(Initial page layout)
Development and Calibration of a new model for Designing and Optimising a new Passive Aeration System for the Treatment of Municipal Wastewater Noelle Jones1,*, Edmond O’ Reilly1,2, Eoghan Clifford1,2 1
Civil Engineering, College of Engineering & Informatics, NUI Galway, University Road, Galway Ryan Institute, NUI Galway (Email:
[email protected],
[email protected]) 2
Abstract Biofilm-based passive aeration systems (PAS) have attracted recent attention as alternative energy efficient and low maintenance technologies for the treatment of municipal wastewater. However the modelling of biofilm-based PAS offers unique challenges for modellers particularly where new technologies are not easily modelled using existing commercial modelling software. However, if the modeller is concerned only with modelling the effluent from the system it may be possible to model these technologies using “surrogate” unit process systems (e.g. using an activated sludge process to model a biofilm process). The pumped flow biofilm reactor (PFBR); a batch biofilm technology, is one such example of a passive aeration system. The PFBR is a two reactor technology that employs a unique hydraulic regime and enables aerobic, anoxic and anaerobic conditions to be sequenced. Biofilm, growing on plastic media modules within the two reactors, is aerated passively as wastewater is moved alternately between the reactors during an aeration sequence. Thus as the two reactors empty and fill a number of times during a typical aeration sequence, the biofilm is exposed, in turn, to atmospheric air and wastewater. Furthermore while the PFBR has many of the features of a sequencing batch reactor the fill and discharge from the system typically take place in Reactors 1 and 2 respectively. GPS-X is modular, multi-purpose commercial software that can be used to simulate various wastewater treatment processes. As a relatively new technology a predictive model for the PFBR system has yet to be developed. Indeed for novel passive aeration systems in general, it can be time consuming and difficult to develop new models that accurately describe performance. In this study, using the PFBR as a case study of a biofilm-based PAS, a predictive model, developed using the modelling package GPS-X (Hydromatis Inc.), is presented. The model, which aims to simulate effluent quality only, was developed by modifying standard activated sludge sequencing batch reactor processes. The model was calibrated using experimental data obtained from a laboratory PFBR study. The model was then applied to a second set of independent laboratory data, obtained by operating the PFBR under an alternative wastewater loading regime. The results showed excellent correlation between experimental and modelled effluent results. While future work will focus on developing a unique model for the PFBR (and similar technologies) this study presents an alternative means to efficiently develop initial predictive models for novel passive aeration systems. Keywords Passive Aeration Systems, Pumped Flow Biofilm Reactor (PFBR), Wastewater Treatment Technology, Batch Biofilm Reactor, GPS-X, Oxygen Patch Sensor, Dissolved Oxygen
INTRODUCTION Wastewater treatment plant modelling is a useful tool for performing plant capacity assessments and improving plant operations. This can lead to improved plant performance, along with reduced energy and chemical costs. Devisscher et al., (2006) noted that mathematical modelling can optimise plant efficiencies resulting in cost savings of (i) aeration energy in the range from 10 to 20% and (ii) chemical dosing up to 30%. Pena-Tijerina & Chiang (2007) also stated that once implemented and calibrated, a model can offer many advantages such as (i) determining maximum flow conditions, (ii) optimisation of daily operation, (iii) energy saving evaluations and (iv) preparing existing wastewater treatment facilities for upcoming regulations such as the Water Framework Directive in Europe.
While wastewater treatment modelling is advancing and making important contributions to practice, it must be remembered that these are complex systems and further challenges to modellers remain. Areas requiring greater focus include (i) increased use of metabolic modelling, (ii) characterisation of the hydrodynamics of suspended and biofilm biological treatment processes, and (iii) the integration of biofilm and suspended growth process modelling (Daigger, 2011). Successful process modelling requires good knowledge of process variables, such as the most influential kinetic and stoichiometry quantities, and the resulting biomass composition (Drolka et al, 2001). Biofilm-based passive aeration systems (PAS) have attracted recent attention as alternative energy efficient and low maintenance technologies for the treatment of municipal wastewater (Clifford, 2013) (O’Reilly, 2011). However the modelling of biofilm-based PAS offers unique challenges for modellers particularly where new technologies are easily modelled using existing commercial modelling software. However, if the modeller is concerned only with modelling the effluent from the system it may be possible to model these technologies using “surrogate” unit process systems (e.g. using an activated sludge process to model a biofilm process). One such novel PAS is the pumped flow biofilm reactor (PFBR). The PFBR is a two-reactor-tank technology that has been extensively tested at laboratory-scale and field-scale. As a PAS, PFBR model development can be complex due to the difficulty in simulating the rate of oxygen transfer into the biofilm and the lack of existing unit processes in various software. The modelling of the PFBR is further complicated by the fact that it is a 2-reactor tank technology where hydraulic pumps are used to circulate wastewater from one reactor tank to the other. Aeration is achieved by alternately exposing the biofilm (attached to plastic media) in each of the two reactor tanks to atmospheric air, thereby eliminating the need for forced aeration using blower or diffusers etc. Anoxic/anaerobic conditions can be developed by maintaining the biofilm media immersed in the wastewater where required. The models of ASM family (ASM1, ASM2, ASM2d, ASM3) are used in most modelling and simulation studies today (Henze et al., 2002). ASM1 includes nitrogen and organic matter removal with simultaneous consumption of oxygen and nitrate as electron acceptors. ASM1 was developed mainly for municipal activated sludge plants (Henze et al., 2002; Henze et al., 2008). The activated sludge model ASM1 was used for biological processes in the PFBR model as the unit is modelled as an activated sludge plant. There are numerous methods and commercial wastewater simulators used to model waste water plants. One such software tool - GPSX (Hydromantis Inc, Canada) - is a modular, multipurpose modelling environment for the simulation of wastewater treatment systems. GPSX is supplied with over 50 preconfigured layouts covering most of the unit processes found in wastewater treatment plants. Six standard biological models e.g. temperature dependant versions of ASM1, ASM2d and ASM3 are available in GPSX. The biological unit processes include carbon, nitrogen and phosphorus removal, in various suspended growth and fixed film configurations (Makinia, 2010). GPS-X uses over 20 process objects in its models including a trickling filter, rotating biological contactor, and membrane bioreactor. Rapidly developing models for new technologies could help direct research, aid technology development and enable new processes to be modelled without the need to develop be-spoke software/unit processes. The use of existing software and “surrogate” unit processes may enable efficient initial model development. In this study a predictive model of the PFBR using an activated sludge unit process is developed. As this model is only concerned with simulating the effluent and cycle performance from the PFBR an existing wastewater unit process in GPS-X (Hydromantis Inc) is used as a surrogate for the PFBR system. The model is developed and calibrated against extensive data from laboratory studies.
METHODS Typical PFBR operating regime At the start of a typical PFBR treatment cycle, primary settled wastewater is pumped into Reactor1 (Fill stage = t1 mins) while simultaneously, treated effluent is discharged from Reactor 2. The wastewater was quiescent in Reactor 1 (Anoxic stage = t2 mins) and this allowed anoxic conditions to develop. Wastewater is circulated between Reactors 1 and 2 using gravity (a motorised valve) and hydraulic pumps – alternately exposing the biofilm in each reactor to the atmosphere and the wastewater (this is known as a pumping cycle). This comprises the aerobic stage (Aeration stage = t3 mins). During the aerobic stage a typical pumping cycle entails water levels being initially equalise between reactors by gravity using a motorised valve. Once equalised, the remainder of the water is pumped to the receiving reactor, thus exposing the biofilm media in the emptied reactor to the atmosphere. This process is repeated a number of times. This is the main aeration mechanism (no forced aeration is required). After the aeration phase, another rest period began (Settle stage = t4 mins). This settle period allowed solids to separate from the liquid producing a clarified effluent. The solenoid valve opened and allowed the treated wastewater to flow out of Reactor 1 under gravity (Draw stage = t5 mins). The operating strategy of the PFBR system is given in Table 1. Table 1. The operational stages of the PFBR PFBR Stage Duration (minutes) Fill (t1) 8 minutes Anoxic (t2) 150 minutes Aerobic (t3) 106 minutes 60 minutes Settle (t4) Draw (t5) 5 minutes Full details of PFBR operation have previously been published (O’Reilly et al., 2008, O’Reilly et al, 2011). Experimental and Model Wastewater Characteristics Two synthetic wastewaters of varying strengths were used as influents for the PFBR plot scale studies. Details of the influent wastewaters are summarised in Table 2. The influent characteristics of the modelled wastewater were estimated using this measured data. Using Influent Advisor (an influent modelling tool associated with GPS-X) a detailed overview of the influent was developed for the model. As phosphorous removal was not considered in this model and phosphorous concentrations were not limiting this was omitted from the influent model. Table 2. Typical average influent concentrations in the synthetic domestic wastewater. Standard deviations shown in () Modelled Experiment Parameter Experiment Modelled High Strength High Strength Low Strength Low Strength Wastewater Wastewater Wastewater Wastewater (mg/l) (mg/l) (mg/l) (mg/l) Total CODt 1021 (155) 1020 346 (32) 346 Total Nitrogen (TNt) 97 (2.5) 97 33 (1.3) 33 Total Filtered Nitrogen (TNf) 97 (2.5) 97 24 (2.4) 24 62 (8) 63 18 (2.7) 20 Ammonium Nitrogen (NH4-N) Suspended Solids (SS) Influent SS concentrations were negligible in the pilot scale studies due to the synthetic nature of the influent; hence influent CODt and influent filtered COD (CODf) and also TNt and TNf are
similar. Filtered results are those obtained after sample filtration through a 0.45 μm glass fibre filter. The model was initially calibrated against the high strength wastewater (Study 1) and then verified against the low strength wastewater (Study 2). It should be noted that the pilot scale PFBR system operated a similar cycle (as shown in Table 1) for both Study 1 and Study 2. RESULTS AND DISCUSSION PFBR Pilot Plant Model Hydraulic Calibration of Model The initial work focused on developing a model that could accurately represent the hydraulic characteristics of the PFBR. An initial investigation showed the SBR object most accurately modelled the PFBR characteristics. As the PFBR is operated as a SBR incorporating the five typical SBR phases: fill, anoxic, aeration, settle and draw and given it is a two reactor technology it was decided to use two linked SBR units to model the PFBR. This enabled the hydraulics to be accurately represented. There are three different SBR objects in GPS-X that can be chosen from: the Simple SBR object, the Advanced SBR object, and the Manual SBR object. All three objects have the same functionality, appearance and choice of biological models. They differ in the manner in which the user specifies the operation of the SBR unit (Hydromantis, 2006). To date 3 efficient layouts used for the hydraulic calibration of the PFBR have been developed using each of the 3 SBR objects available in GPS-X. The hydraulic model using two manual SBR’s was the model chosen to calibrate the model for carbon removal, nitrification and denitrification. An operational cycle was defined to mimic the PFBR operation. The manual SBR object requires that the entire operational cycle be defined by the user, either by having the liquid flows, air flows, and mixing on interactive controllers, or as file inputs. The manual SBRs require the user to define the SBR objects (either from a file input or input controls) pumping cycles, when settling times, etc. Figure 1 shows the general layout of the modelled system. The layout includes two manual SBR objects, an influent object, two flow combiners and a flow splitter. The two manual SBR units were used to model the different phases of the treatment cycle i.e. SBR1 – Fill, anoxic phase and aerobic phase SBR2 – Aerobic phase and Settle and draw phase
Figure 1: GPS-X layout of the PFBR model using two manual SBR objects Table 3 shows the physical dimensions of the SBR reactors used in the model. Table 3. – Physical Design Parameters: Experimental and Modelled Physical Dimensions Size Surface Area 0.0576 m2 Maximum Water Level Height 0.4 m Feed Point from Bottom 0.1 m
As already stated a challenge in modelling PFBR systems is the unique hydraulic operation of the technology. Proprietary software packages do not directly facilitate the modelling of the hydraulics of the PFBR system however the hydraulics in the PFBR has been successfully modelled using the SBR object. During a typical cycle influent wastewater flowed into Reactor 1 over an 8 minute period (Reactor 1 contained treated wastewater from a previous cycle and Reactor 2 is empty at this stage). The wastewater then settled in Reactor 1 for 150 minutes (which equated to an anoxic period). After the anoxic period the wastewater circulated between Reactor 1 and Reactor 2 for 106 minutes (aerobic period). A rest period of 60 minutes follows in Reactor 2 (Reactor 1 is empty at this stage). Finally 42% of the total volume was decanted from Reactor 2 while 58% was recycled back to Reactor 1 (Figure 2) in preparation for the next treatment cycle.
Flow (l/sec)
Influent Pump 1 Pump 2 Discharge
Time (minutes)
Figure 2: Flows in Reactor 1 and Reactor 2 Using two SBR unit processes it was possible to model the experimental hydraulics exactly. In Figure 2 the alternating sequence of pump 1 and pump 2 indicate water being transferred between Reactor 1 and Reactor 2 during an aeration cycle. Model Calibration and Sensitivity Analyses Average experimental effluent results, over a 90 day steady state period of operation were used to calibrate the model during Study 1. The parameters of interest were COD, SS, Total nitrogen, Ammonium-nitrogen and Nitrate-nitrogen. The model was initially calibrated using the high strength wastewater. After the model optimization the model was tested and run against the low strength concentration wastewater. It should be noted that while results are presented for a 35 day modelled period (for clarity) the GPS-X model was run for a period of 100 days to ensure there was no drifting in the effluent results. The most relevant parameters to be estimated during calibration were determined by completing sensitivity analyses. Using the default values for ASM1 sensitivity analyses were carried out and the default values adjusted to calibrate the model. The calibrated kinetic and stoichiometric parameters were then used for process simulation of the biofilm processes taking place in the PFBR. The sensitivity effluent COD, SS, TNt, NH4-N and NO3-N concentrations to various parameters and process operations were analysed. The analyses indicated that the kinetic parameters most significant for calibration were the heterotrophic maximum specific growth rate (μ and the autotrophic yield (YA). Typical results from the sensitivity analyses are shown in Figures 3 and 4.
μ2/d μ6/d μ12/d
COD (mg/l)
Modelled time (days)
Concentration (mg/l)
Figure 3. Impact of heterotrophic growth rate (μH) on COD, TNt, NH4-N and NO3-N
YA2 gCOD/g N YA6 gCOD/g N YA12gCOD/g N
Modelled time (days)
Figure 4. Impact of autotrophic yield (YA) on modelled effluent COD, TNt, NH4-N and NO3-N As expected effluent COD was significantly affected by varying the heterotrophic growth rate. At a μ of 2/d the COD concentration was 830 mg/l. However this decreased to 104 mg/l and 76 mg/l for a μ of 6/d and 12/d respectively. There was a 500% increase in heterotrophic growth rate from 2 to 12 and this led to an 88% decrease in COD. The increase in heterotrophic growth rate had a limited impact on NH4-N in the range analysed. YA was more significant in calibrating TNt, NH4-N and NO3-N than the autotrophic growth rate. For example as YA increased from 0.2 to 1.2 effluent NH4-N concentrations increased steadily –a six fold increase in YA resulted in a 1700% increase in effluent NH4-N concentrations.
Kinetic and Stoichiometry Results The comparison between calibrated parameters and typical literature ranges, are presented in Table 4. Table 4. Estimated and typical literature values for kinetic and stoichiometry coefficients (Mulas, 2005) Symbol
µH KS KOH KNO ηg bH µA KNH bA KOA kh KX ηh ka ƒp YH YA
Parameter (Units) Kinetic Active Heterotrophic Biomass heterotrophic maximum specific growth rate (1/d) readily biodegradable substrate half saturation coefficient (mg COD/L) oxygen half saturation coefficient (mg O2/L) nitrate half saturation coefficient (mg N/L) anoxic growth factor heterotrophic decay rate (1/d) Active Autotrophic Biomass autotrophic maximum specific growth rate (1/d) ammonia half saturation coefficient for autotrophs growth (mg N/L) autotrophic decay rate (1/d) oxygen half saturation coefficient for autotrophs growth (mg O2/L) Hydrolysis maximum specific hydrolysis rate (1/d) slowly biodegradable substrate half saturation coefficient (g COD/g COD) anoxic hydrolysis factor Ammonification ammonification rate m3(g COD d)-1 Model Stoichiometry fraction of biomass leading to particulate products (g COD/g COD) Heterotrophic Yield (g COD/g COD) Autotrophic Yield (g COD/g N)
Literature ranges
Calibrated Values
0.6-13.2 5-225 0.01-0.20 0.1-0.5 0.8 0.05-1.6
12 20 0.2 0.5 0.8 0.62
0.2-1.0 1.0 0.05-0.2 0.4-2.0
0.8 1.0 0.04 0.4
3 0.03
3 0.03
0.4
0.4
0.08
0.08
0.08 0.38-0.75 0.07-0.28
0.08 0.666 0.99
Dissolved Oxygen Dissolved oxygen concentrations were modelled by using diffused air and entering a KLa (mass transfer coefficient) value of 122/d. During the aerobic period DO concentrations peaked after each pumping event and decreased during rest periods between pumping events. Similar profiles were obtained for modelled and experimental results though minimum oxygen levels during the aeration cycle were over estimated by the model. It should be noted that the input KLa was simply modelled to ensure that oxygen profiles from experimental and modelled data correlated well. This value is not necessarily indicative of the in-situ KLa. Experimental and Modelled Results Tables 5 and 6 summarise the experimental results and the GPS-X model results for both the high strength and the low strength synthetic wastewater. As previously stated, and described above, the model was calibrated against the pilot scale experimental data for the high strength wastewater. The calibrated model was then applied to the experimental data gathered using low strength wastewater. The average measured removal efficiency for COD was 93% and 86 % for the high and low strength wastewaters respectively. This compared well with a modelled removal efficiency of 92% and 89% for the high strength and low strength waste water respectively. The experimental efficiency for TNt was 74% and predicted 58% for both high strength and low strength wastewater.
Similarly, acceptable agreement between measured and predicted waste treatment efficiency was observed for NH4-N (84% and 81% for high strength wastewater and 98% and 97% for low strength wastewater). Table 5. Experimental and modelled results for high strength wastewater (Study 1) High Strength Wastewater Concentration Experimental (average of a 100 day study) Model Removal Removal Oulet Efficiency (%) Outlet Efficiency (%) CODt (mg/l) TSS (mg/l) TNt (mg/l) NH4-N (mg/l) NO3-N (mg/l)
76 (20) 10 (6.2) 25 (7) 10 (8) 12 (5)
93 84 74 84 -
86 (9) 15 (1) 25 (2) 12 (3) 10 (2)
92 76 74 81 -
Table 6. Experimental and modelled results for low strength wastewater (Study 2) Low Strength Wastewater Concentration Experimental (average of a 121 day study) Model Removal Removal Oulet Efficiency (%) Outlet Efficiency (%) CODt (mg/l) TSS (mg/l) TNt (mg/l) NH4-N (mg/l) NO3-N (mg/l)
48 (32) 3 (2.8) 14 (5.6) 0.4 (0.4) 13 (2.5)
86 95 58 98 -
38 (11) 0.6 (0.1) 14 (0.9) 0.5 (0.3) 10 (0.5)
89 99 58 97 -
Tables 5 and 6 show excellent agreement between experimental and modelled results both for the calibrated model (Study 1) and for Study 2 (used to verify the model). Figure 5 shows the relationship between the high strength waste water experimental results and model results for CODt, TSS, TNt, NH4-N and NO3-N. The average experimental COD concentration was 76mg/l and the average predicted COD was 87mg/l. Experimental effluent TSS concentrations averaged 10mg TSS/l and modelled concentrations averaged 15mg TSS/l. The experimental and predicted effluent for TNt and NO3-N both averaged 25mg/l and 10mg/l respectively. Figure 6 shows similar data for Study 2.
Concentration (mg/l)
--- Experimental Average ----Model Average Modelled Daily Data
Time (day)
Concentration (mg/l)
Figure 5. Measured and predicted effluent concentrations for high strength wastewater for CODt, TSS, TNt, NH4-N and NO3-N (Study 1).
--- Experimental Average ----Model Average Modelled Daily Data
Time (day)
Figure 6. Measured and predicted effluent concentrations for low strength wastewater for CODt, TSS, TNt, NH4-N and NO3-N (Study 2). CONCLUSIONS Modelling of biofilm-based PAS offers unique challenges for modellers particularly where new technologies are not easily modelled using existing commercial modelling software. One such example is the pumped flow biofilm reactor (PFBR) - a batch biofilm process that employs passive
aeration. Modelling the PFBR offered a unique challenge as it is a new technology that was not easily built and calibrated using existing commercial modelling software. To model the PFBR an existing SBR unit process in GPS-X was used. An initial PFBR model was calibrated with experimental data using a high strength synthetic wastewater. It was then verified against an independent study using a low strength synthetic wastewater. The main conclusions of the study are as follows: 1. A model of a batch biofilm passive aeration process - the PFBR -was successfully calibrated using an activated sludge process to model the PFBR. The model was calibrated using experimental data and then verified using independent experimental data from the PFBR. 2. The model successfully predicted effluent characteristics for carbon suspended solids and nitrogen. Thus the model can be used to enhance and improve reactor operation and inform future studies. 3. It can be necessary to make assumptions relating to oxygen transfer in the activated sludge model that may not be supported by experimental data in order to model oxygen profiles in the biofilm process. While the approach can have limitations; in particular in relation to oxygen transfer characteristic, modelling of biofilm density and the effect of varying support media characteristics, it can offer a method of rapidly developing and calibrating models for new technologies without the need to develop unit processes for existing software or be-spoke modelling software. REFERENCES Clifford, E., Forde, P., McNamara, S., Rodgers, M., O’Reilly, E. (2013). The performance of a new technology – the Air Suction Flow Biofilm Reactor (ASF-BR) – in treating municipal strength wastewater. Journal of Environmental Engineering. (In Press). Daigger G., (2011) – A practitioner’s perspective on the uses and future developments for wastewater treatment modelling. Water Science and Technology, 63(3):516-526. Devisscher M., (2006) - Estimating costs and benefits of advanced control for wastewater treatment plants – the MAgIC methodology. Water Science and Technology, 52(45). Drolka, M., Plazl, I., Koloini, T. (2001) The Results of Mathematical Model and Pilot Plant Research of Wastewater Treatment. Model and Wastewater Treatment 15 (2) 71–74 (2001) Henze, M., Harremoes, P., Cour Jansen, J. la, Arvin, E. (2008) Wastewater Treatment Biological and Chemical Processes Hydromantis (2006) GPS-X Technical Reference Version 5.0 Jeppsson. U., (1996). Modelling aspects of wastewater treatment processes. Lund Institute of Technology, Dept. of Industrial Electrical Eng. and Automation, ISBN 91-88934-00-4; Makinia, Jacek (2010) – Mathematical Modelling and Computer Simulation of Activated Sludge Systems Mulas, M. (2005) Modelling and Control of Activated Sludge Processes. Chemical and Materials Engineering Department Cagliari O’Reilly, E., Rodgers, M. and Zhan, X.-M. (2008). Pumped flow biofilm reactors (PFBR) for treating municipal wastewater. Water Science & Technology 57(12): 1857–1865 O’Reilly, E., Clifford, E. and Rodgers, M. (2011) Municipal Wastewater Treatment using a FullScale Pumped Flow Biofilm Reactor (PFBR) Water Science and Technology 64 1218-1225 Pena-Tijerina, A.J., and Chiang, W. (2007) What does it take to model a wastewater treatment plant?’ In Proceedings: Texas Water, Fort Worth Convention Centre, Texas, USA Rodgers, M., Zhan, X.-M., and O’Reilly, E. (2006). Small-scale domestic wastewater treatment using an alternating pumped sequencing batch biofilm reactor system - Bioprocess Biosystems Engineering. 28: 323–330
