热能 外文翻译 外文文献 英文文献 火力发电厂先进的蒸汽温度调节控制算法.docx

热能 外文翻译 外文文献 英文文献 火力发电厂先进的蒸汽温度调节控制算法.docx

《热能 外文翻译 外文文献 英文文献 火力发电厂先进的蒸汽温度调节控制算法.docx》由会员分享，可在线阅读，更多相关《热能 外文翻译 外文文献 英文文献 火力发电厂先进的蒸汽温度调节控制算法.docx（31页珍藏版）》请在冰点文库上搜索。

热能外文翻译外文文献英文文献火力发电厂先进的蒸汽温度调节控制算法

英文翻译部分

英文部分：

Advancedcontrolalgorithmsforsteamtemperatureregulationofthermalpowerplants

A.Sanchez-Lopez,G.Arroyo-Figueroa*,A.Villavicencio-Ramirez

InstitutodeInvestigacionesElectricas,DivisiondeSistemasdeControl,ReformaNo.

113,ColoniaPalmira,Cuernavaca,Morelos62490,Mexico

Received5February2003;revised6April2004;accepted8July2004

Abstract

Amodel-basedcontroller（DynamicMatrixControl）andanintelligentcontroller（FuzzyLogicControl）havebeendesignedandimplementedforsteamtemperatureregulationofa300MWthermalpowerplant.Thetemperatureregulationisconsideredthemostdemandedcontrolloopinthesteamgenerationprocess.BothproposedcontrollersDynamicMatrixController（DMC）andFuzzyLogicController（FLC）wereappliedtoregulatesuperheatedandreheatedsteamtemperature.TheresultsshowthattheFLCcontrollerhasabetterperformancethanadvancedmodel-basedcontroller,suchasDMCoraconventionalPIDcontroller.Themainbenefitsarethereductionoftheovershootandthetighterregulationofthesteamtemperatures.FLCcontrollerscanachievegoodresultforcomplexnonlinearprocesseswithdynamicvariationorwithlongdelaytimes.

Keywords:

Thermalpowerplants;Powerplantcontrol;Steamtemperatureregulation;Predictivecontrol;Fuzzylogiccontrol

1.Introduction

Currenteconomicandenvironmentfactorsputastringerrequirementonthermalpowerplantstobeoperatedatahighlevelofefficiencyandsafetyatminimumcost.Inaddition,thereareanincrementoftheageofthermalplantsthataffectedthereliabilityandperformanceoftheplants.Thesefactorshaveincreasedthecomplexityofpowercontrolsystemsoperations[1,2].

Currently,thecomputerandinformationtechnologyhavebeenextensivelyusedinthermalplantprocessoperationandcontrol.Distributedcontrolsystems（DCS）andmanagementinformationsystems（MIS）havebeenplayinganimportantroletoshow

theplantstatus.ThemainfunctionofDCSistohandlenormaldisturbancesandmaintainkeyprocessparametersinpre-specifiedlocaloptimallevels.Despitetheirgreatsuccess,DCShavelittlefunctionforabnormalandnon-routineoperationbecausetheclassicalproportional-integral-derivative（PID）controliswidelyusedbytheDCS.PIDcontrollersexhibitpoorperformancewhenappliedtoprocesscontainingunknownnon-linearityandtimedelays.ThecomplexityoftheseproblemsandthedifficultiesinimplementingconventionalcontrollerstoeliminatevariationsinPIDtuningmotivatetheuseofotherkindofcontrollers,suchasmodel-basedcontrollersandintelligentcontrollers.

Thispaperproposesamodel-basedcontrollersuchasDynamicMatrixController（DMC）andanintelligentcontrollerbasedonfuzzylogicasanalternativecontrolstrategyappliedtoregulatethesteamtemperatureofthethermalpowerplant.Thetemperatureregulationisconsideredthemostdemandedcontrolloopinthesteamgenerationprocess.Thesteamtemperaturedeviationmustbekeptwithinatightvariationrankinordertoassuresafeoperation,improveefficiencyandincreasethelifespanoftheequipment.Moreover,therearemanymutualinteractionsbetweensteamtemperaturecontrolloopsthathavebeenconsidered.Otherimportantfactoristhetimedelay.Itiswellknowthatthetimedelaymakesthetemperatureloopshardtotune.ThecomplexityoftheseproblemsanddifficultiestoimplementPIDconventionalcontrollersmotivatetoresearchtheuseofmodelpredictivecontrollerssuchastheDMCorintelligentcontroltechniquessuchastheFuzzyLogicController（FLC）asasolutionforcontrollingsystemsinwhichtimedelays,andnon-linearbehaviorneedtobeaddressed[3,4].Thepaperisorganizedasfollows.AbriefdescriptionoftheDMCispresentedinSection2.TheFLCdesignisdescribedinSection3.Section4presentstheimplementationofbothcontrollersDMCandFLCtoregulatethesuperheatedandreheatedsteamtemperatureofathermalpowerplant.TheperformanceoftheFLCcontrollerwasevaluatedagainsttwoothercontrollers,theconventionalPIDcontrollerandthepredictiveDMCcontroller.ResultsarepresentedinSection5.Finally,themainsetofconclusionsaccordingtotheanalysisandresultsderivedfromtheperformanceofcontrollersispresentedinSection6.

2.Dynamicmatrixcontrol

TheDMCisakindofmodel-basedpredictivecontrol（Fig.1）.Thiscontrollerwasdevelopedtoimprovecontrolofoilrefinementprocesses[5].TheDMCandotherpredictivecontroltechniquessuchastheGeneralizedPredictiveControl[6]orSmithpredictor[6]algorithmsarebasedonpastandpresentinformationofcontrolledandmanipulatedvariablestopredictthefuturestateoftheprocess.

TheDMCisbasedonatimedomainmodel.Thismodelisutilizedtopredictthefuturebehavioroftheprocessinadefinedtimehorizon（Fig.2）.Basedonthispreceptthecontrolalgorithmprovidesawaytodefinetheprocessbehaviorinthetime,predictingthecontrolledvariablestrajectoryinfunctionofpreviouscontrolactionsandcurrentvaluesoftheprocess[7].Controlledbehaviorcanbeobtainedcalculatingthesuitablefuturecontrolactions.Toobtaintheprocessmodel,thesystemis

perturbedwithanunitarystepsignalasaninputdisturbance（Fig.3）.

Thismethodisthemostcommonandeasymeantoobtainthedynamicmatrixcoefficientsoftheprocess.Thecontroltechniqueincludesthefollowingsprocedures:

（a）ObtainingtheDynamicMatrixmodeloftheprocess.Inthisstage,astepsignalisappliedtotheinputoftheprocess.Themeasurementsobtainedwiththisactivityrepresenttheprocessbehavioraswellasthecoefficientsoftheprocessstateintime.Thisstepisperformedjustoncebeforetheoperationofthecontrolalgorithm

intheprocess.

（b）Determinationofdeviationsincontrolledvariables.Inthisstep,thedeviationbetweenthecontrolledvariablesoftheprocessandtheirrespectivesetpointsismeasured.

（c）Projectionoffuturestatesoftheprocess.Thefuturebehaviorofeachcontrolledvariableisdefinedinavector.Thisvectorisbasedonpreviouscontrolactionsandcurrentvaluesoftheprocess.

（d）Calculationofcontrolmovements.Controlmovementsareobtainedusingthefuturevectoroferrorandthedynamicmatrixoftheprocess.Theequationdevelopedtoobtainthecontrolmovementsisshownbelow:

whereArepresentsthedynamicmatrix,ATthetransposematrixofAXthevectoroffuturestatesoftheprocess,faweightingfactor,ItheimagematrixandDhefuturecontrolactions.FurtherdetailsaboutthisequationarefoundinRef.[5].

（e）Controlmovements’implementation.Inthisstepthefirstelementofthecontrolmovements’vectorisappliedtomanipulatedvariables.ADMCcontroller

allowsdesignerstheuseoftimedomaininformationtocreateaprocessmodel.Themathematicalmethodforpredictionmatchesthepredictedbehaviorandtheactualbehavioroftheprocesstopredictthenextstateoftheprocess.However,theprocessmodelisnotcontinuouslyupdatedbecausethisinvolvesrecalculationsthatcanleadtoanoverloadofprocessorsandperformancedegradation.Discrepanciesintherealbehavioroftheprocessandthepredictedstateareconsideredonlyinthecurrentcalculationofcontrolmovements.Thus,thecontrollerisadjustedcontinuouslybasedondeviationsofthepredictedandrealbehaviorwhilethemodelremainsstatic.

3.Fuzzylogiccontrol

Fuzzycontrolisusedwhentheprocessfollowssomegeneraloperatingcharacteristicandadetailedprocessunderstandingisunknownorprocessmodelbecomeoverlycomplex.Thecapabilitytoqualitativelycapturetheattributesofacontrolsystembasedonobservablephenomenaandthecapabilitytomodelthenonlinearitiesfortheprocessarethemainfeaturesoffuzzycontrol.TheabilityofFuzzyLogictocapturesystemdynamicsqualitativelyandexecutethisqualitativeschemainarealtimesituationisanattractivefeaturefortemperaturecontrolsystems[8].TheessentialpartoftheFLCisasetoflinguisticcontrolrulesrelatedtothedualconceptsoffuzzyimplicationandthecompositionalruleofinference[9].

Essentially,thefuzzycontrollerprovidesanalgorithmthatcanconvertthelinguisticcontrolstrategy,basedonexpertknowledge,intoanautomaticcontrolstrategy.Ingeneral,thebasicconfigurationofafuzzycontrollerhasfivemainmodulesasitisshowninFig.4.

Inthefirstmodule,aquantizationmoduleconvertstodiscretevaluesandnormalizestheuniverseofdiscourseof

variousmanipulatedvariables（Input）.Then,anumericalfuzzyconvertermapscrispdatatofuzzynumberscharacterizedbyafuzzysetandalinguisticlabel（Fuzzification）.Inthenextmodule,theinferenceengineappliesthecompositionalruleofinferencetotherulebaseinordertoderivefuzzyvaluesofthecontrolsignalfromtheinputfactsofthecontroller.Finally,asymbolic-numericalinterfaceknown

asdefuzzificationmoduleprovidesanumericalvalueofthecontrolsignalorincrementinthecontrolaction.Thisisintegratedbyafuzzy-numericalconverterandadequantizationmodule（output）.

ThusthenecessarystepstobuildafuzzycontrolsystemareRefs.[10,11]:

（a）inputandoutputvariablesrepresentationinlinguistictermswithinadiscourseuniverse;

（b）definitionofmembershipfunctionsthatwillconverttheprocessinputvariablestofuzzysets;

（c）knowledgebaseconfiguration;

（d）designoftheinferenceunitthatwillrelateinputdatatofuzzyrulesoftheknowledgebase;and

（e）designofthemodulethatwillconvertthefuzzycontrolactionsintophysicalcontrolactions.

4.Im