Feedforward aeration control of a Biocos wastewater treatment plant B. Wett and K. Ingerle Institute for Environmental Engineering, University of Innsbruck, Technikerstraße 13, A-6020 Innsbruck, Austria Abstract The Biocos strategy as a cyclical time controlled activated sludge system shows a great variability in operation and control. One topic such a type of treatment plant has to deal with is the optimum relation between aerobic and anoxic conditions. The aeration control has to adapt the length of the nitrification phases to the current constraints in order to save operational costs and maximise nitrogen elimination. Since wastewater treatment plants up to a certain size are usually not equipped with on-line nitrogen probes, influent flow and temperature can be taken as control parameters for the aeration system. The defined relation between influent flow and ammonia load is based on measurements and the relation between ammonia load and required aeration time is model based. Keywords Activated sludge; aeration control; Biocos; cyclical wastewater treatment; nitrogen elimination Introduction As is commonly known biological nitrogen elimination is a two step process with each step requiring different boundary conditions. Presuming intermittent aeration the length of the aeration periods is the key parameter to define aerobic conditions. The lower limit of the required aeration time is fixed by the given NH 4 N effluent standards. The upper limit of aeration time is defined by the goal to promote denitrification during anoxic conditions and not to waste energy. Therefore both steps - nitrification and denitrification should be kept in the right balance by a proper aeration control. Since aeration aims at defined NH 4 N concentrations, an ammonia based aeration control would be the first choice. The appropriate way is to employ a feedback control loop in order to manipulate the aeration device according to the measured ammonia effluent values (Marlin, 1995). Nevertheless on-line measurements of ammonia are not available at any plant because of high investment costs and effort for maintenance. But on-line measurements of influent flow and temperature are usually installed even at smaller treatment works and could be used in a feedforward control. The aim is to analyse the interactions between feed flow, temperature, aeration and nitrogen effluent concentrations and to formulate a fixed control function for the aeration system. Biocos strategy The aeration control is applied to the Biocos strategy which will be introduced first and fitted into the categories of cyclic activated sludge systems. Irvine suggested at the last SBR conference in Munich (Irvine et al., 1997) that all cars should not be given the name of one single car. In this way he tried to explain as simple as he could that all periodic processes should not be classified as SBRs. Two main activated sludge systems can be distinguished Single Tank Technology with no separate clarification and continuous flow systems with clarification in a separate tank. Continuous flow systems provide a constant water level as an additional characteristic whereas single tank systems can be operated either with constant or variable volume. Between the two main systems with time or space control so-called combined systems are situated which try to combine features from both strategies (Figure 1). Biodenitro and Water Science and Technology Vol 43 No 3 pp 85 91 IWA Publishing 2001 85
B. Wett and K. Ingerle Figure 1 Classification of activated sludge systems with Biocos combining features from Single Tank Technology and continuous flow systems Figure 2 Flow- and operational scheme of a Biocos plant 86 Biocos are two examples for combined systems both operated periodically. The Biodenitro system (Isaacs, 1997; Christensen, 1975) shows alternating flow patterns and the aeration is switched from one reactor to the other. So the biological reactors are operated periodically and the settler continuously. A Biocos system is operated just the other way round: continuous operation of the biological reactor and periodic operation of the settling compartments. A Biocos plant is a cyclical activated sludge system with a continuous influent flow and a continuous effluent flow and is therefore operated at constant water levels. These operational features are achieved by a configuration of three reactors. The influent flow is fed to an aerated reactor the B-reactor, which is followed by two parallel SU-reactors. The SUreactors are operated according to the single tank-technology. The time control of the Biocos system provides a settling phase and a discharge phase in order to withdraw
B. Wett and K. Ingerle Figure 3 Daily variation of influent flow, concentration and load at the WWTP Längenfeld supernatant water from the SU-reactors. Due to alternating operation the effluent valve of one of the two SU-reactors is open and enables the influent flow to displace supernatant water. During this period activated sludge is disposed from the B-reactor to one SU-reactor. Therefore the sludge concentrations need an equalisation after each discharge phase. During the mixing phase the content of the B-reactor is pumped into the SU-reactor near the bottom causing a return flow at the surface until the circulation has balanced the concentrations. In the SU-reactors endogenous denitrification takes place beneath the settling sludge blanket because of high sludge concentrations and a lack of easily degradable carbon. Additionally the denitrification process is promoted in the B-reactor by an interruption of the aeration during the mixing phase and eventually at the beginning of the settling phase. The duration of this interruption is the manipulated variable of the aeration control. Measured correlation between flow and ammonia load A conventional SBR system is fed discontinuously from a storage tank or the influent flow is parted between several parallel SBRs. Thus usually always the same amount of wastewater is treated within every operation cycle. In contrast to this strategy the Biocos system employs no devices for hydraulic equalisation of the influent flow. The actual influent flow determines the required length of the aeration phase, i.e. the oxygen demand depends on the influent organic and nitrogen load. Especially if the WWTP is fed by a separated sewer system with hardly any contribution of stormwater a sufficient correlation between flow and load can be observed. The following example considers the Biocos plant Längenfeld (10,000 PE) in western Austria 87
B. Wett and K. Ingerle Figure 4 Correlation between NH 4 N load and influent flow (Ingerle, 1998). The profiles of influent flow and concentrations show the characteristic daily variations of smaller communities, both with a maximum about 2.5 times higher than the minimum. Thus the load peak as the product of flow and concentration is about seven times higher than the load minimum (Figure 3). Since flow and concentration peaks superimpose each other, influent flow and load show an empirical nonlinear correlation derived from representative measurement data (Figure 4). In general the ammonia load can be expressed as a exponential function of the influent flow: LNH = c Q 4 LNH ammonia load 4 Q influent flow ck, fitted parameters. k (1) 88 Model based correlation between ammonia load and aeration While the relation between flow and load is directly measurable the relation between load and required aeration depends on the system and the boundary conditions. A model as a mathematical system description has to be developed which is able to regard the nonlinear system behaviour and the effects of control actions under dynamic boundary conditions. This requirement means a discrete consideration in time and space in order to take biokinetic processes, sedimentation processes and transport interactions between different reactors into account. For these reasons a three layer settler model was linked with the appropriate IAWQ ASM No.1 and modified according to the characteristic system features (Wett and Rauch, 1996; 1999). At the considered day of measurement in Figure 4 no setpoint of the aeration controller was adjusted. Therefore the oxygen concentration in the B-reactor reached values (9 mg/l) near saturation and respiration hardly decreased the O 2 level to zero during the nonaerated periods. That is the reason for the observed high nitrate concentration and the ammonia concentration near zero which facilitated the calibration of the transport parameters. The profile of the nitrate concentration in the top layer of a SU-reactor during one operation cycle outlines the characteristics of the system. During the settling phase the nitrate concentration decreases because of endogenous denitrification in the thickener layer and the compression zone. Nitrate rises again due to NO 3 transport from the B-reactor during the discharge phase. At the beginning of the mixing phase the nitrate drops down because of an
B. Wett and K. Ingerle Figure 5 Model calibration. Daily variations of ammonia and nitrate in the top layer of the SU-reactor and the uncontrolled oxygen concentration in the B-reactor (2.9.1998 at WWTP Längenfeld; 5800 PE load; Temp.=14,4ºC; MLSS=3.8 g/l) Figure 6 Model verification. Results of an 11 days simulation period compared to measurements (26.9.1998 at WWTP Längenfeld; oxygen setpoint at 2.25 mg/l; 5100 PE load;temp.=13.7ºc; MLSS=4.1 g/l; sludge retention time SRT=23 days) instantaneous concentration balance of all three layers in the SU-reactor. Due to the circulation between the B- and the SU-reactor the nitrate concentration increases again to the equalised level. During the following weeks the oxygen level in the B-reactor was adjusted to a setpoint of 2.25 mg/l. The interruptions of the aeration still lasted 30 minutes. With this fixed adjustment nitrification and denitrification showed a satisfying balance. In order to verify the fitted model parameters a simulation period of 11 days was calculated. The daily average influent loads measured in composite samples had been varied according to the distributions of Figure3 and formed input files for the model calculation. The simulated profile of ammonia and nitrate concentrations showed a sufficient correspondence to the measured values of grab samples as demonstrated in Figure 6. Optimised aeration control After verification of the model predicted effluent concentrations calculated from given influent values a variation of the aeration control could be analysed. The oxygen setpoint was not changed but the duration of the nonaerated periods was. At the beginning of each operation cycle the actual ammonia load was calculated from the influent flow during the 89
preceding cycle according to Eq. 1) (resp. Figure 4). The required aeration time T air depends on this ammonia load with no regard to the COD-load. The organic load influences the oxygen consumption but is not a limiting factor for the aeration time. At constant oxygen level and ammonia concentrations not too close to zero a linear relation between aeration time T air and nitrified ammonia load L NH4 can be assumed: B. Wett and K. Ingerle Tair = l LNH4 l aeration time per kg NH k k Tair = l c Q = m Q mk, fitted parameters. 4 N This exponential function expresses the demand for aeration of the considered system at a fixed oxygen level. The model based parameter fitting of m and k of Eq. (3) aimed at constant ammonia effluent values around 2 mg/l and led to shorter aeration intervals during night-time (see Figure 7). Additionally the operation cycle time was shortened from 160 to 120 minutes in order to reduce concentration discontinuities. Beside daily load variations especially weekly or seasonal load variations are a matter of concern. For example in the tourist catchment area of Längenfeld the weekly average load varies between 4000 and 10,000 PE. Moreover the water temperature shows a range of variation from 5 to 15ºC. To implement the influence of temperature in the control system, the temperature dependency of autotrophic biomass has to be determined. The autotrophic growth rate µ(temp) at the actual temperature can be derived from the growth rate µ(20ºc) at a temperature of 20ºC by following expression: (2) (3) ( *( Temp 20)) µ ( Temp)= µ ( 20º C) e m (4) Figure 7 Performance of a Biocos plant employing a flow dependent aeration control (same boundary conditions as considered in Figure 6) 90 Figure 8 Scheme of the feed forward control loop of the aeration system
Temp actual temperature m coefficient about 0.098 (Gujer, 1985) This expression can be adapted to a temperature coefficient k Temp of the nitrification period T air of the system considered in Figure 6. The temperature coefficient k Temp at the actual temperature (Temp=13.7ºC) is set to 1 and the coefficient k 20 at 20ºC can be calculated: k = = k e Finally the temperature coefficient k Temp of any given temperature can be implemented in Eq. (3): T Temp air mq = k ( m*(temp 20)) 1 20 k Temp aeration time depending on flow and temperature. After evaluation of the control strategy at different loads in model reality it can be applied in full-scale reality. According to the general feed forward control scheme (Stephanoupoulos, 1984) the measured disturbance is represented by the influent flow and the water temperature. The control unit calculates the required aeration time and the duration of the aeration period as the manipulated variable is set (Figure 8). (5) (6) B. Wett and K. Ingerle Conclusions Three issues or statements can be extracted from this presentation: An example for the application of a model based aeration control is given. The model is not applied for real time control purposes but for supporting the development of a fixed control law. Also treatment works with only basic monitoring equipment are not necessarily excluded from flexible control and automation. For instant flowmeter and temperature measurement can be involved in a feedforward control loop for the aeration system. Cyclical time controlled wastewater treatment systems are well suited for adjusting to varying boundary conditions. References Christensen, M.H. (1975). Denitrification of sewage by alternating process operation. Prog. Water Technol., 7(2), 339 347. Gujer, W. (1985). A dynamic model to simulate complex activated sludge systems (in German). Habilitation at the ETH Zürich. Henze, M., Grady Jr., C.P.L., Gujer, W., Marais, G.v.R., Matsuo, T. (1987). Activated sludge model No.1. IAWQ Scient. and Techn. Reports No.1. Ingerle, K. (1998). Biocos-activated sludge system, wwtp Längenfeld and Pielenhofen (in German). gwf, Abwasser Special, 139(4), 32 35. Ingerle, K. (1999). Biocos-sewage treatment plants, description and dimensioning (in German). Korresp. Abwasser, 46(8), 1221 1230. Irvine, R.L., Wilderer, P.A. and Flemming, H-C. (1997). Controlled unsteady state processes and technologies an overview. Wat. Sci. Tech., 35(1), 1 6. Isaacs, S. (1997). Automatic adjustment of cycle length and aeration time for improved nitrogen removal in an alternating activated sludge process. Wat. Sci. Tech., 35(1), 225 232. Marlin, Th.E. (1995). Process Control, Designing Processes and Control Systems for Dynamic Performance. McGraw-Hill Inc., Singapore. Stephanopoulos, G. (1984). Chemical process control. Prentice-Hall Inc., Englewood Cliffs, New Jersey. Wett, B. and Rauch, W. (1996). Simulation of discontinuous treatment strategies. Preprint of 1st IAWQ SBR-conference, Munich, 589 595. Wett, B. (1999). Simulation analysis of a Biocos-plant, Cyclical secondary clarification or continuous flow SBR (in German). Korresp. Abwasser, 46(7), 1068 1074. 91