Evaluation of Different Nitrogen Control Strategies for a Combined Pre- and Post-Denitrification Plant

Transcription

1 Proceedings of e 7 World Congress The International Federation of Automatic Control Evaluation of Different Nitrogen Control Strategies for a Combined Pre- and Post-Denitrification Plant A. Stare, N. Hvala, D. Vrečko and S. Strmčnik Department of Systems and Control, Jožef Stefan Institute, Jamova 39, SI- Ljubljana, Slovenia ( aljaz.stare@ijs.si; nadja.hvala@ijs.si; darko.vrecko@ijs.si; stanko.strmčnik@ijs.si) Abstract: In e paper different nitrogen control strategies are proposed and tested on a simulation model of a combined pre- and post-denitrification plant. The plant configuration corresponds to e Domžale- Kamnik wastewater treatment plant at will be upgraded for nitrogen removal using MBBR (moving bed biofilm reactor) technology. The aim of e study is to find an optimal control strategy in terms of required effluent quality and operating (i.e. carbon and aeration) costs. The tested control strategies address aeration control, internal recirculation control, and external carbon dosage control, and are based on PI and feedforward control algorims. Simulation results indicate at e nitrate PI controller at manipulates external carbon flow-rate and e ammonia PI controller at manipulates oxygen concentration in e aerobic reactors give e best performance wi respect to e effluent quality and operating costs. In addition, it was shown at e control auority of e internal recirculation flow-rate was raer limited as e internal recycle flow-rate on e real plant can be increased only up to 2% of e average influent flow-rate. Hence, no improvement of effluent quality could be achieved wi internal recycle control. While e improvement of effluent quality wi e proposed overall control scheme is small if compared to e basic control scheme wi optimal set-points, e energy savings are quite significant reaching up to 4%.. INTRODUCTION Because of e introduction of stricter legislation for nitrogen removal many wastewater treatment plants (WWTPs), which were primarily built to remove organic matter, need to be upgraded to remove also nitrogen and phosphorus compounds from wastewater. This is also e case in Domžale-Kamnik WWTP, Slovenia. Based on an extensive set of real-plant experiments and a simulation study (Hvala et al., 22) e moving bed biofilm reactor (MBBR) technology was chosen for plant upgrading raer an a conventional activated sludge process. Because ere is a very strict total nitrogen requirement in e effluent, a combined pre- and post-denitrification plant was proposed. The aim of is paper is to propose e control system for e upgraded plant so at not only effluent requirements are met but also e operating costs are minimised. The control strategies considered are based on on-line nutrient (ammonia, nitrate) measurements, which are planned to be installed on e upgraded plant for monitoring and control purposes to improve e dynamic operation of nitrification and denitrification processes. The control loops considered address aeration control, nitrate recirculation control and external carbon dosage control at have been already extensively studied in oer papers (Yuan et al., 22; Stare et al., 27). As e more advanced model-based predictive control concepts did not prove to give a significant improvement of plant operation (Stare et al., 27), only more simple feedforward and PI control algorims were considered in e study. The main contribution of e paper is e evaluation of e control system as a whole wi respect to bo quality and energy costs while considering different alternatives of e chosen control variables and e chosen control algorim. This paper is organized as follows. In e following section, e process configuration is presented. Then, e applied control strategies are described. Next, e simulation analyses are shown. At e end some conclusions are drawn. 2. PROCESS CONFIGURATION The upgraded plant will be built for nitrogen removal and consists of ree identical parallel lines wi seven reactors in each line (Fig. ). A combination of pre- and postdenitrification is used because of strict effluent requirement on nitrogen removal. The effluent total nitrogen (TN) and ammonia nitrogen ( ) concentration should be below mg/l and 3mg/l, respectively. The chosen technology is MBBR, i.e. reactors filled wi small free floating plastic carriers on which a fixed biofilm is formed. From Fig. it can be seen at e first two reactors are anoxic reactors where nitrate is converted to atmospheric nitrogen wi e organic matter in e influent as a carbon source. Nitrification takes place in aerobic reactors 3 and 4. The reactor can be also aerated, but its main purpose is to reduce e dissolved oxygen concentration in e water at is recycled back to e inlet or flows forward to e postdenitrification reactor. The post-denitrification takes place in e reactor where an external carbon source (in our case meanol) can be added to reduce e nitrate concentration /8/$2. 28 IFAC /287--KR-.2792

2 The external carbon at is not removed in e reactor is degraded in e last aerobic reactor. After e biological stage e water enters e separation stage, where flocculation and flotation are used for particle separation. influent AN AN AE AE AE AN 27m 3 243m 3 282m 3 279m m 3 internal recycle Fig.. Scheme of e MBBR plant one line. 2. Simulation model m 3 AE 89 m 3 effluent to e separation stage The performance of e control strategies was examined by GPS-X simulation software (Hydromantis, 2). To simulate e MBBR process a Hybrid-System model was used, which combines standard plug-flow tank configuration wi suspended grow biomass, and e biofilm model representing fixed film grow on e media inserted into e tank. In e model, e reactor contents is represented by layers, e first layer representing e bulk liquid, while e remaining five layers represent e biofilm formed on e carriers. In each layer, all e process components are subject to biological reactions and are modelled wi e Mantis model, which is similar to e well known Activated sludge model no. (ASM) (Henze et al., 2), except wi some minor modifications (Hydromantis, 2). The default kinetic and stoichiometric GPS-X parameters were used in our study, providing a satisfactory agreement between simulated and real plant dynamic data as measured on e MBBR pilot plant. 2.2 Influent data Simulation analyses were performed based on real plant influent data as measured in e plant from In ese analyses only influent flow-rate (Fig. 2), ammonia concentration (Fig. 3) and readily biodegradable substrate concentration (Fig. 4) where changing dynamically. All oer components were set constant and defined according to prior wastewater characterization (Hvala et al., 22). For influent flow-rate (Q in ) e actual measurements were used, only e average flow-rate was increased from approximately (as measured in e plant) to 8333m 3 /day, which was e value used in plant design for one line. The influent ammonia concentration was measured, while influent readily biodegradable substrate concentration was computed from influent total organic carbon measurements. The given data is a good representative of real plant influent conditions wi typical daily ammonia variations, and periods of variable influent carbon source at are typical for e plant (low from day -4, higher from day - in Fig. 4). Simulated at different temperatures, e data is a good representative for e whole year operation. In our case, e analyses were performed at C, which is approximately e average temperature of e wastewater over e year. Flow rate (m 3 /day) Fig. 2. Influent flow-rate. Ammonia nitrogen Fig. 3. Influent ammonia nitrogen. Readily biodegradable substrate Fig. 4. Influent readily biodegradable substrate. 2.3 Evaluation criteria In is study, average energy costs (i.e. aeration) and external carbon dosage costs were used to evaluate e control strategies. The operating costs (OC) were defined as follows: OC = AC + CC, () where AC means average aeration costs ( /day), while CC means average external carbon costs ( /day). The electric energy ( /day) required for aeration of aerobic reactors is calculated using a pre-defined formula inside e GPS-X simulation software (Hydromantis, 2): AC = E T ( t) Q air head ρ H 2O dt, (2) 8. 4 η price t T p = t = 7 p pump where E price is e electricity price, T p is e simulation period, Q air is e airflow rate, head is e hydraulic head, ρ H2O is e density of wastewater and η pump is e pumping efficiency. In our case e following values of parameters 3

3 were used: head = 4. m, ρ H2O around kg/m 3 (depends on e temperature) and η pump =.7. For e E price, an average electricity price in Slovenia (. /kwh) was used. The average external carbon costs ( /day) are calculated as e average external carbon mass flow (kg/day) multiplied by e price of e carbon source ( /kg): C COD t T price S p CC = Qcarb ( t) dt, T = (3) t = p where COD S is e carbon source concentration (.8 mg/l), Q carb is e carbon flow-rate, T p is e simulation period, while C price is e price of carbon (meanol) source. The price of e meanol solution at was used in our case was set to.7 /kg, which is an estimate of e meanol price in Slovenia. Additionally, e effluent quality was also considered for e evaluation of control performance. Effluent requirements for is plant are to achieve effluent TN and chemical oxygen demand (COD) below mg/l and 7mg/l, respectively. However, as e separation stage (flocculation and flotation) was not modelled, only e soluble total nitrogen ( ) concentration (i.e. e sum of ammonia, nitrate, and soluble organic nitrogen concentration) in e last reactor was used to evaluate control strategies. Beside is also ammonia nitrogen and readily biodegradable substrate concentrations in e last reactor were considered in e evaluation of control strategies. 3. CONTROL STRATEGIES Different aeration, internal recirculation flow-rate, and external carbon dosage control strategies were proposed and compared on e simulated upgraded MBBR plant. For e internal recirculation flow-rate control, a nitrate PI controller can be used. In is strategy, e nitrate concentration in e 2 nd reactor is controlled at a desired set-point by manipulating e internal recycle flow-rate (Yuan et al., 22). This control strategy maximises e usage of influent organic matter for denitrification, but has a limited effect on effluent nitrate concentration, as e effluent nitrate concentration will vary wi e influent flow-rate and composition, in particular wi e influent carbon to nitrogen ratio (C/N). Because e internal recycle flow-rate on e real plant can be increased only up to e double value of e average influent flow-rate, e control auority of e internal recirculation flow-rate was limited. In fact, no improvements of nitrogen removal could be gained by using a PI nitrate controller as e internal recycle flow-rate is most of e time at its maximum value (7m 3 /day). Therefore, a constant (maximum) internal recycle flow rate was applied in all control strategies. The following control strategies were tested: (I) The basic control strategy. In is case, a constant internal recycle flow-rate, a constant carbon dosing in e reactor and PI oxygen control were used. In e latter case, e oxygen concentrations in e aerobic reactors are controlled at desired set-points by manipulating e air flow-rates. The oxygen set-point values for e 3 rd, 4,, and 7 reactor were set to 2mg/l, 3mg/l, mg/l, and 3mg/l, respectively. It had previously been determined at satisfactory ammonia nitrogen removal could be achieved even in e case when reactor was not aerated. The chosen carbon flow-rate was.m 3 /day. (II) External carbon dosage control. In is strategy e carbon dosing is performed in one of e two ways described below, while e internal recirculation flow-rate and oxygen control are e same as in e basic control strategy. The two different control strategies for carbon dosing are: (a) Feedforward (FF) control of e external carbon flow-rate in e reactor, where carbon flow is proportional to e influent flow-rate Q in : Q carb = kqin n, (4) The feedforward controller parameters k and n were determined by trial and error. Proportional factor k was set to -4 and n was set to.8m 3 /day. To prevent external carbon from being unnecessarily dosed during high influent flowrates, e Q carb was limited between and.7m 3 /day. (b) PI control of e nitrate concentration in e reactor by manipulating external carbon addition in e reactor (Fig. ). A natural strategy is to control e nitrate concentration at e end of e anoxic reactor at a low set-point (Yuan and Keller, 23). In our study, e nitrate-set point was set to 3mg/l, while Q carb was limited between and.7m 3 /day. Nitrate set-point + _ External carbon flow-rate Nitrate PI controller WWTP Fig.. Control scheme for nitrate PI control. Nitrate concentration in e reactor (III) Ammonia control. In is strategy, e concentration in e reactor was controlled wi a cascade PI controller (Lindberg and Carlsson, 99). In e cascade control (Fig. ), e outer (ammonia) controller adjusts e oxygen setpoint values in e 3 rd and 4 reactor based on desired and actual value, and e two inner (oxygen) controllers manipulate e air flow-rate values based on desired and actual dissolved oxygen concentrations (Stare et al., 27). The ammonia set-point for e reactor was set to mg/l. The oxygen concentrations in e and 7 reactor were controlled as in e basic control strategy wi DO set-points set to mg/l and 3mg/l, respectively. Internal recycle flowrate and carbon dosing were e same as in e basic control strategy. Ammonia set-point + _ Oxygen set-point for e 3 rd and 4 reactor Ammonia PI controller + _ Air flow-rate in e 3 rd and 4 reactor Oxygen PI controller WWTP Fig.. Control scheme for ammonia PI control. 4. EVALUATION OF CONTROL ALGORITHMS 4. External carbon dosage control Ammonia in e reactor Oxygen in e rd 3 and 4 reactor Simulation results of e external carbon control strategies are shown in Fig. 7 and Fig. 8. From Fig. 8 it can be seen at wi FF flow-rate control better (soluble) total nitrogen 3

4 removal was achieved in e period from day 3 to 8 compared to e basic control strategy, while e average (soluble) total nitrogen removal over e whole period was similar (Table ). In addition, more external carbon was dosed in FF control (Table 2). Besides, e parameters of e FF controller (4) have to be modified wi changing operating conditions. For example, at higher temperatures (2 C) a satisfactory removal of e nitrate can be achieved wiout using external carbon dosage. This means at if e parameters of e FF controller are not changed during higher temperatures e carbon would be unnecessarily dosed, while at lower temperatures (e.g. C) too little carbon would be dosed. External carbon flow rate (m 3 /d) Feedforward control Nitrate PI control Fig. 7. External carbon flow-rate for different control strategies. Soluble total nitrogen concentration in reactor Feedforward control Nitrate PI control Fig. 8. Soluble total nitrogen concentration in e 7 reactor This problem can be solved by using e nitrate PI controller, which adds carbon only when e nitrate concentration in e reactor is high. When e nitrate concentration is low, e external carbon is not added. From Fig. 8 it can be seen at by using e nitrate PI controller a better removal of (soluble) total nitrogen is achieved in e period from day to 4. In fact, over e entire simulation period e soluble total nitrogen concentration was kept below 9mg/l. However, e average (soluble) total nitrogen concentration was a little higher when using e nitrate PI controller (Table ) as less carbon was dosed (Table 2). Table. Average concentrations in e last reactor FF control Nitrate PI control Table 2. Carbon and aeration costs Feedforward control Nitrate PI control Ammonia control In Fig. 9, Fig., and Fig. e ammonia control strategy results are compared wi e basic control strategy. It can be seen at e ammonia controller increases DO concentrations during high load periods (see Fig. 9) to enhance nitrification and to successfully remove ammonia nitrogen (Fig. ), while during low loads it decreases e DO concentrations to save aeration energy. Wi e introduction of e ammonia controller a considerable reduction (about 3%) in aeration costs was achieved (Table 4) because of variable DO setpoint. Oxygen concentration in reactor Ammonia PI control Fig. 9. Oxygen concentration in e 4 reactor. Soluble total nitrogen concentration in reactor Ammonia PI control Fig.. Soluble total nitrogen concentration in e 7 reactor. Ammonia nitrogen concentration in reactor Ammonia PI control Fig.. Ammonia nitrogen concentration in e 7 reactor. In Fig. and Table 3 it can be seen at e ammonia controller does not remove more nitrogen an in e basic approach. This happens because e DO set-points in e basic control strategy were set relatively high. Wi lower 32

5 DO set-points in e basic control strategy (e.g. 2,mg/l, which is e average of ammonia controller) e energy consumption in bo cases would be similar, but wi considerably higher effluent ammonia peaks and higher average ammonia concentration in e basic control strategy. Table 3. Average concentrations in e last reactor Ammonia control Table 4. Carbon and aeration costs Ammonia control Overall control In Figures 2 to 4, e performance of e basic control is compared wi e overall control strategy. For e overall control strategy, e PI ammonia controller and e nitrate PI controller were used. As seen in Fig. 3, a better removal of was achieved wi e overall control in e period from day to 4, while in e period from day - a poorer removal was achieved as no external carbon was added (Fig. 2). Ammonia removal in Fig. 4 was slightly better in overall control during e whole simulated period. From Table it can be seen at similar effluent quality was obtained in bo cases, but wi considerable reduction of external carbon dosage costs and aeration costs in e overall control (Table ). In fact, e operating costs (e sum of CC and AC) decreased by around 4% in e overall control. External carbon flow rate (m 3 /d) Overall control Fig. 2. External carbon flow-rate for different control strategies. Soluble total nitrogen concentration in reactor Overall control Fig. 3. Soluble total nitrogen concentration in e 7 reactor Ammonia nitrogen concentration in reactor Overall control Fig. 4. Ammonia nitrogen concentration in e 7 reactor Table. Average concentrations in e last reactor Overall control Table. Carbon and aeration costs Overall control Evaluation of control strategies at low temperatures The basic and overall control strategies were also evaluated at a wastewater temperature of C, which is e lowest temperature at which e required effluent quality should be met but is most difficult to achieve. Due to a lower wastewater temperature e oxygen set-point values for e 3 rd, 4,, and 7 reactor were set to 3mg/l, 4mg/l, mg/l, and 3mg/l, respectively. The chosen carbon flow-rate was increased to.3m 3 /day, while e ammonia and nitrate nitrogen set-points were set to 2mg/l and 3mg/l, respectively. The simulation results indicate at problems related to nitrate removal can be expected during low influent COD concentration and low temperature. Namely, during such periods e nitrate removal in e pre-denitrification reactors ( st and 2 nd reactor) is limited. For is reason, e denitrification process is active only in e postdenitrification reactor ( reactor) where e readily biodegradable substrate from e external carbon source is available. However, in post-denitrification not all e nitrate could be removed because e reactor is too small. Hence, e PI nitrate controller increases external carbon flow-rate to its maximum value (Fig. ) because of high nitrate concentration, but is consequently leads to a high readily biodegradable substrate concentration in e effluent (Fig. ). It has turned out at overall effluent quality results can be improved by using a PI controller at adds carbon to e reactor. In is way, readily biodegradable substrate in all anoxic reactors is available, which improves e removal of nitrate and total nitrogen. From Table 7 it can be seen at e nitrate, readily biodegradable substrate, and (soluble) total nitrogen removal were improved in spite of lower external carbon addition (Table 8). It should be mentioned, however, 33

6 at is control strategy can be applied only in e case when e oxygen concentration in e reactor is low or zero, oerwise oxygen is used for carbon removal, which leads to considerably higher operating costs. To solve is problem, carbon should be added separately to e st and reactor. External carbon flow rate (m 3 /d) Overall control (dosing in reactor ) Overall control (dosing in reactor ) Fig.. External carbon flow-rate for different control strategy. Readily biodegradable substrate in reactor Overall control (dosing in reactor ) Overall control (dosing in reactor ) Fig.. Soluble total nitrogen concentration in e 7 reactor Table 7. Average concentrations in e last reactor Overall control * Overall control ** * Carbon is dosed in e reactor, ** Carbon is dosed in e reactor Table 8. Carbon and aeration costs Overall control * Overall control ** * Carbon is dosed in e reactor, ** Carbon is dosed in e reactor. CONCLUSIONS In is paper, several control alternatives addressing nitrogen removal in a combined pre- and post-denitrification plant were evaluated using a simulation model. Control strategies for external carbon dosing, internal recycle control and e ammonia control were proposed and compared wi respect to effluent quality and operating (i.e. carbon and aeration) costs. The presented simulations clearly indicate at e nitrate PI controller at manipulates external carbon flowrate in e reactor and ammonia PI controller at manipulates oxygen concentration in e aerobic reactors give only slightly better effluent quality if compared to e basic control scheme wi optimally selected set-points, while e energy savings are quite significant reaching around 4%. It was also shown at e control auority of e internal recirculation flow-rate was raer limited as e internal recycle flow-rate on e real plant could be increased only up to 2% of e average influent flow-rate. Hence, no improvements of effluent quality could be achieved wi internal recycle control. The simulation results also indicate at problems related to nitrate removal can be expected during low influent COD concentration and low wastewater temperature. In is case, e nitrate removal in e anoxic reactors of e predenitrification stage is limited. Consequently, e denitrification process is active only in e anoxic reactor of e post-denitrification stage, which is too small for all nitrate to be removed. For is reason, e external carbon should be dosed in e instead in e reactor so at readily biodegradable substrate is available in all anoxic reactors (i.e. in e pre- and post-denitrification stages). In is way, e removal of nitrate and total nitrogen can be improved, while also reducing e operating costs considerably. ACKNOWLEDGEMENT The auors would like to ank to Benchmark group in COST 24 and IWA Task Group on Benchmarking of Control Strategies for WWTPs from where e work was inspired. REFERENCES Henze, M., W. Gujer, T. Mino and M. van Loosdrecht (2). Activated Sludge Models ASM, ASM2, ASM2d and ASM3. IWA Publishing, London, England. Hvala, N., D. Vrečko, O. Burica, M. Stražar, and M. Levstek (22). Simulation study supporting wastewater treatment plant upgrading. Water Science and Technology, 4(4-), Hydromantis (2). GPS-X Technical Reference, GPS-X Version 4. Ontario, Canada. Lindberg C.-F. and B. Carlsson (99). Nonlinear and setpoint control of e dissolved oxygen concentration in an activated sludge process. Water Science and Technology, 34(3-4), Olsson, G. and B. Newell (999). Wastewater Treatment Systems, Modelling, Diagnosis and Control. IWA Publishing, London. Stare, A., D. Vrečko, N. Hvala and S. Strmčnik (27). Comparison of control strategies for nitrogen removal in an activated sludge process in terms of operating costs: A simulation study. Water research, 4(9), Yuan, Z. and J. Keller (23). Integrated control of nitrate recirculation and external carbon addition in a predenitrification systems. Water Science and Technology, 48(-2), Yuan, Z., A. Oehmen and P. Ingildsen (22). Control of nitrate recirculation flow in predenitrification systems. Water Science and Technology, 4(4-),