Proc Pe

Development of High-Performance Control and Parallel Processor Implementation for Induction

Motor Dr ive s

by

@Wali

u,

B.Eng.

Athesis submitt edtothe Schoo lofGraduate Studiesinpartial fulfillm entofthe

requireme ntsforthedegr eeof MuterofEngineeri ng

Facult y of Engine eringandAppliedScien ce Mernod al UniversityofNewfoundland

February 1992

St.John', Newfoundl and Canada

1\1+1

Nalionallibrary

erceoaca BibliotMquc nalioolalt' du""'"'"

Tho authorhasgrall1edan irrevocablenon- excicsive licenceallowingthe NationallJbrary ot Canadatoreproduce.soen.<fLSbibute⁰⁽sea copies of his/her thesisby anymeans and in anyform or format,makingulisUlesIsavailable to interes ted perso ns,

The authorretains ownersh ipofthecopyright inhiS/herthesis.Neitherthe thesis nor substantial extracts IromitmaybeprintedOf otherwise reproduced without hislherper.

mission.

L'auleur a accoroe unelicenceirrCvoc.1hlc ct non exclu sive permettantaIn Bibllc tneq uo natooaie duCanada dereorodulre.pr~te(, disfribuer ou vend redescopies de sathe se deqaeloue menlereatsous quelquaform e que ce soilpour mouredes cxom otaeosde ceuethe seiltadisposition desncrsonnc s interessecs.

L'ooleurconservetapropri cl cdlldroitd'OltJlcuf qui pro tegesameso.HitamesoIlides Cldr;)ils scbstannetsde ccso-ct no doivcnl OttO imprimes au euuemco treproduils s:n.son eutor eeuco.

ISBM 0- :)1 5- 73 31 2·8

Canada

Abstract

Implementationsofthemostpopular schemeof indirect field-orientedcontrol forACinduc tionmotor drives still sufferfrom difficultiesdueto the sensitiv ityof rotor parameters in decouplingtorqueandflux andcontrol algorithm complexity in meeti ngreal-timecontr ol requirements.

This thesis present s soluti onsto someoftheseproblems. A new param eter ada pt at ioncontrolscheme, nam edrotor flux orientationcontro l,is proposedand verifiedby digitalsimulations which show avery goodresultin dynamicallycce- reet ing the mismat ched value intheslipfrequencycalculat orthatusually occur whentemperature rises orflux satu rates. Thi sme thodis basedon theprinciple of feedbackcontrolwhich uses thedetected rotorfluxorient ation asafeedbackcontr ol signal.

For processingcomplex cont rol algorithms in real-Lime,a newparallel processor architecture, usingT800Transpu ters,is proposed.Implementa tionresult s show thatwit h this parallelprocessingcontrollercomplex cont rolalgorithms suchas field-oriented contr oland parameteradaptive controltoget her can be handled in very short processingtimesthusmeetingfast dynamic controldemands.

In addition,thedesignofindirectfield-oriented controlofAC induc tion motoris studied.Adesignmeth od forlinearcontrolof AC inductionmot or controlbased on theindirectfield-orientedcontrolscheme isproposed.Decouplingof torquecurrent and fluxcurre ntis achieved by applyingfeed-forward controllers.

A new controlscheme,namedextended indirectfield-orien tedcontrol , is pro- posedand verified bysimulatio n results whichshow anexcellentcontrol perfor- mancethatcan not be achieved by alinearcontrolscheme.

Furthermore . the inccrpcrat icnofIault tolerance(orTran sputerco m mulli"atinn linkfailu res and processorr.llilu r~is st udied and someslmula rlcntestshave 1'1...·1\

carried out.As a resultof thefaulttole ra ncemecha nismsinco rpo rate..l,coru rol systemrd ia bility is improved.

iii

Acknowledgement

Theauthorwishesto acknowledgethe financial support from the Schoolof Or.idu- aleStudies ,Dr. R.Venkatesan'sresear ch grantand the graduatesupport from the Facultyof Engineering and AppliedScienceof MemorialUniversityofNewfound.

landwhichmadethis workpossible.

Special thanks and sincreappreciationtoDr.R.Venkatesan ,foract ingas my researc h supervisor andproviding advice,guidance and encouragementthroughout the study.

Many thanksto Dr.J.E.Qualcoe andDr. C.R.Moloney,for many valuable discussionsand usefulsuggestions.

Ialsowish to thankDr. R.Venkatesa nand Dr. C. R. Moloney fortheir assist ance in preparingthis thesis.

Helpfromthe technicalstaffat ComputerAid EngineeringCenterfor using Iacilltlcsis gratefully appreciated.

Finally,the dedication isdueto mywife and my parents,(ortheirconstant encouragementand support.

iv

Contents

Abstract

Acknowledgement

List ofFigures

ListofTablel

Notat ion

1 Int roduction

2 Reviewof IndirectPleld-O zlen t ed Cont rol

or

Ind uct ion Motor 2.1 PrineipleaoC("directFleld-DeientedControl

2.2 Fteld-OrleetedControlUsingMicroprocesson 2.3 ParameterldenrificerionandAdept iveControl.. .

3 Modelingand Simula t ion ofIndirectField-Orient ed Control Sys- lem

3.1 Introduction... .. .. . 3.2 SystemAnalysisandModellinll.,

iv

ix

ltvi

"

21

2.

21 2'

3.2.1 AnalysisandModelling of Flux Currentand TorqueCurrent

ControlLoops . . 27

a.2.2 Analysis and Modelling ofSpeed Control Loop.. 32 3.2.3 ModellingofInd irect Field-Orient edControlof Voltage-Fed

Squirrel-CageInductionMotor. 32

:1.:1 Simulation Results of Indirect Field-OrientedControlof AC lnduc-

tionMotor. 37

:lA ExtendedIndirect Field-OrientedControlof AC InductionMotor ·12

.1.5 Discussion s 45

3.5.1 Condit ionsforaTrulyLinear Control System of ACInduc- tion MotorDrives..

3.5.2 Extension of Field-OrientedControl

3.6 Summary

4 RotorFluxOrien t ati on Contro l .1.1 Introducti on . .

·1.2 ThePrinciple of RotorFluxOrientationContro l.

·1.2. 1 Cont rol Strategy

·1.2.2 StabilityAnalysis . 4.2.3 Detectionof the DepartureAngle . 4.3 Realizationof FluxOrientationControl. ..

4 .

,16

46 72 72 7,1 74 76 77 79

·1.5 Discussion .

4..1 SimulationandResults. ... .. 82

8 '

·1.5.1 Advant ages ofthe RotorFlux OrientationCont rol . 85

-1.5.2 Limitati onsand Solutions 86

4.6 Comparisons..

5 Pa rall elPro ceuin gand Transpute n 5.1 Introducti on.... ... . . ..• . • . 5.2 Fundmental Principles orParallelPrecessing 5.3 Tra nsputers T800

ss

10 3 10:)

10·1 illS

6 Par all elProcelllin gCont r ollerIcr the IndirectField-Orien te dCon- trolorACIn duction Moto r

6.1 Int roduction .

II2 II:!

6.2 Two Tra nspu terBasedSimu lationofField-OrientedContro lSj-stem 11·\

6.2.1 Hardware Implementation

6.2.2 SOftareDes;sn .

6.2.3 ExecutionTimes

us

118 I'''' 6.3 '-fulti-TransputcrHued Simulation of Field-OrientedCont rol S)·5tcfTlI25

6.3. 1 Hard...areImp lementa t ion

6.3.2 Parallel ProcessingSoftwareDesisn... 128

6.3.3 Execution Times ... ...• . IJI

6.4 Fault Tolerance .... ... ... . I:I.S

6.4.1 Link Failure..•.. .. .... la6

6.4.2 TransputerFailure 1012

6.5 Effects orSam pling TimeOnControlSystemPerformance 1·16

7 Conclusio ns 150

8 Suggest ion .ferFutureWork 153

vii

Refe re nces

Bibliogr aphy

u,

161

AppendixA:CProgramSimulationorIndir ectPleld-Orle nted

Contr olo( AC Motor 16 '

Appe ndixB:CProgra mSimulationof IndirectField-O riente d

Controlof AC Motor 172

Appendi x C:Simulati onof Indir ectField-O riented Cont ro lof AC

Mot orusing TwoTransp ut ers 178

Appen di xD:Simulationof ParallelProcessing byFiveTra nspu t- ers (or Ind ire ctField-OrientedCor.tr olof ACMot or 188

App endixE: CProgr am SimulationforCh.n'. sehe m e 20,.

viii

List of Figures

1.1 Mo tioncontrolsystem -An interdiscipli narytechnology

2.1 Phasordiagramforindirect field-orientedcontrol

2.2 d-qreferenceframeequivalentcircuits .a) q-axia equivalentcircuit,

b) d-axis equivalentcircuit 10

2.3 Block diagramofmachinemodel withdecouplingcontrol 13 2.4 Positionservosystem with indirect field-orientedcontrol[35/. 1-1

2.5 8086·Microcomputercontrolof AC moto r 16

2.6 Hardwareconfiguratio nof experimentalsystem 18 2.7 Contigurat ionof speedregulator for AC motor 18 2.8 Transputer-basedim plementations ofinduction motorcontra!' (3)

V.type ,(b) l-typc 20

3.1 Fluxcurrent and torquecurr entclosed-loopcontrol.(a)Fluxcurrent

control.(b)Torquecurrent control. 30

3.2 Compound control.a) Fluxcurrent withfeed forwardcontroller, b) Torquecurrentcontrolwith feedforward controller 31 3.3 Blockdiagram ofspeed loop control end pole location.(a) Speed

control loop, (b)Pole location 33

3.4 Time domainsimulationmodelofindirect field-oriented controlsystem 36 3.5 Systemresponse (or ste pchang~in the reference speed. a)Rotor

speed, b)Motor torque... ... .. ••.• •. .. . 18 3.6 System responsefor stepchange inthereference speed,a.)Torque

current, b)Flux current . .. . .•. • • . . . .. . .. . . ..• -19 3.7 Systemresponseforstep changein the reference speed,a.)Stator

current in alphaaxis,b) Statorcurrent in beta axis. • .•... SO 3.8 Systemresponseloratepchangein thereferencespeed,a] Stator

voltagein alphaaxis, b)Statorvoltage in beta axis. . .. .51 3.9 Rotor8uxresponseforstepcha nge ill the referencespeed. 52 3.10System response for stepchangeinthelo&d,e)Rotor speed,b)

Motortorque .. . . •....• .. •. •... •... . . .53 3.11System response for stepchangeln the load, a.) Torquecurrent,b)

F!uxcurrent.... .. .. .. .. . ... ... •. . ... M 3.12 System response forst ep changein the load,a.) Statorcurrent in

alphaaxis,b)Stator currentinbeta axis.• •... ... .. .. 55 3.13Systemresponseforstep ch&ngeintheload, a.) Statorvoltagein

~phaaxis,b) StatervoltJ.&e inbeta&Xis. • .. .. •. .. .•.•. 56 3.14Rotorftuxresponse(or stepchangein the load. • • . ...•. ... 57 3.1.5Syste m responsefor speedreversal. a) Rotorspeed,b)Motortorq ue 58 3.16Systemresponse forspeedreversal. a) Torquecurrent,b)Fluxcurrent59 3.17 Sy!tcm response for speedreversal. a)Stator currentin alphaaxis,

b) St a torcurre nt in betaaxis ... . .•.... ... ... . ...• 60

3.18 System response forspeedreversal. a) Statorvoltageinalphaaxis, b) Stat orvoltageinbeta axis

3.19 Rotor fluxresponse Cor speedreversal.. . . .•..

3.20 System responsefor stepchangeinthereference speed.a)Rotor speed,b) Moto rtorque.. ..

62

63

6·\

us 3.21Syst emresponsefor stepchangeinthe referencespeed, a)Torque

current,b) Fluxcurrent

3.22Syste mresponse for stepchange in thereference speed,a) Stator curre ntinalphaaxis, b)Stat orcurrent inbetaaxis 3.23System response forstepcha nge in the referen ce speed,a)Stator

volt ageinalpha.axis, b) Stato rvolt ageinbetaaxis ... 66 3.24 Rotor fluxresponsefor stepcha.ngeinthe reference speed. 67 3.25Systemresponse with extended field-orien ted control, a)Rotor speed,

b) Moto rtorque, . ... ..

3.26 System response withextendedfield-orientedcontrol,a) Torquecur- rent ,b)Flux curr ent

3.27Syste m response with exten dedfield-orie nte d contro l,a) Stator cur- rent inalphae-ts, b) Rotorflux..

3.28 System response wit hfield-oriented contr ol .

68

10 11 4.1 Flux andtorque current vectors ind-qaxes.a)Flux orienta t ion

alignedtod-axis, b) Fluxorientationdeparted fromd axis 7.'i 4.2 Stato r-oriente d model forestimat ingtorq ue 79 4.3 Par am et er ada.ptationcontro lforindirectfield-orie ntedcontrolof

ACinduc tion motor

xi

80

H f1lJxorienu tionCOlJlrollcr... .•. ..•. .c.5Motor rotor m istancevarjations.• .•.. .

H Response o(indirect field-orientedcontrol,(a)Rotor speed,(b)Rotor speed.. ..

4.7 Responseorindirectfield-orientedcontrol, (a)Mo(ortorque,(b)

~fotortorque•. . .." ... •.

81 iW

91

92 U Responseofindirectfield-oriented control,fa)Torque current,(b)

To"l.ue current •.. • ..• • .•.• •.• • . .•. .... .. . .. 93 .e.g Retporaeorindired field·oriented control,(a)Flux CUrmlt,(b)nux

current••.... ...•.... . . ••..•.••. . . • . 4.10Heepcnseorindirectfield-orientedtontrol,(a)Rotor flux,(bJRotor

flux.

4.11Response ofindirectfield-oriented control withfluxorientationcon- trol,(a)Rotorspeed,(b)Motortorque...

4.12ikspoose orindired field-orieeted centre!withfluxorientationton- 95

trol,a)Torquecurrent,(b)Fluxcurrent•. . .. . .•. ... 97 U3RcspolUeofindirectfield-orem edcontrolwithfluxorientationcon-

trol,(a)RotorFlu, (b)Correctedrctcrresbtence.•.•..••. 98 4.14Response ofindirectfid d-orien ledeeetrcl,(a)Identified stator resis-

tance,(b) Rotor speed.. .•.•.•• •.•...•.. 99 USReepcnseoCindirectfield·orientedccnercl,(a)Motortorque,(b)

Fluxcurrent•.•.• • • , , • 100

4.16 SimulationmultsbyChan a)Rotorspeedresponse,b)Estl- mated valueof rotor resistance ...•.

xii

101

4.17 Simulation resultsbyChan ,a)Torque,b)Torquecurrent 102

5.1 Distri butedand sharedresourcesystems 105

5.2 INMOS T800 archite cture 10!)

6.1 Parallel processinghardware configuration for simulationof ACmo-

tor control. 116

6.2 Transputerbased simulation ofindirectfield-or ientedcontrolsystem ofACinductionmotor.a) Realmotorcontrol,b)Real-time emula-

tion of motorcontrol . . Iii

6.3 Software structurefor simulationofACmotorcontrolsystem 119 6.4 Five Transputerbased parallelprocessing architecture fOTreal-time

simulationof AC motorcontrol 126

6.5 Parallelprocessingprogram struct urefor simulationofAC motor

control. .. . 129

6.6 Controltask dependencyin AC motorcontrol .. 132 6.7 Reconfigurab lecommunicat ion channelsin Transputernetwork. 137 6.8 Recovery of multi-communicationchannel failures. 139

6.9 Reliablecommunicat iontimes.. . 140

6.10 Communicationtime with auxiliarychannel 14\

6.11 Parallelprocessingprogram structur eincor porated withprocessor faulttolerance. . .

6.12Transputer 1failure. 6.13Transpute r2 failure. 6.14Transputer3 failure.

xiii

11.1 144 144 1<5

6.15Transputer4failure, IU6 Transputer5 failure

6.17Multi-votingfortransputerfailure . 6.18 Motortorquecurrentresponse .

xiv

145 146 147 149

List of Tables

3.1 Paramet ersofcontroller . :18

3.2 Parameters of Cont rollerandSlip Frequ ency . .. . .. . .. .... j.j

6.1 Executiontim es in single Transputer

. .

12J

6.2 Exeeuti on times incon venti onalsingleprocenor 12·1

6.3 Execu t iontimesin five transpute rs

...

¹³³

6.4 Executiontimesinmultiprocessors(microsecond s). 13-1

i" ,io :

u"o:

lIo,Uo:

Notation

direct-andquadrature -axisstator currentsinthe synchronousreference frame.

direct-andquadrature-axis rotor currentsin thesyn chronou s reference frame.

direct- and quadrature-axis stator voltagesinthe synchronousreference frame.

direct- and quadrat ure.axia rotor voltagesinthe synchronousreference frame.

alp ha-bet a axescurrentin thestat io naryre ference frame

alpha'andbeta-axis statorcurrents inthe stationar y refere n ce frame.

alpha-betaaxeevoltage!in thestationalreferenceframe

alpha- andbeta-axisstatorvoltages in the stationaryreferenceframe .

[~(k).19(k) : discretedire ct-an dquadra tu re-axisstator cur rent s inthe synchronous referenceframe

U..(k), U,(k) :discretedir ect- andquadratu re-axisstatorvoltages inthe synchrono us reference frame

w. : statorangular frequency

xvi

w, : rotorangular frequency

W,f: slipa.ngula r frequency

1/J~: rotorflux

>Pd. : rotor flux componentindaxis

1/Jq. : rotorflux co mponentinqaxis

tPoa: stator fluxinQ-fJaxis

W_a: stator flux component in a axis

>Po: stator flux component in13axis

tPa.: rotor fluxcomponent inQaxis

.,pOr: rotor flux component in13axis

fJ, : statorelect rica langle

8. : rotor angle

xvii

0.1: slip an gle

suffix denoting the refe rencevalue

11., rotor resistance

R., statorresistan c e

L . ,

rotor inducta nce

L. ,

statorinducta nce

M: mutualinductan ce

D, viscousfrictioncoefficient

J, wtalinertia

T, motortorque

TL : load tor que

T:

eetimatedterq u e

xviii

n,.: nu mberof pole pairs

N: nu mberof pole

p: differential operato r

Laplace op erator

mmf:

T~:

1( , :

rnegnetcrno uveforce

u:

R.

controllerproporti on al gain

ftuxorient a t ioncon troller proport ionalgain

con trollerintegraltime const ant

controller input

y: co ntroller output

xix

To: samp linglime

rpm: rcvolution minuteper

Nm: Newlonmeter

6: depart ure angle

Chapter 1 Introduction

Electricmachineshave been a workhorseof indus tryformanyyears. Thethree basicelec tr ic machines .DC,ind uctio n, andsynchronous.have serve dind u strial needsfor neArlya.century [I). Althoughtheinduc tionmachine issuperior tothe DCmachinewithrespectto site ,weight ,rotorinertia, efficiency, maximum !lpt't.·d.

reliability, cost. ete.,becauseof theforme r',highlynon-linear dyn a.mic str ucture withstrong dynamic interactions,itrequires morecomplexcontr olschemesthlR a sepa.rately excited DC machine .AsforDC muhines,torquecont rolisachieved bycont rol1ingthe motorcurrent. However,incontra.sttoDCmachines,inAC machines ,both the phaseanglean d themodulusofthe currenthavelobecon- trolled.orin otberwords,thecurrentvectorhas to be controlled.Furtherm ore, in DCmachines,the orienta.tionof the field flux Andarmaturemmfisfixedbythe commu t ato r andthe brus hes,while inACmachinesthefieldflux andthe spatial angleofthearmaturern.m.Lrequireexternal cont rol.In theabse n ceof this con- trol,the spatialanglesbetween thevariousfieldsinACmachinesvary withthe load andyield an unwant ed oscilla ti ng dynamicresponse [2).Withfield-oriented controlof anACmachine, the torq ue-end flux-producingchar acteristiC!are similar

tothose of a separately excitedDC machin eand thesyst em will adapt to a.nyload distur-bancesand/orreferencevaluevariationsasfast as a DC machine.

In the pastsuchcontrol techniqueswouldhave not been possiblebecauseof the complexha rdware andsoftwarerequ iredto solvethe complexcontrol problem . Field-o rientedcontrol techniquesincorpo rating fastmicrop rocessorshavemadepos- sible the applicationof AC motordrive sfor highnerfcrmenceapplicationswhere traditionallyonlyDC motordriveswereapplied.

Fie ld-orientedcontrol techn iquesarenow being accepted.almostunive rsallyfo r high. perform a ncecontrolof AC machines.Suchcontrol methodsweredeveloped inGe r manyinthe early1970's.Blaschke[3) developedthe direct method of field- orientedcont rolandHasse

14J

inventedtheindirectmethod.Atpresent,the field- oriente dcontroltechniques havebeen implementedinterms of fourtypes(rotor flux-orientedcontrol,magnetizing-flux-orientedcontrol,rolor-orientedcont rol, and staterflux-orienledcontrol). Themostpopula rmethodis the rotor flux-oriented controlbecauseof its easy implementat ion,robustnessand fastdynamicresponse {5/.Inrotor Bux-oriented control, thereare two schemes: direct and indirectfield- orientedcontrol.Asithas easy implementation,practicaluse and robustness, indire ctfield-oriented control has been givenextensiveattent io nin recentyears (6117J(81191·

Ho wever.the syet-rmresponseofindirect field-orientedcontro lis limitedbythe accuracy wit h whichthe rotortime constant is known.Unfor tunately,the roto r time constan tvarieswith bothairgapflux andtemperaturesothat itis difficult to keepthe controllerin tune.Means fo r estimatingand adjust ingtherot or time constantin theslip frequencycalculatoris current lyan areaofintense activity[6].

Theresearche rs are a.ttemptingto find a suitableparameter-adapt ationscheme for optimumdecoupllngtorque and fluxunder variousoperatingconditions.

Onthe basisof field-orientedcontrol,moderncontrol techniqueshave beenem- ployedfor obtainingahigherdynamic control performance.Moderncont roltheory iscert.alnly more difficultto apply for ACdrives. However,fortheACdriveswith field-orientedcontrolused in the innercore,there isno differenceinthe dynamics between theAC and DC drives.Therefore,itisexpectedthat moderncontrol theorywill be appliedto ACdrives withfield-orientedcontrolintheinnerloop {2]. Applyingmoderncontroltheory toAC motor drivecontrolremainsa challenge because of the largetimecrit ica lcomputationworkrequirement.Increased compu- tation a l speedis,ofcourse, the primary benefit o( parallelprocessing.Thisallows the systemto be controlledmorequickly and gives the choiceof added complexity inthe cont ro lalgorithm.

Fau lttolerance canberealizedin a parallelprocessing systembyorganizing comp utationa l operations in a distribute dsense, so that anoperationfailureresults in performance degradationratherthana completebreakdownofthe con troller,

Today,motion controlis an areaof technologythat embraces many diversedisci- plines,such as electricalmachines,powersemiconductordevices,converter circuits , dedicatedhardwaresignal electrcnice, controltheory,microcomputers,LSI/ VLSI circuits,sophisticatedcomputer-aiddesi gntechniques andsystem reliabili tytech- uiquea(10][F tg.Ll], Each of the compo nentdisciplinesis undergoinganevolution- aryprocess,andiscontributingtothe totaladvancementof machinedrivecontrol technology.

Thea.imof thisthesisis to investigatethecharac teristics of indirectfield-

• ^{- - - '== -_. - S - '"::' "-=' - -}

Figure 1.1:Mot ion controlsyste m-An Inter disclplinerytechnology (10]

orientedcontro lofan AC inductionmotorandtodevelopnewschemes forsolving problemsin indited field-oriented.controlof AC inductionmotor.

Due to thecomplexity of algorithms(Ofindirect field-ceiented control, most rcceutly developed. pu a.:neler adaptat ion cont rolschemes are restrictedtomotor

..teadyoperation(or correct ingthe mismatch ed.valuein the slip calcula tor.

In thisthesis,a new easily implemented rotorparame teradaptat ion control is expected tocorrectthemismatched valueinbothmotorsteadyoperation end transient opera tion.

Consideringthecomplexit y ofindirectfield-oriented cont rolandmodem con- trolin ACdrives,anew VLSI produ ct,knownasaTransputer, ischosen.uthe parallelprocessing controller forthereal-tim e controlrequirement.Simulationsof indirect field-orientedcontrolon multi-transp uternetwork' havebeen carriedout toexaminetheparallelprocessingeffectiveness.The corresponding executio ntime s

for controlalgorithms and some faulttolerancetestresultsforimproving control systemreliabilityhavebeenobtained.

A linearcontrolmodel of an ACinduction motorbasedon indirectfield- oriented controlhavebeen studiedforthepurposeofdesignofPIorPIDcontrollers byusing theclassical methodssuch asroot-locusapproach or frequencydomainscheme.

This thesisisorganizedineight chapters.In Chapter2,the principle ofin- direct field-oriented controlisdescribed. An overview ofthestate-or-the-art of microprocessor-basedimplernentationeoffield-orientedcontrolofAC inductionmo- tor andthe developmentofrotorparameteradapta tioncontrolforindirectfleld- oriented controlofAC inductionmotor ispresented.

InCha pter 3, in order to investiga tethecharacterist icsof indirectfield-oriented contr ol,ananalyticalmodel of indi rectfield-oriented cont rolof

A e

indu ctionmotors hasbeensetup and aseriesof simulations havebeencarriedout.Onthe basis of field-orient ed control,thecondit ionsforobt aining atrulylinearcontrolmodelor the ACinductionmotorare studied. Finally,a newproposalfor achieving excellent propert iesfrom field -oriented controlis madeand thecorrespondingsimulation hllll beencarried outfor verifying this proposal.

InCh apter 4,anewparameter adapta tion scheme forindirect field-oriented controlof ACinductionmotors isdevelopedandsimulationresults are presented, Finally,resultsarediscussed .

In Chapter5,a briefintroducti onto theprinciples ofparallelprocessing and the transputerispresented.

InChap ter6,a parallel processing cont rollerusingfiveTranspute rsforindirect field-oriented controlofACinduction motoris implementedandcorresponding

sim ulat ions erecarried out.Executiontimesa.ndsomefault tolerancetestsare examined.

Chapt er7 presentsconclusions.

TheIMtChapter proposes suggestions forluture study.

Chapter 2

Review of Indirect Field-Oriented Control of Induction Motor

2.1 Prin ciples of Indir ect Field-O riented Co n- tro l [35]

Inrecent yea rs,the problemof obtain ingfast torque responsefrom A.C .machi nes hasreceived considerable attention. However , unlike theD.C. machines,thevolt- age,current,torqueandspeedofanA.C.motorareall interde pen dent ,which resultsin ahighly non-line ardynami c structur e.Thepossibilit y ofobta ininghigh performance from the A.C.motoris verydesirab le. Acontr olstrategy forachiev- ingthis is termedvectororfield.orientedcont rol. Thiswasoriginallyderived from Blackshc; extendedand genera lized earlier workdoneby Hasse;similarideaswere expressed byseveral oth erresearchers:Abbondan ti , Nabeeetal.,Plunkett,Bose andLeonhard[jI].

In principle,thestrat egy involvesthe transformat ionof the machinedynami cs int othose o£apseudo.DCmachine equivalent in which atorqueandfieldcomponent o£machinecurr entare obtained. These currentcom ponent s are indepen dent,i£they are measuredinthe so-calledfidd co·ordinales.In thisway,the AC motormay

becontrolled in the samemannera.stheDC motor,thusleadingtoanimproved transientrepense.Fis ure2.1 explainsthe indirectfield-orientedcontrol withthe help ofa ph&$Or diat;ram.Thea-Daxesarefixed on thestator whilethed.q axes rotate atsynchronousant;ularvelocityw.asshown.At any instant,theqelectrieal ..xisisatangularpositionD.withrespect tothe/3axis.TheangleD.isgivenby thesumof rotor angularpositionD.andslip angularposition901,where8.

=

^to,I,

O.

=

w.l.and8.,

=

Walt. Therotor flux1/J~.consisting of theairgap fl.uxand therotor leakage flux,isalignedtothedaxis asshown.Therefore.for decoupling control,thesta tornux componentof current

i

Jandthe torque componentofcurrent itare tobe alignedto thedandqaxes,respectively.

We canwritethefollowing equationsfront rotat ingframed-qequivalentcircuits a.sshown inFig.2.2:

cl:; + ~if"

^{+(w, -w.J.p..=0}

d~. +

R.i. _(w,-W.)W'"

= o.

t\t;ain,

From equation(2.3) and(2.4),

i9. = *,p,. - ~i,

i"· = t"'''· - ~i,,.

(2.1) (2.2)

(2.3) (2.4)

(2.5)

(2.6)

Fig ufe 2.1:PhllSOfdillgfilmferindirectficJ(l,ofi(~ntcd^control[3.5J

10

Figure2.2:D·Qre ferenceframeequivalent circuits .iLlQaxisequivalentcircuit, b)Daxis equivalentcircuit[351

II

The rotorcurrent fromequations (2.1)and(2.2) canbeeliminatedh~'subsri- lutingequations(2.5)and(2.6) as

d: _{t -} ~R,.i,

⁺

~"·9.

⁺

W.,l,.~J.

⁼⁰

d: t - ~R.i.r

⁺

I:-'. ".- 10"",.

:=0, where10.1

=

^{W. - 10••}

Fordecouplingcontrol,itisdesirab lethat

¢'9.=~""O

!P.r.=

i.

=constant

~=O.

(2.7)

Substitutin~thenrst one condition,equations (2.1)end(2.8)canbesirnplifit-ua.~

(2.9)

12.101

A!a.in,thetorque a.s a func t ionofrotor !lu xandstatorcu rrent ca nderived as follows: Thesta wrflux linkagerelationscanbewrittenfrom Fig.2.2as:

(2. 1I) (2.12)

Substitutingequations(2.5) and(2.6) inequatio ns(2. 11)and(2.12), the following eq uati ons arcobt a ined:

(2.13)

12

(2.14) Thetorqueequation as a function of stato rcur ren tandfluxe sis

(2.15 )

Equations(2.13) and(2.14) canbesubstitutedinequati on (2.15)to eliminate the statorfluxes, therefore

(2.16 )

Substitu ting"-'v< "" 0 and!/J~.⁼

$"

the torqueexpression is:

(2.17) or,using equ at ion(2.10),

(2.18) Theequat ionsabove,togetherwiththemecha nical equation(2.19)describe the machi ne modelindecouplingcontrolas showninFig. 2.3.

(2.19)

The developedtorqueT.respondsinstantan eously with current;9'but has delayed respon se due toi"in equati on(2.10).Theanalogyof themodel with a separately excite d DCmachineisobvious.

Forimplementation ofindirectfield-oriente dcontrol,itis necessaryto take equat ions(2.9)and(2.18) intoconsideration.Asanexam ple,Fig, 2.4showsa posit ion servosystem using the indi rectmethodoffield-oriented cont rol.The flux

Figur e 2,:.1:Blockdiagrrnn ofll1a(hineuunlelwithd"<"'.)I]pll11.1.( contro l1:1.')1

com po nen t of currentijforIIll'll•..;irl'lll'olol' t1IlX1:7,i.~d"lel'lnilled frolll"'1I1i1Lion (2,10)andismaintaine dnlll~1.1 111lu-re.Till 'lo r 'llIl'rumpnncutof("l lrn' l l l.

i;

is derived from thesp eed control loo pas usual. Till''idvaluouf slip"';1isn~ l;it1'd to cu rren ti~byequat ion (:!,!IJ, Til"slip-nllglevectors .o;i" I):1Mil l<:II,sO;"whh-h determine thedesiredelectrica laxis with respe ct to thero tor-me chnrticulaxis,MP gene ra tedina Ieed.fcrwnrcl11I.1I111,'rfromtil<'

n';,

sigllallhrollgh avea,itcouuu-r, and a ROM- base d.sin./co,.gen era tor.Therotor-poshionvector sco,sO,alld,~ill().

are obtainedfro m ang!...encoderandare i111,le<1withLll'~slipvectors to obtainth,~

cosO.andsinO.signals asfollows:

iJinO;

=

^.sin(O,

+

^O:J)

=

siIlO,cosO:,+cosO,sinO:, f2,21)

II

Figure 2..1:PoshionM:rnlsystem with Indirectfield-orient ed control[351

15

2.2 Field-Orient edControl UsingMic ro p roces- sors

Control ofthe ACmachin eisconsiderab le more complex thancontrol of the DC machine. Thecom plexity increasesforhigherperformancerequirements. TIlt' reas on is that ACmachine dynamicsaremorecomplex.intricatesignalprocessing isrequired,andthe varia blefreq uen cy supplyrequires more complexcont rcl t.han docstheDC converte r.

Recent achievements in LSIandVLSllechno!ogyledto practic a lutiliaationof themicro processo r.Themicroprocessormake sitpossible torealize veryccmpli- caredcont rolstrategythatisdifficultorevenimpossible torealizewith conventio nal at,a lagcircuitry.Whenthemicro processoris used in variablespeed motor drives, itgiveshighaccuracysurmou ntingthe nonlince rltiesinvolvedin motor cha racter- istics, powerconversionequipment and sensingdevices, andprovid ing againstthe driftsof temperatu reandsou rcevoltage.

The control syste mshownin Fig. 2.5 has beenimplemen ted on an Intel 8086 microcompute rtowhichasignalprocessorNEe7720 was attachedas a highspeed arit hm etic unit.This makesit possibl eto comput e thenuxmodel, thetrans forrn a- tion and the innercontrolloops with 125 IJS samplin gtimewhich is adequateeven forhighspeedcontrol[111.

[t is obvious, of course,that with fasterdynami cs ofthe sys temimple ment e- tion of optimal control andadaptivecontrolof multi-variabl esystem s willbecome more and more complex. Inthis respect,theuniprocessorbasedmot orcontrol is limit edin offeringhigher computing speed. In recent years, mu lt iprocessor bas ed motorcont rolsystemshavereceived considerable att ention.Increasedcomput e-

Ilotl1lC""fI'#..

.r ; ... ...

^{d. ...}

I-"-r ---I

Figure 2.5:S086·Microcomputercontrolof AC motor [11\

16

tionalspeedis•of course, theprimarybenefitofp.rallel processing.Thisellc....·s Jester systems tobecontrolled and gives the controlchoiceofadded complexityill the controlals:orithm.Ea.syexpansion,within. uniformhard wareendsoftwareen- vironment ,isanot her featureofconcurrent controlsystembecauseit is possibleto add moreprocessors as required.Parallelproceui ngalsoolfo,!!lI a closerrelationship between the inherentparallelismexpressed atthedesign stageand the hardware implementa tion.

Oneof thetypicalapplicationsusingmultiprocessorapplicationsfor field-ceicntcd controlwasdeveloped in1985by FumioHareshima et a\(121.Tn theirproposa.l, athree microprocessorbaaedAC motor controllerwa.:sused a.:s shownin Fig.2.6.

With the parallelcontr oller, two single-chip microcomputersInle1·8031 (12Mlh) areassigned for36/ 26to26/3~signalconversions,respect ively.These signalcon- versions areinitiat ed by theinterrupt signaleveryIms.One16-bitmicroprocC"Or Inte l·8086(5 MH,,)executes both thevectorcontroland speedcontrolprograms.

The executiontime forone cycleofthe control program isabout2.2m".This multiprocessorconfigurationreducesthework l ~of the mainprocessor Intel·8086 and permitstherealizationof moresophisticatedcontrol suchasadaptive cont rol, opt imalcontrol,andso on.Anotheradvanta gefromthisconfigurationisthatitis veryflexibleto be usedinvariouscontrol types,suchasV-type controlandI-type control.

Inthe sameyear,KenjiKuhoetal(13) proposed&fully digitizedACmotor controllerby usingmultiprocessorsystem.Theconfigurationofthe proposed con- troller isshowninFig.2.7.The motordrivesystemisccmpceedof a powersupply, PWMinverter,induct ionmotor,and speed controller.

Figur e 2.6:Hardware configurat ionofexperimentalsystem[12]

Fig ure2.7:Configu rati(lnofspeed reg ulatorfor AC motor [13]

18

Based on aspeed reference and detectedsi~nalsofmotorcurrentandspeed.

PWM gate pulsesarcgenerated accordingto the controlprocessing. TheP\VM inverteroperatesfollowin~thesegate pulsestoreguliltemotorspeed.Fourmicro- processors &J'e usedinthespeed controllerasindicated by the dotted line inFig.

2.7.Twomicroprocessors areassignedto controlprocessing, andothertwo arc usedfor gate pulse generationanddetection processing,respectively.

The executiontimesofSMC 1,SMC2, SMC3andSMC"were 750,700, 380 and 600ps,respectively113].

WiththedevelopmentofVLSI.the32-bittransputer systems havebeen used as promisingparallelprocessingsystemsforreal-lime digitalcontrolapplications, such as DCmotor or ACmotor cont rol.

Thefirst att empt in using TransputerforAC motor controlwas madebyD.1.

Jonesand P.J .Fleming(141.AnotherTrensput er-besedhighperformanceAC mo- tor controlayetem,Walldevelopedby G.M. Asher etal(15].Intheirproposal, a.parallel processing controllerhas beenexploitedsothatverycomplexcontrol all;orithmscouldbeimplementedinreal-t ime toimprovetheACmolar control perfor mance. Three vector controltypes are presented andimplementedby auni- form Transputer-basedparallel processingsystem.The threevectorcontroltypes are V-type,current-controlled.V-typeandl-rype . Implementa tions using paeal- lei processingsyst em are illustratedin Fig.2.8{a.)and 2.8(b).TheTransputer T1handlesspeed andline currentacquisition,speed control,slipand inverter fre- quencycalculation, open-loopstatordyna mic compensationandcurrentacquisition fermonitoring.Speed issampled every 5m.!lwhilecurrenti,sampled a.l250la. Consequently,we canconclude thatthe functionsof ageneralizeddrive controller

20

.-

^~,

.

, . ' .

::::::=:>_ (pQt)co"'""

_ Dli'>t't S"J",,11

C&l

{bl

Figure 2.8:Treneputer-besedimplementations ofinductionmotorcontrol. (a) V·

type(b)l-type

Itsl

:![

exhibita.nat uralparallelism. In particular, a three-way parallelismexits between control,supervisor/communicationand backgroundcomputing funct ions.

2.3 P a r a m e t er I d entificat ion and Adap t iv e Co n - tro l

Theindirectfield-orientedcontrolbasically involvesa feed-forwardcontrolscheme inwhich the motor slip freque ncy is calculatedfrom the estimated(commanded) stator currentand anestimate of therotor timeconstant.Thedesired slipfrequency is added tothe measured rotor speed toform the frequencycommandwhichis fed to the inverter.When properlyadjusted,theindirectfield-orientedcontrolmethod provides torque responsewhich canexceed the equivalentDCmotor[I].However, the speedofresponse is limited by the accuracywith which the rotortimeconstant varieswit h bothairgapfluxandtemperat ure so thatitisdifficultto keep the controller intune.Meansfor estima tingandadjusti ng therotortimeconstant in the slip frequencycalculatoriscurrently anareaof intense activity.

L.J.Garces[17]and M.Koyama et301.125]proposedthecorrectingfunctionfor progressivelyestimating therotor time constantintransient state.Thedrawback of this metho d is that alongcorrectingtime isneeded forobtainingthe act ual rotortime constant.

K.Oh nishiet 301.(26]and Y.VedaetaL(27]developedthemodel referenceedep- tive controlsystem.Thesetwoschemes involve much morecomputatio ns for ceti- matingtheadaptivegain.The estimatedresultsare affectedbyloaddisturbances.

L. C.ZalendT.A.Lipo [19]proposedamethodwhich used anextende dKalman Fillerapproach to estimatetherotor time constant.Thedraw backof this scheme

22

isthat the actualcomp uta.tionforestimatin gtherot ortim e constant tookabout 30 seconds of CPU time onahie:hprocessingspeedMC68000-bMed microprocessor.

T.Mwuo~dT.A.Lipo 118)developed anothermethodin which&prescribed negative sequence curren tperturb ationsisn aJ.wasinjectedinto the AC motor and then theccerespcndingnegat ivesequencevoltagewu detectedfor obtainine: the rotorparameters .This scheme requheeadditionalhardwareendmayinducea

~trOfigsecondha rmonictorque pulsationdue to the interac tion of the positionand negativerotating componentsofmmf.17}.

The recentlyreported resultshavestillbeenconcentrated onpar amet eridentifi - cationormodel referen ceadaptivecontrol.Asanexample, C,C.Cha n(7) proposed an effective met hodror rotorresistanc eidentification.Thismethod isbased on the proper selection ofcoordina teaxes:thedaxi,oftherotating fram ei,set to be coincident withthe statorcurrentvector.Adirect estimateofrotor resistance can be derived fromthereferenceframe.The advanta geofthisapproach lies inits simplicityand accuracyforroto rresistance identification.However,correctresults from thisidentificationmet hod can onlybeobtainedwhenmote-is workingin steady stateand aloadshouldhe applied.

M.B. Zhou andW.Qu124) developedascheme where the rotortime con,ta nt is deri ved from the vect or diagramofT·Itypeequivalence .This schemeismore or less similar to thator

c.

^{C.Chan whichcanbeappliedonlywhen motoris working}

in steadystate.

R.D.LorenzandD.B. Lawson(8J proposed a simp lified approach of model reference adapt ive cont rolsystemtocontinuousen-linetuning forfield-oriented contr ol. Due toits sim plicity,the prima.ryadva.ntageofthe proposedtechnique is

theease of implement ationofthe adaptivecontrolmethodology given the_This method could be aJrectedbyapp lied load(91 and takeslong timefor controleee- vergency,

Basedon the ahove reviewofmicroprocessor-basedcontrollerimplemcntalion and parameteradapt at ioncontrol forfield-oriented cont rol.it canheconcluded thatcomplex field-orientedcontro lor AC inductionmotorcanbe realized with the help of microprocessors.tnaddition, va.riou! parameteradap tation control!cheme' have beendeveloped an d applied tosolvethe problemof parameter sensitivityin indirect field-oriented control.

However,microprocessor-basedcontrollersreportedtodateinthefield-oriented controlofAC induct ion motor havenot beendeveloped adequately tomeethigher performancerequirements. Meanwhile, the developedparameter adaptationcontrol schemescanonly be appliedwhen motorworksin studystale.

Witb an intent to solve theseproblem" some solutions areproposed inthe subsequentchapters.

Chapter 3

Modeling and Simulation of Indirect Field-Oriented Control System

3. 1 Int r oduct io n

In Chapter2,recent developmentsin the theoryof indirectfield-oriented co nt rol ofACinductionmotor and its applications have beenreviewed.Inthischapter, modeling andsimulationofindirect field-oriented controlofinductionmotorwith squirrel-cagestructureare studied . The modelof indirectfield-oriented control ofACinduction motoris used forthe study of paramete r adaptatio ncontroland parallel processingcontrollerwhichare intended to satisfythe requirementeof high- performancecontrol ofAC motor .

The dynamicsofthe AC machinedrive controlsystem isextremelycomplex and the subjectisreceiving wide attentionin recent years.The complexityarises because ofthe nonli nearityof AC motordynamics.Forthisreason, when " control st ra tegy is developed,it isthe usualpracticeto simulatethedrivesystemonthe digitalcomp ut erandstudytheperfo rmance in detail beforeproceedin g to buil dan

24

25 experiment.The electricaldynamies of an inductionmotor can bere pr esented by a fourth-order nonlinear differentialequation which may be either inastationary referenceframe (Stanley equation),orinasynchronouslyrotat ingreference frame {2}.The advantageofthelatt er model is that the steady statesinusoidallyvary ing parametersappea rasDCquantities.The establishedmodelwithameansof feed- backcontrolcanbe simulatedon computerand control parameters. suchasPIor PID, canbefinetuned until the desiredperformanceis obtaine d.

ModelsofAC inductionmotorand field-orientedcontrolofAC ind uctionmotor have been developedand studiedbyresearchers{7].[29],(31]. 120]. and[301.A common method ofmodelingis tha ttheind irect field-orientedcontro lof AC in- duction motorcan belinea rized abouta steadystate operatingpointusingsmall signal pert urbationprinciple.

Intbis chapter,thedesign andsimulationfor indirectfield-orientedcontrol system arestudied.A linea rcontrolsystem is achievedby applyingfeed-forward controlscheme so thatdesig n ofPIorPID controllersdoesno t depen d011system operat ingpoint.Inaddition.a new controlschemehas been developedto imp rove cont rolperformance.

3 .2 System Analysis and Modelling

TheAC drive withfield-or ientedcontrolisa well-knownalternativetotheDC drive.Toimplementthe field-orientedcontrol,thesufficien t condi tions fe r the quicktorquecont rol or an inductionmotor are summa rizedasfollows (12]:

ie secondaryfluxis controUedtobe consta.nt,which is equivale nttokeeping iuta nt.

26

2. The AC powersupplyfre quencyw.i!determinedby:

(3.1)

.1.The prtme rycurrentsiqandiJmustbead justedinstantaneo uslyan dprecisely accordingtothereference values

i;

andij .

Consequently,the arb it rary quick torq uecont rolwitho utunnec essarytransientsis realizedbyad justi ngthetorque currenti,while keeping the secon daryfluxconsta nt

(ori<l,ccnatanj.]. However,thefollowin g proble ms preve n ting th esatisfactio nofthe

foregoingconditionsshouldbetakeninto conside ration:

I. The secondary flu x level may bechanged bythe paramet e rvariationin the induction motorsuchas thechangeof roto r resistance due tothe tempera tu re risinginthewindings.

2.The slipfreque ncyW,tdetermin edbyequ ation (3.1)mayhave some error when the time constantvalueofthesecon darycircuitusedin th econtroll er is mism atched with theactual value.This errorcauses thecoordin ationof the secondary flux todepart fromthe d-axis.

3. Theins ta ntaneousand precise adjustmen t ofthe primary current necessitates some control means, suchasthe currentfeedback control loops.

Toachieve instant aneousand preciseadjustment of the primarycurrent ,design for theinn ercurr entcontrolloopsplays an impor tantrole.

As usual, forthe co nsiderat ion ofsy stemsta bility,theinnerloop (curr ent control loop) designshould be carriedout firs t.Afterthe in n ercont r olloop shavebeen

27

designedsatisfacto rily,the designofoutercontrolloop (speed controlloop)CAnbe ca.rried ou t.

3.2 .1 Analysis an d ModellingofFluxCu rrentandTorque CurrentControlLoops

Toau lysethe indirectfiel d-orien t edcont ro lforACinductionmotor,some neces- sa.ry aJsumptionsare livenbelow [38J:

ASSU M PT I ONS, 1.Symmetricalarmatures 2.Cons ta ntlI.irgll.p

3. Sinusoidalind uction repart iti on 4.Satu ra tion.hysteresisand ed dyeffectsn~ligible

5.Indirectfield-orientedcontrolled(thismeans that thedirectionor therotor fluxis alignedto the synchronousrot atingaxis0').

BesedODthe aboveusum ptions,th edyna.micequa tionsor asquirrel-cageinduction motor inthed-qreferenceframeare given17):

[

" 1 [R . ⁺ p L . -w ,L . p M

II, _ w. R.+pL, w. M

o -

pM 0 R..+pL.

o

wo/M 0 wJLp

(3.2)

Accord ingtothed-q equivale ntcircu ih,shown inFig. 2.2, the following equationsart' derived:

(3.3)

28

(3.4) Bysolvingequations(3.1)·(3.4),the followingequationsareobtained:

I'd==R,id

+ f ( I:

^{T.p)id -w.pL.i , (3.5)}

. . M2 ill .

u,== R. l,+pL.m,

+

w.(L;(I

+

T.p))+pL.W,11l (3.6)

Uol

=

^R.(l

+

T.p\

:P~~)id

^{- w.pL.i, (3.7)}

u,==R.(l

+

pT.p)i,+L,( \

:P~;)w.iJ'

^(3,8)

In ordertocont rolthe flux currenti"andtorque current

i,.

fee d back control is usedfavorably. The purpose

or

^{feedbac kcontrolisto}improve the dynamic per formance ofcontrolledcurrentssuch thatthe torquecur rentiffollows theref- erence valuesas quicklyas possible.whitethe flux currentillkeepsconstant.From theequations (3.7)- (3.8),itis dear that even though thefield-orientedcontrol scheme linearizes and decouplestherelationsbetweenrotorflux and electromag- netictorque, the coupledrelationsbetweend-exiscurrentandq-a x ia curre ntor d-q curre ntsand rotorspeedtvr st illexist,which present sanon-linear andmulti- variablestruct ur eto thecurrentcontrol.

Inord erto app roachthedesig n of currentcontrollersbyusing classicalmethod suchasroot-locus or frequency do m ainana lysis, itis proposed here to usein te rac- tive cancellationtechniquesfordeco uplingtheint er actions betweeni,andiiiorif andi"andtv,.

It isnoted thatiCtheinteractivevaria b lestv,i,andw,iilasshow nin equations (3.7)-(3.8) Meconsideredas external distu rbances to the controlotidandif,the lI.ux

:.!9

current andtorquecurrentcontrolcanbe configuredas a linearcontrol structure and, therefore,thetransferfunctioncan be usedto analyzethe closed-lo op cont rol.

asshowninFig.3.1,where:

Gd(.J)'=pT,TrR,.J2

+ /T~:r:

^f,R,).J

⁺

^{R, (3.9)}

G. (8)'=R.O :PT,.J) (3.10)

Dds)=pL, (3.11)

D2(8)=L,(\: p;'·ss). (3. 12)

Itis obvious thatthe disturbanceslIJ, iqandw,idareeasilycomputablefrom the measuredquantitiesid,i₉andw. ;therefore,an effectivecontrolapproachCor cancelling thiskind of disturbances is touse thefeed-forward controltechnique.

Feed-forwardcontrolmeansthecontrolof undesirableeffects ofmeasurabledis- turbenceby approximatelycompensatingforit before it materializes.This feature is advantageous,because, inausual feedback controlsystem,thecor rective act ion st a rtsonly after theoutput has been affected. Thistechnique achievesthe objcc- tive of instantaneouscompensationcontrolofexternaldisturbancesi"w.and i9w, andminimi zes the transient error. Whenfeedback controliscombi ned with the feed-forwardcontrol,it is expected that the feedback controlcompensatesfor any im perfectionsinthe functioning ofthe feedbackcontroland providescorrections forunmeasurable disturbance, and meanwhilethefeed-forwardcontrol minimize! thetransienterrorcaused bymeasurabledisturbancessuch as thew, i9andw.iJ.

The blockdiagram ofthiscompoundcontrolisshown in Fig.3.2 , where:

(3.13)

30

<aJ

(b)

Figure3.1:Fluxcurrentand torquecurrentclosed-loopcon t ro l.(a.)Flux current control.(b)Torquecurrent control.

"

ij

(a )

(b)

Figure 3.2:Compo undControl. a)Fluxcurrentwithfeedforwardcontrolle r,Ii) Torquecurrent controlwithfeedforward controller

32

(3.141 Itcanbe seen thatthedisturbancestV.illonillandw,i.. oni,l J.Obecancelled completely bythefeed-forward controllersG/(s )andGl(s),whilethe PIcontrollers effectthe currentloop control.

3.2.2 Analysi s an d Modellingof Speed ControlLoop Bytaking advantage ofthe feed-forwardcontrol, the disturbancescan beca ncelled sothat alinear cont rolstructure ofAC motorspeed control canbe simplifiedas shown in Fig.3.3.(a ),

where:

(3.15) Todesig nthespeedloop,severalmethodsarc available.The root-loc usbased ap- proachispopular. In general,according todesignspecificat ions, dominantcomplex poles(-a

+

hj)and(-a- hj)with afar awayreal pole(-c)can he specified as sho wn in Fig.3.3{b).Theparam eters ofPIcontrollers can beeasily obtainedfrom theloca tionsof dominant poleswhich provide a desirablecontrol performance.

3.2.3 Mod elling of Indirect Field-Oriented Control of Voltage- FedSquirrel- Cag eInductionMotor

Forthe purposeof analysis ofindirect field-orient edcontrol ofAC ind uctio nmotor, it is essent ialtorun simulatio ns

o r

an overall AC inductionmotorcont rolsystem model instead ofa simplified controlmodel.Asa.result,thesimu lat ion resultsdo not only givecontrolvariablesinsynchronous referencefra me, but also givethe cont rolva riables in stationaryreferenceframe whichprovides analysisof alternate

33

(, )

(b)

Figure3.3: Block <!iagralll of slwc-dloo p contrcl illl,lpoleIOCil.t io n.(a)Spf!<:d ((," lr ol loop,(bl Polelocat io n

variablessueh as statorcurrents,voltagesandrotor flux.

In the limula.tx:lD set up,the voltage-fedPWM invert er is definedas an ideal sinusoidal powersupply.The gain oftheinvert erisassu medasKo

=

30.The currentfeed.b«k coefficient ADd thespeedfeedbackcoefficient aredefinedI.lIK, andK••respectively,

Asusual,themotor model isrepresentedin the two-axissystem,oopreference frame. An N-poleinductionmotorwithashort-circuitedrotorcircuit is character- izcdby thefollowingcqua~ionll:

[

.. ] [R.+

_{Up 0}

pL.

_R,+⁰_pL.

o

⁼ pM w.M

o

-IL,.M pM

o ][;. ]

pM io

w.L. i12•

~+ pL. ip.

(3.16)

!....~""T-TL -D~

npJt np

T=iMnp(i,i...-i..i,.)

~., ..=

^1.1..- R.i12

~J/;

^{...=-R..i...-W.¢II.}

f,rPtJ.=-R,itJ.+w.!/J...

~rPtJ =l.ltJ -R.ill'

where:

(3.l i) (3.18) (3.19)

(3.2<1)

(3.21)

(3.22)

(3.23)

Based on equati ons(3.16)-(3.23),the timedomainsimulationmodel for indirect field-orientedcontrolofACinductionmotorcanbe constructediL!Ishown inFig.

3.4.TheVAX8530digita lcompute r is chosen for st udyingthe dynam icsofAC motor control.Underthe indirect field-oriented control, therequiredstator voltages u.,andUQare obtained bythefollowing coordinatetransformat ions:

(3.241

ThreepheeevariablesinACmotorcan beeasilytransformed Irom twoaxes variables by the following transformati ons:

[: :_«,] =

[ ~l ~ _{- t -If} ] [ ::] .

^{(3.25 )}

Thestatorcurrentioandioin thea-fJreference framecan bc transformed tofield-orientedunidirect ion quantitie si,Jandi,with the followingcoordinate transformations:

and

[ ;

_iQ

0] ₌ [' 0 01[ ;']

0

~ - * : :

[i,]

=

[co~D,

'in O , ][;.].

I, -sm9. coslJ. 117

(3.26)

(3.27)

Thefrequency and phase of the voltage componentsUoandUQin thestation- aryreferenceframeforest ablishingtherequired rotorflux andstato r currentare controlled by theslipfrequencyW,Iwhich canbe calculated from equat ion (3.1).

The controllers applied intheindirect field-oriented control ofACinduction motorinclude speed and current feedback PI controllersas wella.sflux andtorque current feed-forwardcontrollers.The transfer function that approximate s the Action

b~d,IIlllt h "1,1"

,I:":"r."':.~!"!"_~. luuhllllitl

---r- - --+-'-' , '.

r,lIIfwuli..

FigureJ..l:Time-,!omai llmodelorindi rect field-oriented control system 3S

:17

ofafeed back continuousPIcontrollerisgiven by:

(3.28) whereTisthetime consta nt,K,is thecont rollergain.

A differen ceequatio nof thistype of controlleris characterizedby:

Un

=

^Yn_1+K,(tn -tn_I)+To*K,*t... (J.2!l) whereKi

=

~;Toisthe samp lingtime.

All the feed back quant ities arecomp aredwit hthe corr espondi ngreferencequanti- ties. andthe error quantityeformstheinput tothe controll er.Thefeed-forward contr oller ofthefluxcurre nt isexpressedby the followingequation:

(3.30)

Thefeed-forwardcont roller oftorquecurre nt isexpressedby thefollowing equatio n:

(3.31 ) Adifferen ceequat ionofthetorquecurrentfeed-forwardcont roller iseharaeteeized by:

Yn=(T~ .Yn_1

+

To

*

L.

*

p.T.(t"-t,,_d )! (1

o+

T.). (3.32) Fordifferent loadsimulat ions, theload-sett ing unit isused tochange theapplied load tothe motor.

3.3 Simulation R esults of Indirect F ield-Orie nt ed Control of AC Induction Motor

ACprogram for the simula t ionoftime domainmodelWMdesigned forsimul ation purposes (see append ixA-I),The parametersof theACmotor given in[7J are

38

Table:}.I:Parametersolcon/roller

type"icontrolJer P I

Speedcontroller 8 0.003 Fluxcurrentcontr oller O.S 0.01 Torqul' currentconlroller 0.6 0.01

Tab le3.1:Parametersof cont roller

listed in appendixA·2.UsingtheMatLab to simulate thelinea rized andsimpli- lied ACmot orcontrol system asshown inFig. 3.3{a),theproporti onal gain and the integral timeconstantof PIcontrollerscan be determined.Thepar ameters

or

currentcontroJlen andspeed controller are listedin Table3.1. For studyi ngthe dynamic properties ofindirectfield-oriented.cont rolofAC mot or,a aeries ofsim- ulations havebeencarriedout.By viewing thesimulationresults,itcan beseen that, when the compoundcont rol of feed-back andreed forwiU'dcont rollers were applied,thefluxcurrentwas controlled betterthan thatoConly feedback control.

In these!simulations, thedynamic responseoCthe control syste mfor theCollowing disturba.nceswerestudied :

I.Stepchangein the referen ce speed

2.Stepchangeintheload

3. Speed reversal.

Step Change in theRefer e n ce Speed

Thestepresponseof acontro lsystem is usuallyappliedto evaluatethecontrol systemperformance in itsrespon se time, overshoot,transient orsteadyerroreAnd sta bilit y.

Nine motor variableswere chosenin thissimulationfor stud yingthedynamic performance .They arethe d -q coordina tecom ponents, suchasthe torque current and flux current,a-/1coord ina tecom po nents,such asstato rcurrents andvolt agc9, andmotor st ate variable s, such aselect romagnetictorq ue, rotorflux and speed.The behaviourof the field coordinatesand the cont rolvariablesare plottedin Figs. :t 5

·3.9.

Fig.3.5(30)shows therot orspeed response withstarting speedof1040 rpm.

Att:::23,the speed referencewas changed tothe speedof 512rpm. Hcan he seenthat the rotorspeedisaccelerated or decelerat ed sm oot hly by contr ollingthe motortorque whichis proportional to torque current.Fig.3.5(b) shows the motor torque response.At the beginning,it isobviousthat the torquewas not linearly proportionalto the torquecurrent, because theflux current wasnotconsta nt due to dynamictransience as shown in Fig.3.6(b ). At I:::0.63,a2Nmload waa applied ,the correspondingtorqueandtorque current ,as illust ratedin Fig.3.5(b) and 3.6(30),show the linear response tothe input loadbecausetheconstantnuxhas been setup at tbistime. Thisproves that underthe field-orientedcontrolifthe fluxis constant, the machinetorqueislinearly proportional to thetorqu e current. In thissimulation,as thefeed-forwardcontroller was oot applied, the fluxcurrent

40 was disturbedby changesinspeed orload,asindi cated inFig.3.6(b ).

Figs.3.7·3.9showthecorresponding responses of statorcurre nts,statorvolt- agesandrotor fluxwhen inputspeed andload werechanged.In ordertosatisfy the conditions of field-orientedcontrol.the correspondingstator current orvoltages changedtheirmagnitude and phase, and therotorflux keptconst ant.

StepCha:Igein theLoad

The ability to withstanddistu rbancesin a controlsystem isanother importa nt feature.Aste p change intheload is consideredas a typicalexternaldisturbance . A highperforma ncecontrolsystemshould haveafu tdynamic responsein adjusting its cont rolvariab lessotha t thesystemoutputsaffectedbythedistu rba nceswill recover totheiroriginalstatusas soonaspossible.The dropped apti tude ofsyste m outp utsuchasrotor speed and itsrecovering timearetheimportantperformance specifications.

Figs.3.10 ·3.14showthe control system response when astep loadwas applied.

Inthissimulatio n,the feed-forwardcontrollerwasapplied.In Fig.3.l 0(b),a2N m step loadwas appliedtothemoto ratt

=

0.6". The correspondingmotor torque has emergedto bal ance theload(orkeepingspeedconstant asshowninFig. 3.II (a).

When theloadwaschangedfrom 2Nm to therated valueof8Nm att

=

25,the motortorquequicklyfollowed the load appliedto attaina new equilibriumpoint . Therotor speed, as shownin Fig.3.10(a),droppe dalittl e att=0.6" and!

=

2" , andthen raised quicklyto its originalvalue. In thistime, duetotheIeed-Iorward control.the Buxcurrentremainedccnst ent withou t beingdisturbed by changesin load.

II

Figs. 3.12 -3.14 show the corres pondi ngresponsesofstatorcurrents, stetcr volt ages androtor fluxwheninputloadwerechanged.

SpeedReversal

With the field-oriented contro l.ACmotorcanbeoperatedinfourquadrAnt!like theDC motor.

Figs.3.15 ·3.19show the system response forspeedreversal. Accordingto inputcommand,therotor speedreversed all=23from positi veratedspeed to negative tilled speeda.sshownin Fig. 3.15(&).Thestillercurrents and voltegce, asshowninFigs. 3.17•3.19,changedtheirphase,frequencyand amplitudewhen speed re-vcued.A nega tivecurre nt,asshown inFig. 3.16(&),occurredtoproduce negativetorque,a.sshowninFig. 3.15(b), forspeed revcrsel.Thenuxcurrent and rotorDux arekeptconstantwithspeedreversal.Inthis simulation,thefeefl forwardcontrolwas&Iso applied,therefore,theflux currentremainedcons~AlIt withoutbein&disturbedbyspeed.reversal.

Figs. 3.5 -3.19demonstratethefield-orientedcontrol mechanism ofthelin- ea rizedcontrolofdecouplcdfluxandtorque currents in anACinductio n motor.

The control performance ofAC induct ionmotor isfound tobelike thatofDC motor.Themotorspeedwasbeencontrolled to regain theoriginalspeedaftera smalldipfor ashort durati onwhen load wu applied.Italsoshowsthat anAC driveofthis kindpermitsoperation infourquadran ts andallows the motortobe loadedcontinuouslywith ratedtorque at standstill.Becauseof the slowresponse of the rotorflux,thereisasmalldipin magnit udeduringspeedreversal whichis regained subsequently.

Forthepurposeofcomparisonbetweenfeedbackcont rolandcompo undcontrol, Fig.3.20 · 3.24presentthecorrespondingsimulation results.

fig.3.20(11.)showsthe rotor speed responsea.ta sta rtingspeedof512rpm.At L

=

2" ,the speed reference was changed torated speedof 1710rpm.Fig. 3.20(b) showsthemotortorque responsewhen a 2Nmloadwasappliedatt=0.6".The correspondingtorqueand torque current,as illustrate din Fig. 3.20(b)and 3.21(11.), showthe linear response tothe inputlced, In thissimulation,a.sthe feed-forward controllerwas applied,theflux current remained constant withoutbeing disturbed bychanges inspeed orload,as indicat edinFig. a.21(b).

Figs. 3.22·3.24showthe correspondingchangesof stato rcurrents,stat or voltages androtor flux when inputspeedand load werein changes.

It canbeseen that withfeed-forwardcontroller,the /lux currentis controlled more constan t,as shownin Fig.3.21(h),thanthatwit houtIeed-Iorwsrdcontroller, as shownin Fig.3.6(b).

3.4 Extended Indirect Fi eld-Oriented Control of AC Induction Motor

Theconceptof series-shunt excited cont roliswell-known inDC motor control techniques.Theseries-shuntexcitedcontroltakes advantagesofcoupledtorque andflux currentcontrolto enhancethe ability ofwithst andingdisturbances andto increasethe maximumtorque.Implement ation ofthe series-shuntexcite dcontrolin DC motorismade by connecting fluxwindingandtorquecurrent circuit toletflux currentcontainsomeor torque current.TheDC motorcontrol theoryhas proved that a series-shunt excited controlsystem canprovidehigher torqueresponse th an

'"

separ atelyexcitedcontrol.Thisperforma nceshows a strong rc bus rncss and fi\.'ll contro lresponse.

InACmo tor control, thenaturallycoupledtorqueand flux currentcouldhe easily realizedas a series-shuntexcited controlmodel. ifapropervalue of slipis found. Howe ver, thecoupled torq ue andfluxcurrentcharacteristics makes the analysisofcontrolperforma nceverycomplex becauseofits non-linea rand multi.

variablestruct ure.

To exploitthe possibilityofseries-shuntexcited cont rolin AC motor,first let the ACmotorbe controlledin field-orientedcoordinates, which presentsa separately excited typeand thenthe slipis red uce dto a.smallervaluethanthat forcompletely decouplingtorque andflux.In thiscase,we canexpecttoobt aina series-shunt excited controlin AC motor like in DC motor.

The proposedseries-shuntexcited controlin AC motorhasbeensimulatedina digitalcompu terto verifythevalidityandfeasibility ofthis method.The simulation programs andmoto rpara.meters[ 13] are listedin Appendix B-t andAppendixB·

2,respecti vely.Theparamete rsofPIcontrollerandslip frequency arelisted in Table3.2.Simulation resultsarcillustr a tedin Figs.3.25·3.27,whichindicatean excellent robustness andfMtresponsewith almost no overshoot.The featuresof thiscontrolperformance canbe summa.rizedasfollows:

1.There is almost no overshoot duringmotortransientspeed response (sec Fig.

3.25(.».

2.Thetransient speedresponse timeis equaltothat ofsepa rately excited cont rol of AC motor(compared Fig.3.25(a)wit hFig.3.IO(a)).

d SI' F T bl Ja ea.;..:) pnmnuiersoIeon/ro erfl1I 'p ^{requf 1lcy}

lype P I Normal Sma ll

Speed controller 8 0,003 Flux curre ntcontroller 0.8 0.01 Torque currentcontroller 0.6 0.01

Slip Frequency lOa/.' IO:{

Table 3.2: Parametersof ControllerandSlip Frequency

3.Thereisalmost no drop or increaseinspeed(Fig. 3.25(a))whenratedload was appliedorremoved(20Nm, see Fig. 3.25(b)),

'1.Rotorflux currentisnot constantwhen controlsystem isworkingintransient stat us (Fig.3.26(b ),whichis similarto the series-shuntexcitedDC motor control).

5. Thestator current is still sinusoidal which can be providedby powerelectr onic devices(Fig, 3.27(a» .

Forcomparison,Fig.3.28 showsthe linear controlresponse byregular field- oriented controlmethod.The motorisstarted at a speed of 680rpmand changed tothe ratedspeedof1710rpmatt""2".Itcan be seen that,with 20Nmrated loadapplied att "" 0.7s and removedatt"" 3s, the roto rspeed dropped much loweror higherthan thatofseries-shunt excitedcontrolscheme.

3.5 Discussions

3.5.1 Condition sfor a TrulyLinear Control SystemofAC InductionMotor Drive s

Generally, anAC inductionmotorcan be controlled linearlybyapplying theind irect field-orientedcontrolscheme .However ,it isafacttha~the indirectIiold-orionted control is onlya necessary condition in realizing alinearACinductionmotorcont rol sys t em.Inother words,the indirectfield-oriente d cont rol onlyprovides thesolutio n thatifthe fiuxcurrent isconstantandtherotornuxorienta tionisaligned tothe d-axis in synchronous reference frame,alinearcont rol of ACmotor torquecanhe realizedby controllingthetorque curren t .

Butthe field-orientedcontrol scheme docsnotensurethat the JInxandtorque current canhe controlled linearlyandseparately.

Infact ,underindirect field-orientedcontrol. the torque current isproducednot onlybythevoltage inthe q-axlsbutalsobythe voltageinthed-axis andby the rotor speed.The8ux currentis alsoproducedbybothvoltagesinthedandtheq axes as wellas by rotorspeed.

The coupled relationsbetweenflux currentand torque currentareexpressed in equations(3.5)•(3.6).Accordingto thestudy inthischapter, a linearcontrol modelfor an

Ae

inductionmotorunder theindirect field-orientedcontrolcanbe obtainedif thefeed-forward controller is applied, which resultsindccouplingthe torque currentfrom fluxcurrent as shownin Fig.3.2.Therefore,it can besaid thatthe necessaryand sufficientconditionsfor obtaininga linearcontrolsystem of AC inductionmotor is toemploy bothfield-orientedcontrol andfeed-forward control.