Anda di halaman 1dari 10
ve upgrading of the 1 the project plan- spgrading the acti- {allows treatment nitrifiers grown in 1 waste sources. ale tests and full- ch (STOWA). s, >nal co-operation based evaluation 128), 12-118, (1998), The astewater, Wat sdge model No, -zoltenabatch sale biologi 1. Be Eng heslans. aby oititcation itiieaton by rents Experience and extensions to the ASM2 family of models 4}: Dudley’, G. Buck’, R. Ashley" and A. Jack {NR Fraskind Road, Svndon SNS BYF, UK. Email: duleyureplecoud) (Erna pach it & Encarta Engen, User of Brats, We! Yori, 80710 (Email Reshley@bradtordacuk) mat cnt Serie Delve, UPM Team, There Hote, Ligey Mere, Waregon WABSLW (Ema ancrow ack @uuple. co.uk) cera nthe dt colecin stam tequry, and ment ony aparece '# Swedish sewage works. Thi paper shows that ASM wer storage on sewage, Serraino rg quanttes of Vasantha eto on he cowage treatment processes, was unfoundac, tie patted she: ASME: ation modeing hosp removal spars dt ‘STOAT; LUKWIR: VEA Introduction Taga WA Activated Sludge Model No.2 was published in 1994, with formal publication in [R25 (Henze eta, 195) and updated in 1998 (Henze ta, 2000) Sins its publication there fs anecdotal evidence thatthe model has been widely accepted hg there have been Few papers published giving experience ofits use Dynamic modeling, which isthe area most considered by the users of ASM 2.is com, monly seen as requiring reasonable (or even large) quantities of acevrane, high. y pirtionrae a Dilly, hourly or beter dataon COD profiles inan aration tan oe arte messy ne Many Wastewater treatment plants arent setupto provide there ara Srahe measurement frequency. Many sewage works collec data sing daily samples, either hing enPosit, and sample the effluent quality and, possibly with reduced frequency {ris Paper thus introduces two unrelated aspects of ASM 2, based on UK experience wa ine te model. These are (1) the application of ASM 2 where date is collected ‘only fo outne plan operation, rather than forthe specifics of model calibea and (2) z : § 2 z é = 3 fi | g 8 5 g ‘modifications of ASM 2 to provide a better expression forthe heterotrophic usage of VEAs and fermentable COD. Results from the particular project that led to the ASM 2 model being checked for how COD competition should be expressed are also given, although this ‘lates more tothe issue of hydraulic disturbances and control of sewage treatment works, “Sewage works” data sets ‘The first application was to biological phosphorus removal at a Swedish sewage works, This works operates four activated sludge lanes. At the time of this study two lanes were using chemical phosphorus removal, with the other two lanes modified t0 investigate bio- logical phosphorus removal, using a UCT configuration. The data collected was only for routine monitoring ~ composite samples of the influent and effluent, but little additional data, There had been no intent o model the phosphorus removal system until the STOAT. ‘model was introduced as part ofa separate project, at which point it was of interest to see hhow well the model could be calibrated using the existing historical records. The basic details of the sewage works are given in Table | An existing model was available, based on ASM 1, and the kinetic parameters used for {that model, calibrated by the site staff, were used in the ASM 2 model. The ASM 2 model was calibrated only for the biological phosphorus removal, and for the solid setting properties. ‘Model calibration The model was calibrated by eye, adjusting the parameters until a best fit was judged to hhave been found. This procedure has been used successfully in the past. The reasons for this, compared to more formal procedures (such as least squares assessments) are: *+ Where little is known about the underlying error distribution in the data then graphical fits are more robust; + Outliers are more readily handled through graphical comparison; + Cutoff data (e.g., ammonia data recorded only as “<2 mgil") is better handled through _ztaphical comparisons, ‘The resulting set of parameters is given in Table 2. Only the italicised items were cali- brated for biological phosphorus removal ~ the other parameters were calibrated in previ ‘ous years for previous models used by the plant, based on ASM 1. Two values are given for the autotroph growth rate as the practice at the works was 10 use several values for diferent parts of the year, as they had a repeatable annual cycle of inhibition caused by some lunknown discharge within the catchment. Tablet Ste detis Volum 8,200 m? ‘Surface area 1,220? Anaerobic action: 124% Depth 38m ‘Anoxic racton: 97.9% RAS ow 1,000 Nitrate recycle 2.450" WAS ow 300 m°h ‘Anaacrecyee 810 9! MLSSe.4.000mq/ Flow 760.2 600 200mg/ TSS90m! ‘Ammonia 18 malt Phosphorus 3m Tab Ae Hee Hote Ae Hye Hy Hyer Frat Hate Frac Neos Neo nox ‘rae Pua ‘Autot Sim Ther avail inter othe predi there 02m sage of VEAs SM 2 model although this ment works. wage works. 10 lanes were vestigate bio- was only for le additional il the STOAT Us, The basic ters used for ASM 2 mode! solid settling was judged to ve reasons for hen graphical adled through ims were cali- ated in previ- sare given for for different ised by some Table2 Calration parameters Aeon onpmatre cou! 176) 2.101 0.108 rtrogh ath rate @ 15°C (1h) 0018 0.012 rotroph death romperature coefcient (1/"0) ona 0.063 ‘Autowoph death temperature coetfcient (1/"C) ona onto Hyaroysie rate @ 15°C (1h) 0.072 o.102 Hyaroyistomporaturecosfcint (1/°C) o10 ost Hyaroyishaltvate constant @ 18°C 0.0007 o.cor2 HalFrata constant tomperaturecoetcent (1 on08 “0.10 haat coeficin! autorophs) (mg Oa!) os 08: INH haltzatecoefcent(auttrophs) (ma NI) 03 1 Fractional hydrolysis rate, anoxic contons -) oa. os Hteretoph ial ig COD/mg COD} oer 068 Fraction of nat COD in partculate svbetrte mg CODImg COD) 008 ° Neantet of biomass (mq Ning COD) 0.088 07 Noantent ont particulate COD (mg Nimg COD) 0.08 0.03 Anove hycrolys tector 08 o4 “Anaerobic hyralysiafactor os o4 PHA uptake rate @ 18°C (1/h) 02 on ‘Autatroph growth rate 15°0 (1h) 00138 ‘Autotroph grout rate@ 15°0 (1/5) 0.0206 ‘Simulation results ‘The results using the calibration data set are given in the following graphs. Data was usually available for the period Monday-Friday, with weekend data being estimated by linear interpolation between Friday and Monday. Not all weekdays had data, and missing data ‘was again estimated by linear interpolation between the available measurements. This adds to the level of uncertainty in the model and explains part ofthe issue of goodness of fit. It can readily be seen that for nitrate and ammonia there was good match between the model predictions and the data. The ammonia results also show the effect of censored data, where there appear to be at least two different laboratory interpretations: one presenting data as 0.2 mel, the other as 0 mg/l, when presumably the measurements indicated thatthe result was “less than 0.2 mg/l.” Shine Sida Ge ker Celay Won GAM OLA Ghw OID OIAe” OTe 310K Figure 1 Eiuent vate ee Kopnat ve tapea't | ene * yu Figure2 Etiuent ammonia results are not quite as good, as can be seen in Figure 3. is representing the basis ofthe dynamics in he response, For phosphorus the modell However itis clearthat the mod indicating when there are periods of high phosphorus inthe effluent. We assume that accu rate modelling of phosphorus is more sensitive to inadequacies in the data set than is the case with nitrogen ~ but that the model will give reasonable warning of likely problem periods. The model was reasonably calibrated by increasing the anoxic and anaerobic hydrolysis rates, which by default are 40% and 10% of the aerobic values, Using values that were 80% ‘of the aerobic values produced a large improvement in the model predictions. There was further, smaller, improvement, in increasing the PHA uptake rate from 3/d to 6/d (at 20°C) Thisis in line with the findings of Larrea et al. (2001), Other parameters tried, but with litle effect, were! + Fermentation rate. This had little effect because the rate-limiting step was hydrolysis of particulate COD to fermentable COD, or the provision of sufficient quantities of partic- ulate COD. + High heterotrophic death rates, to release COD into the anaerobic zone. While this worked the levels required were much too high, around 10 times the default values, + High polyphosphate storage concentrations. This produced the best results but required tian ada GE O2eq CBtky Cdn Pak OI Oem OLDE ott DACee Figures Ettuent total phosphows 1 in Figure 3 the response, sme that accu- et than is the sely problem ichydrolysis| hat were 80% 1s. There was sid (at 20°C). bout with biele hydrolysis of sies of partie: 2. While this «values. but required that the biomass comprise about 90% non-volatile solids, compared to a more common Value of around 25%, + High growth rates of PAO bacteria, This had litle effect because of limitations in avail- able COD for growth, As is common, MLSS is usually well represented. The greater level of detail given to representing total solids in ASM 2 has made matching MLSS easier than was the case with ASM I ~this has been our experience with other sites as well In-sewer storage systems ‘The issue of the correct representation of how heterotrophs should use VFAS and fermentable substrate arose from UK concerns over the adoption of large in-sewer storage volumes to minimise sewer overflows. With large storage volumes, potentially stored for long periods, there was a possibility of anaerobic conditions leading to the formation of high concentrations of VFAs. When these VFAS are discharged and reach the sewage works they may cause odour problems and, being perceived as more degradable than other COD forms, may also lead to extensive near-anaerobic conditions near the stat of aeration lanes or the inlet part of filters. This concern led to a UK Water Industry Research-funded Project to examine the effects of large amounts of in-sewer storage on sewage works (Ashley era, 2001). This project looked atthe potential effects that large volumes ofin-sewer storage could have on three example systems: nitrifying trickling filters; nitefying/denitrifying activated sludge; and a three-stage Bardenpho system, The in-sewer storage volumes used covered two scenarios, a 44-fold and a 90-fold increase over the “traditional” storage provision in UK sewer networks. Heterotrophic usage of VFAs Because some of the activated sludge systems being considered were traditional carbon- nitrogen removal systems only there was a need to model the heterotrophic competition for ‘FAs and fermentable COD consumption. ASM 2 and 2d (Henze et al., 1995) have two ‘equations to describe the usage of VFAs and fermentable COD by heterotrophis. The essen \ial aspects of these equations are that the choice of the usage of VFA or fermentable sub- strate is made using the following switching functions Figures miss ev hopng Usage of VFA: Usage of fermentable substrate: Ky +5; 5 +5 ‘The defaut values forth switching constants Ky and K, are Ky'=K, = 4mg/lCOD. With these default values, and with a biological interpretation ofthe afi constants K and Kg there sno difeeniation between the consumption of VFAs and fermentable COD, ‘Whether he substateisassgned solely 8, ory crspitbeteenS, and, sould make no diferencetotbekinetic sul ‘The resting soluble COD consumption equation has the form SnSe_ SS, Se Se Ss Kivi 5 Ray E The standard form should be FS Assuming ,=S, (tomaximige sensiiviy ofthe fesling ouput hen the ASM24sutching functions the coneetauesathe ints >> Kt undepreitfrSK, ithe equation simpli ingto TKS A better form for the switching function has been known in the standard biochemical engi neering literature (Bailey and Olli, 1986)! and has the form fF i Syl Ky eer olyee T¥5,) Kat Sp) Kp : Se Ky Usageof termentablesubsirate Se STR Ss For the special ease K,, = Ky this equation reduces to 3, a8 required, For general mod: clling the default values for the switching constants were made Ky = 4 mg/l and Ky = 20 ‘mg/l. The value for Kis in line with the Value used in the previous ASM | and isa value that has commonly been quoted for heterotrophic usage of soluble COD. Penya-rojaet al (2001 also concluded that better value for Ky, was around 20 mg/l, based on respirometric calibration of ASM2d on Spanish wastewaters. The revised switching function has not bbeen tested with experimental data but is required to achieve consistency with abiological interpretation of the Monod-style equations. Results twas necessary to mode! long term rainfall events to study prolonged loading and dilution effects and also shorter, more intense storms with foul flushes. Selection of long-term ‘model rainfall inputs was based on the method proposed by Jack and Asbley (2001), known asthe “Total Emission Analysis Period” (TEAP). A critical sequence of wet and dry weath- cer was selected based on historical annual data (drawn from a 10-year series). An existing HydroWorks sewer flow model was used fora verified system for which storage has recent- ly been introduced. This model was adjusted to simplify the level of detail in the sewer "One other popular phosphorus removal model that of Barker an Dold (2000), uses diferent set of *witeing function o distinguish berweenheterotophic usage of VPA and fermentable COD. The switching functions used inthe Dold model als fail to produce the corret equation frm forthe special 38 K, Ko and the Dold model should similarly be modified to use the Bailey and Ollis expression, In the ete ofthe Dold mode the switching function duces to 2 §/(2 K +S) with the same assumption of 5425p. AtTow S this simpliiesto,S/K (compared to ASM 2ofS/2K, while at high Sitsimplifis to 2, compared to ASM2 andcorect, | tion of the bination o oxygen tra sewage tre The Ws VFAs wen concentrat from WAT flows entes over a leny upstream e smaller tha the trunk s small peak The short a When th quantities ¢ still atcone results for « \VFAs subst Tecan ea Final seein the effects ¥ minimal. W ceaused by point the ae: Had the me 2 Theanalys tine storage for longer, stillbe 8 jneral. ometric has not logical dilution og-term known ‘weath- xisting recent tof The special sion, Figure § Production and consumption of VAsin ‘network, and to investigate the effects of different storage volumes and chanting the loca tion of the storage position within the sewer. Sewer flow quality was modelled using acom- bination of HydroWorks (solids and nitrogen) and the Danish WATS model (carbon and ‘oxygen transformations) (Hvitved-Jacobsen eral, 1998), STOAT™ was used to model the sewage treatment works. The WATS modelling discovered that unless storage was close to the sewage works the ‘VFAs were rapidly oxidised during transport through the sewer, and were never present at concentrations high enough to be an odour or other problem? Figure S illustrates the results from WATS, looking at a section of a sewer line. The rapid increase in VFAs is caused by Flows entering the trunk sewer from lateral storage tanks. The VFAs are oxidised linearly over a length of approximately 900 metres. The first peak is from a storage tank at the lupstream end of the catchment. The second peak, which is from a larger storage volume, ‘smaller than the first because of oxidation in the lateral sewer connecting the storage tank to the trunk sewer and because ofthe diluting effects ofthe flows in the trunk sewer. The final small peak is due to the flows from a third, smaller, storage tank on a downstream branch, ‘The short aerobie distances needed to oxidise VEAs can be seen in this example. When the storage was considered as being almost next to the sewage works then larger Quantities of VFAs (and hydrogen sulphide) were present atthe sewage works inlet — but Still at concentrations low enough that they were not an issue. As an example, the summary results for a comparison between running with VFAs present in the influent, and with the \VEAs subsumed into the fermentable substrate, are given in Table 3 Itcan easily be seen that there is no effect on suspended solids — hydraulic effects on the final setling tank affect this more, so unless the VFAs produce a large change in the MLSS the effects will be negligible. For COD there has been asmall change, but again the effect is ‘minimal. With ammonia there is a larger difference ~ the mean value is almost doubled, caused by the greater affinity for VFA over fermentable substrate, so that at one critical Point the aeration system operated for a longer period with lower DO than the no-VFA case. Had the modelling used higher values for the K,a, or experimented with the controller ‘The analysis assumed that he storage was provided in-line, that i spat f the main ewer low. Off line storage, where the low is ivered oa separate holding tank, may potentially have sewage anding for longer. producing more VAs, and may adjust slightly the conclusions reached her. There would sil be rapid consumption of VFAs and hydrogen sulphide afterdischarge, provided that some part of ‘he sewer was aerobic, oat the general conelason~ that VFAs produced through anaerobic storage conditions are usually "ested" by the sewer system itself is probably il valid wetoora't | Table 3 Ettuentqualty impact of VFA inthe infuent Mean s2g_ sa 128 atc Minar raneueeas cM eeet Oo 40. 104 On Maximum ms 80779, wo7 819 Standard deviation Nae ta Te ta settings to provide a more responsive control system, then this affect may have been less marked, ‘The rest of the modelling for this project focused on the use of control strategies to reduce the hydraulic effects of the prolonged periods of storm flow caused by the sewer storage. The aeration system used was set up 10 be approximately 8 hours retention at average flow, 8 day sludge age, 4 stages of which the first was 10% of the volume and anox- jcand the other three all erated and 30% of the volume. The results can be summarised as follows: ‘operating conditions and storm Flows, however, this produced a worse effluent quality than operating at fixed RAS flowrate + Step feed was a viable solution and did not affect the ammonia values much’, probably because of the dilution of the ammonia in the influent during the storm periods. The results are similar to those of Chen and Beck (2001). They found that step-feed produced better effluent quality than flow-proportional RAS when looking at TSS and COD, but, worse when looking at ammonia. The comparative results from our modelling exercise ate given in Table 4 + Running at higher DO setpoint didnot have much effect. + Bypassing the primary tanks during storms, or having a storm overflow after the primary tanks (so tha lf storm flows went through the primary tanks) did not have much effect. + Having a sacrificial lane to handle a large share ofthe storm flow did not produce any significant improvement in ether effluent quality or recovery time following the end of the storm + Modelling single storm events did not stress the modelled works as much as having a sequence of storms, Although the effluent quality indicated a quick return to normal operation after the first storm event, there was stil disturbance of the intemal states. Because of this, a second storm event following in quick suecession produced a larger effect on effluent quality than the fist. The storm events being used were taken from a particularly wet period for the catchment area being used. + Atickling filter model was also used. Although this did not produce as good an effluent ualty asthe activated sludge systems (something which canbe addressed by changing the filter bed area, depth or packing characteristics) there was very litle effectof the dit ferent storage scenarios on effluent quality, unlike the activated sludge system. The tlter system appears tobe robust o hydraulic upsets, moreso than activated sludge systems. Conclusions ‘Complex models such as the ASM 2 can be calibrated against data that has been collected Such datais typically low for routine site monitoring rather than to assist model calibrati fed reduces the acratin volume available for nitrification, and this edetion can ead to reduced ition, Tables F Mean Minimum Macmum Stordarae Frequene with mis duality — predictio regardles ‘Asus. reasonab tively, to ues forth that lead Thest mentable well-kno adding Altho. logical p anaerobi 2isunlik Of ser two, ster effluent: Acknow ‘The autt ‘work rep here. JD this pape ple, Bre (Odense Voller. UK) for Referer Ashley. Proce Recei Bailey. ISBN! Busksr.P been less utegies to he sewer cention at and anox rarised as bone set of at quality probably iods. The produced COD, but ve primary cheffect. duce any the end of shaving a to normal nal states, cd a larger en from a aneffluent vy changing of the dif- 1 The filter systems. collected pically low voreduced ‘Table 4 Performance wih variable RAS and stoped coo we cS is ese ws 08 wos 8812 Minimur. 30 108 Ot 3398 o4 Maximum 199 78278 es e992 ‘Standard deviation 788 81 103 tt frequency typically, daily — with many holes ~ typically, week-days only and even then with missing days. Consequently the driving function for the model - the influent flow and {quality ~has to make use of much interpolated data. The goodness of fit between the model predictions and the measured effluent quality will usually be reduced because of this, regardless ofthe quality ofthe calibration or the adequacy of the model representations. "As used at one site, we found that it was possible to get a good match on nitrogen and a reasonable match on phosphorus. The phosphorus results are possibly better used qualita- tively, to indicate likely periods of with- and without-consent effluent, The parameter val- ues forthe calibrated model are usually close to the recommended default, or within a range that leads to litle comment about the values. "The standard representation of the equations for heterotroph utilisation of VEAs and fer mentable COD is incorrect, and fails to meet simple analytical requirements. A simple, ‘well-known, correction to the equations produces the correct theoretical behaviour while adding no additional complexity to the ASM 2 family of models, ‘Although there is this theoretical weakness, in practice ASM2is normally used with bio- logical phosphorus removal systems, where most of the VFAs would be consumed in the anaerobic zone. Consequently, for practical purposes this theoretical weakness of the ASM is unlikely to require that any previous modelling work would need to be revisited. (f several control strategies tried to improve performance under storm conditions, only ‘wo, step feed and variable RAS flows, were found to provide any useful improvements in cffluent quality. Acknowledgements, ‘The authors would like to thank the UK Water Industry Research for funding part of the work reported in this project, and the European Union for funding for other work reported here. JD and GB would also like to thank WRe ple for permission to publish the material in this paper. The views expressed here are not necessarily those of UKWIR, the EU, WRe ple, Bradford University or United Utilities. 3D would also like to thank David Cecil (Odense Water, Denmark), Hansruedi Sigriest (EAWAG, Switzerland) and the paper reviewers for their comments about this paper during the Kollekolle conference, and Jes ‘Vollerisen (University of Aalborg, Denmark) and John Blanksby (University of Bradford, UK) for the WATS modelling material References Ashley, RM. Dudley J Volersen 1, Sal, A, Jack, A. and Blanksby J.R (2001). The effect of extended inmsewerstorage on wastewater eatment plant performance, Water Seience and Technology (accepted Proceedings ofthe 2nd International Conference on Inseractons benween Sewers, Treatment Plans and Receiving waters(INTERURBA 1) Bailey, LE and Os, .F (1986), Biochemical Engineering Fundamentals, 20d edition, MeGraw Hil ISBN 007 Y66691 5. Barker, PS. and Dold, PL (1997) “Genera model for bologieal nutrient removal tivated sludge ss tems: model presentation”, Water Environment Research 645) 69 eve hapa rwistapras | sonal control af storm sewage at a setivated sludge process, Water (Chea J and Beck, MLB. (2001). Oper ‘Science and Technology 487) 131-137. Henze. M., Gur, W., Mino, T, Matsuo, T, Wentzl, Mand Marais, GR, (1995). Activated Stadge “Mvdel No, 2, IAWO Scientific and technica report No 3, ISBN 1 900222 000. Henze, ML. Gojer.W..Mino,T. and Loosdrecht, M, (2000). Acivared Siudge Models ASM ASME, ASM24 “and ASM, TWA Scientific and Technical Report No, 9 ISBN 1 900 222 248, Hvitved-Jeobsen,T.Vollertsen J. and Tanaka N. (1998) Wastewater quality changes during tansportin ‘seers an integrated aerobic and anaerobic model concept for carbon and sulphur microbial transfor mations, Water Science and Technology 310) 257-264, Jack, A.and Ashley, RM. (200). The impact ofthe contolled emptying of n-sewer storage on wastewater treatment plant performance, Water Science and Technology (accepted): Proceeding ofthe 2nd Imernational Conference on Interactions berween Sewers, Treatment Plans and Receiving waters ean (INTERURBA I. peal Larrea, L rit I and Hidalgo, ME, (2001). lmpeoving the predictions of ASM24 through modeling in | ot practice, prevented athe Fifh Kolleolle Activated Sludge Modeling Seminar, Klleoll, Denmark pone ‘Suimized to Water Science and Technolog treatment Penyatoja,J.M. Seco, A. Fee, and Seat, J. (2001). Calibration and validation of wctivated sludge cstaael model No. 24 for Spanish municipal wastewater, submited1o Environmewa Technology Letters udecie process tot ‘woatment optimizator their apps Keywords Optim algorit Introduce’ When de: engineers process © Teally, th are met r increasing design a ina virtwa ing activi process of future effl Genera proposes « modeled a ‘operation: is then ites tion ofthe designens probleme are 10 des 1024, pos: ‘means tha designern ficult by t between t ance. The design pro erable m0 |

Anda mungkin juga menyukai