Environ Sci Pollut Res (2011) 18:192–198 DOI 10.1007/s11356-010-0362-7

RESEARCH ARTICLE

Simulation of the influence of industrial wastewater on a municipal sewage treatment plant—a case study Ákos Rédey & Viola Somogyi & József Ányos & Endre Domokos & Péter Thury & Tatiana Yuzhakova

Received: 27 October 2009 / Accepted: 11 June 2010 / Published online: 30 June 2010 # Springer-Verlag 2010

Abstract Purpose Industrial wastewater flow caused operational difficulties in the wastewater treatment plant in Debrecen, Hungary. Bioaugmentation was successfully applied to maintain effluent quality in the periods when wastewater of high starch content was accepted, but, at the end of 2008, the nitrification capacity of the plant decreased considerably due to improperly pre-treated pharmaceutical wastewater. Methods and material Dynamic simulations were carried out in a prototype programme developed by the Environmental Expert System Research Group at the University of Pannonia, Hungary. Several parameters for heterotrophic biomass were adjusted in function of time, and the specific growth rate of autotrophic biomass was altered in function of time and temperature in order to describe the effects of inoculation and toxic influence. Simulations were carried out with both constant and adjusted parameters. Results Though results on effluent COD of the different modelling versions were similar, the ammonia concentration fitted the measured data only when modified parameters were used. The study revealed that the autotrophic biomass had slowly adapted to the toxic compound. Different control strategies of aeration and decreased excess sludge removal

Responsible editor: Hailong Wang Á. Rédey (*) : V. Somogyi : E. Domokos : P. Thury : T. Yuzhakova University of Pannonia, Egyetem Str. 10, 8200 Veszprém, Hungary e-mail: [email protected] J. Ányos Debrecen Waterworks Inc, Hatvan Str. 12-14, 4025 Debrecen, Hungary

rate were tested to enhance the nitrification in the critical time intervals. The amount of ammonia and inorganic nitrogen decreased in all cases while the oxygen demand increased to a maximum of 10.1%. Conclusions Reducing excess sludge removal rate gave satisfactory results even without changing aeration. Further improvement could be achieved by introducing aeration into the post-denitrification reactor. The combination of the two modifications can compensate for the effect caused by toxicity. Keywords Bioaugmentation . Toxicity . Dynamic simulation . Nitrification . Wastewater treatment

1 Background, aim and scope Operating a municipal wastewater treatment plant (WWTP) is a complex and challenging task. The WWTP has inherently dynamic characteristics of high temporal variability in terms of both the flow and the concentrations of components in the incoming wastewater. In addition to that, it is common to accept pre-treated wastewater from industrial sources which may present difficulties in complying with the desired contaminant removal rate. The municipal WWTP in Debrecen, Hungary, has a capacity of 60,000 m3/day or 400,000 population equivalent. The primarily clarified wastewater is fed to the biological stage consisting of four parallel activated sludge cascades. There is no extra substrate addition in the postdenitrifying stage. The average sludge age is kept relatively low (4–6 days). The primary sludge after gravity thickening is pumped to the anaerobic digesters. The excess sludge from the secondary settlers is fed to the raw

Environ Sci Pollut Res (2011) 18:192–198

193

water to improve the removal of the organic load during clarification. Twenty percent of the incoming wastewater is of industrial origin from sources such as the pharmaceutical (33%), dairy (8%), canning and meat processing industries (24% and 7%, respectively). Health care contributes to this with 28%. The sweet corn processing plant had just finished the start-up of an up-flow anaerobic sludge blanket reactor to treat its technological wastewater during the 2008 campaign. In previous years, the starch content of the incoming flow from the factory had caused serious problems in the operation of the WWTP. Due to the increased biological load, the dissolved oxygen (DO) concentration fell in the aerated tank which resulted in sludge bulking and decrease in nitrification. To solve these problems, several modifications were carried out. In addition to improving aeration, the management chose to inoculate mixed microbial cultures that were able to use starch as a substrate very fast (Ditrói et al. 2008). These microorganisms:

&

& &

Since there is no biological excess phosphorous removal at the plant, the activated sludge model ASM No1 (Henze et al. 1987; 2000) was chosen to describe the biological processes of the plant. The operations of the parallel lines do not differ from each other; therefore, it was decided to simplify the model to a single biological row (Fig. 1). Influent data were measured after the primary settlers. Unfortunately, these values do not include the effect of the recirculated reject water. Figure 2 shows how the concentrations varied in the reject water. In the second part of 2008, both dissolved organic matter and ammonia concentrations increased in the reject water. Since it was thought to be the reason for decreasing nitrification, the management of the WWTP decided not to accept the sludge originating from the rendering plant in October, but the ammonia concentration of the reject water only decreased in December. With sludge retention time of 20–25 days in the digester, the effect of the hydrolysed sludge affected the next month as well. Unfortunately, the assumption proved to be false; the ammonia concentration remained well above the desired value. These data were included in the model with the following assumptions: the flow of the reject water was taken to 2% of the main stream, only the dissolved COD and ammonia concentration was taken into consideration, and concentration values were taken as constants until the next measured values. Characterisation of wastewater was done according to data in the relevant literature (Benedek 1990; Roeleveld and van Loosdrecht 2002). Soluble inert organic matter (SI) was determined according to Roeleveld and van Loosdrecht (2002) and ratios for readily biodegradable substrate (SS),

&

&

Are able to utilise starch as a specific substrate; Reproduce under aerobic conditions with relatively small generation time and are also able to reproduce in anoxic circumstances; Utilise the dissolved oxygen better than the microbes that cannot feed on starch directly even in cases of lower sludge concentration, therefore the specific sludge yield decreases; Have their concentrations driven by the concentration of the contaminant, thus if the processing season ends, the microbes are gradually washed out.

Phosphorous removal is achieved by adding poly-aluminium chloride (PAC). The amount of PAC is increased in the critical periods (i.e., corn processing season) to meet the effluent limit values. This improves the settling of the sludge as well. Starting from October 2008, nitrification efficiency fell significantly due to the toxic impacts of wastewater from the pharmaceutical factory. However, the source of this toxicity was unknown for several months, and the desired effluent quality could not be restored until spring 2009. Modelling and dynamic simulation of wastewater systems is commonly used for conceptual understanding of systems (Henze et al. 1987; Henze et al. 2000). In this paper, modelling was used to study bioaugmentation, a biotechnical modification. Additionally, by changing the model parameters for fitting the measurement data, the duration of the toxic effect was also estimated. The main aims set for the research were as follows: &

To prepare and run dynamic simulation based on longterm data of the WWTP of Debrecen;

&

To model the effect of bioaugmentation and toxicity by altering model parameters; To suggest an adequate control strategy for the future in cases of nitrogen removal deficiency.

An additional goal was to test new software for modelling wastewater treatment processes. The simulations were carried out with the prototype of a programme developed by the Environmental Expert System Research Group at the University of Pannonia in the frame of the ÖKORET project (RET_06 PEKHIT). The software creates the customised programme code according to the technological description, input data and control strategy provided. The work presented in this paper also served to test and validate the programme, mainly in terms of its capability for using kinetic parameters that are functions of another variable such as time or temperature.

2 Methods and material

194

Environ Sci Pollut Res (2011) 18:192–198

Fig. 1 Schematic representation of the plant

slowly biodegradable substrate (XS) and inert particulate organic matter (XI) were taken from Benedek (1990). Though XI is usually the bottleneck of the characterisation process (Hulsbeek et al. 2002), the results to be introduced later showed that this simplification was appropriate. Soluble organic nitrogen (SND) was calculated from SS and particulate biodegradable organic nitrogen (XND) from XS with conversion factors of 0.02 and 0.04, respectively (Roeleveld and van Loosdrecht 2002). Ammonia concentrations (SNH) were retrieved directly from measurement data while alkalinity (SALK) was calculated with Eq. 1 (Benedek 1990). ½SALK  pH ¼ pKs þ lg ½CO2 

ð1Þ

where: [SALK] [CO2] pKs

value of oxygen was multiplied with a factor given in Eq. 2. The set-point values were estimated by fitting simulation results to measurement data. Recirculation and excess sludge removal rates were defined as a percentage of the influent flow rate, 140% and between 4–6%, respectively, in accordance with the practice of the WWTP operation.    SO;2 F ¼ max 1  ;0 ð2Þ b where: F SO,2 b

−

3

alkalinity (mole HCO3 /m ) concentration of carbon dioxide, with a value of 0.141mole HCO3−/m3 dissociation constant of carbonic acid, with a value of 6.4.

Aeration in the second reactor was adjusted by changing the set-point of the applied DO controller. The input Fig. 2 Concentrations of COD and ammonia-nitrogen in the soluble fraction of the reject water

the factor applied to control the dissolved oxygen concentration the DO concentration in the second (aerated) reactor (g/m3) DO set-point, function of time (g/m3).

After the steady-state simulation two tests were carried out, one with a constant set of parameters, as used in the work of Henze et al. (2000; named: “with constants”). In the second test, certain parameters were adjusted in function of time (Table 1) or time and temperature (Table 2 and Eq. 3; referred to as: “modified”). Another possibility would have been to introduce new processes in the model,

Environ Sci Pollut Res (2011) 18:192–198

195

μA20

Table 1 Parameters that are a function of time Time, day

1–145 146–195 196–280 281–366

μH, d

6 5 4.2 6

–1

kh, g XS/(g COD*day)

ηh

YH, g COD/g COD

3 4 6 3

0.4 0.6 0.8 0.4

0.67 0.49 0.32 0.67

3 Results and discussion

but since the effect of the specialised biomass was temporary and the properties of the toxic matter were unknown, this alternative was rejected. The parameters were chosen according to preliminary knowledge of the effects of the inoculated bacteria culture (Ditrói et al. 2008); their values were changed in two steps in accordance with the dates of inoculation processes. Since these microorganisms are able to utilise starch as a substrate, the maximum specific hydrolysis rate (kh) was increased. Their generation time is comparable to the hydraulic retention time of the plant (cca. 10 h); therefore, the maximum specific growth rate for heterotrophic biomass (μH) was reduced. Their reproduction in anoxic circumstances was described by the increased correction factor for μH under anoxic conditions (ηH), implying that the ratio of bacteria able to reproduce under such conditions increased. Finally, yield for heterotrophic biomass (YH) was adjusted to smaller values due to the utilisation of DO which resulted in lower specific sludge yield. Pambrun et al. (2008) modelled the effect of specific toxic materials to nitrifying bacteria by altering the value of the decay rate (b A ). In the present study, it was decided to adjust the value of maximum specific growth rate for autotrophic biomass in function of time to fit measured values (Table 2) since the toxic substance was unknown and therefore a posteriori assumptions had to be drawn. In this case, all negative effects were described by decreasing μ A20 while the decay rate was kept constant. The implemented equation was taken from Reichert (1994). mA ¼ mA20  expð0:0981  ðt  20ÞÞ

ð3Þ

where: μA

maximum specific growth rate for autotrophic biomass (d–1)

Table 2 Values of μA20 in the function of time and temperature from day 260

Time, day μA20, d–1 Temperature, °C

t

maximum specific growth rate for autotrophic biomass at 20°C (d–1) temperature (degrees Celsius).

256–270 0.8 23

The effluent COD concentrations of the two simulations were very similar to each other and to the measured data (Fig. 3). Simulation with default constants would also have been satisfactory for a quick estimation of effluent COD. The average difference between the values of the first and second tests was 3.04 g COD/m3 after the first inoculation process, while, after the second, this difference was 5.38 g COD/m3 with standard deviations of 0.68 and 1.31 g COD/ m3, respectively. Contrary to that the curves of ammonia concentration differ greatly (Fig. 4). The simulation with constant values did not correlate with measured data starting from the beginning of July 2008. This indicates that, in the processing season, without bioaugmentation, the oxygen would be utilised primarily by the heterotrophic biomass, thus the nitrification would decrease. That had been the situation before inoculation was applied. The critical part of modelling the ammonia concentration in 2008 was the time interval of October and November. The curve of the “modified” version after altering μA20 fitted the measurement data. From previous tests not shown in this paper, it became clear that neither the decreasing temperature nor the increased ammonia and COD concentration of reject water could cause the low nitrifying capacity. The simulation gave adequate results only when the specific growth rate was adjusted in function of time. The values of μA20 that were adjusted to fit measurement data suggest that the effect of the toxic compound became less significant as time passed. It cannot be excluded that the amount of toxic material decreased gradually, but it can be stated with confidence that the autotrophic biomass adapted to the new circumstances. Contrary to the problems in modelling ammonia, the nitrate concentrations showed better fitting in case of simulating with constant values and fits measured data with modified values (Fig. 5). This and the fact that the values of COD concentration correlated with the measurement data suggest that heterotrophic bacteria were not affected by the unknown compound.

271–291 0.51 23

292–315 0.52 22

316–325 0.55 21

326–343 0.63 18.5

344–353 0.71 18.5

354–366 0.8 17

196

Environ Sci Pollut Res (2011) 18:192–198

Fig. 3 Effluent COD concentrations in simulation sessions compared with measured values

Fig. 5 Effluent nitrate concentrations in simulation sessions compared with measured values

As the sludge age of the plant was particularly low, there is a risk that similar events could occur in the future. In order to mitigate the consequences, several control strategies were tested. The goal was to keep the effluent ammonia concentration below 5 g N/m3 and the effluent inorganic nitrogen (sum of ammonia, ammonium, nitrite and nitrate, henceforward, dissolved inorganic nitrogen (DIN)) concentration below 20 g N/m3. To decide which control version would provide the best results, ammonia and DIN amounts above the permissible level were taken into consideration starting from the 260th day. Additionally, numbers of violations were also reported (Table 3). In the first case, dissolved oxygen concentration in the second (aerated) tank was controlled by changing the setpoint according to the ammonia concentration measured in the third tank. In the second case, the original control in the aerated reactor was not changed, and aeration was introduced into the post-denitrifying reactor. Control was achieved by adjusting the set-point according to the ammonia concentra-

tion measured in the third tank. The third case was a combination of the previous two; aeration was controlled in the second and the third tanks according to effluent ammonia concentration. The level of aeration in the third reactor in the second and third cases was kept low in order to maintain the denitrification; the DO concentration did not exceed 0.22 g/m 3 , and the average value was 0.013 g/m3. Figure 6 shows the change of effluent ammonia concentration. Due to the control strategies, the ammonia concentration decreased significantly but DIN concentration did not change considerably in either case. Therefore, in addition to aeration control, the excess sludge removal rate was decreased in October, November and December uniformly to 4% from 6, 4.8 and 4.4%, respectively. This resulted in lower ammonia and DIN concentrations as well (Table 3). Unfortunately, in neither case could the desired effluent quality be achieved. In the first case, without modifying the excess sludge removal rate, the amount of ammonia above the limit decreased by 61%, while DIN was lower by 24%. That suggests that the denitrification capacity reached its maximum: only a small fraction of the produced nitrate could be converted into nitrogen gas. The oxygen demand in the first case was only higher by 5% of the original control strategy. The second and the third cases gave similar results. The ammonia level was near the desired value but the DIN content was, in both cases, about the same as in the first case. The oxygen demand was 8.5% and 8.6% higher than without modification, respectively. This gives the management the freedom to decide which strategy is to be chosen for such situations. The second case, where aeration in the second tank was controlled according to the original practice and aeration was introduced in the third tank according to ammonia concentration, is more practical since the control devices in the second tank do not have to be reprogrammed.

Fig. 4 Effluent ammonia concentrations in simulation sessions compared with measured values

Environ Sci Pollut Res (2011) 18:192–198 Table 3 Results of the different control strategies

Values indicate amounts above the permissible limit

197 Aeration only

Original First case Second case Third case

Aeration and modified excess sludge removal

NH3 (kg N/day)

DIN (kg N/day)

Limit violation NH3/DIN

NH3 (kg N/day)

DIN (kg N/day)

Limit violation NH3/DIN

26,819 10,456 198 86

17,511 13,269 13,248 13,243

84/72 73/66 3/61 3/61

11,166 4,892 56 39

10,582 9,397 9,784 9,755

54/60 49/57 3/55 2/54

It also became clear that nitrogen removal efficiency could have been improved by keeping the excess sludge removal rate lower than in practice even if aeration control was not modified. This resulted in the amount of ammonia and inorganic nitrogen above the permitted level to decrease by 58.4% and 39.6%, respectively, compared with the original control strategy, while oxygen demand only rose by 5.8%. When enhanced sludge removal rate and modified aeration control was simulated the results showed even more improvement though the difference between the values was less significant. The oxygen demands in each case were 108.0%, 110.1% and 109.2% of the actual value. Considering effluent quality and aeration costs, the best result was achieved in the third case though the outcome is very similar to the values of the second case.

4 Conclusions As the results show, the wastewater treatment plant operates on the verge of its capacity and is very sensitive to external influences in terms of nitrification. In order to handle such problems, it is recommended that the

Fig. 6 Comparison of different control strategies in terms of ammonia concentration

wastewater treatment plant monitors all flows including inner ones. The management of the plant plans to implement on-line ammonia and nitrate sensors in the aerated and post-denitrifying tanks. That allows new control strategies to be applied for which this paper gives suggestions. The sludge age of the plant is somewhat low to achieve full nitrification. In normal operation, this did not cause problems because the average temperature of the influent was satisfactory even in winter time (not lower than 17°C). However, coupled with toxic impacts, nitrification efficiency fell drastically, and the system could not restore itself. In order to achieve better nitrification performance in such cases, several possible solutions were considered. If aeration was introduced into the post-denitrifying reactors, ammonia concentration could be decreased, but the inorganic nitrogen did not change significantly since the nitrate yielded could not be removed. It can be concluded that in cases when nitrification efficiency falls decreasing the excess sludge removal rate can compensate for the negative effect better since both the amount of ammonia and inorganic nitrogen decreased in contrast to the situation when only new aeration controls were applied. The applicable value of excess sludge removal has to be determined based on the efficiency of the primary clarifiers since excess sludge is used to improve settling properties. Further improvement can be achieved if aeration is introduced to the post-denitrifying tank allowing simultaneous nitrification and denitrification. Since the equipments of aeration in the third reactor are already present, additional investments costs do not arise. These interventions do, of course, increase the operation costs, but to a lesser extent than they improve the effluent quality. The research was carried out using the prototype simulator developed by the research group of ÖKORET. According to that, it can be stated that the prototype was proved to be suitable for solving the problems and tasks presented in this paper.

198 Acknowledgement The financial support of the University of Pannonia ÖKORET Knowledge Centre (RET_06 PEKHIT) is gratefully acknowledged.

References Benedek P (ed) (1990) Biotechnology in environmental protection. (Biotechnológia a környezetvédelemben) Műszaki Könyvkiadó, Budapest Ditrói J, Pandur J, Thury P (2008) Precluding sludge bulking in the period of corn processing at the Debrecen wastewater treatment plant. (Iszapduzzadás megakadályozása a csemegekukorica feldolgozás időszakában a Debreceni Szennyvíztisztító Telepen) In: Proceedings of the National Conference on Environmental Engineering 16–18 September 2008, Siófok, Hungary pp14–22 Henze M, Grady CPL, Gujer W, Marais GvR, Matsuo T (1987) Activated sludge model no. 1. IAWPRC Scientific and Technical Report No. 1. IAWPRC, London

Environ Sci Pollut Res (2011) 18:192–198 Henze M, Gujer W, Mino T, van Loosdrecht MCM (2000) Activated sludge models ASM1, ASM2, ASM2d and ASM3. IWQA Scientific and Technical Report No. 9. IWA Publishing, London Hulsbeek JJW, Kruit J, Roeleveld PJ, van Loosdrecht MCM (2002) A practical protocol for dynamic modelling of activated sludge systems. Water Sci Technol 45:127–136 Pambrun V, Marquot A, Racault Y (2008) Characterization of the toxic effects of cadmium and 3.5-dichlorophenol on nitrifying activity and mortality in biologically activated sludge systems —effect of low temperature. Environ Sci Pollut Res 15:592– 599 Reichert P (1994) Concept underlying a computer program for the identification and simulation of aquatic systems. Swiss Federal Institute for Environmental Science and Technology, Dübendorf, pp 180–192 Roeleveld PJ, van Loosdrecht MCM (2002) Experience with guidelines for wastewater characterisation in The Netherlands. Water Sci Technol 45:77–87