Real-time control-oriented quality modelling in combined urban drainage networks

Real-Time Con ol-O ien ed Quali y Modelling in Combined U ban D ainage Ne wo ks ? Congcong Sun ∗Be na Joseph-Du an ∗∗ Thibaud Ma uejouls ∗∗∗ Gab iela Cemb ano ∗,∗∗ Jo di Mesegue ∗∗ Vicenc¸ Puig ∗Xa ie Li ico ∗∗∗ ∗Ad anced Con ol Sys ems G oup a he Ins i u de Rob`o ica i In o m`a ica Indus ial (CSIC-UPC), Llo ens i A igas, 4-6, 08028 Ba celona, Spain, (e-mail: [email p o ec ed]). ∗∗ CETaqua, Wa e Technology Cen e, C a. d’Esplugues 75, Co nell`a de Llob ega , 08940 Ba celona, Spain, (e-mail: [email p o ec ed]). ∗∗∗ LyRE, Lyonnaise Resea ch Cen e , Domaine du hau -Ca ´e, Bˆc −C4, 33400 Talence, F ance, (e-mail: [email p o ec ed]). Abs ac : U ban d ainage ne wo ks (UDN) ca y u ban was ewa e o was ewa e ea men plan s (WWTP) in o de o egene a e i be o e eleasing i o he en i onmen . Combined UDN (CUDN) ca y bo h ain and was ewa e oge he , which can o e load he UDN and p oduce combined sewe o e lows (CSO) ha pollu e he en i onmen . Managemen o CUDN is ecei ing inc easing a en ion om bo h esea che s and wa e manage s, in o de o mee he high quali y s anda ds equi ed o wa e and en i onmen acco ding o EU Wa e F amewo k Di ec i e. Due o he complex dynamics o wa e quali y, in eg a ed con ol o CUDN and WWTP conside ing bo h lows and quali y o he con eyed was ewa e is a di icul p oblem. In o de o design a eal- ime con ol (RTC) aking in o accoun hyd aulic and quali y a iables, he use o concep ual quali y models is conside ed as a sui able op ion. This pape mainly p esen s a simpli ied concep ual quali y modelling app oach o ep esen he dynamics o suspended solid in sewe s o CUDN o ien ed o eal- ime con ol. A sewe simula o implemen ed in SWMM (S o m Wa e Managemen Model) in eg a ed wi h a lumped concep ual model o o al suspended solid (TSS) is used o calib a ion and alida ion. A eal example o Pe ino sewe ne wo k is used as a case s udy. Discussions abou RTC implemen a ion in CUDN a e also p o ided in his pape , whe e Model P edic i e Con ol (MPC) is p oposed as he sui able me hod o con ol he in eg a ed wa e and quali y models in CUDN as u u e mo i a ion. Keywo ds: CUDN, WWTP, CSO, RTC, concep ual quali y model, SWMM, MPC. 1. INTRODUCTION Combined u ban d ainage ne wo ks (CUDN) a e gene ally de- signed o con ey was ewa e lows o ea men acili ies in d y wea he . Du ing hea y- ain e en s, as soon as capaci ies o u - ban d ainage ne wo ks (UDN) and Was ewa e T ea men Plan (WWTP) a e exceeded, mixed wa e is by-passed o ecei ing bodies p oducing combined sewe o e lows (CSO), which a e iden i ied as pollu ing haza dous o biological species and eco- logical s a us as explained in Becouze e al. (2009); Gaspe i e al. (2008). In o de o op imize o e all objec i es o he comple e u ban d ainage sys em, including he CUDN, he WWTP and he ecei ing en i onmen , and p e en pollu ion o he ecei ing wa e s, an in eg a ed con ol o bo h UDN and WWTP sys ems is equi ed. Since 1970s, he po en ial o using eal- ime con ol (RTC) wi hin CUDNs has been discussed in Bu le e al. (2005); Schilling (1989). Re e ences ha e p o ed ha RTC is a eliable and cos -e ec i e solu ion o CUDNs, which can imp o e he pe o mance o CUDNs minimizing looding and CSO ol- umes, hus p o ec ing he en i onmen as p esen ed in Fu e al. (2010); Bul e e al. (2010); Bu le e al. (2005); Cemb ano ?This esea ch is unded by EU unding o he p ojec LIFE EFFIDRAIN LIFE14 ENV/ES/00080 e al. (2004); Xu e al. (2013); Beeneken e al. (2013); Ga c´ ıa e al. (2015). Among eal- ime con ol me hods, model p edic- i e con ol (MPC), which can compu e op imal con ol ac ions aking in o accoun no only he cu en measu emen s bu also p edic i e beha io s in a ce ain ho izon, has been success ully es ed in wa e supply and also in he con ex o ad anced u ban d ainage ne wo ks by Cemb ano e al. (2004); Puig e al. (2009); Bu le e al. (2005); Pleau e al. (2005). Ne e heless, he majo i y o RTC applied in CUDN ha e only ocused on hyd aulic model and con ol objec i es wi hou conside ing he pollu ing quali y load inside he ca ied wa- e , e.g. Cemb ano e al. (2004). The ew e e ences which conside quali y models du ing he con ol p ocess mos ly use simula ion ools o o e come dealing wi h he complexi y o modelling wa e quali y di ec ly. In Ra hnayake (2015), con ol o CUDNs minimizing CSO is achie ed by using a non-so ed gene ic algo i hm (NSGA) linked wi h s o m wa e manage- men model (SWMM) 5.0. Bu le e al. (2005) op imizes in- eg a ed modelling o he ope a ion and con ol o an in eg a ed u ban d ainage sys em by de eloping a simula ion package SYNOPSIS which in eg a es he sewe sys em (simula ed by KOSIM), ea men plan (simula ed by IAWPRC) and i e model (simula ed by DUFLOW) using di e en so wa es. Fo he pu pose o ope a ing RTC o in eg a ed u ban d ainage sys ems aking in o accoun CSO pollu ion, app op ia e quali y models a e needed. Such models allow o e alua ion o wa e s acco ding o wa e quali y c i e ia as in Ahye e e al. (1998). Because o he inpu da a unce ain y and di icul y in calib a- ion, modelling he gene a ion and anspo a ion o pollu ion in CUDN du ing a s o m e en is complex. The e indeed exis some physically-based models which can p esen quali y dy- namics in CUDN, bu he ma hema ical equa ions a e di icul o be implemen ed wi hin RTC as shown in Rouse (1937); Van Rijn (1984); Macke (1980); Acke s and Whi e (1973). In o de o con on hese challenges, a simpli ied concep ual quali y modelling app oach o ien ed o RTC o dynamics o o al suspended solid (TSS) is p esen ed. The SWMM so - wa e (Rossman (2015)) in eg a ed wi h a lumped concep ual TSS model which depends indi ec ly on physical pa ame e s h ough low and compu a ion o Sain -Venan model, is used o calib a e and alida e he modelling app oach. A eal sewe in Bo deaux, Pe ino sewe ne wo k is used as a case s udy. The emainde o his pape is o ganized as ollows: Sec ion 2 p esen s he SWMM simula o in eg a ed wi h a lumped con- cep ual TSS model, which ep esen s cha ac e is ics o solid dynamics in he sewe . Sec ion 3 p esen s he simpli ied con- cep ual quali y modelling app oach o TSS in sewe s, a discus- sion abou RTC applied in CUDN is also p o ided in he end o Sec ion 3 as mo i a ion o u u e wo k. In Sec ion 4, calib a ion and alida ion o he p oposed simpli ied modelling app oach using SWMM-TSS and he eal-li e case Pe ino sewe ne wo k is p esen ed. Sec ion 5 p esen s he conclusion o his pape . 2. SWMM-TSS BASED ON LUMPED CONCEPTUAL MODEL As a ep esen a i e example o sewe pollu an s, models o solid concen a ion and load a e p esen ed in his pape , which gene ally p esen h ee dynamic beha io s by Be and- K ajewski (2006); Rossman (2015): •Accumula ion o solid sedimen s o e u ban ca chmen ; •Washo o solid sedimen s by ain all; •T ans e , e osion, deposi ion o solids in sewe ne wo ks and e en ion anks. 2.1 Concep o SWMM-TSS Since solids in sewe a e highly d i en by hyd aulics, a new quali y model has been de eloped based on he SWMM 5 sim- ula o ( e sion SWMM5.1.011) ha al eady includes a de ailed desc ip ion o hyd aulics using ull Ba ´ e de Sain Venan equa- ions. The model lib a y has been modi ied by adding equa ions desc ibing solids beha io o he ollowing phenomena: •Hyd ology: The exponen ial build-up model on he ca ch- men was modi ied o ake in o accoun only he impe i- ous a ea a he han he o al a ea. The washo model was eplaced by he one p oposed by M´ e adie (2011); •Sewe : Accumula ion and e osion phenomena a e de- sc ibed using Wiu (1985) ene gy balance based model used by Be and-K ajewski (1993); •S o age uni : Mixing and se ling p ocesses a e inspi ed om he wo k o B ia (1995); Ma u´ ejouls e al. (2012). Fig. 1. Modi ica ions made in he SWMM 5 lib a y model Figu e 1 is he scheme illus a ing he modi ica ions made in he SWMM 5 lib a y model. Whi e boxes a e o exis ing modules in SWMM 5 and g ey boxes a e o added quali y module. 2.2 Sewe Equa ions in SWMM-TSS As p e iously explained, he main addi ion in he model lib a y a e ega ding he sewe accumula ion and e osion model. This model is based on Wiu (1985) ha calcula es a anspo ca- paci y o he low CT depending on he wa e ρeand pa icle densi ies ρn, he wa e eloci y Umand pipe slope I, he pa - icles’ se ling eloci ies Wnacco ding o pa icles’ ac ions Fpn o each pa icle classes and he yield coe icien ηndi ec ly dependen on hyd aulics cha ac e is ics. CT =ρeUmI PN n=1(Fpn(ρn−ρe)Wn ρnηn) (1) The CT is compa ed wi h he TSS a he pipe inle o calcula e he quan i y o TSS ha can se le. The e osion model is based on a i s o de equa ion dependen on he mass o sedimen and a speci ic e osion coe icien o each pa icle classes. 3. SIMPLIFIED CONCEPTUAL MODELLING APPROACH Lumped complex hyd ology models wo k well o simula ing eal dynamics o TSS in sewe ne wo ks, bu o he RTC pu - pose o CUDN, a simple model s uc u e should be p esen ed acco ding o he ollowing p inciples p esen ed by No eys and Cluckie (1997); Cluckie e al. (1999); Puig e al. (2009): •Rep esen a i eness o he main dynamics; •Simplici y, lexibili y, expendabili y and speed; •A ailabili y o on-line calib a ion and op imiza ion. The modelling app oaches p esen ed in his pape a e designed o be used by RTC o p edic he e olu ion o TSS pa ame e s o e a sho ho izon, which p e en he limi a ion o classical modelling app oaches o empo al measu emen campaign. In o de o e alua e TSS discha ges o he ecei ing wa e , pe o ming measu emen s a he ou le o sewe is he mos sui able s a egy. The simpli ied dynamic model o TSS in CUDNs will conside TSS in sewe s and mass balance equa ion in he junc ions. TSS beha io s in de en ion anks, wei s and also WWTP will be pa o u u e esea ch. The leas squa e unc ion is used o measu e pe o mance o hese app oaches. 3.1 Simpli ied Modelling App oaches o Sewe O igina ing om he hyd aulic model o CUDNs as in Cem- b ano e al. (2004); Puig e al. (2009), a sewe unk in a CUDN can be assumed as a wa e ank con aine wi h capaci y o collec wa e based on he di e en be ween ups eam (Qin: m3/s) and downs eam (Qou :m3/s) lows as in Figu e 2. The anspo model o wa e olume (V) inspi ed om linea ank model o hyd aulic anspo will be exp essed as: V(k+1) =V(k)+ ∆ (Qin(k)−cV(k)) (2a) Qou (k)=cV(k) (2b) V Qin, TSSin Qou , TSSou Fig. 2. Tank model o a sewe In a sewe unk, mass conse a ion does no necessa ily apply o he solids because o solid se lemen and e osion. In o de o gene alize anspo model o TSS, an in e media e a iable Xwhich has no di ec physical meanings is used and he in-ou ela ion o TSS can be de ined as: X(k+1) =X(k)+ ∆ (TS S in(k)−cX(k)) (3a) TS S ou (k)=cX(k) (3b) whe e TS S in : TSS (g/m3) en e in o a sewe TS S ou : TSS (g/m3) ou o a sewe k: he cu en ime ∆ : sampling ime c: pa ame e s need calib a ion A e combining he equa ion (3a) and (3b), he ollowing dynamic model o TSS in a sewe is p oduced: Model 1 TS S ou (k+1) =(1 −c)TS S ou (k)+cTS S in(k) (4) By le ing he wo coe icien s o TS S ou and TS S in be inde- penden , a gene aliza ion o Model 1 eadily comes o mind, leading o: Model 2 TS S ou (k+1) =c1TS S ou (k)+c2TS S in(k) (5) Fu he mo e, om he physical cha ac e is ics, in a sewe , he dynamic o TSS is a ec ed by he TSS sedimen , e osion and also ime delays (Figu e 3). A e ha , a new linea delayed exp essions o TSS and wa e olume can be de ined as: Model 3 TS S ou (k)=c cT S S in(k−d)+ep(6a) Qou (k)= cQin(k−d0) (6b) whe e d: delay o TSS inside a sewe d0: delay o low inside a sewe c c, c,ep: pa ame e s need calib a ion Fig. 3. TSS model in a sewe Using hese modelling me hodologies, simpli ied models o TSS inside a sewe a e c ea ed, which allow on-line model calib a ions and eal- ime con ol o sewe ne wo ks. Fo each sewe , he pa ame e s c,c1,c2,c c,d, c and epneed o be calib a ed by GAMS using his o ic o eal- ime da a om eleme y sys ems. Emphasized on TSS modelling app oach, c, c1,c2,c c,dand epa e will be calib a ed. 3.2 Simpli ied Modelling App oaches o Junc ion Junc ions in a sewe ne wo k co espond o he poin s whe e wa e lows and also TSS a e me ged o spli . These elemen s i mass balance ela ions. When TSS comes in o a junc ion in he sewe ne wo k, he dynamics o TSS is modelled as equali y cons ain s ela ed o ups eam (TSS mass en e s in o his junc ion) and downs eam (TSS mass exi s om his junc ion). Fo di e en junc ions, he e could exis di e en numbe s o b anches (mo e han wo). Take a h ee b anches junc ion as an example (as shown in Figu e 4), he exp ession o he mass conse a ion o TSS can be w i en as (i is conside ed ha he junc ion does no add a signi ican delay): TS S in3(k)Qin3(k)=TS S ou 1(k)Qou 1(k)+TS S ou 2(k)Qou 2(k) (7) Fig. 4. TSS model in a junc ion 3.3 Model Pe o mance E alua ion In o de o alida e and compa e implemen a ions o hese p o- posed modelling app oaches, he leas squa e unc ion is used o e alua e model pe o mance o calib a ed alue using hese simpli ied model compa ing wi h simula ed esul s p oduced by lumped model o SWMM-TSS. FC = K X k=1 (V(k)−R(k))2(8) whe e V: calib a ed alue using simpli ied model R: simula ed alue using SWMM-TSS K: ime s eps Besides ha , he i ing a e be ween he calib a ed alue wi h simpli ied model and he simula ed alue wi h SWMM-TSS is de ined using Nash Su cli e model e iciency coe icien as explained in Nash (1970): V=PK k=1V(k) PK k=1k(9a) FCTS S =100             1−qPK k=1(R(k)−V(k))2 qPK k=1(R(k)−V(k))2             (9b) 3.4 Real- ime con ol o CUDN While s a e-o - he-a implemen a ions o RTC in UDN a e based only on hyd aulic a iables, he in eg a ed con ol o CUDNs and WWTP mus ake in o accoun quali y a iables in o de o minimize he o e all pollu ing load o he ecei ing en i onmen . The main ideas behind his in eg a ed con ol a e: •con olling he wa e de en ion and di e sion in he u ban d ainage ne wo k, so as o minimize TSS in he e luen s o he sys ems; • aking in o accoun he a iabili y in WWTP capaci y, acco ding o he in luen low and TSS concen a ion. The p ocess o CUDN con ol is also highly dependen on ain all scena ios which equi e ain p edic ions. Conside ing hese cha ac e is ic o CUDN con ol, MPC has been accep ed o ha e ad anced ad an ages in con olling u ban d ainage wa e sys ems as in Cemb ano e al. (2004); Xu e al. (2013). MPC can gene a e op imal con ol ac ions by op imizing he objec i e unc ion a e e y con ol s ep. Besides ha , MPC can p edic he u u e beha io o he sys em h ough using an in e nal model o e a ini e p edic ion ho izon. In CUDN, s a e o space o MPC can be p esen ed as in Xu e al. (2013): x(k+1) = (x(k),u(k),d(k)) (10) whe e xis a ec o o ne wo k s a es (e.g. wa e olume and TSS mass in a ank); uis a ec o o con ol a iables such as low ac oss a commanded ga e; dis a ec o o dis u bances ela ed o ain in ensi y and uno . Con inued e o s a e ocused on he in eg a ed con ol o CUDN which in ol es wa e quan i y and quali y using easible modelling app oaches o imp o ing pe o mance o CUDN like in Bu le e al. (2005); Van olleghem e al. (2005). 4. MODEL CALIBRATION AND VALIDATION The eal case s udy o he sewe ne wo k o Pe ino in Louis Fa gue ca chmen o Bo deaux Me opole co e s a o al a ea o 260 ha ha is mainly esiden ial. The sewe leng h is 3 km wi h an a e age slope o 0.007, qui e cons an o e he whole ca chmen . I includes a e en ion ank sepa a ed in h ee hyd aulically connec ed bodies o a o al s o age olume o 35000 m3. E en i he slope is gene ally low, he e is no sedimen issues on he sewe epo ed om he ope a o s. Rain scena ios used o calib a ion and alida ion come om he eal ain all measu ed a F ance in he yea o 2007 as show in Figu e 6. Rain scena ios T1, T2 in di e en ime s age o he yea 2007 which ep esen di e en ain densi ies a e selec ed o p oduce TSS aining da a using SWMM-TSS in o de o calib a e he p oposed simpli ied models. A e achie ing calib a ed TSS models, ano he wo di e en ain scena ios V1, V2 a e applied in SWMM-TSS o p oduce TSS da a o alida ing he wo king o hese p oduced simpli ied models. Fig. 5. SWMM con igu a ion o he Pe ino case s udy Fig. 6. Rain Scena io o Pe ino in he yea o 2007 The selec ed ain scena ios a e he ain all in he ollowing ime s ages (whe e he ime use o ma MM/DD/YYYY HH : MM :S S ): T1 : 10/10/2007 00:00:00-10/11/2007 00:00:00 T2 : 12/02/2007 00:00:00-12/03/2007 00:00:00 V1 : 02/10/2007 00:00:00-02/11/2007 00:00:00 V2 : 07/08/2007 00:00:00-07/09/2007 00:00:00 4.1 Model Calib a ion Fo each elemen s in CUDN, on-line calib a ion is needed. Se he sampling and epo ime o SWMM-TSS as 5 minu es, an eceden d y days as 10. A e applying T1, T2 ain scena ios in o Pe ino case, he TSS beha io s du ing hese ain all a e p oduced, which is he aining da a o calib a e he concep ual models. Taken condui APERI2.1 (as shown in Figu e 5) in he ups eam o Pe ino ne wo k as an example, whe e TS S in ep esen s TSS en e s in o condui APERI2.1, TS S ou ep esen s TSS ou o his elemen . Table 1 p o ides de ail pa ame e s calib a ed o Table 1. Calib a ed Pa ame e s c c1c2c c epd T1 0.26 0.73 0.26 0.87 45.11 2 T2 0.20 0.80 0.20 0.80 72.77 2 he h ee models o condui APERI2.1 in T1 and T2 ain sce- na ios. Figu e 7 and Figu e 8 show he pe o mances compa - isons be ween calib a ed models (TS S ou Model in he igu es) and simula ed alue using SWMM-TSS(TS S ou S WMM in he igu es) in T1 and T2 ain all scena ios. These esul s con i m ha , all he models ha e conside able pe o mance. Model 2 wo ks be e han o he wo models. Model 3 has low i ing pe o mance bu he changing ends o TSS is simila wi h ha in SWMM-TSS, which has meaning du ing he con ol p ocess. 0 50 100 150 200 250 200 300 400 500 600 700 Time S eps [5min] TSS Concen a ion [mg/l] Model 1 i ing = 93.3504%, Model 2 i ing = 93.4646%,Model 3 i ing = 62.1974% TSSin SWMM TSSou SWMM TSSou Model 1 TSSou Model 2 TSSou Model 3 Fig. 7. Calib a ion pe o mance o T1 0 50 100 150 200 250 200 250 300 350 400 450 500 Time S eps [5min] TSS Concen a ion [mg/l] Model 1 i ing = 98.0939%, Model 2 i ing = 98.2986%,Model 3 i ing = 89.0995% TSSin SWMM TSSou SWMM TSSou Model 1 TSSou Model 2 TSSou Model 3 Fig. 8. Calib a ion pe o mance o T2 4.2 Model Valida ion In o de o alida e u he he p oposed calib a ed models, ain scena ios V1 and V2 a e applied o he models calib a ed in T1 and T2 sepa a ely. Figu e 9, 10, 11 and 12 p o ide he alida ion pe o mance using V1 and V2 ain scena ios o he p oduced models by T1 and T2. The i ing compa isons p o- ided om hese igu es shows ha , compa ing wi h simula ion esul s om SWMM-TSS, all he p oposed models a e wo king well using he calib a ed models. Table 2 wi h de ailed i ing pe o mances o bo h he calib a- ion and alida ion scena ios con i m hese conclusions. 5. CONCLUSIONS This pape p oposed simpli ied concep ual modelling ap- p oaches o ep esen ing dynamic beha io s o TSS in sewe ne wo ks, which a e o ien ed o be in ol ed in o he in eg a ed 0 50 100 150 200 250 200 300 400 500 600 700 800 900 Time S eps [5min] TSS Concen a ion [mg/l] Model 1 i ing = 96.5059%, Model 2 i ing = 96.4982%,Model 3 i ing = 91.213% TSSin SWMM TSSou SWMM TSSou Model 1 TSSou Model 2 TSSou Model 3 Fig. 9. V1 alida ion pe o mance o T1 0 50 100 150 200 250 500 1000 1500 2000 2500 3000 3500 4000 Time S eps [5min] TSS Concen a ion [mg/l] Model 1 i ing = 65.5939%, Model 2 i ing = 65.4938%,Model 3 i ing = 54.6964% TSSin SWMM TSSou SWMM TSSou Model 1 TSSou Model 2 TSSou Model 3 Fig. 10. V2 alida ion pe o mance o T1 0 50 100 150 200 250 200 300 400 500 600 700 800 900 Time S eps [5min] TSS Concen a ion [mg/l] Model 1 i ing = 95.5693%, Model 2 i ing = 95.5597%,Model 3 i ing = 90.8331% TSSin SWMM TSSou SWMM TSSou Model 1 TSSou Model 2 TSSou Model 3 Fig. 11. V1 alida ion pe o mance o T2 0 50 100 150 200 250 500 1000 1500 2000 2500 3000 3500 4000 Time S eps [5min] TSS Concen a ion [mg/l] Model 1 i ing = 63.6791%, Model 2 i ing = 63.6013%,Model 3 i ing = 44.358% TSSin SWMM TSSou SWMM TSSou Model 1 TSSou Model 2 TSSou Model 3 Fig. 12. V2 alida ion pe o mance o T2 RTC con olle o CUDN and WWTP o en i onmen al p o- ec ion. The SWMM is used o p oduce ealis ic simula ed da a o calib a ion and alida ion. The Pe ino sewe ne wo k, which is a eal li e example, is used as a case s udy. The applica ion and alida ion esul s o he p oposed modelling app oaches ha e p o ed ha , he simpli ied concep ual model P ep in s o he 20 h IFAC Wo ld Cong ess Toulouse, F ance, July 9-14, 2017 Table 2. Pe o mance Compa isons Tes /Valida e Scena ios Model 1 Model 2 Model 3 T1 93.35% 93.46% 62.20% V1 96.51% 96.50% 91.21% V2 65.59% 65.49% 54.70% T2 98.09% 98.30% 89.10% V1 95.57% 95.56% 90.83% V2 63.68% 63.60% 44.36% can cap u e he main cha ac e is ics o TSS e olu ion in sewe ne wo ks wi h equa ions ha a e simple enough o be used o in eg a ed RTC con ol o CUDN. This wo k is pa o a esea ch p ojec aimed o design Model P edic i e Con ol o bo h wa e quan i y and quali y models in CUDN. ACKNOWLEDGEMENTS The au ho s wish o hank he suppo ecei ed by he Eu o- pean Commission esea ch g an o p ojec LIFE EFFIDRAIN (LIFE14 ENV/ES/000860), hank Aig¨ ues de Ba celona o he inancial and echnical suppo . Au ho s also hank Bo deaux Me opole and SGAC o echnical suppo and inancial sup- po h ough he esea ch con en ion signed wi h he LyRE. REFERENCES C. Becouze, J.-L. Be and-K ajewski, A. Demb´ el´ e, C. C en- Oli ´ e and M. Coque y. P elimina y assessmen o luxes o p io i y pollu an s in s o mwa e discha ges in wo u ban ca chmen s in Lyon. P oceed. o he 13 h IWA in e n. con e . on Di use Pollu. and In eg. Wa e s. Manage., Seoul, Sou h Ko ea, 2009. J. Gaspe i, S. Ga naud, V. Roche and R. Moille on. P io i y pollu an s in was ewa e and combined sewe o e low. J. Scien. o he To . En i on., 407(1): 263–272, 2008. D. Bu le and M. Sch¨ u ze. In eg a ing simula ion models wi h a iew o op imal con ol o u ban was ewa e sys ems. J. En i on. Modell. and So w., 20(4): 415–426, 2005. G. Cemb ano, J. Que edo, M. Salame o, V. Puig, J. Figue as and J. Ma ´ ı. Op imal con ol o u ban d ainage sys ems. A case s udy. J. Con . Engin. P ac ., 12(1): 1-9, 2004. B. Joseph-Du an, M.N. Jung, C. Ocampo-Ma ´ ınez, S. Sage and G. Cemb ano. Minimiza ion o sewage ne wo k o e - low. J. Wa . Resou . Manage., 28(1): 41-63, 2014. M. Pleau, H. Colas, P. La all´ ee, G. Pelle ie , R. Bonin. Global op imal eal- ime con ol o he quebec u ban d ainage sys- em. J. En i on. Modell. and So w., 20: 401-413, 2005. M. Xu, P.J. an O e loop, N.C. Van de Giesen. Model educ ion in model p edic i e con ol o combined wa e quan i y and quali y in open channels. J. En i on. Modell. and So w., 42: 72-87, 2013. L. Ga c´ ıa, E. Ba ei o-Gomez, E. Escoba , D. T´ ellez, N. Qui- jano and C. Ocampo-Ma ´ ınez. Modeling and eal- ime con ol o u ban d ainage sys ems: a e iew. Ad an. in Wa . Resou ., 85: 120-132, 2015. D. Bu le and J. Da ies. U ban D ainage., Den e , CRC P ess, 2010. M. Ahye e, G. Chebbo, B. Tassin and E. Gaume. S o m wa e quali y modelling, an ambi ious objec i es? J. Wa . Sci. Tech., 37(1): 205-213, 1998. W. Schilling. Real- ime con ol o u ban d ainage sys ems. The s a e-o -a . London: IAWPRC Task G oup on Real- ime Con ol o U ban D ainage Sys ems, 1989. T. Beeneken, V. E be, A. Messme , C. Rede , R. Rohl ing, M. Schee , M. Schue ze, B. Schuma che , M. Weiland and M. Weyand. Real ime con ol (RTC) o u ban d ainage sys ems - A discussion o he addi ional e o s compa ed o con en ionally ope a ed sys ems. J. U b. Wa ., 10(5): 293- 299, 2013. G. Fu, S. Khu and D. Bu le . Op imal dis ibu ion and con ol o s o age ank o mi iga e he impac o new de elopmen s on ecei ing wa e quali y. J. En i on. Engine., 136(3): 335- 342, 2010. P.A. Van olleghem, L. Benede i and J. Mei laen. Modelling and eal- ime con ol o he in eg a ed u ban was ewa e sys em. J. En i on. Modell. and So w., 20: 427-442, 2005. J.-L. Be and-K ajewski. Modelling o sewe solids p oduc ion and anspo . J. Modell. o Sedi. T ans. and P oce., INSA de Lyon, 2006. L.A. Rossman S o m Wa e Managemen Model Use ’s Manual Ve sion 5.1., U.S. En i . P o . Agn., 2015. W.C. Hube De e minis ic modelling o u ban uno quali y. U ban Runo Pollu ion, Be lin, Sp inge -Ve lag, 1986. L.C. an Rijn Sedimen anspo , pa II : suspended load anspo . J. Hyd a. Engine., 110(11): 1613-1641, 1984. V. Puig, G. Cemb ano, J. Rome a, J. Que edo, B. Azna , G. Ram´ on and J. Cabo . P edic i e op imal con ol o sewe ne wo ks using CORAL ool: applica ion o Rie a Blanca ca chmen in Ba celona. J. Wa . Sci. Technol., 60(4): 869- 878, 2009. R. No eys and I. Cluckie A no el app oach o eal- ime modelling o la ge u ban d ainage sys ems J. Wa . Sci. Technol., 36(8-9): 19-24, 1997. I.D. Cluckie, A. Lane and J. Yuan. Modelling la ge u ban d ainage sys ems in eal ime. J. Wa . Sci. Technol., 39(4): 21-28, 1999. H. Rouse. Nomog am o he se ling eloci y o sphe es. J. Comm. on Sedim., 57-64, 1937. P. Acke s and W.R. Whi e. Sedimen anspo : a new app oach and analysis. J. Hyd a. Di i., 99(11): 2041-2060, 1973. E. Macke Ve gleichende Be ach ungen zum Fes s o ans- po im Hinblick au ablage ung eie S ¨omungszus ¨ande in Regen- und Schmu zwasse kan¨alen, B aunschweig, Deu sch- land, pages 109-233, 1980. U. Ra hnayake. Mul i-objec i e op imiza ion o combined sewe sys ems using SWMM 5.0. J. Ci il Eng. A chi ec . Res., 2(10): 985-993, 2015. J.L. Be and-K ajewski. Mod´ elisa ion du anspo solide en ´ eseau d’assainissemen uni ai e: le mod` ele HYPOCRAS. La Houille Blanche, 4: 243-255, 1993. R. Wiu . T anspo o suspended ma e ial in open and sub- me ged s eams. J. En i on. Eng. ASCE., 111(5): 774-792, 1985. M. M´ e adie . T ai emen e analyse de s´ e ies ch onologiques con inues de u bidi ´ e pou la o mula ion e le es de eje s u bains en emps de pluie. PhD hesis a Ins i u Na ional des Sciences Appliqu´ees Lyon, INSA, Lyon, F ance, 2011. T. Ma u´ ejouls, P.A. Van olleghem, G. Pelle ie and P. Lessa d. A phenomenological e en ion ank model using se ling eloci y dis ibu ions. Wa . Res., 46: 6857-6867, 2012. P. B ia . Lyonnaise des eaux in e nal Resea ch and De elop- men p og ams, 1995. J.E. Nash and J.V. Su cli e. Ri e low o ecas ing h ough concep ual models pa I-A discussion o p inciples. Jou nal o Hyd ology, 10(3): 282-290, 1970.