Improving Nutrient Removal While Reducing Energy Use at Three Swiss WWTPs Using Advanced Control

Transcription

1 Improving Nutrient Removal While Reducing Energy Use at Three Swiss WWTPs Using Advanced Control Leiv Rieger 1,2 *, Imre Takács 3, Hansruedi Siegrist 1 ABSTRACT: Aeration consumes about 60% of the total energy use of a wastewater treatment plant (WWTP) and therefore is a major contributor to its carbon footprint. Introducing advanced process control can help plants to reduce their carbon footprint and at the same time improve effluent quality through making available unused capacity for denitrification, if the ammonia concentration is below a certain set-point. Monitoring and control concepts are cost-saving alternatives to the extension of reactor volume. However, they also involve the risk of violation of the effluent limits due to measuring errors, unsuitable control concepts or inadequate implementation of the monitoring and control system. Dynamic simulation is a suitable tool to analyze the plant and to design tailored measuring and control systems. During this work, extensive data collection, modeling and fullscale implementation of aeration control algorithms were carried out at three conventional activated sludge plants with fixed pre-denitrification and nitrification reactor zones. Full-scale energy savings in the range of 16 20% could be achieved together with an increase of total nitrogen removal of 40%. Water Environ. Res., 84, 170 (2012). KEYWORDS: Biological nutrient removal, aeration control, activated sludge model, energy savings, denitrification, control system design, ammonia control. doi: / x Introduction Aeration contributes about 60% to the total energy consumption of WWTPs (Ingildsen, 2002). Reducing aeration by innovative control to the minimum level required to guarantee a desired effluent quality helps saving costs and significantly reduces the carbon footprint of the plant. Costs savings however must not increase the risk of effluent quality violations in terms of total nitrogen and phosphorus content nor must it increase the carbon footprint of the plant through elevated greenhouse gas emissions. The use of online sensors for automated process control has been the topic of various studies since the early seventies 1 Eawag, Swiss Federal Institute of Aquatic Science and Technology, Environmental Engineering, Ueberlandstr. 133, PO Box 611, CH-8600 Duebendorf, Switzerland; hansruedi.siegrist@eawag.ch. 2 *EnviroSim Associates Ltd., McMaster Innovation Park, 175 Longwood Rd S, Suite 114A, Hamilton, Ontario, L8P 0A1, Canada; rieger@envirosim.com. Corresponding author. 3 EnviroSim Europe, 15 Impasse Fauré, Bordeaux, France; imre@dynamita.com. (compare Ingildsen, 2002; Olsson and Newell, 1999); first with a focus on DO control and later also on biological nutrient removal (Ingildsen, 2002; Olsson et al., 2005; Yuan et al., 2002). Although several publications claimed that the reliability of sensors improved significantly (e.g., Jeppsson et al., 2002), data quality and signal availability can still be seen as the limiting factor for the use of process control in practice. Worldwide, the number of full-scale implementations of advanced process control in wastewater treatment is still scarce (Ingildsen, 2002; Jeppson et al., 2002; Palmer et al., 2007; Weijers, 2000). During the present work, extensive measuring campaigns, simulation studies and full-scale implementations of aerationcontrol concepts were carried out. All three WWTPs evaluated are conventional activated sludge plants with fixed volumes for pre-denitrification and nitrification and cover a range of plant sizes from 35,000 population equivalents (p.e.) (13,000 m 3 /d), 130,000 (40,000 m 3 /d) and 600,000 p.e. (205,000 m 3 /d). This should guarantee the applicability of the study to other plants. The primary goal of this project was to demonstrate that increasing nitrogen removal in parallel with a reduction of the energy consumption is viable using control technology. The excess capacity recovered in the optimized nitrification process, particularly during summer time, is used to introduce anoxic or even anaerobic phases and therefore allows increased biological nitrogen and phosphorus removal. Thanks to the increased biological phosphorus removal, chemical doses can be reduced and the sludge production decreases. The objectives were to test different control strategies at a variety of full-scale plants to find the best control concepts for specific conditions. Procedure A general procedure for design and implementation of process control strategies (Figure 1) was followed in all three case studies. The objectives should be defined together with the responsible plant manager and the operator-in-chief. Data quality evaluation is essential for a successful design and should begin directly after the start-up of the optimization study and should accompany all measuring campaigns. The dynamic model provides the platform to design, evaluate, adapt and finally select the optimal control concept(s), which will then be implemented and tested. Finally, it is proposed to evaluate the success on the basis of a cost-benefit calculation. Definition of Objectives. The goals of the studies were to increase the nitrogen removal in parallel with a reduction of energy consumption. The constraints were to keep the effluent 170 Water Environment Research, Volume 84, Number 2

2 Figure 1 Procedure to design and implement process control strategies. concentrations below the specified regulation limits. The Swiss ammonia limit of 2 mg N/L in the 24-hour composite sample was the main boundary for the aeration control concepts tested. Priority was on nitrogen removal, but if enough free capacity exists in the plant, it should be used for enhanced biological phosphorus removal (Bio-P) to reduce the dosage of P- precipitants. The reduction in sludge production leads to lower sludge disposal costs. Special emphasis was placed on the implementation of the measuring and control system. This required understandable and adaptable control concepts, suitable sensors with low maintenance requirements and tools for monitoring the data quality of the measuring devices. Data Quality Evaluation and Control. Good data quality is essential for reliable modeling results as well as for effective control systems. A concept for data collection, quality evaluation, and reconciliation in optimization studies (Rieger et al., 2010; see Figure 2) was used in this work: After analyzing existing data and locating gaps in routine measurements, a few additional measurements were taken to calculate mass balances on the basis of redundant information. The combination of different mass balances allowed systematic measuring errors to be detected, isolated, and identified (Thomann, 2008). Additional experiments were carried out to obtain values for the precision of the equipment and analytical methods and to evaluate the response times of the measuring devices (Rieger et al., 2003) and the actuators (Rieger et al., 2006). Intensive measuring campaigns were planned to include redundant information for quality checks (e.g., mass and flow balances). The implementation and success monitoring data sets were evaluated for data quality too. Simulation-supported Controller Design. The goals of process control in wastewater treatment can be summarized as follows: N To meet effluent discharge requirements. N To achieve good disturbance rejection. N To optimize operation to minimize the operating cost. In this work the main goal was to reduce operating costs (including effluent taxes) with the constraint of meeting the effluent limit requirements. The behavior of the control system during peak loading was tested and special feed-forward controllers designed to increase the control authority against influent variability. The specific control goal is to keep the ammonia concentration at the end of the aerated zone near a defined set-point. This is done by increasing or decreasing the amount of air which is blown into the reactor. Due to the Monod dependence, the modeled nitrification rate reaches ca. 80% of the maximum at DO concentrations of 2 mg/l (depending on the half saturation constant, which in this study was chosen as 0.5 mg/l). Consequently, the ammonia control loop has been put on top of the conventional DO control loop. This guarantees DO concentrations below a limit of 2 3 mg O 2 /L, even if the ammonia is above the set-point. Below the ammonia set-point, the oxygen input will be decreased and capacity for an enhanced denitrification arises in parallel with a reduction in energy consumption. Whether the oxygen input can be switched off completely or only reduced to a minimum depends on the kind of aeration system used. If ceramic aerators are used, a minimum airflow Figure 2 Concept of data collection, evaluation, and reconciliation for simulation studies (Rieger et al., 2010). February

3 Figure 3 DO control loop. rate has to be maintained, whereas membrane aerators can normally deal with the air supply being completely switched off. To guarantee a sufficient degree of mixing, additional mixers have to be installed or an overlying control scheme has to restart aeration after a certain period of time. The length of the unaerated interval depends on the sludge settling behavior and should be selected based on an experimental analysis. If the ammonia sensor is installed at the end of the aerobic zone (feed-back concept), the signal may be too late for a control action to manage peak loading. At plants with a highly dynamic influent combined with plug flow behavior and small reactor volumes, this can lead to violation of the effluent limits. This effect is increased if sensors with a high response time (. 30 min.) are used (Alex et al., 2003). One solution is to combine feed-back and feed-forward concepts. An additional ammonia sensor upstream of the aeration zone (e.g., in the primary effluent or in one of the unaerated reactors) provides the required advance signal to prepare the plant for a peak load. The evaluated aeration control concepts enhance the existing DO control loop (Figure 3) by an ammonia control loop. The following basic concepts have been tested: N On-off control of the blowers based on the ammonia concentration (Figure 4) N DO set-point adjustment based on the ammonia concentration (Figure 5) N Combination of feed-back and feed-forward control using a maximum condition (Figure 6). Feed-forward Controller. On the basis of an additional ammonia sensor in the effluent of the primary clarifier and a flow measurement, the feed-forward controller calculates the time when full aeration is necessary. A simplified model is used Figure 5 DO + ammonia control loops. which compares the ammonia load in the effluent of the primary clarifier with an estimated nitrification capacity (eq 1). where: L NH 4,Inf nit state ~ ðr nit : VolBR Þ L NH4,Inf 5 Measured NH 4 -N influent load [g/d] r nit 5 Nitrification rate [g/(m 3 *d)] Vol BR 5 Volume aerated bio-reactors [m 3 ] The nitrification rate is estimated according to eq 2 for the aeration strategy with the lowest air demand (this could mean intermittent aeration or not all tanks aerated). where: r nit ~ X ANO : m max,t : taer Y ANO ð1þ ð2þ X ANO 5 Concentration of nitrifiers (from steady state and low aeration strategy, see Eq. 3) [g/m 3 ] m max,t 5 Actual max. growth rate at given temperature 5 m max * exp(0.105*(t-20uc)) [1/d] m max, [1/d] (Max. growth rate at 20uC) Y ANO [g COD/g N] (Autotrophic yield) t aer [-] (reduction factor for intermittent aeration and low DO conc.) T 5 Actual temperature [uc] (in this study 15uCis used) The concentration of nitrifiers depends on the mean influent load of ammonia (eq 3): Figure 4 DO control loop + ammonia on-off control. Figure 6 DO + ammonia feed-forward/feed-back control. 172 Water Environment Research, Volume 84, Number 2

4 Figure 7 Process scheme WWTP Werdhoelzli (one of twelve lanes). where: X ANO ~ L : NH 4,removed YANO : SRT Vol BR : 1zbANO,T : ð3þ ð SRTÞ L NH4,removed 5 Mean NH 4 -N influent load 2 NH 4 feed-back controller set-point [g/d] SRT 5 Sludge retention time [d] b ANO,T 5 Max. decay rate at given temperature 5 b ANO,20 * exp(0.105*(t-20)) [1/d] b ANO, [1/d] (Max. decay rate at 20uC) If the load is higher than the nitrification capacity, increased aeration is triggered after a delay time. The delay is necessary to take the hydraulic retention time in the anoxic zone into account. The combination of feed-forward and feed-back is implemented by a selector using a maximum condition, that is the controller with the higher air demand is selected. Simulation studies. The plant models were calibrated and validated on the basis of intensive measuring campaigns at WWTP Morgental and WWTP Thunersee. In addition, the hydraulic behavior of the plants was evaluated by tracer experiments using sodium bromide. For WWTP Werdhoelzli, a representative data set was created by analyzing the available routine data and 11 diurnal variations. The simulation studies were carried out to evaluate the potential of the plants and to design, test and finally select the best controller for each specific plant. The evaluation of the simulation results at WWTP Thunersee was based on the existing effluent load tax of the canton of Berne, whereas a slightly modified effluent load tax is used for the studies at WWTP Morgental and WWTP Werdhoelzli, where no official tax exists. Effluent quality tax in the canton of Berne (KGSchG, 1996): N NH 4 -N 5 4 CHF/kg N (<4$/kg N) N NO 3 -N 5 1 CHF/kg N (<1$/kg N) N PO 4 -P 5 30 CHF/kg P (<30$/kg P) Effluent quality tax (used at WWTP Morgental and WWTP Werdhoelzli): N NH 4 -N 5 1 $/kg N (for concentrations, 2 mg N/L) N NH 4 -N 5 4 $/kg N (for concentrations. 2 mg N/L) N NO 3 -N 5 1 $/kg N N PO 4 -P 5 30 $/kg P The energy costs were calculated on the basis of 0.15 $/ kwh. The costs of the phosphorus precipitants were set to 3 $/kg P precipitated and the costs for the precipitation sludge disposal to 7 $/kg P. Full-scale Implementation. On the basis of the simulation studies, the selected control concepts were adapted to several constraints of the plants and implemented on a full scale at WWTP Morgental and WWTP Thunersee. Due to financial restrictions, no additional mixing devices were installed in the normally aerated zones, and consequently the sludge settled during phases without aeration. To prevent problems with permanently settled sludge, the settling behavior was tested and as a result the maximum duration of non-aerated phases were set to min in both full-scale implementations followed by a short re-aeration phase. Control of Success. Measuring campaigns were carried out at WWTPs Morgental and Thunersee to validate the simulation results. The cost-benefit calculations included annual costs for additional investments and operation of the control systems under evaluation and compared them against savings in terms of energy and chemicals consumption, sludge disposal and effluent quality (based on the effluent tax). Case Studies Two full case studies were carried out for WWTP Morgental and WWTP Thunersee. For WWTP Werdhoelzli, two simulation studies on ammonia-based control and the implementation of the AI-process (alternating influent, intermittent aeration) scheme were performed during this work. WWTP Werdhoelzli. WWTP Werdhoelzli treats the wastewater from the city of Zurich and several nearby villages. With a current capacity of 550,000 p.e. (205,000 m 3 /d), WWTP Werdhoelzli is the largest treatment plant in Switzerland and must remove substantial quantities of nitrogen because the receiving river Limmat flows into the Rhine with special demands on nitrogen limits. The plant was initially designed for nitrification only. In 1997 two anoxic compartments comprising 28% of the total volume were implemented in each of the twelve aeration tanks (Figure 7). This allowed 55 60% of the nitrogen to be removed (Siegrist et al., 2000). Definition of Objectives. WWTP Werdhoelzli expressed interest to further improve nitrogen removal and reduce aeration energy. This work on ammonia-based control is part of a major plant upgrade planning, which also includes separate treatment of digester supernatant. Therefore, scenarios for separate treatment and controlled dosing of the supernatant were included in this study. Data Quality Evaluation and Control. Since only a relative comparison is to be made, only basic data quality control was carried out. The typical weekly variation used is based on the analysis of the routine measurements of the plant laboratory of the years 2000 and 2001 and eleven measured diurnal variations. Simulation-supported Controller Design. A simulation study for WWTP Werdhoelzli was carried out to compare the costs of the aeration energy, the phosphate precipitation and an effluent quality tax for different aeration control concepts. An important February

5 Table 1 Modeled control concepts at 156C with parameter settings (WWTP Werdhoelzli). No Control concept Parameter V0 Actual control concept DO set-point 5 2 mg/l, return sludge 5 25,000 m 3 /d (1 lane) Valve for last two aeration registers minimum 70% open V1 DO set-point control based on NH 4,out at the end of the aeration tank (NH 4 feed-back control) NH 4,eff. 1.8 mg/l 5. DO set-point 5 2mg/L NH 4,eff, 1.6 mg/l 5. min. airflow m 3 /(h * aerator), valve of last aeration register minimum 70% open V1a V1 + control of valve for last aeration register based on DO at the end of aerated zone According to V1 + possible reduction of valve opening to minimum 10% to reach minimal airflow V2 DO set-point control based on combination of NH 4 feed-back and feed-forward control According to V1a + feed-forward control based on an NH 4 sensor in the primary effluent (Eq. 1 3) V3 Feed-forward control of digester supernatant (DS) dosage According to V1a + control of DS dosage based on the ammonium load in the primary effluent V3a Separate treatment of digester supernatant According to V1a + separate treatment of digester supernatant cost factor is the reduction of the phosphate precipitants and the sludge production due to enhanced biological phosphate removal (Bio-P). The ASM3 with the EAWAG Bio-P module (Rieger et al., 2001) was selected as the biological process model. Only a basic calibration of the biological model was carried out. The energy consumption and amount of air used were calibrated based on a six-month data set with daily averages. Besides the current DO control system operating at 15uC, four main versions were investigated (Table 1). Version 1 includes a feed-back control of the DO set-point based on ammonium sensors at the end of the aeration tanks, whereas V1a comprises additional oxygen feed-back control of the valve serving the last two aeration zones. Version 2 combines the feed-back control of version 1a with a feed-forward control based on an ammonium sensor in the primary effluent. Version 3 comprises of a controlled dosage of digester supernatant based on the NH 4 load in the primary effluent and version 3a is a separate treatment of the supernatant. The feed-forward controller was simulated according to equations 1 to 3. Table 2 shows the step-wise increase in aeration intensity for the feed-forward controller for WWTP Werdhoelzli. Note that only a ratio of 1.4 (NH 4 load/nitrification capacity) triggers an action by the feed-forward controller. This is due to the fact that the disturbances should mainly be dealt with by the feed-back controller. The feed-forward controller should provide a fast reaction to sudden ammonia peaks in the influent. The simulated results show that there is an enormous potential for measuring and control technology to reduce operating costs and simultaneously improve effluent quality. The best control version without separate supernatant treatment (V3) reduces the energy cost by 25% and chemical addition by more than 50% compared to the current version. If an effluent load tax is also taken into account, a total cost reduction of about 30% in relation to the actual situation can be Table 2 Settings for feed-forward controller WWTP Werdhoelzli. Control loop nit State DO set-point Aer ,1.8 0 Aer ,1.2 0 attained. Looking at the benefits, the savings in energy of about $425,000/yr greatly exceed the additional annual costs of about $50,000 for the measuring and control system and its operation. Net savings increase to about $1,400,000/yr if the effluent tax ($225,000/yr) and the reduction of precipitants and sludge disposal ($800,000/yr) due to the biological P-removal are included (Figure 8). The version (V3a) with separate digester supernatant treatment (Anammox) increases net energy savings by another $80,000/yr (difference of aeration energy for the oxidation of 300 t NH 4 /yr supernatant separately instead of in the activated sludge treatment). Total net savings then increase to about $2,400,000/yr (effluent tax: $550,000/yr; precipitants and sludge reduction $1,300,000/yr due to biological P-removal) from which about $400,000/yr have to be deducted for separate supernatant treatment (Siegrist et al., 2004). The total nitrogen removal efficiency improves to better than 80% (Figure 9). The examined feed-forward control did not improve the effluent conditions because all control scenarios fall below the effluent limit of 2 mg NH 4 /L. However, feed-forward control decreases the peaks in energy consumption caused by ammoniabased feedback. This is an advantage because the energy price takes into account the maximum electrical power uptake. Uncertainties in the simulation results are mostly due to the input data with typical diurnal and weekly variations and general model limitations. In particular, the model prediction for denitrification and enhanced biological P-removal during low oxygen concentration is of limited accuracy. Other limitations are the response times of the sensors and the actuators, which were not taken into account in this study. Low oxygen concentrations could also increase the nitrite concentration and thus inhibit the biological P-uptake. Low DO concentrations may also favor the growth of filamentous organisms and consequently can lead to bulking and or foaming. The potentially negative effects of low oxygen concentrations were therefore investigated also. Implementation. Based on a second simulation study, WWTP Werdhoelzli decided to implement the A/I process scheme including NH 4 -based control. The plant upgrade has been started but not finalized yet. WWTP Morgental. The WWTP Morgental (Figure 10) is a single-stage activated sludge plant built in 1975 for the removal of organic matter of 83,000 p.e. (31,000 m 3 /d). Since 1994, the 174 Water Environment Research, Volume 84, Number 2

6 Figure 8 Modeled ammonia and nitrate load (left) and annual costs (right) for the investigated control concepts (WWTP Werdhoelzli). plant has had to nitrify the wastewater of 42,000 p.e. (15,500 m 3 /d). During the upgrade, two reactors were assigned for predenitrification. Approximately 35,000 p.e. (13,000 m 3 /d) are currently connected to the plant. In 1999 and 2000 additional digester supernatant from a nearby WWTP was treated. The final effluent is discharged into Lake Constance. WWTP Morgental is not well suited to aeration control. The plant consists of six lanes, and each lane in the biological stage is divided into two sub-lanes. The aeration system consists of four identical root-type blowers whose output cannot be reduced below a minimum of 70%. This leads to step changes in the aeration capacity during the speed-up or shutdown of blowers. Moreover, the blower capacity cannot be throttled to the required minimum during night hours or other low loading periods. The air distribution is designed as a one-collector system. Since no stabilizing pressure control is possible (due to the step changes), the aeration of the single lanes is quite different and variable. The calculation of the aeration capacity is based on the average DO concentration of all the lanes. The valves of the single lanes are used for DO control in the lanes. The intention was to use the valves only for fine-tuning the air distribution, but in reality the control loops do not have enough control authority to reach the set-point, thus resulting in a varying oxygen input in the different lanes. Moreover, the wastewater distribution is not equalized. Lanes one to three are supplied with less particulate matter than lanes four to six. The outermost left- and right-hand lanes are also supplied with more particulate compounds. Figure 9 Total net savings compared to V0 including energy costs, effluent load tax and P-precipitation (WWTP Werdhoelzli). February

7 Rieger et al. Figure 10 WWTP Morgental with measuring locations. The waste activated sludge is withdrawn from the last aerated reactor and pumped back to the primary clarifier to be finally removed with the primary sludge. Due to several shortcomings of the primary clarifier, this leads to a significant re-seeding with activated sludge and consequently to an accumulation of inert particulate matter in the system. The plant is equipped with membrane aerators so that the aeration can be switched off completely. The DO control loop uses a DO probe in the middle of the first aerated reactor. Only one of the two sub-lanes is equipped with a DO sensor. Definition of Objectives. The main goal at WWTP Morgental was to reduce the energy consumption, since no effluent quality tax and no other limits for total nitrogen exist. Enhanced biological phosphorus removal was also outside the scope of the study. Data Quality Evaluation and Control. An extensive data quality control was performed at WWTP Morgental. It included the evaluation of all flow and concentration measurements. The complete operating program for the various measurements in the laboratory was analyzed, including of testing the micropipettes, scales etc. Finally, the mass rates were evaluated by combined mass balances (Thomann, 2008). Numerous in- and ex-situ devices were installed on the plant to monitor its processes and provide an input for the control loops. A new monitoring concept was developed on the basis of this experience (Thomann et al., 2002; Rieger et al., 2004a). In November 2001, six different ex-situ analyzers and in-situ sensors for ammonia were tested according to ISO (2003) and a field testing protocol (Rieger et al., 2005). Simulation-supported Controller Design. It could be shown in the simulation study that the WWTP Morgental has great potential for increasing the nitrogen removal and reducing the 176 energy consumption through the use of suitable measuring and control concepts. On the basis of intensive measuring campaigns of 10 and 11 days respectively, the plant model was calibrated and validated using ASM3 as the biokinetic model (Gujer et al., 1999) with a parameter set for Swiss municipal wastewater (Koch et al., 2000). The simulated control concepts take the technical and process constraints of the plant into account. Table 3 shows the investigated control concepts. All the concepts were tested with the actual blower limitation due to the step changes and with an optimized blower unit where a continuous change in capacity was possible (Figure 16). The comparison resulted in a difference of 1 to 2% reduction in energy consumption if the blower unit was optimized. The absolute cost reduction is between 350 and $1,000/yr. For the base case, a higher reduction of $1,800/yr can be attained due to insufficient control authority for throttling the aeration capacity to a minimum except with intermittent aeration. The evaluation shows that the on-off concept combined with feed-forward control on 3 mg DO/L (concept 4) gives the best results with respect to effluent concentrations and energy consumption. The feed-back part guarantees aeration according to the requirements and the feed-forward part provides an early signal for increased aeration during peak loadings. Use of a DO set-point of 3 mg/l during peak loadings guarantees that enough oxygen is available in both aerobic reactors (Figures 11, 12, and 13). The reduction potential of the energy consumption reaches $26,000/yr for the selected concept, which is 70% of the actual energy requirement. If an effluent quality tax is applied in addition, the increased nitrogen removal yields a reduction potential of $65,000/yr (Figure 14) or an increase in the nitrogen removal by 48% compared with the actual capacity (Figure 15). Water Environment Research, Volume 84, Number 2

8 Table 3 Investigated control concepts in the simulation study for WWTP Morgental. No. Control concept Parameter 0 Actual control concept DO set-point 5 2 mg/l 1 DO set-point adjustment due to the ammonia concentration If NH 4,BB, 1.5 mg/l, than DO set-point mg/l If NH 4,BB. 1.8 mg/l, than DO set-point mg/l 2 On-off of the blowers due to the ammonia concentration If NH 4,BB 5 Min., 1.5 mg/l 5. Blower 30 min off If NH 4,BB within time interval. 1.8mg/L 5. Blower on 3 Concept no. 2 plus feed-forward control Time delay FF control 5 25 min Relation NH 4 influent load/nitrification capacity (Eq. 1 and 2) X AUT 580gCSB/m 3, Volume m 3,t int 5 Part of aerated time at intermittent aeration Relation set-point aeration on Relation set-point aeration off Combination feed-back and feed-forward by maximum condition 4 Concept no. 3, but with increased DO set-point for feed-forward control DO set-point at feed-forward 5 3 mg/l For a cost-benefit calculation, the annual costs were estimated at $12,000/yr including investment and operational costs for a lifetime of 10 years with an interest rate of 5%. In relation to the energy reduction, this results in a net benefit of $14,000/yr. If an effluent quality tax is taken into account, the net benefit would increase to $53,000/yr. Implementation. The control concepts were applied in stages due to mainly technical constraints. After the first measuring campaigns and an extensive data analysis and evaluation, the starting situation could be characterized as follows: N The aeration capacity could not be reduced to the required minimum at low loading situations (the blower unit is significantly oversized after changing to fine-bubble aeration). N Dosing of digester supernatant in one pulse resulted in effluent peaks of up to 10 mg/l (or more than 70 mg/l in the influent to the biological stage). N Unequal distribution of wastewater (flow and particulate matter). N Unequal distribution of air. N DO sensors only in one of two sub-lanes. First Optimization Step. The first step was to equalize the dosing of digester supernatant and to introduce intermittent aeration on the basis of the average DO concentration in order to prevent DO saturation during the night hours. During the period of intermittent aeration and due to the unequal air and wastewater distribution certain lanes do not get enough oxygen and others run into DO saturation. In the aeration phase, the blower control loop tries to reach the DO set-point of 2 mg/l. However, the actual value for the controller is the average DO concentration of all lanes with the assumption of equal DO concentrations in all lanes. Since the DO controller of the single lanes has insufficient control authority, this leads to Figure 11 DO concentrations in the second aerated reactor for different feed-forward control concepts (WWTP Morgental). February

9 Figure 12 Ammonia concentration in the second aerated reactor for different feed-forward control concepts (WWTP Morgental). a DO deficiency in some lanes and consequently to increased ammonia concentrations in the effluent from the entire plant. A first intensive measuring campaign was carried out to obtain data for calibrating the dynamic simulation model. A tracer experiment was used to characterize the hydraulic behavior of the plant. The ammonia control concepts were also tested in one pilot lane during this phase. The plant model was validated with a second intensive measuring campaign (with a DO set-point adjustment due to the ammonia concentration). Accordingly, it was decided to implement the ammonia control at the whole plant. A simulation study was carried out to predict the reduction potential of the energy consumption and the nitrogen discharge. In the middle of this phase, the butterfly valves in lanes 1 and 2 were replaced by iris valves in order to test the effect of more suitable valves on the control authority of the DO controller. The result was sufficient control authority if the DO concentration was higher than the set-point because the influence of the other lanes predominates at DO concentrations below the set-point. Second Optimization Step. To reduce the investment costs, only three ion-sensitive ammonia sensors were installed in the lanes with the higher loads. During this phase, all the ammonia feed-back control concepts modeled in the simulation study were tested. The DO set-point adjustment was applied using an intermittent aeration concept. Due to financial restrictions, the feed-forward concepts were not applied to the plant. The Figure 13 Ammonia effluent concentration for the modeled control concepts (WWTP Morgental). 178 Water Environment Research, Volume 84, Number 2

10 Figure 14 Reduction potential of the modeled control concepts (WWTP Morgental). average of the three ammonia sensors was used as the input to the control loops. The result was that the average ammonia concentration could be kept near the set-point of 1.5 mg N/L, but the concentrations in the lanes varied strongly. Moreover, the DO concentrations in the lanes varied so greatly that one lane had DO concentrations close to zero during the intermittent aeration (Figure 17). The DO concentration also varied between the two sub-lanes due to the lack of a second DO probe and additional actuators to control both lanes. Figure 15 Comparison of the ammonia and nitrate effluent load for the modeled control concepts (WWTP Morgental). February

11 Figure 16 Reduction potential for energy consumption for the modeled control concepts (WWTP Morgental). The conclusions of this phase were: N Due to the averaging, the control loop was unable to equalize the varying ammonia concentrations N The air distribution during the intermittent phase was insufficient N The air distribution between the two sub-lanes was insufficient Third Optimization Step. In this step, the following measures were taken to overcome the problems identified: N Automatic equalization between the lanes equipped with ammonia sensors through a decrease of the DO set-point in the lanes with a higher nitrification rate. N Increase of the sludge retention time in the lanes without ammonia sensors. N Replacement of the remaining butterfly valves by wedgegate valves (lanes 3 to 6). After fine-tuning of the valve control parameters, the result was acceptable, although the air distribution during intermittent operation is still critical. Due to ongoing problems with the ammonia sensors, the benefit resulting from the measures could not be evaluated over a longer period within the schedule of this work. The last measuring campaign carried out by EAWAG followed the second optimization step. Within this period and when the sensors worked, the reduction of the energy consumption was less than the potential savings calculated in the simulation study, but in the range of 20%. At the same time, the nitrogen removal increased by approximately 40%. Control of Success. It could be shown that ammonia control leads to a significant reduction in energy consumption and in an increase of nitrogen removal even with unfavorable plant designs. The cost-benefit analysis for WWTP Morgental (based on the simulation study) results in a ratio of 0.46 between the annual costs (amortization of investment and operational costs Figure 17 DO concentrations in the different lanes during the second optimization step (WWTP Morgental). 180 Water Environment Research, Volume 84, Number 2

12 Rieger et al. Figure 18 A2O process scheme and measurement locations of WWTP Thunersee. for service, maintenance and chemicals) and the potential energy reduction. WWTP Thunersee. WWTP Thunersee is an activated sludge plant mainly treating municipal wastewater of 130,000 p.e. (40,000 m3/d) (actual and design capacity) with an industrial influent of approximately 30,000 p.e. After the upgrading of the biological stage in 1998, the plant operates according to the A2O process scheme for enhanced biological phosphorus removal (Figure 18) and consists of two completely independent lanes with separate sludge systems. Each lane is divided into two sublanes in the biological stage and is connected to four secondary clarifiers. The return activated sludge is mixed per lane before it is pumped back to the first reactors of the two sub-lanes (Figure 19). An annual effluent quality limit for total phosphorus of 0.5 mg P/L is mandated by the authorities. The plant was already able to control the aeration in each reactor in several ways on the basis of DO and ammonia. Two blower units per lane are installed and each of the blower units serves one pair of parallel reactors in the two sub-lanes (Figure 20). One blower unit is controlled by a DO control loop based on the average DO concentration of the two sub-lanes. The air distribution is also controlled by the DO concentration. The valve of the sub-lane with the higher DO concentration is throttled. The air supply to the first reactor can be reduced by intermittent aeration. The second reactor can be controlled by means of a DO set-point adjustment in five steps based on the NH4 concentration. Definition of Objectives. The goal of the study at WWTP Thunersee was to reduce the operational costs. Because of the effluent quality tax levied by the canton of Bern (see above), the main optimization goal is to increase the biological phosphorus removal and the nitrogen removal. The energy consumption also represents a significant share of the operational costs and its reduction is consequently the second goal of the study. Data Quality Evaluation and Control. At WWTP Thunersee as well, great efforts have been made to check the data quality. Unlike the case of WWTP Morgental, however, EAWAG only supplied the methods and acted as a contact in the event of questions. This approach started a whole campaign of plant analysis by the plant staff. The acceptance of their own measurements increased greatly after the campaign, although an erroneously calibrated photometer and imprecise test kits had Figure 19 Biological stage of WWTP Thunersee. February

13 Figure 20 Allocation of the blower units to the reactors in one lane (WWTP Thunersee). been detected. A photometer was consequently acquired from another manufacturer and used for the subsequent measurements. The response times of all important on-line devices were measured with special experiments. All on-line sensors were monitored using a software tool specifically developed for this purpose (Rieger et al., 2004). Simulation-supported Controller Design. The plant model based on the ASM3 with the EAWAG Bio-P module (Rieger et al., 2001) was calibrated for a data set of 10 days in November 2000 and validated based on online and lab data from November 2002 to June In March 2003 a tracer test using sodium bromide revealed insufficient mixing and back flows between the non-aerated reactors. Changes to the separation walls, closing of a short circuit, and re-direction of the internal recycle to reactor 5 were done in the test lane before the full-scale tests were started. For the controller design study the 2000 data set was used as model input but as the base of comparison the internal recycle was redirected to reactor 5 and complete mixing in all tanks was assumed. The control concepts were investigated in Table 4. After evaluation of the results, concept 6 (a 4-step controller combined with a feed-forward part, see Figure 21) was selected as the best choice with respect to effluent-quality tax, energy consumption, dosing of precipitants and sludge disposal. The controller reduces the air supply into the two reactors in steps on the basis of an ammonia measurement at the end of the last aerated tank. The feed-forward controller assures greater safety with respect to peak loading and allows the feed-back controller to be run at the optimum without additional risk factors due to the delayed signal. The annual costs arising from the need for additional measuring devices and control approaches were estimated for the cost-benefit calculation. They consist of investment and operational costs for a lifetime of 10 years with an interest rate of 5%. The reduction potential for effluent-quality tax, precipitants and precipitation sludge disposal is 40%, or $260,000/yr in absolute numbers (Figure 22). The energy reduction potential was calculated as 30% or $120,000/yr. The reduction in energy consumption alone would be much higher than the annual costs of the control system of $16,000/yr. The total resulting net revenue was calculated to be $360,000/yr. Regarding the environmental benefit, the nitrogen removal could be increased by 135 t N/yr or 60% compared to a full aeration in both reactors (Figure 23). The reduction potential for phosphorus is also high due to better denitrification (Figure 24). Precipitation was not directly modeled in the simulation study, but in the analysis all Table 4 Investigated control concepts in the simulation study for WWTP Thunersee. No. Control concept Parameter 0 Prior situation DO set-point in both reactors 5 2 mg/l (IR in fourth reactor) Internal recycle in 4 th reactor 1 Base case for study DO set-point in both reactors 5 2 mg/l (IR in fifth reactor) Internal recycle in 5 th reactor 2 Reactor aer1+aer2: DO set-point adjustment based on NH 4 aer1: NH 4,BB, 1.6 mg/l 5. DO set-point 5 0.5mg/L NH 4,BB. 1.8 mg/l 5. DO set-point 5 2mg/L aer2: NH 4,BB, 1.3 mg/l 5. DO set-point 5 0.5mg/L NH 4,BB. 1.5 mg/l 5. DO set-point 5 2mg/L 3 aer1: On-off control aer1: NH 4,BB, 1.6 mg/l 5. intermit. aeration: 30 min on/30 min off NH 4,BB. 1.8 mg/l 5. DO set-point 5 2mg/L aer2: Base case aer2: DO set-point 5 2 mg/l 4 aer1: On-off control aer1: NH 4,BB, 1.6 mg/l 5. intermit. aeration: 30 min on/30 min off NH 4,BB. 1.8 mg/l 5. DO set-point 5 2mg/L aer2: DO set-point adjustment based on NH 4 aer2: NH 4,BB, 1.3 mg/l 5. DO set-point 5 0.5mg/L NH 4,BB. 1.5 mg/l 5. DO set-point 5 2mg/L 5 4-step controller see Figure step controller combined with feed-forward control Concept 5 + feed-forward control based on NH 4 -N PC (Eq. 1 3) Time delay feed-forward control 5 15 min Relation NH 4 influent load/nitrification capacity Relation set-point aeration on aer1 5 2: DO set-point 2 mg/l Relation set-point aeration off aer : acc. to 4-step controller Relation set-point aeration on aer : DO set-point 5 2 mg/l Relation set-point aeration off aer : acc. to 4-step controller 182 Water Environment Research, Volume 84, Number 2

14 Figure 21 Scheme of the 4-step controller (WWTP Thunersee). phosphorus above 0.5 mg/l was calculated as precipitated. An increase in the phosphorus eliminated by Bio-P leads to a lower consumption of precipitants and thus to less chemical sludge to be disposed of. When the 4-step controller was used, the biological phosphorus removal increased by 50% or 12 t P/yr. One part of this reduction is caused by enhancement of the anaerobic zone achieved by returning the internal recycle into the 5th instead of the 4th reactor. Implementation. The findings of the simulation study provided the basis for discussing implementation of additional control concepts with the plant manager and the operator-inchief. It was decided to test three different control concepts against the current situation (Table 5 and Figure 25). The base case includes an increase of the anaerobic zone to four instead of three reactors. Based on the results from the tracer study, the partition walls between the non-aerated reactors were modified in order to achieve better mixing and prevent back flows. In the test phase, the first of the two lanes was used for the experiments while the second lane was operated with full aeration and served as a reference. The comparison of the experimental and reference lanes shows a reduction in energy consumption of 10% for control concept no. 1, 11% for no. 2 and 16.5% for no. 3 (Figure 26) compared with the base case (DO set-point of 2 mg DO/L in both reactors). The reason for the limited reduction is the insufficient control authority of the implemented control concepts (Figure 27) to reduce the oxygen concentrations to very low levels. The mean value of ammonia in the last aerated reactor during the test of the third control concept was 1 mg N/L, whereas a concentration of 2 mg N/L should be maintained by the ammonia controller. Nevertheless, the total nitrogen elimination was increased by 40% with the third control concept (Figure 28). The results were calculated by comparing the 24-h composite samples of the experimental and reference lanes. A reduction potential for phosphorus precipitation cannot be given because a basic precipitation took place in both lanes. The precipitant dosage is controlled manually by the operators on the basis of the on-line measurements for phosphate and the 24-h composite samples. Control of Success. The reduction in energy consumption and total nitrogen discharge achieved is less than the results obtained in the simulation study for the best scenario, but still represents a significant improvement with 16% (energy) and 40% (total nitrogen). The better results obtained in the simulation are Figure 22 Reduction potential of the modeled control concepts (WWTP Thunersee). February

15 Figure 23 Comparison of the ammonia and nitrate effluent load for the modeled control concepts (WWTP Thunersee). due to a higher reduction potential of the air supply and consequently a better control authority for reaching the ammonia set-point. The cost-benefit analysis for WWTP Thunersee results in a ratio of 0.1 between the annual costs (amortization of investment and operational costs for service and maintenance) and the reduction in energy consumption and effluent-quality tax for nitrogen. Practical Aspects of Aeration Reduction in WWTPs The control of aeration in an activated sludge system by ammonia can lead to a significant reduction of the total energy Figure 24 Phosphorus effluent load for the modeled control concepts without chemical P-precipitation (WWTP Thunersee). 184 Water Environment Research, Volume 84, Number 2

16 Table 5 Implemented control concept and duration of the experiments in WWTP Thunersee (schemes in Figure 25). No. Control concept Duration of experiment 1 aer1: Intermittent aeration Nov Feb aer2: DO set-point 5 2 mg/l 2 aer1: Intermittent aeration (2 states) Feb March 2003 aer2: DO set-point 5 2 mg/l 0 Base case (DO set-point 2 mg/l in both reactors) March 2003 April 2003 and July aer1: Aeration based on blower capacity of aer2 June 2003 aer2: DO set-point adjustment based on NH 4, BB consumption in conjunction with decreased effluent concentrations of total nitrogen. However, this is only one part of the cost calculation. The need for reliable and accurate sensor values as an input for the control concepts requires more monitoring and service operations to be performed. Moreover, there are concerns that low oxygen concentrations can lead to serious problems in plant operation. During several measuring campaigns at WWTP Thunersee, the practical and cost aspects of a significant reduction of aeration were evaluated. Hypothesis. Two hypotheses for the effects of low oxygen concentrations were formulated and examined: N Low oxygen concentrations lead to increased growth of filamentous bacteria resulting in the production of foam and scum. N Low oxygen concentrations lead to higher concentrations of nitrite in the effluent. In addition, a long-term evaluation should evaluate practical problems occurring as a result of greatly reduced oxygen concentrations. Methods. The first hypothesis was evaluated using fluorescence in-situ hybridization (FISH) to measure the abundance of M. parvicella and nocardioform actinomycetes. A rapid quantification method was applied which was suitable for practical use in wastewater treatment systems (Hug et al., 2005). In addition, the foam coverage of the reactors was monitored and the foaming potential evaluated using a batch test method developed by Ho and Jenkins (1991). The accumulation of nitrite was monitored with a submerged spectrometer sensor (Rieger et al., 2004b). Figure 25 Schemes of the tested control concepts for WWTP Thunersee. Maximum blower capacity is 300%, 100% means that one of three blowers operates at maximum output. February

17 Figure 26 Measured reduction in energy consumption (WWTP Thunersee). First Results and Discussion. Hypothesis 1. Microthrix parvicella always occurred at very high concentrations although it decreased during one time period. This trend was identical in both independent lanes (no sludge mixed) and is consistent with the commonly observed seasonal variations of this microorganism. Hence, the reduction does not seem to be caused by the changed conditions due to the control strategies implemented. With regards to nocardioform actinomycetes, none of the typical branched filaments were found, but rather non-filamentous types that did not vary over time. The variations of the foam coverage and foaming potential could be linked to the type and amount of the P-precipitants (FeClSO 4, AlCl 3 ) but not to the implemented control strategies. Hypothesis 2. The nitrite measurements (Figure 29) showed increased concentrations if the aeration is reduced only in the first aerated reactor. When a reduction was implemented in both reactors, the concentrations increased at first, but decreased to normal values of less than 0.1 mg/l after one week. The differences between laboratory and sensor values can be related to ongoing reactions in the 24 h-composite sample. Although the concentrations in the reference lane also increased, the nitrite concentrations in the pilot lane were higher. Long-term Evaluations. Experience from these and other sensor implementations and evaluations showed that the maintenance and monitoring outlay using advanced aeration control plays a major part in the cost calculations. Depending on the measuring device, this outlay is between 0.5 and 3 h/week and device on average. This means that at most plants the staffing requirements will account for the main part of the annual costs. At WWTP Thunersee, a significant part of the cost reduction due to lower average aeration was lost again due to short peaks of high energy consumption, since the energy price structure for this plant takes into account the maximum electrical power consumption in a quarter of an hour. After three years of applying the control strategies at WWTP Morgental, corrosion problems of the concrete were detected at the end of the second aerated reactor. This was due to a too excessive manual reduction of the airflow in the last sector of Figure 27 Ammonia concentration during the test of the third control concept (WWTP Thunersee). 186 Water Environment Research, Volume 84, Number 2