conversion rates

Biological parameter values
They corrspond approximatively to a temperature of 15°C.

• Stoechiometric parameters

  YA	g cell COD formed (g N oxidized)-1	0.24
  YH	g cell COD formed (g COD oxidized)-1	0.67
  fP	dimensionless	0.08
  iXB	g N (g COD)-1 in biomass	0.08
  iXP	g N (g COD)-1 in endogenous mass	0.06

• Kinetic parameters

  µH	day-1	4.0
  KS	g COD m-3	10.0
  KO,H	g O2 m-3	0.2
  KNO	g NO3-N m-3	0.5
  bH	day-1	0.3
  hg	dimensionless	0.8
  hh	dimensionless	0.8
  kh	g slowly biodegradable COD (g cell COD . day)-1	3.0
  KX	g slowly biodegradable COD (g cell COD)-1	0.1
  µA	day-1	0.5
  KNH	g NH3-N m-3	1.0
  bA	day-1	0.05
  KO,A	g O2 m-3	0.4
  ka	m3 . COD(g.day)-1	0.05



• Detailed plant layout
• Reactor
  Number of compartments: 5
  Non-aerated: compartments 1-2
  Aerated:
  • - compartments 3-4, fixed k_la = 10 h^-1
  • - compartment 5: DO controlled by manipulation of k_la at 2 mg.l^-1
    

  
  For each compartment:

  • Flowrate: Q_k
  • Concentration: Z_k
  • Volume:
    • Non-aerated compartments: V₁ = V₁ = 1000 m³
    • Aerated compartments: V₃ = V₄ = V₅ = 1333 m³
  • Reaction rate: r_k

  

  
  
  Mass balances (general formula)
  

  • For k = 1 (unit 1)

    
    

  • For k = 2 to 5

    
    

  • Special case of oxygen (S_O,k)

    
    where the concentration at saturation of oxygen is

  • Miscellaneous

    
    
    
    
    



• Secondary settler
  - Number of layers: 10
  - Feed layer: 6^th layer (from bottom)
  
  • Geometry:
    A: Area = 1500 m²
    z_m, m=1 to 10: Height of layer m (all equal)
    Total height : 4 m
    Settler volume = 6000 m³
    There is no reaction in the settler
    
  • Solid flux due to gravity sedimentation J_s
    

    

    where X is the total sludge concentration
    

  • Double-exponential settling velocity function (Takács et al., 1991)

  
  

  	Parameter	Units	Value
  Maximum settling velocity		m.d-1	250.0
  Maximum Vesilind settling velocity		m.d-1	474
  Hindered zone settling parameter		m3.(g SS)-1	0.000576
  Flocculant zone settling parameter		m3.(g SS)-1	0.00286
  Non-settleable fraction			0.00228

  • Mass balances for the sludge
  • For the feed layer (m = 6)

  

  • For the intermediate layers below the feed layer (m = 2 to m = 5)

  

  • For the bottom layer (m = 1)

  

  • For the intermediate clarification layers above the feed layer (m = 7 to m = 9)

  

  where

  • For the top layer (m = 10)

  

  with
  The threshold concentration X_t is equal to 3 g.l^-1
  

  • For the soluble components (including dissolved oxygen)
  • For the feed layer (m = 6)

  

  • For the layers m = 1 to 5

  

  • For the layers m = 6 to 10

  
  
  

  • Concentration in the recycle and wastage flow:

  

  • To calculate
    - the sludge concentration from the concentrations in the bioreactor unit 5:

    

    
    as fr_COD-SS = 4/3

    - the concentrations in the recycle and the wastage flows:

    

    
    Similar equations hold for X_P,u, X_I,u, X_B,H,u, X_B,A,u and X_ND,u.

    - Calculation of sludge age (steady state case)

    The sludge age calculation is based on the total biomass present in the system, i.e. the reactor and the settler:
    

    

    where TX_a is the total biomass present in the reactor:

    

    TX_s is the total biomass present in the settler:

    

    f_e is the loss rate of biomass in the effluent :

    

    with m = 10 and f_w is the loss rate of biomass in the wastage

    



• Influent data

  The influent data were initially proposed by Vanhooren and Nguyen ( Gent_OttawaReport.pdf). The time is given in days, the flowrate is given in m³/day and the concentrations are given in g/m³. The data are given in the following order:

  time S_S X_B,H X_S X_I S_NH S_I S_ND X_ND Q₀

  In any influent: S_O = 0; X_B,A = 0., S_NO = 0., X_P = 0. and S_ALK = 7 moles/m³

  • Dry weather

  Download of influent file Inf_dry.txt

  

  

  

  • Storm weather

  Download of influent file Inf_strm.txt

  This file contains one week of dry weather and two storms during the second week.

  

  

  

  • Rain weather

  Download of influent file Inf_rain.txt

  This file contains one week of dry weather and a long rain event during the second week.

  

  

  



• Initials
  

  Initial values can be selected by the user. A 100-days period of stabilization in closed loop with no noise on measurements has to be completed before using the dry weather file (14 days) followed by the weather file to be tested. Noise on measurements should be used with the dynamic files. The dynamic load averages to be used as inputs during the stabilization period are given in the following table.

  Variable	Value	Unit
  Q0,stab	18 446	m3/day
  SS,stab	69.50	g/m3
  XB,H,stab	28.17	g/m3
  XS,stab	202.32	g/m3
  XI,stab	51.20	g/m3
  SNH,stab	31.56	g/m3
  SI,stab	30	g/m3
  SND,stab	6.95	g/m3
  XND,stab	10.59	g/m3

  The system is stabilized if the steady state for these conditions is reached. A simulation period of 10 times the sludge age suffices for that. If for some control strategy the sludge age is influenced, the stabilization period must be adjusted accordingly.

Open-loop assessment

In order to users to verify their simulator open-loop results for the dry weather situation will be posted as soon as possible on the web site. The procedure to assess the open loop case is similar to the closed-loop one: run the plant for a stabilization period of 100 days before using the dry weather file.

For open-loop assessment the control variables have the following values:

The steady state values after 100 days will be found in the text file Steady.txt and the first day of the weather file in the text file First_day.txt (15mn sampled results).

The steady-state and first day values have been provided by Dr Ulf Jeppsson and obtained by implementing the benchmark on Matlab/Simulink.

A first comparison of the steady-state results obtained on three platforms (Matlab/Simulink, GPS-X and FORTRAN code) can be found in Pons et al. (1999).

Table 1

Biological reactor steady-state

For evaluation of the simulation results over a fixed period of time (T= t_f-t₀), average values are to be calculated as follows:
Note: All the integrals for performance assessment are calculated by rectangular integration with a time step of 15 min.

• Flow rate (m³/d):

  

• Concentration for compound Z_k (mass/m³) in flow Q must be flow proportional:

  

Set-up of controllers

• Controller variables
  
  • Measured variable: NO₃-concentration in compartment 2:
    Noise on measured variable: white, zero-mean, with a constant standard deviation of 0.1 mg/l, normally distributed. The minimal value measurable by the sensor is 0.1 mg/l.
    Delay on NO₃-concentration measurement: 10 mn
    Sampling period: 10 mn
    Manipulated variable: internal recycle flow from compartment 5 back to 1
    

  • DO control in compartment 5: Probe model = ideal probe (no delay, no response time, ...)
• Controller types
  Both suggested controllers are of the PI types. Their performance is assessed by (i = 1 for nitrate-PID and i = 2 for oxygen-PID):
  • IAE (Integral of Absolute Error)

    

    where e_i is the error:

    

  • ISE (Integral of Square Error)

    

  • Maximal deviation from setpoint:

    

  • Variance of error:

    

    with

    

    

  • Variance of manipulated variable (u_i) variations:

    

    with

    

    

    

• Control primary objective
  Maintain the NO₃-concentration in the 2^nd compartment at a predetermined set point value: 1 mg.l^-1

  Maintain the oxygen concentration in the 5^th compartment at a predetermined set point value: 2 mg.l^-1

  
  

• Other aspects
  Constraints on recirculation flows:
  - Range for Q_a: 0 to 5 Q_0,stab
  - The external recycle flowrate Q_r is maintained constant and is set to Q_r = Q_0,stab
  Constraints on oxygen transfer in compartment 5: k_la = 0 to 10 h^-1
  
  

Performance assessment

Constraints with respect to the eflluent quality are defined as follows. The flow average values of the effluent concentrations over the three test periods (dry, rain and storm weather: 7 days for each) should obey the following limits (total N = S_NO,e+S_NKj,e):

The percentage of time the constraints are not met must be reported, as well as the number of violations. The number of violations is defined as the number of crossings of the limit (from below to above the limit).

The performance assessment is made at two levels.

• The first level concerns the local control loops, assessed by IAE (Integral of the Absolute Error) and ISE (Integral of the Squared Error) criteria, by maximal deviation from setpoints, and by error variance. Basically, this serves as a proof that the proposed control strategy has been applied properly.
• The second level provides measures for the effect of the control strategy as such on plant performance and it can be divided into four sub-levels:
  • the effluent quality: levies or fines are to be paid due to the discharge of pollution in the receiving bodies:
  The Effluent Quality (E.Q.) (kg pollution unit/d) is averaged over the period of observation T (d) (i.e. the second week or 7 last days for each weather file) based on a weighting of the effluent loads of compounds that have a major influence on the quality of the receiving water and that are usually included in regional legislation. It is defined as:

  

  where
  

  

  

  

  

  and the B_i are weighting factors for the different types of pollution to convert them into pollution units. The concentrations are to be expressed in g/m³.
  

  Factor	BSS	BCOD	BNKj	BNO	BBOD5
  Value(g pollution unit/g)	2	1	20	20	2

  The B_i have been deduced from Vanrolleghem et al. (1996).
  

  • the costfactors for operation
    • the sludge production to be disposed (kg/d) (calculated from the total solid flow from wastage and the solids accumulated in the system over the period considered, i.e. 7 days for each weather file):
      The sludge production, P_sludge, is calculated from the total solid flow from wastage and the solids accumulated in the system over the period of time considered (t_f = 7 days for each weather file)

      Amount of solids in the system at time t : TSS (t)

      

      where TSS_a(t) is the amount of solids in the reactor :

      

      TSS_s(t) is the amount of solids in the settler :

      

      

    • the total sludge production (kg/d) takes into account the sludge to be disposed and the sludge lost at the weir:

      

    • the aeration energy (AE) (kWh/d) and the pumping energy (PE) (kWh/d) (internal and external recycle pumps).

      The pumping energy is calculated as:

      with the flowrates expressed in m³/d.

      The aeration energy AE should take into account the plant peculiarities (type of diffuser, bubble size, depth of submersion, etc ...) and is calculated from the k_la in the three aerated tanks according to the following relation, valid for Degrémont DP230 porous disks at an immersion depth of 4m:

      

      with k_La in h^-1 and where i is the compartment number.
    • controllers output variations: the maximum values and the variance of the manipulated variables variations should be given. This will provide an indication on peak loads and the wear of the pumps and aeration devices.
    
    
  • Finally the increase in capacity which could be obtained using the proposed control strategy should be evaluated. This factor is related to investment costs if the plant would simply be extended to deal with an increased load. This is expressed by the relative increase in the influent flow rate, a, which can be applied while maintaining the reference effluent quality index, E.Q._ref, for the three weather conditions (T=7 days for each). E.Q. _ref is calculated from the above equation in open loop. Q_0,i (t) =a.Q_0,i(t), i = 1 for dry weather, i = 2 for storm weather and i = 3 for rain weather. Operation variables such as Q_w, Q_r and k_La in compartments 3 and 4 remain unchanged.
  Furthermore an Influent Quality (I.Q.) can be defined as:

  

  with:

  

  

  

  

  Tests of performance assessment, in open and closed loops, are proposed by Dr Ulf Jeppsson for a MATLAB/SIMULINK implementation (not updated for new B's values):
  found in Pons et al. (1999).

  Table 2

  	Open loop	Closed loop
  Steady state	openloop_ss.pdf	closedloop_ss.pdf
  Performance	openloop_perf1.pdf	closedloop_perf1.pdf

  A first comparison of the global results obtained on two platforms (Matlab/Simulink and FORTRAN code) can be found in Alex et al. (1999).

  

  Sensors and control handles
  

  Introduction

  When the behaviour of the models has been validated for the open-loop and closed-loop test cases by comparing the results with the ones available on the web-site (See Table 1 and Table 2), you can now implement and test your own control strategy on the benchmark plant. To allow for a wide range of different strategies to be tested (within the confinement of the physical plant layout) a significant number of sensors and control handles are available. Their mathematical descriptions focus on simplicity rather than completely accurate reproductions of their true behaviour. The reason for this choice is fourfold:

  • increased computational burden due to sensors and control handles should be limited;
  • actual implementation of the available sensors and control actions should not be a hinder to inexperienced benchmark users;
  • descriptions of sensors and control handles should not reduce the number of possible simulation platforms to be used for the benchmark;
  • main focus of the benchmark is to promote new control strategies rather than details and problems related to individual sensors and control handles.

  It is believed that the characteristics of the sensors and control handles provided below represent a fair compromise between realistic behaviour and the overall purpose of the benchmark project. If you feel that some type of sensor class (Table 3) or control handle are lacking in order for you to implement your favourite control strategy then you are requested to contact the COST 624 benchmark group . Your suggestions can then be evaluated, added to the benchmark description and made available to all benchmark users.

  The principle for any good control strategy implies that the number of sensors and control actions should be minimised within the framework of the selected control strategy, due to the investment cost, maintenance cost, etc. Consequently, penalties in the evaluation criteria will be added in relation to the number and which types of sensors are being used, the use of external carbon source, etc. (although not yet formulated). High complexity of a control strategy is not a goal by itself but rather an ‘unwanted’ consequence of stringent control objectives.

  It is of the utmost importance that you describe in detail the type of sensors and control handles you have used to implement your specific control strategy and the characteristics of these sensors (which sensor class, detection limit and noise level) when you publish papers, reports, etc. Enough information should always be provided to allow other benchmark users to duplicate your results and allow for a fair comparison with other suggested control strategies for the benchmark system. A dedicated area within the COST Benchmark site will later be made available where the results from different benchmark exercises can be downloaded and viewed for the purpose of comparison.

  Sensors

  To enhance the simplicity and flexibility of the benchmark the sensors are defined by a number of sensor classes. The main problem with describing every type of sensor in an exact way is that it limits the flexibility for the future use of the benchmark and also lumps all particular sensors into one specific group, which may not always be applicable (not all sensors used to measure a specific variable are of the same type). As new types of sensors appear on the market a user can simply change a sensor to a different class with more appropriate characteristics but the basic benchmark description is still valid and does not have to be re-written. Furthermore, this approach simplifies the programming since each class description will only have to be coded once and then any signal can be manipulated based on the properties defined for that specific sensor class.

  Table 3: Sensor classes for the benchmark (UD = user defined).

  Class	Sample time (min)	Delay time (min)	Low detection limit	Measurement noise (s)
  0 (ideal)	continuous	0	0	0
  1	continuous	0	> 0, UD	0
  2	continuous	0	0	> 0, UD
  3	continuous	0	> 0, UD	> 0, UD
  4	10	10	0	0
  5	10	10	0	> 0, UD
  6	10	10	> 0, UD	> 0, UD
  7	30	30	> 0, UD	> 0, UD
  UD	UD	UD	UD	UD

  Note that the sensor classes with their selected characteristics should only be used for implementing your control strategy (i.e. to get access to the required measurements for the controllers). When calculating the various benchmark performance indices the exact values provided by the internal mathematical model should be used. The class UD is provided to allow you a simple way of defining your own sensor class if the need exist. By simply setting values to the four characteristics of the user-defined sensor class, you can in a clear way propagate the characteristics of a specific sensor to your colleagues. However, the user-defined class should be used with care and only when none of the predefined sensor classes are even close to the sensor characteristics you need to describe. Note that a low detection limit of zero means that there is no limit that has any relevant influence on the benchmark plant. For example, a temperature sensor certainly has a low limit, which is less than zero but such values are not of interest for WWTP applications, and the sensor should still (for reasons of simplicity) be classified as a low detection limit of zero. The measurement noise indicated in the table should in all ‘normal’ cases be white zero-mean normally distributed noise.

  As an example of how to use the sensor classes given in the table the oxygen and nitrate sensors described for the default closed-loop test case can now very easily be described as:

  • oxygen sensor: Class 0;
  • nitrate sensor: Class 6 with low detection limit = 0.1 mg N/l and measurement noise (s) = 0.1 mg N/l.

  A large number of different sensors exist for WWTP applications and their characteristics vary significantly due to functionality, brand, etc. If you do not have any detailed information with regard to the sensors you want to use, some rough guidelines are given below (Table 4) for some common sensor types. Obviously, sensors for pH, temperature, phosphate and alkalinity are not very useful for the current representation of the simulation benchmark since these variables are not included or do not influence the behaviour of the models. They are simply provided here to allow for a more complete overview and future extensions of the benchmark. Moreover, the classification of a sludge blanket level sensor as ideal may not be exactly valid. However, due to the discrete layering (each 0.4 m in height) used to model the settler, any delay or noise related to the sensor will be insignificant in comparison to errors introduced by this fact.

  Table 4: Typical sensor characteristics.

  Measured variable	Class	Low detection limit	Measurement noise (s)
  Flow rate (m3/h)	0	–	–
  Water level (m)	0	–	–
  Temperature (°C)	0	–	–
  pH	0	–	–
  SO (mg O2/l)	0	–	–
  Sludge blanket level (m)	0	–	–
  SNO (mg N/l)	6	0.1	0.1
  SNH (mg N/l)	6	0.2	0.2
  SS (mg COD/l)	7	5	1
  Phosphate (mg P/l)	6	0.1	0.1
  SALK (mol HCO3/m3)	5	–	0.1
  Mixed-liquor suspended solids (mg/l)	6	100	100
  Total COD (mg COD/l)	6	100	100
  OUR (mg/(l·d))	6	25	50

   

  Control Handles

  For reasons of simplicity all available control handles are considered to be ideal with regard to their behaviour. In the closed-loop test case only two control handles were used: the internal recirculation flow rate (Q_a) and the oxygen transfer rate in reactor number 5 (K_La5). The following control handles are considered to exist for the implementation of new control strategies at the benchmark plant:

  • internal flow recirculation rate (Q_a);
  • return sludge flow rate (Q_r);
  • wastage flow rate (Q_w);
  • anoxic/aerobic volume – all five biological reactors are equipped with both aerators and mechanical mixing devices, i.e. in a discrete fashion the volumes for anoxic and aerobic behaviour can be modified;
  • aeration intensity individually for each reactor (K_La1, K_La2, K_La3, K_La4, K_La5);
  • external carbon source flow rate (Q_{external carbon 1}, Q_{external carbon 2}, Q_{external carbon 3}, Q_{external carbon 4}, Q_{external carbon 5}) where the carbon source is considered to consist of readily biodegradable substrate, i.e. S_S;
  • influent distribution by use of step feed (fractions of the influent flow to each of the five biological reactors: f_Qinflow1, f_Qinflow2, f_Qinflow3, f_Qinflow4, f_Qinflow5);
  • distribution of internal flow recirculation (fractions of the internal recirculation flow to each of the five biological reactors: f_Qa1, f_Qa2, f_Qa3, f_Qa4, f_Qa5);
  • distribution of return sludge flow (fractions of the return sludge flow to each of the five biological reactors: f_Qr1, f_Qr2, f_Qr3, f_Qr4, f_Qr5);

  The above selection gives you about 30 individual control handles with which to manipulate the defined benchmark plant and dramatically increases its flexibility. Such a number of available control handles may not be realistic for a real plant but is defined for the benchmark plant in order to allow for basically any type of general control strategies and this is the main purpose of the COST benchmark. The defined limitations for the different control handles are given in the Table 5.

  Table 5: Available control handles and their limitations.

  Control handle	Minimum value	Maximum value	Comments
  Qa (m3/d)	0	92230	Max = 500% of Q0,stab
  Qr (m3/d)	0	36892	Max = 200% of Q0,stab
  Qw (m3/d)	0	1844.6	Max = 10% of Q0,stab
  KLa1 (d-1)	0	360	Reactor 1
  KLa 2 (d-1)	0	360	Reactor 2
  KLa 3 (d-1)	0	360	Reactor 3
  KLa 4 (d-1)	0	360	Reactor 4
  KLa 5 (d-1)	0	360	Reactor 5
  Qexternal carbon 1 (m3/d)	0	5	Reactor 1 Carbon source conc. 400000 mg COD/l available as SS (e.g. 25% ethanol solution)
  Qexternal carbon 2 (m3/d)	0	5	Reactor 2 Otherwise same as above
  Qexternal carbon 3 (m3/d)	0	5	Reactor 3 Otherwise same as above
  Qexternal carbon 4 (m3/d)	0	5	Reactor 4 Otherwise same as above
  Qexternal carbon 5 (m3/d)	0	5	Reactor 5 Otherwise same as above
  fQin1, fQin2, fQin3, fQin4, fQin5	0	1	Part of the influent flow rate distributed to each biological reactor Note: the sum of all five must always equal one
  fQa1, fQa2, fQa3, fQa4, fQa5	0	1	Part of the internal recirculation flow rate distributed to each biological reactor Note: the sum of all five must always equal one
  fQr1, fQr2, fQr3, fQr4, fQr5	0	1	Part of the sludge return flow rate distributed to each biological reactor Note: the sum of all five must always equal one

  

  Bibliography
  

  Alex J., Beteau J.F., Copp J.B., Hellinga C., Jeppsson U., Marsili-Libelli S., Pons M.N., Spanjers H., Vanhooren H. (1999) Benchmark for evaluating control strategies in wastewater treatment plants, ECC'99 (European Control Conference), Karlsruhe, Sept. 1999.
  

  Jeppsson, U. (1996) Modelling Aspects of Wastewater Treatment Processes, PhD thesis, Lund Institute of Technology, Sweden.
  

  Henze M., Grady Jr C.P.L., Gujer W., Marais G.v.R., Matsuo T. (1986) Activated Sludge Model n° 1, IAWQ Scientific and Technical Report n°1, IAWQ, London, Great-Britain
  

  Pons M.N., Spanjers H., Jeppsson U. (1999) Towards a benchmark for evaluating control strategies in wastewater treatment plants by simulation, Escape 9, Budapest, June 1999
  

  Spanjers H. Vanrolleghem P.A., Nguyen K., Vanhooren H., Patry G.G. (1998) Towards a simulation benchmark for evaluating respirometry-based control strategies, Wat. Sci. Tech. 37(12) 219-226.
  

  Takacs I., Patry G.G., Nolasco D. (1991) A dynamic model of the clarification thickening process, Water Research, 25, 10 1263-1271
  

  Vanhooren H. And Nguyen K. (1996) Development of a simulation protocol for evaluation of respirometry-based control strategies. Report University of Gent and University of Ottawa.
  

  Vanrolleghem P.A., Jeppsson U., Carstensen J., Carlsson B., Olsson G. (1996) Integration of WWT plant design and operation - A systematic approach using cost functions, Wat. Sci. Tech., 34 (3-4), 159-171.