- Email: [email protected]
Optimization of a large industrial wastewater treatment plant using a modeling approach: A case study
Journal of Environmental Management 249 (2019) 109436
Contents lists available at ScienceDirect
Journal of Environmental Management journal homepage: www.elsevier.com/locate/jenvman
Research article
Optimization of a large industrial wastewater treatment plant using a modeling approach: A case study
T
Roberta Muoioa, Laura Pallia, , Iacopo Duccia, Ester Coppinib, Elena Bettazzib, Daniele Daddib, Donatella Fibbib, Riccardo Goria ⁎
a b
Department of Civil and Environmental Engineering, University of Florence, Via Santa Marta 3, Florence, Italy G.I.D.A. SpA, Via di Baciacavallo 36, Prato, Italy
ARTICLE INFO
ABSTRACT
Keywords: Sewage sludge Wastewater treatment plant Case study Cost optimization West by DHI Modeling
The objective of this paper is to find the optimum solid retention time (SRT) of a wastewater treatment plant (WWTP), which minimizes operating costs, using a modeling approach with WEST software by MIKE DHI®. For the determination of the kinetic and stoichiometric parameters (used for the correct calibration of the model implemented), respirometric and kinetic batch tests were carried out. Each Oxidation ditch was modeled by a sequence of four aerated activated sludge units (ASUs) and four anoxic ASUs with recirculation. The model is able to simulate the separation efficiency of the secondary settler, which is generally quite low: in fact, the industrial origin of the wastewater induces the formation of small flocs, the dimensions of which can be further reduced by the presence of surface aerators. The nitrification/denitrification process is also accurately predicted. Using data obtained from the model, mass balances at the steady state for COD and N were made and compared to the ones obtained using measured data. After calibration and validation of the model, steady-state simulations were carried out by increasing and decreasing the SRT of the system under two different operational conditions used by the managing company and by evaluating the costs related to the water treatment line and the sludge treatment line for each scenario. It is interesting to note how the total costs are lower in summer than in winter (7.2 €cent/m3 in summer and 8.7 €cent/m3 in winter, in scenario 0). In general, the increase in the SRT led to a decrease in the total management costs. In fact, differences between scenario 0 and the scenario with the lowest total treatment costs (corresponding to an SRT of 11.4 d in winter and 10.0 d in summer) could give rise to total savings of about 44·000€/year in summer and 93·000€/year in winter.
1. Introduction Nowadays, activated sludge (AS) is affirmed as the most common process for wastewater treatment. The quantification of environmental and economic performance has been of major interest in developing control strategies for the AS process (Benedetti et al., 2010; FloresAlsina et al., 2014); in fact, one of the main challenges in the management of a wastewater treatment plant (WWTP) is to minimize the operating costs while satisfying the effluent discharge limits. Optimization of the process, which is the key for successful management of a WWTP, can be achieved by seeking optimal process conditions such as solid retention time (SRT), aeration rate, and internal recycle flow rate (De Ketele et al., 2018; Kim et al., 2015; Newhart et al., 2019; Qiao and Zhang, 2018). For WWTPs, the optimization methods cannot be applied accurately without a mathematical modeling approach (Hreiz et al., 2015). Moreover, once developed, these models can be used for design
⁎
and control purposes, process upgrades and the evaluation of greenhouse gas emissions (Baalbaki et al., 2017). The two essential constituents for modeling an activated sludge plant are the biological reactors and the hydraulic component. As far as the former is concerned, many advanced mathematical models have been developed and proposed in the past, such as those developed by IWA (International Water Association), known as ASM1, ASM2, ASM2d, and ASM3 (Hauduc et al., 2013). These models are considered good solutions for correlating the complexity of the activated sludge processes and predicting the biological treatment efficiency in dynamic conditions (Wu et al., 2016). The hydraulic component describes the tank volumes and behavior of the flow rates. The secondary settler can be modeled in different ways, from a simple ideal separator, with no retention time, to more complex models such as reactive settlers (Gernaey et al., 2006) or CFD-aided models (Karpinska and Bridgeman, 2016). Modeling is an important part of the WWTP design and operation because it makes it
Corresponding author. E-mail address: [email protected] (L. Palli).
https://doi.org/10.1016/j.jenvman.2019.109436 Received 21 May 2019; Received in revised form 14 August 2019; Accepted 18 August 2019 0301-4797/ © 2019 Elsevier Ltd. All rights reserved.
Journal of Environmental Management 249 (2019) 109436
R. Muoio, et al.
List of abbreviations AOB AS ASM BOD CFD COD DO F/M HRT IWA MLSS NOB OUR SRT SVI TN TP TSS WWTP
bA YH YA μH,max μA,max μAOB,max
Ammonium Oxidizing Bacteria Activated sludge Activated Sludge Model Biochemical Oxygen Demand Computational Fluid Dynamics Chemical Oxygen Demand Dissolved Oxygen Food-to-Microorganism Ratio Hydraulic Retention Time International Water Association Mixed Liquor Suspended Solids Nitrite Oxidizing Bacteria Oxygen Uptake Rate Solid Retention Time Sludge Volume Index Total Nitrogen Total Phosphorous Total suspended solids Wastewater Treatment Plant
μNOB,max XBH XAOB XNOB SS SI XS XI N–NH4+ N–NO2 N–NO3 N-NOx sbN snbN pbN pnbN
List of symbols bH
Endogenous decay coefficient of autotrophic biomass Heterotrophic biomass yield Autotrophic biomass yield Heterotrophic maximum specific growth rate Autotrophic maximum specific growth rate Maximum specific growth rate of ammonium-oxidizing bacteria Maximum growth rate of nitrite-oxidizing bacteria Active heterotrophic biomass concentration Active AOB biomass concentration Active NOB biomass concentration Soluble Biodegradable COD Soluble Non-Biodegradable COD Particulate Biodegradable COD Particulate Non-Biodegradable COD Ammoniacal nitrogen Nitrous nitrogen Nitric nitrogen Sum of nitrous and nitric nitrogen Soluble Biodegradable Nitrogen Soluble Non-Biodegradable Nitrogen Particulate Biodegradable Nitrogen Particulate Non-Biodegradable Nitrogen
Endogenous decay coefficient of heterotrophic biomass
possible to better understand some of the process mechanisms and to predict the behavior of the plant in different scenarios (Hreiz et al., 2015). In this context, the objective of this work is to optimize the operating conditions of the Baciacavallo WWTP, located in Tuscany and managed by G.I.D.A. S.p.A. In particular, the objective of the work is to
find the optimum SRT for minimizing the operating costs. The SRT has been chosen as the optimization parameter because it can shift costs between the water treatment line (for aeration, water pumping and chemicals if dosed) and the sludge treatment line (related, for example, to energy for dewatering, chemicals, and disposal) (Dionisi and
Fig. 1. Flowchart of the WWTP. Light grey line: wastewater and recycled flow; dark grey line: sludge. ST: Septic Tank; Se: external Sludge; M: flowmeter; S: Sampling. Sections from 1 to 11 represent water treatment line. Sections from 13 to 14 represent sludge treatment line. 1: Coarse bar screen; 2: pumping; 3a and 3b: fine bar screen, line a and b; 4a and 4b: grit removal line a and b; 5a and 5b: flocculation line a and b; 6a and 6b: primary sedimentation line a and b; 7: flow equalization; 8a and 8b: Oxidation ditches line a and b; 9a and 9b: secondary sedimentation line a and b; 10a and 10b: clariflocculation line a and b; 11: Ozonation; 12: thickening; 13: centrifugation; 14: incineration. 2
Journal of Environmental Management 249 (2019) 109436
R. Muoio, et al.
N–NO3-, TP, surfactants, color and metals.
Rasheed, 2018; Drewnowski et al., 2019; Fan et al., 2017; Liu and Wang, 2015). Indeed, with an increase in the SRT, more solids are retained in the oxidation tank and more oxygen is required (increasing costs related to the water treatment line), but at the same time, more solids are oxidized in the oxidation tank, leading to a minor amount of solids entering the sludge treatment line and minor costs related to the same. On the other hand, a reduction in the SRT would lead to an increase in these costs but with a reduction in oxygen requirements in the water treatment line. The Baciacavallo plant (900.000 PE) treats urban wastewater (from combined sewer) and textile wastewater (about 70% of the total flow rate) jointly. The current treatment train includes pretreatment, primary settler, secondary and tertiary treatment (clariflocculation and ozonation), while the sludge treatment line includes thickening, centrifugation and lastly, incineration. The unusual sludge treatment line (due to the absence of stabilization) gives rise to high costs related to sludge management. For this reason, it is of paramount importance to optimize the process conditions in order to contain the sludge management costs. In this paper, an attempt has been made to find the optimum SRT using a modeling approach with WEST software by MIKE DHI®.
2.2. Experimental In order to determine the kinetic and stoichiometric parameters, respirometric and kinetic batch tests were carried out. The former were conducted to determine the heterotrophic biomass parameters (bH, YH, μH,max), and the latter, the autotrophic biomass (μA,max). The results were used to correctly calibrate the model implemented. 2.2.1. Respirometric tests Three respirometric tests were performed using a modified version of a previously reported protocol (Coppini et al., 2019; Palli et al., 2014). For each test, an experimental 1.2L reactor was filled with AS collected from the four oxidation tanks (mixed in even parts), and preaerated for 24 h in order to assure endogenous conditions. During the test, the temperature was maintained at 20 °C, and the pH at 7.5 by dosing NaOH or HCl, and mixing was assured via use of a magnetic stirrer. In each test, 2 mL sodium acetate (24 gCOD/L) were spiked in the reactor as a rapidly biodegradable carbon substrate, maintaining a food-to-microorganism ratio (F/M) of 0.05 (Andreottola et al., 2002). All the tests were performed with the addition of 15 mL of ATU (1 g/L) to inhibit nitrification. The respirometric system developed enabled remote-controlled and automated online monitoring of the O2 consumption, pH, temperature and oxygen uptake rate (OUR).
2. Methodology 2.1. Description of the case study The Baciacavallo WWTP, located in Prato (Central Italy), treats both industrial and municipal wastewater with an average inflow of 100·000 m3/d. Industrial wastewater mostly originates from the textile manufacturing processes and represents 70% of the total influent flow, while the remaining amount (30%) is represented by domestic wastewater and sludge from septic tanks (0.07%). The WWTP has a flow capacity of 144·000 m3/d and is designed for 900·000 PE. The treatment process is divided into two lines (line a and line b), with biological treatment consisting of four oxidation ditches (two per line) for the organic matter removal and nitrification. The treatment train (Fig. 1) is composed by coarse bar screens, pumping, fine bar screens, grit removal, primary sedimentation (where the residual water from the sludge treatment unit is recycled), equalization (in order to balance the flow variations), then biological oxidation/nitrification, clariflocculation and final ozonation. Sludge from septic tanks is introduced upstream of the biological treatment. The biological treatment consists of four oxidation ditches (with a total volume of 29·000 m3) and four secondary sedimentation tanks (with a total volume of 24·308 m3). The underflow from the secondary sedimentation tanks flows back to the oxidation ditches at a ratio of 1.2:1.4 in order to supply the nutrient requirements of the bacteria and maintain an adequate solid concentration. During the process operation, the concentration of dissolved oxygen (DO) is maintained at approximately 1.5 mg/L. Oxygen transfer is carried out by six surface rotor-brushes for each oxidation ditch. The sludge resulting from secondary and tertiary treatments can be recycled in primary sedimentation and removed together with the primary sludge, otherwise it can be sent into the sludge treatment line separately. The sludge is processed by means of gravity thickening (with 3 thickeners), centrifugation with pre-conditioning by cationic polyelectrolytes (by 2 dewatering units), and finally, incineration. In this section the sludge is treated together with the sludge coming from another WWTP managed by the same company. Thickener overflow and centrate are recycled into the primary sedimentation unit. Fig. 1 also shows the position of the flow meters and the sampling points located all along the treatment train. All samplers are automatic except for the sampling points in the oxidation tanks, where manual samples are taken three times a day in order to control the MLSS (Mixed Liquor Suspended Solids) and SVI (Sludge Volume Index). The analysis of wastewater entails the determining of the concentration of specific constituents each day by G.I.D.A S.p.A.‘s internal laboratory, including pH, COD, BOD, TSS, TN, N–NH4+, N–NO2-,
2.2.2. Biomass yield, YH, estimation For the estimate of the yield factor the following formula (Equation (1)) was used:
YH (mgCODcell / mgCODsub) = 1
O2 COD
(1)
where ΔCOD is the COD consumed (calculated as the difference between the soluble COD in the system after spiking and at the end of the test), and ΔO2 is the oxygen consumed due to substrate degradation. This was evaluated as the difference between the total oxygen consumption (the area below the OURs measured) and the oxygen consumed due to endogenous respiration. 2.2.3. Endogenous decay coefficient estimation The procedure for estimating the endogenous decay coefficient, bH according to a previously reported protocol (Avcioglu et al., 1998; Spanjers and Vanrolleghem, 1995), considers the values of OURmax (maximum value of OUR during exogenous respiration) determined on different days after the spiking of 2 mL of sodium acetate solution. The plot of the natural logarithm of the recorded OURmax values vs. Time shows an exponential decrease in the biomass as a straight line with the slope bH. 2.2.4. Maximum specific heterotrophic bacterial growth-rate estimation For estimating the maximum specific heterotrophic bacterial growth rate (μH,max) the following formula (Equation (2)) was used:
µH , max =
OURmax ·YH (1 YH )·XBH
(2)
where XBH (gSSV/L) represents the active heterotrophic biomass concentration in the reactor, evaluated by (Equation (3)):
XBH = YH ·
SRT ·(S0 S ) HRT ·(1 + bH ·SRT )
(3)
where SRT (d) is the sludge retention time; S0 (mg/L) is the average daily concentration of COD entering the biological treatment; S (mg/L) is the average daily concentration of COD leaving the biological treatment; HRT (h) is the hydraulic retention time. 3
Journal of Environmental Management 249 (2019) 109436
R. Muoio, et al.
2.2.5. Batch tests For the characterization of the autotrophic biomass, a kinetic batch test was performed. For the respirometric tests, an experimental 1.6 L reactor was filled with activated sludge collected from the four oxidation tanks (mixed in even aliquots), and pre-aerated for 24 h in order to assure endogenous conditions. During the test, the pH was maintained at 7.5 by dosing NaOH or HCl, and mixed conditions were assured via use of a mechanical stirrer. Oxygen was inflated by means of a fine bubble diffuser and at the beginning of the tests, ClNH4 was dosed until reaching a final concentration of 20 mg/L. Samples were then taken at regular intervals (every 15 min) and immediately filtrated with 0.45 μm syringe filters. The ammonium, nitrate and nitrite were analyzed in each sample.
2.3.2. The set-up the WWTP layout The treatment train was simulated from the primary sedimentation up to the clariflocculation steps. The primary settler was modeled by the Primary Point Settler Block: the settler is a phase separator and has no real volume or hydraulic retention time. In this unit, the effluent particulate concentration is determined as a fraction (f_ns) of the influent concentration. The underflow concentration is based on a mass balance. This fraction was set to 0.54, according to an experimental mass balance carried out in the primary settler (considering all the solids entering from the influent, secondary and/or tertiary sludge, supernatants from the sludge treatment line, and all the solids exiting with the effluent). In WEST, each Oxidation ditch has been modeled by a sequence with four aerated Activated Sludge Units (ASUs) and four anoxic ASUs with recirculation, as reported by other authors (Alaya et al., 2010). The choice to implement the anoxic ASUs, despite the absence of a proper anoxic section in the plant, was driven by the experimental data, which showed how part of the influent nitrogen was lost, probably due to denitrification occurring inside the oxidation tank. In fact, experimental measures of DO in the oxidation tanks showed how the lower part of the tank was actually anoxic, causing a partial denitrification of the nitrates. An anaerobic ASU was installed downstream from each secondary settler in order to simulate the spontaneous denitrification processes that take place during secondary sedimentation. The secondary settlers were modeled by the Takács-SVI Block: they predict the solid concentration profile in the settler by dividing the settler into a number of layers of constant thicknesses and creating a balance of the solids around each layer (Takács et al., 1991). This model is an extension of the Takács model in which the hindered settling parameter (r_H) is estimated via SVI measurement (Daigger and Roper, 1985). The parameters r_P (low concentration settling parameter (m3/ g)) and v0 (Maximum theoretical settling velocity (m/d)) were calibrated using the Parameter Estimation (PE) function of the software and were equal to 0.00089 m3/g and 306 m/d respectively. The clariflocculation unit was modeled by the Efficiency Thickener Block: the separation of particulate fractions is described through a mass balance and the definition of the separation efficiency (e_X), that is, the fraction of solids going into the underflow sludge. This fraction was set to 0.87 according to an experimental mass balance made in the clariflocculation unit (considering the solids entering from the biological treatment outlet and those exiting with the effluent). The sludge treatment unit, consisting of thickeners and centrifuges, was modeled as a single Efficiency Thickener Block similar to the clariflocculators.
2.2.6. Maximum specific autotrophic bacterial growth-rate estimation For estimating the maximum specific autotrophic bacterial growth rate (μA,max) the following formulas (Equations (4)–(6)) were used:
µAOB, max = YAOB ·
µNOB, max = YNOB·
d (N
d (N
NH4+) 24 · dt XAOB
(4)
NO2 ) 24 · dt XNOB
(5)
µA, max = µAOB, max + µNOB, max
(6)
where, μAOB,max (d−1) and μNOB,max (d−1) are the Ammonium Oxidizing Bacteria (AOB) and Nitrite Oxidizing Bacteria (NOB) maximum specific growth rates respectively; d(N–NH4+)/dt and d(N–NO2-)/dt (mg/l·h) represent the ammonium and nitrite consumption rates respectively; XAOB and XNOB (gSSV/L) represent the AOB and NOB active biomass concentration in the reactor respectively; YAOB and YNOB (mgSSV/mgN) are respectively the AOB and NOB yield factors, considered equal to 0.13 and 0.04 (Randall and Buth, 1984). XAOB and XNOB were evaluated through a mass balance at the steady state, considering an ammonium consumption for the AOB equal to the difference between N–NH4+ in the influent and the effluent of the plant, and a nitrite consumption for the NOB equal to the difference between the ammonium consumption (considered equal to the nitrite production) and the N–NO2- in the effluent of the plant. 2.2.7. COD and nitrogen fractionation The fractions of wastewater COD and N were determined in order to obtain more in-depth knowledge of the influent characteristics and ensure more accurate results from the model. The fractions of COD and N were evaluated following the protocol reported in the Supplementary Material.
2.3.3. Scenario simulations Steady-state simulations were carried out by varying the waste sludge flow rate of the biological treatment and consequently, by increasing and decreasing the SRT in order to find the optimum SRT for minimizing the treatment costs. The company running the WWTP uses two different strategies to manage the secondary waste sludge: in winter this is sent directly to the sludge treatment line, while in summer it is recirculated in the primary settler and then removed together with the primary sludge. For this reason, two different sets of scenarios were simulated (by varying the waste sludge flow rate), one with data from January 2017 (representative of winter management) and one with data from June 2017 (representative of summer management). As far as winter season is concerned scenario 0 was the one with SRT 9.0 d and then 4 other scenarios were simulated, with 7.7 d, 7.2 d, 10.7 d and 11.7 d, respectively. As far as summer season is concerned scenario 0 was the one with SRT 9.0 d and the 4 other scenarios were simulated with 7.9 d, 7.3 d, 9.9 d and 10.1 d, respectively. The aeration efficiency (gO2/kWh) was calculated by dividing the oxygen consumption (in kgO2/d), obtained from the model in scenario 0 for the mean daily consumption of electric energy related to the
2.3. Simulation in WEST® by DHI software In this study, the WWTP's configuration was implemented in WEST 2016 ® (MIKE DHI), using the available data on plant characteristics, wastewater quality, and respirometric and kinetic tests results. 2.3.1. The WEST environment In this study, an ASM1Temp model was considered for the configuration studied. This is an extension of the original ASM1 model published by the International Association on Water Quality (IAWQ) Task Group on Mathematical Modeling for Design and Operation of Biological Wastewater Treatment Processes (Henze et al., 1987), which furthermore considers temperature correction and ammonium limitation for the aerobic and anoxic growth of heterotrophs. An input file was generated containing the characteristics of the influent wastewater. All the plant data used in this study were taken from routine operating records. The COD and N components were determined via fractionation as reported above. 4
Journal of Environmental Management 249 (2019) 109436
R. Muoio, et al.
aeration system (kWh/d). The mean daily kWhs were calculated separately for the two simulated seasons by using real data from January 2017 and June 2017. These aeration efficiencies, which were 706 gO2/ kWh in January and 593 gO2/kWh in June, were kept constant in the other scenarios. From the simulation scenarios, the total oxygen required from the water line and total solids entering the sludge line were evaluated in order to calculate the total management cost of each scenario.
implemented in the model determined during the respirometric tests, compared to the ASM1Temp default values. The biological treatment compartment was initially modeled with an aerated ASU for each WWTP line. This configuration allowed for an accurate simulation of the COD and TSS removal, however the simulated N–NH4+ and N-NOx concentrations were very different from the data measured. For this reason, a modified configuration of the WWTP layout was implemented considering the denitrification contribution that takes place on the bottom of the oxidation tanks in addition to the hydrodynamics of the oxidation ditches. In the modified configuration, each oxidation ditch was modeled with a sequence of four aerated ASUs (12,428 m3) and four anoxic ASUs (16,572 m3). After calibrating the model it was validated using data from the January-February-March 2017 trimester. As regards the concentrations of the effluent TSS, COD, N–NH4+ and N-NOx, the ASM1Temp model was running and a good agreement was found in the values from the model prediction and the observed data as shown in Fig. 3. It can be noted how the model is able to simulate the separation efficiency of the secondary settler (Fig. 3a), which is generally quite low. The industrial origin of the wastewater may play a role in this low efficiency, since it induces the formation of small flocs with low settling velocity; moreover, surface aerators can cause floc breakup, affecting floc size and settleability. As far as the COD is concerned, the same variability observed in TSS is present in the total COD values (Fig. 3b), whereas when considering the soluble COD (Fig. 3c) the model predictions are quite accurate. Furthermore, the nitrification/denitrification process is predicted accurately (Fig. 3d–e). In fact, with few exceptions, the model is able to accurately predict peaks of N–NH4+ and N-NOx. The peaks in the N-NOx concentrations (and those in soluble COD concentrations) are due to unusual conditions occurring on the weekends when most industries are closed. Given the industrial nature of the wastewater, the COD and ammonium loads entering the plant are much lower than during the weekends and this leads to higher DO concentrations in the oxidation tank (as confirmed by the data observed) and consequently, better performance of the oxidation/nitrification tank. A similar situation occurred in the first week of simulation, when the N–NH4+ was particularly low and the NNOx particularly high: this was the first week of January when probably most industries were closed for the Christmas holydays, with consequently very low loads entering the plant (as confirmed by the COD observed in the influent). Such particular situations are impossible to simulate for the model. Another important difference between observed and simulated data was found during the last two weeks, when a significant increase in the N–NH4+ was observed in the real data but not predicted by the model. This phenomenon is quite unusual and may be due to the presence of inhibiting substances in the influent, which cannot be explained by the model. Using the data obtained from the model, mass balances at the steady state for COD and N were carried out and compared with those obtained using measured data. The results are presented in Fig. 4. As can be observed, the majority of the influent COD ends up in the sludge (in primary, secondary and tertiary sludge), 22% is oxidized to CO2 and the remaining 10% is released with the final effluent. These results differ from the ones obtained in another plant managed by the same company
2.4. Cost evaluation In the aim of finding the optimum SRT for minimization of the total treatment costs, costs related to each scenario were identified. For the water treatment line, only the costs related to the aeration were considered because all the others (such as those for water pumping or for reagents in the primary and tertiary treatments) were not considered to be dependent on the SRT. By increasing the SRT, and the solid load to the secondary clarifiers, an increase in the solid content of secondary effluent is expected; this could lead to an increase in the coagulant dosage in the tertiary treatment, however this aspect has not been taken into consideration in this work. As far as the sludge treatment line is concerned, the following costs were taken into account: i) consumption of flocculants (polyelectrolyte) used for chemical conditioning; ii) energy for centrifugation; ii) auxiliary fuel (methane) consumption and iv) costs related to ash disposal. The parameters and unitary costs used in this work were obtained from the effective plant data (year 2017) and are reported in Table 1. In particular, the prices for energy, polyelectrolyte, methane and ash disposal are obtained by dividing the total costs borne by the company in 2017 for the total assets used in the same period; the polyelectrolyte dosage is the mean dosage used by the company in 2017; the methane consumption is obtained by dividing the total methane used in 2017 by the total load of solids entering the incinerator in the same period. 3. Results and discussion 3.1. Influent characterization The wastewater entering the WWTP is characterized by a significant industrial contribution from the textile industry of Prato. The mean parameters of the wastewater are reported in Table 2. In order to run the model properly and obtain reliable results, it was necessary to gain more in-depth knowledge about the characteristics of the influent; for this reason, a fractionation of the influent COD and the influent nitrogen was carried out according to the methods described above. The results are reported in Fig. 2. From the fractionation of the COD it can be observed how the major COD part is in particulate form (about 63%). This is in contrast with Dulekgurgen et al. (2006) and Yetilmezsoy and Sapci-Zengin (2009) who found 72% and 57% respectively of the influent COD in soluble form in textile wastewater. As far as the biodegradability of the influent is concerned, it is important to underline that almost half the influent COD was nonbiodegradable (about 47%), with a significant percentage of soluble inert COD. Also in this case, it is in contrast with the results reported by Yetilmezsoy and Sapci-Zengin (2009) who found about 25% of nonbiodegradable COD.
Table 1 Parameters and unit costs used for evaluating the total treatment costs in different scenarios.
3.2. Model calibration and validation The calibration of the model was carried out using historical data for the January-February-March 2016 trimester. This step foresaw changes both of the parameters characterizing the various blocks (separation efficiencies, kinetic and stoichiometric parameters, etc.) and variations in the biological treatment configuration of the WWTP. Table 3 shows the kinetic and stoichiometric parameter values 5
Parameter
Value
U.M.
Energy price Polyelectrolyte price Polyelectrolyte dosage Methane price Methane consumption Ash disposal
0.134 1.50 21 0.29 0.45 0.14
€/kWh €/kg g/KgSS €/Nm3 Nm3/KgSS €/kg
Journal of Environmental Management 249 (2019) 109436
R. Muoio, et al.
Table 2 Characteristics of the influent of the Baciacavallo WWTP (obtained from a 24-h composite sample). Q, pH, TSS, COD and color are obtained from daily observation since year 2017; other parameters are obtained from three observations per week since year 2017. Parameter
Inflow Q pH TSS COD TN N_NH4+ N_NO3− N_NO2− TP Surfactants Color
Value
Table 3 Comparison between model parameter estimated by respirometric test and ASM1Temp default values.
U.M.
Mean ± St. Dev.
Maximum
Minimum
92,041 ± 24,852 7.8 ± 0.13 142 ± 84 284 ± 134 22 ± 5.3 12 ± 2.7 0.9 ± 0.9 0.13 ± 0.12 2.4 ± 0.8 11 ± 2.9 0.162 ± 0.072
153,526 8.1 898 1215 45 18 5.2 0.66 6 20 0.387
24,994 7.3 14 40 9 3 0.1 0.05 0.66 2.6 0.02
m3/d – mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L abs 420 nm
Parameter
Implemented value
Default value
Units
bA bH μA μH YA YH
0.09 0.76 0.68 4.2 0.25 0.76
0.15 0.62 0.8 6 0.24 0.67
[d−1] [d−1] [d−1] [d−1] [gCOD/gN] [gCOD/gCOD]
denitrification process is slightly over-estimated, probably because a specific denitrification section is not present in the actual plant: in this way process conditions (such as anoxic volumes within the oxidation tank and/or in the secondary clarifiers) change with the incoming loads and are unknown. 3.3. Optimization of the SRT
(Coppini et al., 2018), where the majority of the COD was oxidized to CO2. Such differences could be due to the different SRT of the two plants, which is quite low in this study (around 9d), while Coppini et al. (2018) worked with a very high SRT (over 100d). In any case, the COD mass balance evaluated is in accordance with the fractionation presented in paragraph 3.1, where it was assessed that about 63% of the inflow COD was in particulate form and about 12% was soluble inert COD. The model predicted very similar percentages, even though it slightly under-estimated the oxidation of the COD in the biological tank. As far as N is concerned, it is interesting to note how a nonnegligible percentage (about 8%) is lost due to denitrification, another 25% is released in the effluent, and about 67% ends up in the sludge. This last percentage is predicted well by the model, while the
Five steady-state simulations for each period (January and June) were carried out in order to evaluate the costs related to management of the water treatment line and sludge treatment line, according to the methods described in 2.4. The results are presented in Fig. 5. It is interesting to note how the total costs are lower in summer than in winter (7.2 €cent/m3 in summer and 8.7 €cent/m3 in winter, in scenario 0). In fact, while the costs for aeration were slightly higher in summer, the costs related to sludge treatment were lower due to the different management of the sludge flow in summer, which led to generally lower solid fluxes entering the sludge treatment line. As regards the range of SRTs tested, it is important to underline that it is not possible to increase the SRT over 11 d because, in the plant studied, the TSS in the secondary effluent increased with the longer SRT, making it
Fig. 2. a) Fractionation of influent COD; b) Fractionation of influent N among organic and inorganic; c) Fractionation of organic N; d) Fractionation of inorganic N.
6
Journal of Environmental Management 249 (2019) 109436
R. Muoio, et al.
Fig. 3. Results of the simulations (solid line) with respect to observed data (dots) for TSS (a), COD (b), Soluble COD (c), N–NH4+ (d) and N-NOx (e) in the effluent of the secondary treatment after calibration and validation.
impossible to retain solids in the biological compartment for more than 11 d. In addition, such a significant load of solids in the secondary effluent would result in an inacceptable increase in the chemicals to be dosed in the clariflocculation compartment. At the same time, it would not be possible to reduce the SRT below 7 d, as lower times would compromise the nitrification process. Nevertheless, a comparison of the results obtained at different SRTs shows how an increase in the SRT (in an acceptable range for the plant management) generally gave rise to a decrease in the total management costs. In fact, differences between scenario 0 and the scenario with the lowest total treatment costs (corresponding to an SRT of 11.4 d in winter and 10.0 d in summer) produce total savings of 44·000€/year in summer and 93·000€/year in winter. However, the differences among scenarios are not that relevant and the total costs do not vary much in the range tested. This is probably due to the type of influent, which, depending on the fractionation, has an important load of solids, and also to the presence of primary sedimentation which allows for the separation of these solids before entering the biological sections. In this way, a significant part of the
solids entering the sludge treatment line is represented by the solids present in the inflow, which do not vary with the SRT. 4. Conclusions The model developed in WEST® helped to identify, for the WWTP studied, the optimal SRT for minimizing the total costs of the wastewater treatment, which were 10d for summer and 11d for winter. In fact, by increasing the SRT from the actual 9d there would be total savings of 44·165€/year in summer and 93·110€/year in winter. In general, the model showed how the settleability of the sludge was quite low: in fact, the industrial origin of the wastewater induces the formation of small flocs, the dimensions of which can be further reduced by the presence of surface aerators. Such small flocs can hardly be retained inside the secondary settler and this causes an important loss of solids through the secondary settler with an increase in the SRTs. Indeed, the model showed that it is impossible to increase the SRT above 11d.
7
Journal of Environmental Management 249 (2019) 109436
R. Muoio, et al.
Fig. 4. Mass balances of a) COD, using measured data; b) COD, using results from the model; c) N using measured data and d) N, using results from the model. The percentages represent the destiny of the fractions of entering COD and nitrogen.
Fig. 5. Costs evaluation of different scenarios in winter (a) and summer (b).
Acknowledgements
heterotrophic endogenous respiration rate under aerobic and anoxic conditions. Water Sci. Technol. 38, 95–103. https://doi.org/10.1016/S0273-1223(98)00682-9. Baalbaki, Z., Torfs, E., Maere, T., Yargeau, V., Vanrolleghem, P.A., 2017. Dynamic modelling of solids in a full-scale activated sludge plant preceded by CEPT as a preliminary step for micropollutant removal modelling. Bioproc. Biosyst. Eng. 40, 499–510. https://doi.org/10.1007/s00449-016-1715-5. Benedetti, L., De Baets, B., Nopens, I., Vanrolleghem, P.A., 2010. Multi-criteria analysis of wastewater treatment plant design and control scenarios under uncertainty. Environ. Model. Softw 25, 616–621. https://doi.org/10.1016/J.ENVSOFT.2009.06.003. Coppini, E., Palli, L., Antal, A., Del Bubba, M., Miceli, E., Fani, R., Fibbi, D., 2019. Design and start-up of a constructed wetland as tertiary treatment for landfill leachates. Water Sci. Technol. 79, 145–155. https://doi.org/10.2166/wst.2019.030. Coppini, E., Palli, L., Fibbi, D., Gori, R., 2018. Long-term performance of a full-scale membrane plant for landfill leachate pretreatment: a case study. 2018. Membrane 8https://doi.org/10.3390/MEMBRANES8030052. Page 52 8, 52. Daigger, G.T., Roper, R.E., 1985. The relationship between SVI and activated sludge settling characteristics. J. Water Pollut. Control Fed. 57, 859–866. De Ketele, J., Davister, D., Ikumi, D.S., 2018. Applying performance indices in plantwide modelling for a comparative study of wastewater treatment plant operational strategies. Water S.A. 44, 539–550. https://doi.org/10.4314/wsa.v44i4.03. Dionisi, D., Rasheed, A.A., 2018. Maximisation of the organic load rate and minimisation of oxygen consumption in aerobic biological wastewater treatment processes by
The authors wish to sincerely thank all the Baciacavallo WWTP personnel. Appendix A. Supplementary data Supplementary data to this article can be found online at https:// doi.org/10.1016/j.jenvman.2019.109436. References Alaya, S. Ben, Haouech, L., Cherif, H., Shayeb, H., 2010. Aeration management in an oxidation ditch. Desalination 252, 172–178. https://doi.org/10.1016/J.DESAL.2009. 11.001. Andreottola, G., Foladori, P., Gelmini, A., Ziglio, G., 2002. Biomass active fraction evaluated by a direct method and respirometric techniques. Water Sci. Technol. 46, 371–379. https://doi.org/10.2166/wst.2002.0504. Avcioglu, E., Orhon, D., Sözen, S., 1998. A new method for the assessment of
8
Journal of Environmental Management 249 (2019) 109436
R. Muoio, et al. manipulation of the hydraulic and solids residence time. J. Water Process Eng. 22, 138–146. https://doi.org/10.1016/J.JWPE.2018.02.002. Drewnowski, J., Mąkinia, J., Szaja, A., Łagód, G., Kopeć, Ł., Aguilar, J.A., 2019. Comparative study of balancing SRT by using modified ASM2d in control and operation strategy at full-scale WWTP. Water 11, 485. https://doi.org/10.3390/ w11030485. Dulekgurgen, E., Doğruel, S., Karahan, Ö., Orhon, D., 2006. Size distribution of wastewater COD fractions as an index for biodegradability. Water Res. 40, 273–282. https://doi.org/10.1016/J.WATRES.2005.10.032. Fan, H., Liu, X., Wang, Hao, Han, Y., Qi, L., Wang, Hongchen, 2017. Oxygen transfer dynamics and activated sludge floc structure under different sludge retention times at low dissolved oxygen concentrations. Chemosphere 169, 586–595. https://doi.org/ 10.1016/J.CHEMOSPHERE.2016.10.137. Flores-Alsina, X., Arnell, M., Amerlinck, Y., Corominas, L., Gernaey, K.V., Guo, L., Lindblom, E., Nopens, I., Porro, J., Shaw, A., Snip, L., Vanrolleghem, P.A., Jeppsson, U., 2014. Balancing effluent quality, economic cost and greenhouse gas emissions during the evaluation of (plant-wide) control/operational strategies in WWTPs. Sci. Total Environ. 466–467, 616–624. https://doi.org/10.1016/J.SCITOTENV.2013.07. 046. Gernaey, K.V., Jeppsson, U., Batstone, D.J., Ingildsen, P., 2006. Impact of reactive settler models on simulated WWTP performance. Water Sci. Technol. 53, 159–167. https:// doi.org/10.2166/wst.2006.018. Hauduc, H., Rieger, L., Oehmen, A., van Loosdrecht, M.C.M., Comeau, Y., Héduit, A., Vanrolleghem, P.A., Gillot, S., 2013. Critical review of activated sludge modeling: state of process knowledge, modeling concepts, and limitations. Biotechnol. Bioeng. 110, 24–46. https://doi.org/10.1002/bit.24624. Henze, M., Grady, C., Gujer, W., Marais, G., Matsuo, T., 1987. Activated Sludge Model No.1. IAWPRC Scientific and Technical Report No.1, London. Hreiz, R., Latifi, M.A., Roche, N., 2015. Optimal design and operation of activated sludge processes: state-of-the-art. Chem. Eng. J. 281, 900–920. https://doi.org/10.1016/J. CEJ.2015.06.125. Karpinska, A.M., Bridgeman, J., 2016. CFD-aided modelling of activated sludge systems –
a critical review. Water Res. 88, 861–879. https://doi.org/10.1016/J.WATRES.2015. 11.008. Kim, D., Bowen, J.D., Ozelkan, E.C., 2015. Optimization of wastewater treatment plant operation for greenhouse gas mitigation. J. Environ. Manag. 163, 39–48. https://doi. org/10.1016/j.jenvman.2015.07.005. Liu, G., Wang, J., 2015. Modeling effects of DO and SRT on activated sludge decay and production. Water Res. 80, 169–178. https://doi.org/10.1016/J.WATRES.2015.04. 042. Newhart, K.B., Holloway, R.W., Hering, A.S., Cath, T.Y., 2019. Data-driven performance analyses of wastewater treatment plants: a review. Water Res. 157, 498–513. https:// doi.org/10.1016/J.WATRES.2019.03.030. Palli, L., Gullotto, A., Tilli, S., Gori, R., Lubello, C., Scozzafava, A., 2014. Effect of carbon source on the degradation of 2-Naphthalenesulfonic Acid polymers mixture by Pleurotus ostreatus in petrochemical wastewater. Process Biochem. 49, 2272–2278. Qiao, J., Zhang, W., 2018. Dynamic multi-objective optimization control for wastewater treatment process. Neural Comput. Appl. 29, 1261–1271. https://doi.org/10.1007/ s00521-016-2642-8. Randall, C.W., Buth, D., 1984. Nitrite build-up in activated sludge resulting from temperature effects. J. Water Pollut. Control Fed. 56, 1039–1044. Spanjers, H., Vanrolleghem, P., 1995. Respirometry as a tool for rapid characterization of wastewater and activated sludge. Water Sci. Technol. 31, 105–114. https://doi.org/ 10.1016/0273-1223(95)00184-O. Takács, I., Patry, G.G., Nolasco, D., 1991. A dynamic model of the clarification-thickening process. Water Res. 25, 1263–1271. https://doi.org/10.1016/0043-1354(91) 90066-Y. Wu, X., Yang, Y., Wu, G., Mao, J., Zhou, T., 2016. Simulation and optimization of a coking wastewater biological treatment process by activated sludge models (ASM). J. Environ. Manag. 165, 235–242. https://doi.org/10.1016/J.JENVMAN.2015.09.041. Yetilmezsoy, K., Sapci-Zengin, Z., 2009. Stochastic modeling applications for the prediction of COD removal efficiency of UASB reactors treating diluted real cotton textile wastewater. Stoch. Environ. Res. Risk Assess. 23, 13–26. https://doi.org/10.1007/ s00477-007-0191-5.
9
Contents lists available at ScienceDirect
Journal of Environmental Management journal homepage: www.elsevier.com/locate/jenvman
Research article
Optimization of a large industrial wastewater treatment plant using a modeling approach: A case study
T
Roberta Muoioa, Laura Pallia, , Iacopo Duccia, Ester Coppinib, Elena Bettazzib, Daniele Daddib, Donatella Fibbib, Riccardo Goria ⁎
a b
Department of Civil and Environmental Engineering, University of Florence, Via Santa Marta 3, Florence, Italy G.I.D.A. SpA, Via di Baciacavallo 36, Prato, Italy
ARTICLE INFO
ABSTRACT
Keywords: Sewage sludge Wastewater treatment plant Case study Cost optimization West by DHI Modeling
The objective of this paper is to find the optimum solid retention time (SRT) of a wastewater treatment plant (WWTP), which minimizes operating costs, using a modeling approach with WEST software by MIKE DHI®. For the determination of the kinetic and stoichiometric parameters (used for the correct calibration of the model implemented), respirometric and kinetic batch tests were carried out. Each Oxidation ditch was modeled by a sequence of four aerated activated sludge units (ASUs) and four anoxic ASUs with recirculation. The model is able to simulate the separation efficiency of the secondary settler, which is generally quite low: in fact, the industrial origin of the wastewater induces the formation of small flocs, the dimensions of which can be further reduced by the presence of surface aerators. The nitrification/denitrification process is also accurately predicted. Using data obtained from the model, mass balances at the steady state for COD and N were made and compared to the ones obtained using measured data. After calibration and validation of the model, steady-state simulations were carried out by increasing and decreasing the SRT of the system under two different operational conditions used by the managing company and by evaluating the costs related to the water treatment line and the sludge treatment line for each scenario. It is interesting to note how the total costs are lower in summer than in winter (7.2 €cent/m3 in summer and 8.7 €cent/m3 in winter, in scenario 0). In general, the increase in the SRT led to a decrease in the total management costs. In fact, differences between scenario 0 and the scenario with the lowest total treatment costs (corresponding to an SRT of 11.4 d in winter and 10.0 d in summer) could give rise to total savings of about 44·000€/year in summer and 93·000€/year in winter.
1. Introduction Nowadays, activated sludge (AS) is affirmed as the most common process for wastewater treatment. The quantification of environmental and economic performance has been of major interest in developing control strategies for the AS process (Benedetti et al., 2010; FloresAlsina et al., 2014); in fact, one of the main challenges in the management of a wastewater treatment plant (WWTP) is to minimize the operating costs while satisfying the effluent discharge limits. Optimization of the process, which is the key for successful management of a WWTP, can be achieved by seeking optimal process conditions such as solid retention time (SRT), aeration rate, and internal recycle flow rate (De Ketele et al., 2018; Kim et al., 2015; Newhart et al., 2019; Qiao and Zhang, 2018). For WWTPs, the optimization methods cannot be applied accurately without a mathematical modeling approach (Hreiz et al., 2015). Moreover, once developed, these models can be used for design
⁎
and control purposes, process upgrades and the evaluation of greenhouse gas emissions (Baalbaki et al., 2017). The two essential constituents for modeling an activated sludge plant are the biological reactors and the hydraulic component. As far as the former is concerned, many advanced mathematical models have been developed and proposed in the past, such as those developed by IWA (International Water Association), known as ASM1, ASM2, ASM2d, and ASM3 (Hauduc et al., 2013). These models are considered good solutions for correlating the complexity of the activated sludge processes and predicting the biological treatment efficiency in dynamic conditions (Wu et al., 2016). The hydraulic component describes the tank volumes and behavior of the flow rates. The secondary settler can be modeled in different ways, from a simple ideal separator, with no retention time, to more complex models such as reactive settlers (Gernaey et al., 2006) or CFD-aided models (Karpinska and Bridgeman, 2016). Modeling is an important part of the WWTP design and operation because it makes it
Corresponding author. E-mail address: [email protected] (L. Palli).
https://doi.org/10.1016/j.jenvman.2019.109436 Received 21 May 2019; Received in revised form 14 August 2019; Accepted 18 August 2019 0301-4797/ © 2019 Elsevier Ltd. All rights reserved.
Journal of Environmental Management 249 (2019) 109436
R. Muoio, et al.
List of abbreviations AOB AS ASM BOD CFD COD DO F/M HRT IWA MLSS NOB OUR SRT SVI TN TP TSS WWTP
bA YH YA μH,max μA,max μAOB,max
Ammonium Oxidizing Bacteria Activated sludge Activated Sludge Model Biochemical Oxygen Demand Computational Fluid Dynamics Chemical Oxygen Demand Dissolved Oxygen Food-to-Microorganism Ratio Hydraulic Retention Time International Water Association Mixed Liquor Suspended Solids Nitrite Oxidizing Bacteria Oxygen Uptake Rate Solid Retention Time Sludge Volume Index Total Nitrogen Total Phosphorous Total suspended solids Wastewater Treatment Plant
μNOB,max XBH XAOB XNOB SS SI XS XI N–NH4+ N–NO2 N–NO3 N-NOx sbN snbN pbN pnbN
List of symbols bH
Endogenous decay coefficient of autotrophic biomass Heterotrophic biomass yield Autotrophic biomass yield Heterotrophic maximum specific growth rate Autotrophic maximum specific growth rate Maximum specific growth rate of ammonium-oxidizing bacteria Maximum growth rate of nitrite-oxidizing bacteria Active heterotrophic biomass concentration Active AOB biomass concentration Active NOB biomass concentration Soluble Biodegradable COD Soluble Non-Biodegradable COD Particulate Biodegradable COD Particulate Non-Biodegradable COD Ammoniacal nitrogen Nitrous nitrogen Nitric nitrogen Sum of nitrous and nitric nitrogen Soluble Biodegradable Nitrogen Soluble Non-Biodegradable Nitrogen Particulate Biodegradable Nitrogen Particulate Non-Biodegradable Nitrogen
Endogenous decay coefficient of heterotrophic biomass
possible to better understand some of the process mechanisms and to predict the behavior of the plant in different scenarios (Hreiz et al., 2015). In this context, the objective of this work is to optimize the operating conditions of the Baciacavallo WWTP, located in Tuscany and managed by G.I.D.A. S.p.A. In particular, the objective of the work is to
find the optimum SRT for minimizing the operating costs. The SRT has been chosen as the optimization parameter because it can shift costs between the water treatment line (for aeration, water pumping and chemicals if dosed) and the sludge treatment line (related, for example, to energy for dewatering, chemicals, and disposal) (Dionisi and
Fig. 1. Flowchart of the WWTP. Light grey line: wastewater and recycled flow; dark grey line: sludge. ST: Septic Tank; Se: external Sludge; M: flowmeter; S: Sampling. Sections from 1 to 11 represent water treatment line. Sections from 13 to 14 represent sludge treatment line. 1: Coarse bar screen; 2: pumping; 3a and 3b: fine bar screen, line a and b; 4a and 4b: grit removal line a and b; 5a and 5b: flocculation line a and b; 6a and 6b: primary sedimentation line a and b; 7: flow equalization; 8a and 8b: Oxidation ditches line a and b; 9a and 9b: secondary sedimentation line a and b; 10a and 10b: clariflocculation line a and b; 11: Ozonation; 12: thickening; 13: centrifugation; 14: incineration. 2
Journal of Environmental Management 249 (2019) 109436
R. Muoio, et al.
N–NO3-, TP, surfactants, color and metals.
Rasheed, 2018; Drewnowski et al., 2019; Fan et al., 2017; Liu and Wang, 2015). Indeed, with an increase in the SRT, more solids are retained in the oxidation tank and more oxygen is required (increasing costs related to the water treatment line), but at the same time, more solids are oxidized in the oxidation tank, leading to a minor amount of solids entering the sludge treatment line and minor costs related to the same. On the other hand, a reduction in the SRT would lead to an increase in these costs but with a reduction in oxygen requirements in the water treatment line. The Baciacavallo plant (900.000 PE) treats urban wastewater (from combined sewer) and textile wastewater (about 70% of the total flow rate) jointly. The current treatment train includes pretreatment, primary settler, secondary and tertiary treatment (clariflocculation and ozonation), while the sludge treatment line includes thickening, centrifugation and lastly, incineration. The unusual sludge treatment line (due to the absence of stabilization) gives rise to high costs related to sludge management. For this reason, it is of paramount importance to optimize the process conditions in order to contain the sludge management costs. In this paper, an attempt has been made to find the optimum SRT using a modeling approach with WEST software by MIKE DHI®.
2.2. Experimental In order to determine the kinetic and stoichiometric parameters, respirometric and kinetic batch tests were carried out. The former were conducted to determine the heterotrophic biomass parameters (bH, YH, μH,max), and the latter, the autotrophic biomass (μA,max). The results were used to correctly calibrate the model implemented. 2.2.1. Respirometric tests Three respirometric tests were performed using a modified version of a previously reported protocol (Coppini et al., 2019; Palli et al., 2014). For each test, an experimental 1.2L reactor was filled with AS collected from the four oxidation tanks (mixed in even parts), and preaerated for 24 h in order to assure endogenous conditions. During the test, the temperature was maintained at 20 °C, and the pH at 7.5 by dosing NaOH or HCl, and mixing was assured via use of a magnetic stirrer. In each test, 2 mL sodium acetate (24 gCOD/L) were spiked in the reactor as a rapidly biodegradable carbon substrate, maintaining a food-to-microorganism ratio (F/M) of 0.05 (Andreottola et al., 2002). All the tests were performed with the addition of 15 mL of ATU (1 g/L) to inhibit nitrification. The respirometric system developed enabled remote-controlled and automated online monitoring of the O2 consumption, pH, temperature and oxygen uptake rate (OUR).
2. Methodology 2.1. Description of the case study The Baciacavallo WWTP, located in Prato (Central Italy), treats both industrial and municipal wastewater with an average inflow of 100·000 m3/d. Industrial wastewater mostly originates from the textile manufacturing processes and represents 70% of the total influent flow, while the remaining amount (30%) is represented by domestic wastewater and sludge from septic tanks (0.07%). The WWTP has a flow capacity of 144·000 m3/d and is designed for 900·000 PE. The treatment process is divided into two lines (line a and line b), with biological treatment consisting of four oxidation ditches (two per line) for the organic matter removal and nitrification. The treatment train (Fig. 1) is composed by coarse bar screens, pumping, fine bar screens, grit removal, primary sedimentation (where the residual water from the sludge treatment unit is recycled), equalization (in order to balance the flow variations), then biological oxidation/nitrification, clariflocculation and final ozonation. Sludge from septic tanks is introduced upstream of the biological treatment. The biological treatment consists of four oxidation ditches (with a total volume of 29·000 m3) and four secondary sedimentation tanks (with a total volume of 24·308 m3). The underflow from the secondary sedimentation tanks flows back to the oxidation ditches at a ratio of 1.2:1.4 in order to supply the nutrient requirements of the bacteria and maintain an adequate solid concentration. During the process operation, the concentration of dissolved oxygen (DO) is maintained at approximately 1.5 mg/L. Oxygen transfer is carried out by six surface rotor-brushes for each oxidation ditch. The sludge resulting from secondary and tertiary treatments can be recycled in primary sedimentation and removed together with the primary sludge, otherwise it can be sent into the sludge treatment line separately. The sludge is processed by means of gravity thickening (with 3 thickeners), centrifugation with pre-conditioning by cationic polyelectrolytes (by 2 dewatering units), and finally, incineration. In this section the sludge is treated together with the sludge coming from another WWTP managed by the same company. Thickener overflow and centrate are recycled into the primary sedimentation unit. Fig. 1 also shows the position of the flow meters and the sampling points located all along the treatment train. All samplers are automatic except for the sampling points in the oxidation tanks, where manual samples are taken three times a day in order to control the MLSS (Mixed Liquor Suspended Solids) and SVI (Sludge Volume Index). The analysis of wastewater entails the determining of the concentration of specific constituents each day by G.I.D.A S.p.A.‘s internal laboratory, including pH, COD, BOD, TSS, TN, N–NH4+, N–NO2-,
2.2.2. Biomass yield, YH, estimation For the estimate of the yield factor the following formula (Equation (1)) was used:
YH (mgCODcell / mgCODsub) = 1
O2 COD
(1)
where ΔCOD is the COD consumed (calculated as the difference between the soluble COD in the system after spiking and at the end of the test), and ΔO2 is the oxygen consumed due to substrate degradation. This was evaluated as the difference between the total oxygen consumption (the area below the OURs measured) and the oxygen consumed due to endogenous respiration. 2.2.3. Endogenous decay coefficient estimation The procedure for estimating the endogenous decay coefficient, bH according to a previously reported protocol (Avcioglu et al., 1998; Spanjers and Vanrolleghem, 1995), considers the values of OURmax (maximum value of OUR during exogenous respiration) determined on different days after the spiking of 2 mL of sodium acetate solution. The plot of the natural logarithm of the recorded OURmax values vs. Time shows an exponential decrease in the biomass as a straight line with the slope bH. 2.2.4. Maximum specific heterotrophic bacterial growth-rate estimation For estimating the maximum specific heterotrophic bacterial growth rate (μH,max) the following formula (Equation (2)) was used:
µH , max =
OURmax ·YH (1 YH )·XBH
(2)
where XBH (gSSV/L) represents the active heterotrophic biomass concentration in the reactor, evaluated by (Equation (3)):
XBH = YH ·
SRT ·(S0 S ) HRT ·(1 + bH ·SRT )
(3)
where SRT (d) is the sludge retention time; S0 (mg/L) is the average daily concentration of COD entering the biological treatment; S (mg/L) is the average daily concentration of COD leaving the biological treatment; HRT (h) is the hydraulic retention time. 3
Journal of Environmental Management 249 (2019) 109436
R. Muoio, et al.
2.2.5. Batch tests For the characterization of the autotrophic biomass, a kinetic batch test was performed. For the respirometric tests, an experimental 1.6 L reactor was filled with activated sludge collected from the four oxidation tanks (mixed in even aliquots), and pre-aerated for 24 h in order to assure endogenous conditions. During the test, the pH was maintained at 7.5 by dosing NaOH or HCl, and mixed conditions were assured via use of a mechanical stirrer. Oxygen was inflated by means of a fine bubble diffuser and at the beginning of the tests, ClNH4 was dosed until reaching a final concentration of 20 mg/L. Samples were then taken at regular intervals (every 15 min) and immediately filtrated with 0.45 μm syringe filters. The ammonium, nitrate and nitrite were analyzed in each sample.
2.3.2. The set-up the WWTP layout The treatment train was simulated from the primary sedimentation up to the clariflocculation steps. The primary settler was modeled by the Primary Point Settler Block: the settler is a phase separator and has no real volume or hydraulic retention time. In this unit, the effluent particulate concentration is determined as a fraction (f_ns) of the influent concentration. The underflow concentration is based on a mass balance. This fraction was set to 0.54, according to an experimental mass balance carried out in the primary settler (considering all the solids entering from the influent, secondary and/or tertiary sludge, supernatants from the sludge treatment line, and all the solids exiting with the effluent). In WEST, each Oxidation ditch has been modeled by a sequence with four aerated Activated Sludge Units (ASUs) and four anoxic ASUs with recirculation, as reported by other authors (Alaya et al., 2010). The choice to implement the anoxic ASUs, despite the absence of a proper anoxic section in the plant, was driven by the experimental data, which showed how part of the influent nitrogen was lost, probably due to denitrification occurring inside the oxidation tank. In fact, experimental measures of DO in the oxidation tanks showed how the lower part of the tank was actually anoxic, causing a partial denitrification of the nitrates. An anaerobic ASU was installed downstream from each secondary settler in order to simulate the spontaneous denitrification processes that take place during secondary sedimentation. The secondary settlers were modeled by the Takács-SVI Block: they predict the solid concentration profile in the settler by dividing the settler into a number of layers of constant thicknesses and creating a balance of the solids around each layer (Takács et al., 1991). This model is an extension of the Takács model in which the hindered settling parameter (r_H) is estimated via SVI measurement (Daigger and Roper, 1985). The parameters r_P (low concentration settling parameter (m3/ g)) and v0 (Maximum theoretical settling velocity (m/d)) were calibrated using the Parameter Estimation (PE) function of the software and were equal to 0.00089 m3/g and 306 m/d respectively. The clariflocculation unit was modeled by the Efficiency Thickener Block: the separation of particulate fractions is described through a mass balance and the definition of the separation efficiency (e_X), that is, the fraction of solids going into the underflow sludge. This fraction was set to 0.87 according to an experimental mass balance made in the clariflocculation unit (considering the solids entering from the biological treatment outlet and those exiting with the effluent). The sludge treatment unit, consisting of thickeners and centrifuges, was modeled as a single Efficiency Thickener Block similar to the clariflocculators.
2.2.6. Maximum specific autotrophic bacterial growth-rate estimation For estimating the maximum specific autotrophic bacterial growth rate (μA,max) the following formulas (Equations (4)–(6)) were used:
µAOB, max = YAOB ·
µNOB, max = YNOB·
d (N
d (N
NH4+) 24 · dt XAOB
(4)
NO2 ) 24 · dt XNOB
(5)
µA, max = µAOB, max + µNOB, max
(6)
where, μAOB,max (d−1) and μNOB,max (d−1) are the Ammonium Oxidizing Bacteria (AOB) and Nitrite Oxidizing Bacteria (NOB) maximum specific growth rates respectively; d(N–NH4+)/dt and d(N–NO2-)/dt (mg/l·h) represent the ammonium and nitrite consumption rates respectively; XAOB and XNOB (gSSV/L) represent the AOB and NOB active biomass concentration in the reactor respectively; YAOB and YNOB (mgSSV/mgN) are respectively the AOB and NOB yield factors, considered equal to 0.13 and 0.04 (Randall and Buth, 1984). XAOB and XNOB were evaluated through a mass balance at the steady state, considering an ammonium consumption for the AOB equal to the difference between N–NH4+ in the influent and the effluent of the plant, and a nitrite consumption for the NOB equal to the difference between the ammonium consumption (considered equal to the nitrite production) and the N–NO2- in the effluent of the plant. 2.2.7. COD and nitrogen fractionation The fractions of wastewater COD and N were determined in order to obtain more in-depth knowledge of the influent characteristics and ensure more accurate results from the model. The fractions of COD and N were evaluated following the protocol reported in the Supplementary Material.
2.3.3. Scenario simulations Steady-state simulations were carried out by varying the waste sludge flow rate of the biological treatment and consequently, by increasing and decreasing the SRT in order to find the optimum SRT for minimizing the treatment costs. The company running the WWTP uses two different strategies to manage the secondary waste sludge: in winter this is sent directly to the sludge treatment line, while in summer it is recirculated in the primary settler and then removed together with the primary sludge. For this reason, two different sets of scenarios were simulated (by varying the waste sludge flow rate), one with data from January 2017 (representative of winter management) and one with data from June 2017 (representative of summer management). As far as winter season is concerned scenario 0 was the one with SRT 9.0 d and then 4 other scenarios were simulated, with 7.7 d, 7.2 d, 10.7 d and 11.7 d, respectively. As far as summer season is concerned scenario 0 was the one with SRT 9.0 d and the 4 other scenarios were simulated with 7.9 d, 7.3 d, 9.9 d and 10.1 d, respectively. The aeration efficiency (gO2/kWh) was calculated by dividing the oxygen consumption (in kgO2/d), obtained from the model in scenario 0 for the mean daily consumption of electric energy related to the
2.3. Simulation in WEST® by DHI software In this study, the WWTP's configuration was implemented in WEST 2016 ® (MIKE DHI), using the available data on plant characteristics, wastewater quality, and respirometric and kinetic tests results. 2.3.1. The WEST environment In this study, an ASM1Temp model was considered for the configuration studied. This is an extension of the original ASM1 model published by the International Association on Water Quality (IAWQ) Task Group on Mathematical Modeling for Design and Operation of Biological Wastewater Treatment Processes (Henze et al., 1987), which furthermore considers temperature correction and ammonium limitation for the aerobic and anoxic growth of heterotrophs. An input file was generated containing the characteristics of the influent wastewater. All the plant data used in this study were taken from routine operating records. The COD and N components were determined via fractionation as reported above. 4
Journal of Environmental Management 249 (2019) 109436
R. Muoio, et al.
aeration system (kWh/d). The mean daily kWhs were calculated separately for the two simulated seasons by using real data from January 2017 and June 2017. These aeration efficiencies, which were 706 gO2/ kWh in January and 593 gO2/kWh in June, were kept constant in the other scenarios. From the simulation scenarios, the total oxygen required from the water line and total solids entering the sludge line were evaluated in order to calculate the total management cost of each scenario.
implemented in the model determined during the respirometric tests, compared to the ASM1Temp default values. The biological treatment compartment was initially modeled with an aerated ASU for each WWTP line. This configuration allowed for an accurate simulation of the COD and TSS removal, however the simulated N–NH4+ and N-NOx concentrations were very different from the data measured. For this reason, a modified configuration of the WWTP layout was implemented considering the denitrification contribution that takes place on the bottom of the oxidation tanks in addition to the hydrodynamics of the oxidation ditches. In the modified configuration, each oxidation ditch was modeled with a sequence of four aerated ASUs (12,428 m3) and four anoxic ASUs (16,572 m3). After calibrating the model it was validated using data from the January-February-March 2017 trimester. As regards the concentrations of the effluent TSS, COD, N–NH4+ and N-NOx, the ASM1Temp model was running and a good agreement was found in the values from the model prediction and the observed data as shown in Fig. 3. It can be noted how the model is able to simulate the separation efficiency of the secondary settler (Fig. 3a), which is generally quite low. The industrial origin of the wastewater may play a role in this low efficiency, since it induces the formation of small flocs with low settling velocity; moreover, surface aerators can cause floc breakup, affecting floc size and settleability. As far as the COD is concerned, the same variability observed in TSS is present in the total COD values (Fig. 3b), whereas when considering the soluble COD (Fig. 3c) the model predictions are quite accurate. Furthermore, the nitrification/denitrification process is predicted accurately (Fig. 3d–e). In fact, with few exceptions, the model is able to accurately predict peaks of N–NH4+ and N-NOx. The peaks in the N-NOx concentrations (and those in soluble COD concentrations) are due to unusual conditions occurring on the weekends when most industries are closed. Given the industrial nature of the wastewater, the COD and ammonium loads entering the plant are much lower than during the weekends and this leads to higher DO concentrations in the oxidation tank (as confirmed by the data observed) and consequently, better performance of the oxidation/nitrification tank. A similar situation occurred in the first week of simulation, when the N–NH4+ was particularly low and the NNOx particularly high: this was the first week of January when probably most industries were closed for the Christmas holydays, with consequently very low loads entering the plant (as confirmed by the COD observed in the influent). Such particular situations are impossible to simulate for the model. Another important difference between observed and simulated data was found during the last two weeks, when a significant increase in the N–NH4+ was observed in the real data but not predicted by the model. This phenomenon is quite unusual and may be due to the presence of inhibiting substances in the influent, which cannot be explained by the model. Using the data obtained from the model, mass balances at the steady state for COD and N were carried out and compared with those obtained using measured data. The results are presented in Fig. 4. As can be observed, the majority of the influent COD ends up in the sludge (in primary, secondary and tertiary sludge), 22% is oxidized to CO2 and the remaining 10% is released with the final effluent. These results differ from the ones obtained in another plant managed by the same company
2.4. Cost evaluation In the aim of finding the optimum SRT for minimization of the total treatment costs, costs related to each scenario were identified. For the water treatment line, only the costs related to the aeration were considered because all the others (such as those for water pumping or for reagents in the primary and tertiary treatments) were not considered to be dependent on the SRT. By increasing the SRT, and the solid load to the secondary clarifiers, an increase in the solid content of secondary effluent is expected; this could lead to an increase in the coagulant dosage in the tertiary treatment, however this aspect has not been taken into consideration in this work. As far as the sludge treatment line is concerned, the following costs were taken into account: i) consumption of flocculants (polyelectrolyte) used for chemical conditioning; ii) energy for centrifugation; ii) auxiliary fuel (methane) consumption and iv) costs related to ash disposal. The parameters and unitary costs used in this work were obtained from the effective plant data (year 2017) and are reported in Table 1. In particular, the prices for energy, polyelectrolyte, methane and ash disposal are obtained by dividing the total costs borne by the company in 2017 for the total assets used in the same period; the polyelectrolyte dosage is the mean dosage used by the company in 2017; the methane consumption is obtained by dividing the total methane used in 2017 by the total load of solids entering the incinerator in the same period. 3. Results and discussion 3.1. Influent characterization The wastewater entering the WWTP is characterized by a significant industrial contribution from the textile industry of Prato. The mean parameters of the wastewater are reported in Table 2. In order to run the model properly and obtain reliable results, it was necessary to gain more in-depth knowledge about the characteristics of the influent; for this reason, a fractionation of the influent COD and the influent nitrogen was carried out according to the methods described above. The results are reported in Fig. 2. From the fractionation of the COD it can be observed how the major COD part is in particulate form (about 63%). This is in contrast with Dulekgurgen et al. (2006) and Yetilmezsoy and Sapci-Zengin (2009) who found 72% and 57% respectively of the influent COD in soluble form in textile wastewater. As far as the biodegradability of the influent is concerned, it is important to underline that almost half the influent COD was nonbiodegradable (about 47%), with a significant percentage of soluble inert COD. Also in this case, it is in contrast with the results reported by Yetilmezsoy and Sapci-Zengin (2009) who found about 25% of nonbiodegradable COD.
Table 1 Parameters and unit costs used for evaluating the total treatment costs in different scenarios.
3.2. Model calibration and validation The calibration of the model was carried out using historical data for the January-February-March 2016 trimester. This step foresaw changes both of the parameters characterizing the various blocks (separation efficiencies, kinetic and stoichiometric parameters, etc.) and variations in the biological treatment configuration of the WWTP. Table 3 shows the kinetic and stoichiometric parameter values 5
Parameter
Value
U.M.
Energy price Polyelectrolyte price Polyelectrolyte dosage Methane price Methane consumption Ash disposal
0.134 1.50 21 0.29 0.45 0.14
€/kWh €/kg g/KgSS €/Nm3 Nm3/KgSS €/kg
Journal of Environmental Management 249 (2019) 109436
R. Muoio, et al.
Table 2 Characteristics of the influent of the Baciacavallo WWTP (obtained from a 24-h composite sample). Q, pH, TSS, COD and color are obtained from daily observation since year 2017; other parameters are obtained from three observations per week since year 2017. Parameter
Inflow Q pH TSS COD TN N_NH4+ N_NO3− N_NO2− TP Surfactants Color
Value
Table 3 Comparison between model parameter estimated by respirometric test and ASM1Temp default values.
U.M.
Mean ± St. Dev.
Maximum
Minimum
92,041 ± 24,852 7.8 ± 0.13 142 ± 84 284 ± 134 22 ± 5.3 12 ± 2.7 0.9 ± 0.9 0.13 ± 0.12 2.4 ± 0.8 11 ± 2.9 0.162 ± 0.072
153,526 8.1 898 1215 45 18 5.2 0.66 6 20 0.387
24,994 7.3 14 40 9 3 0.1 0.05 0.66 2.6 0.02
m3/d – mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L abs 420 nm
Parameter
Implemented value
Default value
Units
bA bH μA μH YA YH
0.09 0.76 0.68 4.2 0.25 0.76
0.15 0.62 0.8 6 0.24 0.67
[d−1] [d−1] [d−1] [d−1] [gCOD/gN] [gCOD/gCOD]
denitrification process is slightly over-estimated, probably because a specific denitrification section is not present in the actual plant: in this way process conditions (such as anoxic volumes within the oxidation tank and/or in the secondary clarifiers) change with the incoming loads and are unknown. 3.3. Optimization of the SRT
(Coppini et al., 2018), where the majority of the COD was oxidized to CO2. Such differences could be due to the different SRT of the two plants, which is quite low in this study (around 9d), while Coppini et al. (2018) worked with a very high SRT (over 100d). In any case, the COD mass balance evaluated is in accordance with the fractionation presented in paragraph 3.1, where it was assessed that about 63% of the inflow COD was in particulate form and about 12% was soluble inert COD. The model predicted very similar percentages, even though it slightly under-estimated the oxidation of the COD in the biological tank. As far as N is concerned, it is interesting to note how a nonnegligible percentage (about 8%) is lost due to denitrification, another 25% is released in the effluent, and about 67% ends up in the sludge. This last percentage is predicted well by the model, while the
Five steady-state simulations for each period (January and June) were carried out in order to evaluate the costs related to management of the water treatment line and sludge treatment line, according to the methods described in 2.4. The results are presented in Fig. 5. It is interesting to note how the total costs are lower in summer than in winter (7.2 €cent/m3 in summer and 8.7 €cent/m3 in winter, in scenario 0). In fact, while the costs for aeration were slightly higher in summer, the costs related to sludge treatment were lower due to the different management of the sludge flow in summer, which led to generally lower solid fluxes entering the sludge treatment line. As regards the range of SRTs tested, it is important to underline that it is not possible to increase the SRT over 11 d because, in the plant studied, the TSS in the secondary effluent increased with the longer SRT, making it
Fig. 2. a) Fractionation of influent COD; b) Fractionation of influent N among organic and inorganic; c) Fractionation of organic N; d) Fractionation of inorganic N.
6
Journal of Environmental Management 249 (2019) 109436
R. Muoio, et al.
Fig. 3. Results of the simulations (solid line) with respect to observed data (dots) for TSS (a), COD (b), Soluble COD (c), N–NH4+ (d) and N-NOx (e) in the effluent of the secondary treatment after calibration and validation.
impossible to retain solids in the biological compartment for more than 11 d. In addition, such a significant load of solids in the secondary effluent would result in an inacceptable increase in the chemicals to be dosed in the clariflocculation compartment. At the same time, it would not be possible to reduce the SRT below 7 d, as lower times would compromise the nitrification process. Nevertheless, a comparison of the results obtained at different SRTs shows how an increase in the SRT (in an acceptable range for the plant management) generally gave rise to a decrease in the total management costs. In fact, differences between scenario 0 and the scenario with the lowest total treatment costs (corresponding to an SRT of 11.4 d in winter and 10.0 d in summer) produce total savings of 44·000€/year in summer and 93·000€/year in winter. However, the differences among scenarios are not that relevant and the total costs do not vary much in the range tested. This is probably due to the type of influent, which, depending on the fractionation, has an important load of solids, and also to the presence of primary sedimentation which allows for the separation of these solids before entering the biological sections. In this way, a significant part of the
solids entering the sludge treatment line is represented by the solids present in the inflow, which do not vary with the SRT. 4. Conclusions The model developed in WEST® helped to identify, for the WWTP studied, the optimal SRT for minimizing the total costs of the wastewater treatment, which were 10d for summer and 11d for winter. In fact, by increasing the SRT from the actual 9d there would be total savings of 44·165€/year in summer and 93·110€/year in winter. In general, the model showed how the settleability of the sludge was quite low: in fact, the industrial origin of the wastewater induces the formation of small flocs, the dimensions of which can be further reduced by the presence of surface aerators. Such small flocs can hardly be retained inside the secondary settler and this causes an important loss of solids through the secondary settler with an increase in the SRTs. Indeed, the model showed that it is impossible to increase the SRT above 11d.
7
Journal of Environmental Management 249 (2019) 109436
R. Muoio, et al.
Fig. 4. Mass balances of a) COD, using measured data; b) COD, using results from the model; c) N using measured data and d) N, using results from the model. The percentages represent the destiny of the fractions of entering COD and nitrogen.
Fig. 5. Costs evaluation of different scenarios in winter (a) and summer (b).
Acknowledgements
heterotrophic endogenous respiration rate under aerobic and anoxic conditions. Water Sci. Technol. 38, 95–103. https://doi.org/10.1016/S0273-1223(98)00682-9. Baalbaki, Z., Torfs, E., Maere, T., Yargeau, V., Vanrolleghem, P.A., 2017. Dynamic modelling of solids in a full-scale activated sludge plant preceded by CEPT as a preliminary step for micropollutant removal modelling. Bioproc. Biosyst. Eng. 40, 499–510. https://doi.org/10.1007/s00449-016-1715-5. Benedetti, L., De Baets, B., Nopens, I., Vanrolleghem, P.A., 2010. Multi-criteria analysis of wastewater treatment plant design and control scenarios under uncertainty. Environ. Model. Softw 25, 616–621. https://doi.org/10.1016/J.ENVSOFT.2009.06.003. Coppini, E., Palli, L., Antal, A., Del Bubba, M., Miceli, E., Fani, R., Fibbi, D., 2019. Design and start-up of a constructed wetland as tertiary treatment for landfill leachates. Water Sci. Technol. 79, 145–155. https://doi.org/10.2166/wst.2019.030. Coppini, E., Palli, L., Fibbi, D., Gori, R., 2018. Long-term performance of a full-scale membrane plant for landfill leachate pretreatment: a case study. 2018. Membrane 8https://doi.org/10.3390/MEMBRANES8030052. Page 52 8, 52. Daigger, G.T., Roper, R.E., 1985. The relationship between SVI and activated sludge settling characteristics. J. Water Pollut. Control Fed. 57, 859–866. De Ketele, J., Davister, D., Ikumi, D.S., 2018. Applying performance indices in plantwide modelling for a comparative study of wastewater treatment plant operational strategies. Water S.A. 44, 539–550. https://doi.org/10.4314/wsa.v44i4.03. Dionisi, D., Rasheed, A.A., 2018. Maximisation of the organic load rate and minimisation of oxygen consumption in aerobic biological wastewater treatment processes by
The authors wish to sincerely thank all the Baciacavallo WWTP personnel. Appendix A. Supplementary data Supplementary data to this article can be found online at https:// doi.org/10.1016/j.jenvman.2019.109436. References Alaya, S. Ben, Haouech, L., Cherif, H., Shayeb, H., 2010. Aeration management in an oxidation ditch. Desalination 252, 172–178. https://doi.org/10.1016/J.DESAL.2009. 11.001. Andreottola, G., Foladori, P., Gelmini, A., Ziglio, G., 2002. Biomass active fraction evaluated by a direct method and respirometric techniques. Water Sci. Technol. 46, 371–379. https://doi.org/10.2166/wst.2002.0504. Avcioglu, E., Orhon, D., Sözen, S., 1998. A new method for the assessment of
8
Journal of Environmental Management 249 (2019) 109436
R. Muoio, et al. manipulation of the hydraulic and solids residence time. J. Water Process Eng. 22, 138–146. https://doi.org/10.1016/J.JWPE.2018.02.002. Drewnowski, J., Mąkinia, J., Szaja, A., Łagód, G., Kopeć, Ł., Aguilar, J.A., 2019. Comparative study of balancing SRT by using modified ASM2d in control and operation strategy at full-scale WWTP. Water 11, 485. https://doi.org/10.3390/ w11030485. Dulekgurgen, E., Doğruel, S., Karahan, Ö., Orhon, D., 2006. Size distribution of wastewater COD fractions as an index for biodegradability. Water Res. 40, 273–282. https://doi.org/10.1016/J.WATRES.2005.10.032. Fan, H., Liu, X., Wang, Hao, Han, Y., Qi, L., Wang, Hongchen, 2017. Oxygen transfer dynamics and activated sludge floc structure under different sludge retention times at low dissolved oxygen concentrations. Chemosphere 169, 586–595. https://doi.org/ 10.1016/J.CHEMOSPHERE.2016.10.137. Flores-Alsina, X., Arnell, M., Amerlinck, Y., Corominas, L., Gernaey, K.V., Guo, L., Lindblom, E., Nopens, I., Porro, J., Shaw, A., Snip, L., Vanrolleghem, P.A., Jeppsson, U., 2014. Balancing effluent quality, economic cost and greenhouse gas emissions during the evaluation of (plant-wide) control/operational strategies in WWTPs. Sci. Total Environ. 466–467, 616–624. https://doi.org/10.1016/J.SCITOTENV.2013.07. 046. Gernaey, K.V., Jeppsson, U., Batstone, D.J., Ingildsen, P., 2006. Impact of reactive settler models on simulated WWTP performance. Water Sci. Technol. 53, 159–167. https:// doi.org/10.2166/wst.2006.018. Hauduc, H., Rieger, L., Oehmen, A., van Loosdrecht, M.C.M., Comeau, Y., Héduit, A., Vanrolleghem, P.A., Gillot, S., 2013. Critical review of activated sludge modeling: state of process knowledge, modeling concepts, and limitations. Biotechnol. Bioeng. 110, 24–46. https://doi.org/10.1002/bit.24624. Henze, M., Grady, C., Gujer, W., Marais, G., Matsuo, T., 1987. Activated Sludge Model No.1. IAWPRC Scientific and Technical Report No.1, London. Hreiz, R., Latifi, M.A., Roche, N., 2015. Optimal design and operation of activated sludge processes: state-of-the-art. Chem. Eng. J. 281, 900–920. https://doi.org/10.1016/J. CEJ.2015.06.125. Karpinska, A.M., Bridgeman, J., 2016. CFD-aided modelling of activated sludge systems –
a critical review. Water Res. 88, 861–879. https://doi.org/10.1016/J.WATRES.2015. 11.008. Kim, D., Bowen, J.D., Ozelkan, E.C., 2015. Optimization of wastewater treatment plant operation for greenhouse gas mitigation. J. Environ. Manag. 163, 39–48. https://doi. org/10.1016/j.jenvman.2015.07.005. Liu, G., Wang, J., 2015. Modeling effects of DO and SRT on activated sludge decay and production. Water Res. 80, 169–178. https://doi.org/10.1016/J.WATRES.2015.04. 042. Newhart, K.B., Holloway, R.W., Hering, A.S., Cath, T.Y., 2019. Data-driven performance analyses of wastewater treatment plants: a review. Water Res. 157, 498–513. https:// doi.org/10.1016/J.WATRES.2019.03.030. Palli, L., Gullotto, A., Tilli, S., Gori, R., Lubello, C., Scozzafava, A., 2014. Effect of carbon source on the degradation of 2-Naphthalenesulfonic Acid polymers mixture by Pleurotus ostreatus in petrochemical wastewater. Process Biochem. 49, 2272–2278. Qiao, J., Zhang, W., 2018. Dynamic multi-objective optimization control for wastewater treatment process. Neural Comput. Appl. 29, 1261–1271. https://doi.org/10.1007/ s00521-016-2642-8. Randall, C.W., Buth, D., 1984. Nitrite build-up in activated sludge resulting from temperature effects. J. Water Pollut. Control Fed. 56, 1039–1044. Spanjers, H., Vanrolleghem, P., 1995. Respirometry as a tool for rapid characterization of wastewater and activated sludge. Water Sci. Technol. 31, 105–114. https://doi.org/ 10.1016/0273-1223(95)00184-O. Takács, I., Patry, G.G., Nolasco, D., 1991. A dynamic model of the clarification-thickening process. Water Res. 25, 1263–1271. https://doi.org/10.1016/0043-1354(91) 90066-Y. Wu, X., Yang, Y., Wu, G., Mao, J., Zhou, T., 2016. Simulation and optimization of a coking wastewater biological treatment process by activated sludge models (ASM). J. Environ. Manag. 165, 235–242. https://doi.org/10.1016/J.JENVMAN.2015.09.041. Yetilmezsoy, K., Sapci-Zengin, Z., 2009. Stochastic modeling applications for the prediction of COD removal efficiency of UASB reactors treating diluted real cotton textile wastewater. Stoch. Environ. Res. Risk Assess. 23, 13–26. https://doi.org/10.1007/ s00477-007-0191-5.
9