污水处理相关文献

Energyefficiencyinwastewatertreatmentsplants:Optimizationofactivatedsludgeprocesscoupledwithanaerobicdigestion

abstract

ThispaperpresentsastudyconcerningtheoptimizationofaWasteWaterTreatmentprocess.Theprocessdealswithcarbonandnitrogenremovalandincludesactivatedsludgereactorscoupledwithananaerobicdigestionreactor.Nitrificationandde-nitrificationbiochemicalreactionsareduetothebiological

activityofheterotrophicandautotrophicmicro-organismsoccurringinsidethereactors.RigorousPlant-WidemodelsthatrepresentthemainbiochemicaltransformationshavebeenconstructedaspertheCEITapproach[1].TheenergyconsumptionforeachPhysicalUnitOperation(P.U.O.)involvedinthe

flow-sheetisevaluatedandafulllinkismadebetweenthebiologicalactivityandtheelectricaldemandorproduction.Steady-statemathematicaloptimizationsarethencomputedandtheinfluenceofprimarysettlingefficiencyonelectricalautonomyisquantifiedanddemonstrated.Theammoniumrecyclingfrom

digestiontoactivatedsludgereactorsisalsodemonstratedtobealimitingfactorfortheoverallenergyefficiency,aswellastheC-substrateavailabilityfordenitrifying.Someconclusionsarethendrawntoimprovetheglobalelectricalefficiencyofthesystem.

1.Introductionandgeneralcontext

Wastewatertreatmentconsistsmainlyofthreemajorprocesses:

biochemicaltreatments,liquid/solidseparationsoperationand

thermalprocessesforsludgetreatmentandvalorization.TheFig.1isanoverviewofaclassicalWasteWaterTreatmentPlant(WWTP),eachunitrepresentingaPhysicalUnitOperation(PUO).Thewastewaterisfirsttreatedbymeanofbiochemicalandsettlingprocessesinthewaterstream.Thiswaterstreamfinallyresultsintwonewstreams:the“purified”wateranda“sludge”streamconcentratedinbiomass.Thesludgestreamprocessesconsisttheninstabilizingandvalorizingthesludgebeforetheremainingmatterisreleasedtotheenvironment.Themostusedtechnologyforliquid/solidsseparationisthesettlingprocess.Duetothephenomenaofgravity,theparticlessettledownandfinallytwostreamsareproduced:aparticleclarifiedstreamandaparticleconcentratedstream.Thesettlersaregenerallyclassifiedintoprimaryandsecondarysettlers.Theprimarysettlingisapplieddirectlytotherawwastewaterandproducesacarbon-richprimarysludge,whilethesecondarysettlingprocessisappliedtoseparatethebiomassproducedintheActivatedSludgereactorsfromthewater(seeFig.1).

Aerobicbiologicaltreatmentconsistsofsupplyingoxygeninside

theActivatedSludgereactorsinordertomaintainandgrowmicroorganisms.Boththecarbon-basedpollutantsandthenitrogenbasedpollutantsarethendegradedinsidethereactorsbythe

combinedbiologicalactivitiesofheterotrophicandautotrophic

bacteria[2].Anaerobicdigestionisabiologicalphenomenonthat

appearswhenoxygenandnitrateconcentrationsareverylowin

thesludge.Underspecifictemperatureandforsufficientresidencetimes,specializedmicro-organismsbecomeactive.Theorganicnutrientspresentinthesludgearethenusedbythesemicroorganismswhichpartiallyconvertitintoamixofmethaneandcarbondioxide[3].Thisanaerobicbiologicaltreatmentisthereforecombinedwithdewateringanddryingprocessesandastabilizedsludgeisfinallyobtained.Thebiogasproducedbythedigestercanbeburnedinsideacogenerationengine.Heatandelectricitycanthereforebeproducedlocallyandcontributetoreducetheenergycosts.

Uptonowthescientificcommunityinvolvedinthefieldof

wastewatertreatmenthasfocusedmainlyonthewaterqualityand

associatedmodelingissues.Inouropinion,effortsmustbedonetolinkwaterqualityandWWTPpollutantsremovalefficiencywith

energyaspects,becausefutureindustrialpracticeinthecontextofglobalwarmingandfossilresourcerarefactionwillbegreatly

impactedbytheenergycosts.Onanotherside,environmental

pressureonaquaticresourcesbecomemoreandmoresevereand

theenergyconsumptiontopurifywastewaterandkeepriversand

lakesinanecologicallyacceptablestatewillprobablyincrease.

Mathematicalmodelsandrigorousoptimizationalgorithmsare

reallyhelpfulinthiscontext,astheycapturethemainfeaturesofeachPUO.Furthermoremodelsbringadeepunderstandingofthe

physicalandbiologicalmechanismsandtheirinteractions.The

mathematicalmodelsusedinthisstudyweredevelopedbasedon

thescientificliteratureandareimplementedonthegPROMS

platform[4].Effortshavebeendonetosystematicallylinkthe

biologicalactivityandthecorrespondingenergyconsumptionor

production,providingbythiswayalinkbetweenwaterquality

andenergyefficiency.TheobjectiveofthestudypresentedinthispaperwasinfacttodeterminehowtheglobalenergyefficiencyofWWTPcouldbeimproved,byactingonprocessdesign.Inthiscontext,thekeyfactorsarethetrendsgivenbythemodels,ratherthanaveryaccuratepredictionofthepollutantsconcentrationattheexitoftheplant.AllthecomputationspresentedinthispaperhavebeendoneusingthenumericalsolversprovidedbyProcessSystemEnterprise.

2.Modelingmethodology

Thefirstsectionintroducesthemainequationsusedtorun

steady-stateoptimizations.Someequationsarenotintroducedas

theyareclassicalmodelsavailableintheextensiveliteratureconcerningWWTPmodeling.Thisisthecaseforthesettlingprocesses,includingone-dimensionalmodelsandpointmodels.Actually,thesemodelsarethebasisforalltheliquid/solidseparationprocessesincludedontheflow-sheetstudiedinthispaper(seeFig.1).Thebiochemicalreactionsmodelinghavebeenthesubjectofspecialattention.Themodelsdevelopedarefullyconservativeintermsofmass,byoppositiontotheoriginalASM1modelthatconserveonlytheTheoreticalOxygenDemand(ThOD),nitrogenandioniccharge[2].Themasstransferbetweentheliquidandgaseousphasesisalsoincludedinthereactor’smodel.Eachchemicalcompoundisdescribedbyastoichiometricformulaandquantifiedbyaconcentrationexpressedasmole,gramsofnitrogenorgramsofThOD.TheThoDisassumedtobeequivalenttotheChemicalOxygenDemand(COD).TheCODisaspecificchemicalmeasurementthatcharacterizestheCarboncontentofthewastewateranditiswidelyusedinthefieldofWWTPengineering.Theplant-widemethodusedtoquantifythebiochemicalreactionsimpliesahomogeneousmattercompositioninallthedifferentPUO.NospecialinterfacesarethenneededtoconnecttheActivatedSludgeprocesswiththeAnaerobicDigestionreactor.ThesemodelsarebasedontheCEITmethodology[1].Thereactorsmodelsaswellasthecompressorsandpumpsmodelsarealsointroducedinthissection.

2.1.Plant-widebiochemicalmodeling

Theplant-widemodelsincludedinthereactorsmodelsarepresentedundertheformofaPetersenmatrixandakineticvector,

similarlytotheASM[2]andADM[3]models.TheyarecalledPWAS

(Plant-wideActivatedSludge)andPW-AD(Plant-WideAnaerobic

Digestion)modelsandareabletoreproduceallthefeaturesof

ASMandADMmodels.Thefulldetailsandthemethodologyused

todevelopthesemodelswillbegiveninadedicatedpaper.The

modelshavebeencomparedsuccessfullywithliteraturebenchmarks.TheFig.2presentsforexampleacomparisonofthePW-ASmodelwiththeclassicalASM1modelinthecontextoftheBenchmarkSimulator1(BSM1)[5].ThequantityplottedonFig.2istherejectedconcentrationofammonium.

2.1.1.Commonsetofspecies

Thefirststeptobuildthebiochemicalmodelsconsistsofdefiningthesetofspeciesthatisincludedintheplant-widemodels.ThissetmustbeabletoreproduceallthefeaturesofASMandADMmodels.Aparticularefforthasbeenmadetoavoidtheuseoflumpedcomponentsandallthechemicalspeciesneededtofullydescribethedifferentbiologicalactivitiesareincluded.Mineralscompoundsandcorrespondingacids/basesequilibriumsarealsotakenintoaccount.Theplant-widemodelsinclude39species(exceptwater)dividedinto25solublespeciesand14particulatespecies.EachcomponentincludedinthemodelischaracterizedintermofC,H,O,N,P,chargeandequivalentThOD.Theelementalcompositionsforeachcompoundarethesameasthosereportedin[1].Somespeciesareconsideredtobeaqueousandareincludedintheliquid/gasmasstransfer.Biologicalinertsarerepresentedbothinsolubleandparticulateforms.Thesefractionsareassumedtobecommontothedifferentbiologicalmodels.Someparticulatescomponentsareconsideredassubstratesandothersasgroupsofmicro-organismswhichdegradeandtransformtheorganicmatter,producinginthesametimemineralsandgaseousspecies.Onecompositecompoundisalsoincludedinthemodelsandrepresentsthebiggestorganicparticlesandmacromolecules.

2.1.2.PetersenmatrixandkineticsvectorforPW-ASandPW-AD

models

Themathematicalequationsandformulasemployedtobuild

theActivatedSludgeandAnaerobicDigestionplant-widemodels

consistsofwritingstoichiometrymatricesandkineticsvectors

(Petersenformalism).Thebehaviorofthedifferentkindsofmicroorganismsconsideredintheaerobic/anoxicandanaerobicreactorsisdescribedbyassumingthatthedifferentgroupsarefullydifferentiated.Thematrixesarethenbalancedforeachelement:C,H,O,N,PandThOD.Thisisdoneforeachprocessjbycomputingcorrectvaluesforthesinkorsourcescoefficientsinthematrices(notedqjk,withke{C,N,P,H,O,ThoD},seeappendixA).Thespecieschosentofulfilltheelementalbalancesaremineralones:dissolvedoxygen,ammonium,protons,phosphates,wateranddissolvedcarbondioxideorbicarbonate(HCO3).

2.2.Completelystirredtankreactors(CSTR)models

ActivatedSludgeprocessesareusuallyoperatedinaeratedtankreactorsandchannels.Partofthemixingisensuredbymechanicalwork(impellers)andtheotherpartbyaerators.ConcerningAnaerobicDigestion,moresophisticatedreactorsexistandparticlescouldbeseparatedfromwaterusingmembranesand/orsettlingeffects.ThereactorsinvolvedinthisworkaremodeledusingtheCSTRhypothesis.Eachreactormodelincludestwophases:aliquidandagaseousphase.ThisisdonemainlybecausealltheenergyaspectsinvolvedinWasteWaterTreatmentarestronglyassociatedtogaseousmasstransfer(oxygentransfertoactivatedsludgeandmethaneproductionbydigestion).

2.2.1.Liquidphasemassbalances

Themassconservationequationsfortheliquidphasearewrittenunderthefollowingformforeachcompoundi,whereCiisaconcentrationreferringtooneofthesolubleorparticulatecomponentsincludedintheplant-widemodels.Somecompoundsarethensubjectedtomasstransferwiththegasphase(butnotallthecompounds).TheHenry’slawcoupledwithanexperimentallawforcomputingtheliquidgasexchangecoefficientquantifiesthemasstransferphenomenonfortheactivatedsludgereactors.TheexchangecoefficientkLiaisthenafunctionofgasflowrate,gascompositionanddiffuserscharacteristics.Forthedigester,theexchangecoefficientissettotheconstantvalueindicated

in[3].Combinedwithplant-widemodels,40OrdinaryDifferentialEquations(ODE)mustbesolvedforeachreactorifthewatercompoundisincluded.

2.2.2.Gasphasemassbalances

Chemicalinteractionsarenotconsideredtooccurinthegasphase

andconsequentlynosourcetermsappearintheEq.(4).Themass

transferbetweenliquidandgasisstilltakenintoaccountby

meansofthekLiaicoefficientandtheassociatedterminvolving

saturationconstantsforaqueouscompound.ThetotalgasflowItmeansthattheaerationsystem(usuallycompressors)mustpressurizethegastotherequiredpressurePgasintoformbubblesinsidethereactor,butitalsoneedtocompensatetheheadlossesduetothenetworkdistributionpipesanddiffusers.

3.Steady-stateoptimizations

ThePUOmodelsarelinkedonaflow-sheettoreproducetheActivatedSludgeprocesscombinedwiththeAnaerobicDigestion

process.Theresultingmodelisasetofequationsthatcouldbe

solvedbothfordynamicorsteady-statecases.TheWWTPconfigurationstudiedisreproducedonFig.1.

3.1.Initializationandsteady-state

Inordertocomputeasteady-statesolution,aninfluentstream

totheprocessisfirstspecifiedbydefiningtheinlettemperature,thevolumetricflowrateofwaterandtheconcentrationofthedifferentspeciesconsideredinthePWmodels.AcompletesetofparametersisalsoprovidedandtheASM1benchmark(BSM1)[5]withtheADM1[3]reporthasbeenusedasreferencesforthebiochemical,aerationandsettlersmodels.TheinfluentstreamtotheWWTPwasalsospecifiedsimilarlytotheBSM1dryweatherconfiguration.ThecharacteristicsofeachpumpsandcompressorsarededucedfromtheenergeticdataconsumptionavailableintheBSM1[5]orintheSwissFederalOfficeforEnergyreport[6].Themodelsaresolvedforsteady-state,byaninitializationprocedurethatconsistsinapplyingastepatinitialtimetothewastewaterinfluenttotheplant.Thesystemisthenrelaxedandthedynamicequationsconvergetothefinalsteady-state,accordinglytothestepandtheparameterssettingschosenforthedifferentmodels.Themassbalancesaresystematicallycheckedtoensurethatasteady-stateiseffectivelyreached(seeFig.3).

3.2.Optimalpointandparametricprocedure

InordertooptimizetheWWTPsomevariablesaresetasdecision

variables.Theoptimizationalgorithmnextcomputesthisset

(withinapredefinedrange)tominimizeormaximizeapredefined

objectivefunction.Someconstraintsarealsoimposedonthe

systemandmustbesatisfiedduringthisprocedure.Theoptimalsetofvariablesobtainedisthencorrespondingtoaminimumor

amaximumoftheobjectivefunctionthatsatisfiestheconstraints

imposed.Themethodologyusedtosolvetheoptimizationproblem

isagradient-basedmethod(SequentialQuadraticProgramming

algorithm)includedinthegPROMS
software.Ifone(ormore)constraint(s)is(are)modified,anewsetofdecisionvariablesandanewvaluefortheobjectivefunctionareobtained.Thisprocedureiscalledparametricoptimizationandaspecialalgorithmhasbeendevelopedinordertoobtainacompleterangeofoptimalpointsthatcorrespondstodifferentconstraintsonwaterquality.TheprocedureisresumedonFig.4.Foreachoptimalsetofdecisionvariables(correspondingtoeachconstraintvalue)asteady-stateoptimalpointisrecomputedandsavedasatextfile(includingallthemodelsvariables).Thefinalresultcouldbe,forexample,theoptimalvalueoftheobjectivefunctionasafunctionofthevaryingconstraint,butanymodelvariable(decisionvariablesornot)couldberepresentedasafunctionofthevaryingconstraint.Thismethodprovidesavaluabletoolforanalyzingtheoptimalresultsandunderstandingwhatthelimitingfactorsare,asshowninthelastsectionofthispaper.ThevariablesselectedasdecisionvariablesarereportedintheTable1.TheseconstraintsdefinethewaterqualitybyspecifyingindexesonCODandnitrogenconcentrationattheexitoftheWWTP(intheclarifiedandtreatedwater),asindicatedonFig.1.Ammoniumandnitratearedifferentiatedandthevaluesareexpressedbycubicmeterofwater.Aconstraintisalsospecifiedforthesludgeage,ensuringatechnicalacceptablerange.ThesludgeageisaquantityrepresentingtheresidencetimeofparticlesintheActivatedSludgeprocessanditisanimportantdesignvariablefortheActivatedSludgeprocess.Constraintsarealsoimposedfortheresidencetimeindigester,forTotalSolidsSuspended(TSS)inActivatedSludgereactorsandintheincomingsludgetothedigesterreactor.ForallcomputationsthepHinthedigestionreactorismaintainedabove7.Thisisdonebyadding

bicarbonatetothereactor(buffereffect).TheTable2summarizes

theconstraintsandtheassociatedunitsandrange.TheconstraintonrejectedNH4isvaryinganditistheobjectoftheparametricoptimizationprocedure,becausetheremovalofammoniumisknowntobeastronglimitingfactortotheoverallenergyefficiencyofWWTP.TheconstraintonrejectedNO3isalsovariedandsettopredefinedvalues(8,12and20gN/m3).ResultsThethermalautonomywasnotdirectlyinvestigatedduringthecomputations.Ifnecessaryitcouldbeincludedintheobjectivefunctionorasanewconstraint.TheelectricalautonomyandtheelectricalconsumptionareknowntobecriticalissuesofWWTPandtheresultspresentedinthispaperfocusonelectricalaspects.Anyway,mostoftheconclusionsdeducedfromthecomputationspresentedinthispaperarestillvalidifconsideringthethermalautonomy.

4.1.Primarysettlingefficiencyasaparameter

Thefirstcomputationspresentedhavebeendonebysettingthe

primarysettlerefficiencyasaparametervaryingbetween0%and

100%.0%efficiencymeansthattheparticlesincomingtotheWWTP

arenotsegregatedbytheprimarysettler.Theparticlesconcentrationsinthemainwaterfluxandintheprimarysludgefluxarethenequal.100%efficiencymeansthatalltheparticlesincomingtotheWWTParesenttotheprimarysludgeandconsequentlyinthedigester.Thethickenerbeforethedigesterreactorensuresthatthemixofsecondaryandprimarysludgeissufficientlyconcentratedinallthecases(theparticlesconcentrationinthesludgetodigestisconstrainedto50g/l).TheresultscorrespondingtotherejectedNO3concentrationequalto8gN/m3arepresentedonFig.5.

4.2.Primarysettlingefficiencyasadecisionvariable

Theresultspresentedintheprevioussectiondemonstrateboth

theimportanceofprimarysettlingandtheavailabilityoforganiccarboninthedifferentreactors:carbonisrequiredfordenitrificationintheASprocessbutitisalsorequiredforbiogasproductioninthedigestionreactor.Itmeansthattheoptimalelectricalautonomyofthewholeprocessisstronglyrelatedtothecarbonandnitrogensplittingbetweentheaerobicandanaerobictreatment.Inthenextcomputations,theprimarysettlingefficiencywassetasadecisionvariable.Theoptimizercannowestimateoptimumvaluesforthisefficiency.Theothersdecisionvariablesandconstraintsareidentical.

5.Conclusionsandperspectives

MathematicalmodelsformostofthePUOinvolvedinWWTPhavebeendevelopedandimplementedonthegPROMS
platform.Theresultingsetsofequationsareso