Modelling the WWTP of Nîmes and Validating the Ammonair Control Algorithm to Ensure Low Energy Consumption and N 2 O Emissions A. Filali 1, E. U. Remigi 2, L. Philippe 3, F. Nauleau 3, S. De Grande 2, F. Claeys 2, S. Gillot 4 1 Irstea, UR HBAN, 1 Rue Pierre-Gilles de Gennes, CS 13, F-92761 Antony Cedex, France 2 DHI, MIKE Powered by DHI, Merelbeke, Belgium 3 SAUR, 1 rue Antoine Lavoisier, 7864 Saint Quentin en Yvelines, France 4 Irstea, UR MALY, 5 rue de la Doua, CS 777, F-69626 Villeurbanne Cedex, France Full-scale; modelling; nitrogen removal; nitrous oxide N 2O; energy control. Background and objective In activated sludge systems, aeration provides the oxygen that is required by the aerobic micro-organisms; ensures mixing and homogenization of the liquor; and facilitates stripping the gaseous by-products of the degradation processes. On the other hand, aeration is generally the single largest contributor to energy consumption in wastewater facilities. With the increasing need for containing operating costs, new aeration control strategies have recently been proposed. Solutions based on the continuous monitoring of nitrogen forms (NH 4 +, NO 3 - ) for instance ensure a sufficient air supply to treat the nitrogen load while maintaining relatively low dissolved oxygen concentrations in the basin which in turns translates into lower energy consumption. Whether such strategies have an impact on nitrous oxide (N 2 O) emissions is yet to be ascertained. Nitrous oxide is a key greenhouse gas, about 3 times more effective than carbon dioxide, and a major sink for stratospheric ozone (IPCC, 27). The wastewater treatment plant of Nîmes (23 PE) located in France consists of two parallel activated sludge lines operated with different aeration strategies. Ammonair, an aeration control logic based on ammonia and DO concentration, was implemented to reduce energy consumption of one treatment line. This work combines field measurements and mathematical modelling and is aimed at investigating the impact of the Ammonair control system on nutrient removal and energy consumption. The model developed was also used to assess the potential for GHGs emissions in relation to the specific aeration regime. Materials and methods The model of the plant was set up in WEST (http://www.mikepoweredbydhi.com; Vanhooren et al., 23) using its standard features with the exception of the Ammonair control logic that was implemented ad-hoc in the MSL modelling language. The algorithm is as follows: aeration is turned on, as soon as ammonia exceeds a given threshold (1.5 mgn/l) or when anoxic conditions have lasted for an extended period of time (1 min); while aerating, if ammonia is within a given range and declining, and as long as dissolved oxygen does not fall below.1 mg/l, aeration intensity may be reduced; if ammonia exceeds a given upper limit (2 mgn/l), aeration intensity is increased by increasing the blower frequency.
The Activated Sludge Model for GHG No.2 (ASMG2d) was used for simulating N 2 O production through both autotrophic and heterotrophic pathways as well as the nutrient (N and P) removal (Guo, 214). Most of the data needed for modelling (e.g. flows, MLSS concentration, airflow, blower frequency, etc.) were provided by the plant operator. A COD on-line sensor placed at the reactor inlet captured the dynamic of influent composition. Influent TKN and NH4 concentrations were estimated based on COD sensor data using average COD to TKN and COD to NH4-N ratios. COD fractionation was estimated using both ultimate BOD-test and physicochemical methods. Sensors data (NH + - 4 /NO 3 and oxygen) used for aeration control by Ammonair controller were used to evaluate the developed model. Off-gas oxygen transfer measurements (ASCE, 1996), carried out at different locations of the aerated zone of the reactor, enabled establishing a relationship between the oxygen transfer rate and the blower frequency that was used to model aeration in WEST. N 2 O emissions were measured over a period of 24 h in one position of the reactor (Filali et al. 213). Figure 1 shows the layout of the simulated treatment line. The ASMG2d parameters were first calibrated using the estimated K L a based; then, the Ammonair control logic was integrated to predict the blower frequency based on ammonia and DO concentrations. The default energy model of WEST 214 was used for the prediction of energy consumption. Since there is currently no consensus on the modelling of N 2 O production and consumption pathways and no default parameter sets provided, no calibration effort was spent and default parameters of Guo (214) were used. Calibration was only performed on conventional nitrogen removal pathways. The study was complemented with the development of a custom graphical user interface (GUI) for operator s training and decision support, on top of the WEST model of the plant. A first approach would be to set up a custom dashboard within WEST itself, using the graphical widget toolbox that the latter provides. The second would concern developing a fully customised user interface using a general-purpose programming language (typically C#) on top of WESTforAUTOMATION (WfA), which is a software development kit (SDK) that provides access to the WEST engine (Claeys, 28). For the Nîmes WWTP, a slightly different approach was followed, as next to the WfA SDK also a number of graphical modules from the WEST source code base where re-used, in order to speed up the development process, however still maintaining full flexibility. For manipulating parameters the Nîmes dashboard consists of sliders and input fields. In terms of output it has gauges, time series plots and tables. A clickable photo of the plant is used to select the process unit for which data output is to be visualized. Finally, the application has the ability of archive and replay scenarios.
Figure 1 WEST layout of the WWTP of Nimes Results and discussion 1. Model validation During the period chosen for model work (almost a month), the daily average NH 4 and NO 3 concentration in the effluent was of 1.3 and 1.5 mg/l, respectively. DO concentration in the bulk liquid was generally below.5 mg/l, except for a limited period of time when a spike of air was used to unclog the membranes (Figure 2 C). This event has not been included in the Ammonair algorithm, which is why it is not described by the model. The developed model managed to describe nitrogen performances and daily average oxygen concentration (Table 1 and Figure 2). With respect to the oxygen and nitrogen dynamics, the model does not seem to be capable of fitting the sensors signals well. This may be due to the influent model not reflecting the increase in nitrogen load which occurred at certain times of the day, because the nitrogen load was estimated using the dynamic influent load and COD concentration but assuming a fixed COD-to-TKN ratio. Since the concentration of ammonium-oxidizing bacteria was stable during the studied period and the average oxygen concentration is well described by the model, most likely the influent ammonium load was underestimated in those days. In this respect, the introduction of a dynamics in the COD-to-TKN ratio (rather than a fixed average value) seems more appropriate to accurately describe the behavior of the Ammonair controller. Table 1 Process parameters and simulation results over the simulation period Aeration tank 1 Measurement Model Influent flow (m 3 /d) 13882 ± 847 COD influent (mg/l) 731 ± 61 TKN influent (mgn/l) 61.3 ± 8.3 NH 4 effluent (mgn/l) 1.3 ±.3 1.1 ±.1 NO 3 effluent (mgn/l) 1.5 ±.2 1.6 ±.2 Figure 3 presents typical N 2 O emission pattern based on modelling result. We can see that most of the N 2 O emissions occur during the aerated phases which can be explained by: (i) the enhanced N 2 O stripping associated with aeration and (ii) the production of N 2 O by autotrophic ammonia-oxidizing bacteria through AOB denitrification pathway. This result is in accordance with several studies indicating that the contribution of AOB to N 2 O production could be significantly higher than that of heterotrophic bacteria (Wunderlin et al. 212). The N 2 O peaks that appear immediately after aeration stops are due to an unbalanced production and consumption rate of N 2 O by heterotrophs. On the other hand, during aerated phases heterotrophic organisms consume part of the N 2 O produced by AOB.
N2O emissions (g/d) kla (d-1) The average N 2 O flux over the entire modelling period is of 53 g/d, corresponding to an emission factor of.65% of incoming nitrogen. The simulated N 2 O flux is 15 times higher than the one measured on this specific site (Filali et al. 213). However, one should consider that without any calibration of the N 2 O pathways and using the default parameter of Guo (214), the emission factor predicted by the model is in the order of magnitude of the lowest recent literature values [.116%-.36% influent nitrogen load] (Rodriguez-Caballero et al. 214, Aboobakar et al. 213). 1 2 N2O kla 9 18 8 16 7 14 6 12 5 1 4 8 3 6 2 4 1 2 2. 2.2 2.4 2.6 2.8 3. 3.2 3.4 3.6 3.8 4. Time (day) Figure 3 Example of N 2O emission pattern A
B C Spike of air Figure 2 Dynamic of NH 4 (A), NO 3 (B) and O 2 (C). For more visibility NH 4 and O 2 concentrations are presented for a limited period whereas NO 3 concentration is given for the entire simulation period. Arrows in (A) indicates periods where increase of NH4 concentration was not described by the model 2. Impact of Ammonair control parameters The model was used to assess the impact of aeration regime on nitrogen removal and the potential for GHGs emissions. Ammonair settings were varied in a range coherent with discharge limits of the studied WWTP (daily average effluent ammonium concentration lower than 2 mgn/l). A first set of simulations indicated that changing only NH4_High had no significant impact on nitrogen removal since when aeration started (at 1.5 mgn/l), nitrification rate was high enough to maintain ammonium concentration below 2 mgn/l. However, simulations indicated that both NH4_Low and NH4_Aeration on have a significant impact on nitrogen removal and energy consumption. Indeed, based on these results, it was decided to change all parameters in the same way to remain coherent with the Ammonair algorithm (Table 2). Results are presented in Figure 4. Table 2. Scenario analysis Variation NH4_High (mgn/l) NH4_Aeration on (mgn/l) NH4_Low (mgn/l)
Default settings % 2. 1.5.8 Simulation #1 +5% 3. 2.25 1.2 Simulation #2 +2% 2.4 1.8.96 Simulation #3 +1% 2.2 1.65.88 Simulation #4-1% 1.8 1.35.72 3.5 3. A 3 6 3 5 B 2.5 3 4 2. NH4 (mg/l) 3 3 1.5 NO3 (mg/l) 3 2 ENG (kwh) 1. TN (mg/l) 3 1.5 3. 1 2 3 Default settings 4 2 9 1 2 3 Default settings 4 515 51 55 C 5 495 49 485 N2O (g/d) 48 475 47 1 2 3 Default settings 4 Figure 4 Impact of Ammonair controller parameters on (A) nitrogen removal, (B) energy for aeration and (C) N 2O emissions Increasing effluent ammonium concentration with Ammonair settings (from simulation # 4 to 1, Fig 4 A) decreases ammonium removal rate from 64 kgn/d to 61 kgn/d which have an impact on nitrate and total nitrogen concentration in the effluent. The reduction of ammonium removal is also reflected by a reduction in oxygen consumption for aeration and energy (Fig 4 B). N 2 O emissions are higher in simulation 4 compared to 1 (Figure 4 C). However emission factor remained the same in both simulations (.8% of NH4 oxidized ). The increase in N 2 O emissions in simulation 4 is explained by a lower consumption of N 2 O by heterotrophs compared its production by ammonium-oxidizing bacteria whereas the opposite was observed in simulation 1 (see Figure 5).
Net N2O production rate by AOB (g/d) Net N2O production rate by heterotrophs (g/d) 4 3 5 Net production rate_simulation 1 3 Net production rate_ simulation 4 2 5 2 1 5 1 5 2. 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3. Time (day) 5 4 Net production rate_simulation 1 3 Net production rate_ simulation 4 2 1 2. -1 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3. -2-3 -4-5 Time (day) Figure 5 Comparison of N 2O production rates by AOB (left) and heterotrophs (right) for simulations 1 and 4 Conclusion In this study, activated sludge model for GHG No.2, calibrated on conventional nitrogen pathways and using default parameters for N 2 O pathways, was applied to a full-scale WWTP. Ammonair control logic was implemented in order to simulate aeration control. Results indicated that the model managed to describe nitrogen performances and daily average oxygen concentration. However, their dynamics were not always well described due to the simplifications adopted for the influent model. Since the primary objective of this study was to gather a better understanding of the impact of the aeration regime on nitrogen removal and the potential for GHGs emissions, the model was judged to be reliable. Ammonair settings were varied in a range coherent with discharge limits of the studied WWTP. Results suggest that, under the studied conditions, reducing ammonia concentration in the effluent will be reflected in an increase of the total nitrogen concentration in the effluent as well as of the energy consumption and would also increase N 2 O emissions. These preliminary conclusions should however be validated over longer periods of operation (at variable influent loads, MLSS concentrations, temperatures). References Aboobakar, A., Cartmell, E., Stephenson, T., Jones, M., Vale, P. and Dotro, G. (213) Nitrous oxide emissions and dissolved oxygen profiling in a full-scale nitrifying activated sludge treatment plant. Water research 47(2), 524-534. ASCE, Standard Guidelines for In-Process Oxygen Transfer Testing, ed. U. New York. 1996. Claeys, F., A Generic Framework for Modeling and Virtual Experimentation with Environmental Systems. PhD thesis. Ghent University, Department of Applied Mathematics, Biometrics and Process Control (BIOMATH), Coupure Links 653, B-9 Gent, Belgium, January 28. Filali, A., Fayolle, Y., Peu, P., Philippe, L., Nauleau, F. and Gillot, S. (213) Aeration control in a full-scale activated sludge wastewater treatment plant: impact on performances, energy consumption and N2O emission, Narbonne, France. Guo, L. (214). Greenhouse gas emissions from and storm impacts on wastewater treatment plants: process modelling and control. (PhD) - Université Laval. IPCC, Changes in atmospheric constituents and in radiative forcing. Solomon, S. et al. (Eds.), Climate Change 27: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, 27: p. 114 143 Rodriguez-Caballero, A., Aymerich, I., Poch, M. and Pijuan, M. (214) Evaluation of process conditions triggering emissions of green-house gases from a biological wastewater treatment system. Science of The Total Environment 493(), 384-391. Vanhooren, H., Meirlaen, J., Amerlinck, Y., Claeys, F., Vangheluwe, H. and Vanrolleghem, P. A., WEST: modelling biological wastewater treatment. Journal of Hydroinformatics 5 (23) 27-5 Wunderlin, P., Mohn, J., Joss, A., Emmenegger, L. and Siegrist, H. (212) Mechanisms of N(2)O production in biological wastewater treatment under nitrifying and denitrifying conditions. Water research 46(4), 127-137.