Observer-Based Direct Field Orientation Analysis and Comparison of Alternative Methods

8/4/2019 Observer-Based Direct Field Orientation Analysis and Comparison of Alternative Methods

1/9

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 30 , NO. 4, JULY / AUGUST 1994 945

Observer-Based Direct Field Orientation:Analysis and Comparison of

Alternative MethodsPa t r i ck L. Jansen , Student Member, IEEE, Robert D. L o r e n z , Senior Member, IEEE, an dD o n a l d W. Novotny , Fellow, EEE

Abstract-This paper focuses on methods of achieving directfield orientation (DFO) of induction machines based on closed-loop, stator, and rotor flux observers which are well suited toboth zero and very high-speed operation. Both observer topolo-gies are dominated by a current model at zero and low speeds ,and a voltage model at high speeds. Application of such rotorand stator flux observers to both stator and rotor direct fieldorientation is presented, including experimental resu lts for threedifferent methods.The influence which flux regulation has on parameter sensi-tivity of the complete DFO system is analyzed. A rotor-flux-reg-ulated and -oriented system is shown to be sensitive to leakageinductance under high slip (Le., field weakened) operation. Botha stator-flux-regulated and -oriented system and a stator-flux-regulated, rotor-flux-oriented system are show n to have reducedparameter sensitivity at hig h speeds.Unlike stator flux orientation using simple voltage integrationstator flux models, excellent zero a nd low-speed operation of a nobserver-based stator-flux-oriented system is demonstrated.

I . INT RO DUCT IO NCHIEVING high-quality torque and flux control inA pplications requiring b oth zero and very high-speedoperation is difficult with existing approaches to inductionmachine field orientation. Indirect field orientation (IFO)utilizing a shaft encoder or resolver is the most commonmeans, especially when zero-speed operation is required[l]. However, controller detuning can result in a substan-tial deterioration of performance. IF0 is particularly pa-rameter sensitive in large machines, high-efficiency ma-chines, and when field weakening a t high speeds [2].Direct field orientation (DFO) based upon estimationof either the rotor o r stator flux from the terminal voltageand current is an alternative approach that is very attrac-tive for high-speed, field-weakened operation. At highspeeds, the voltage model provides an accu rate st ator flux

Paper IPCSD 94-22, approved by the Industrial Drives Committee ofthe IEEE Industry Applications Society for presentation at the 1993 IASAnnual Meeting. Manuscript released for publication March 24 , 1994.This work was supported by the Wisconsin Electric Machines and PowerElectronics Consortium (WEMPEC) of the University of Wisconsin-Madison.P. L. Jansen is with the Morrison Knudsen Corporation, Boise, ID83705.

R. D. Loren2 and D. W. Novotny are with the Department of Electri-cal and Computer Engineering, University of Wisconsin-Madison,Madison, WI 53706.IEEE Log Number 9402641.

estimate because the machine back EMF dominates themeasured terminal voltage. However, at low speeds, thestator IR drop becomes significant, causing the accuracyof the flux estimate to be sensitive to th e estimated s tatorresistance. Due to that sensitivity and to inherent signalintegration problems at low excitation frequencies, DFOsystems based solely upon the voltage model a re generallynot capable of achieving high dynamic performance at lowand zero speeds [3]-[7].A DFO approach was proposed in [8]-[10] that effec-tively combines the best accuracy attributes of IF0 an dvoltage model DFO by utilizing a specific topology ofclosed-loop rotor flux observer. That topology of closed-loop observer provides a smooth, deterministic transitionbetween flux estimates produced by two different rotorflux models. A prior approach to provide a sm ooth transi-tion between models was developed by Takahashi andNoguchi. They combined two stator flux models via asimple first-order lag-summing network [ll]. In addition,rather than performing DFO, their stator flux estimatewas used to select appropriate voltage vectors for directtorque control.While [SI focused on the general design and accuracyanalysis of both open- and closed-loop rotor flux ob-servers, and [9] proposed the closed-loop rotor flux ob-server for DFO in wide speed range applications, thispaper analyzes and compares different approaches toobserver-based DFO. Both stator and rotor flux DFOsystems based upon stator and rotor closed-loop fluxobservers are analytically and experimentally evaluated.Particular emphasis is placed on parameter sensitivity inthe field-weakening region. Unlike in [lo], measured r otorposition is assumed to be available.

11. ROTOR LUX F O SYSTEMSThe rotor and stator flux production in an inductionmachine can be modeled by the block diagram in Fig. 1,where the superscript e denotes an arbitrarily alignedsynchronous frame. Cross coupling in the form of a rotor

flux-slip frequency product is clearly evident. (Note th atthis is the same cross coupling that is recognized as backEM F when modeled in the stationary frame.)As is well known, orientation to the rotor flux (i.e.,using rotor flux as the d-axis reference frame, e + f )0094-9994/94$04.00 0 1994 IEEE


2/9

946 IEEE TRANSACTIONSON INDUSTRY APPLICATIONS, VOL. 30, NO. 4, JULY AUGUST 1994. ...............................................................................i General Svnchronous Frame Induction Machine Model i

........................

................................................................................. Fig. 2. A tuned rotor-flux-regulated rotor-flux-oriented (RFR-RFO )system assuming ideal current regulation. CC,c is a flux regulator, e.g., aFig. 1. Current input, flux output model of the induction machine in an simple PI controller.)arbitrarily aligned synchronous frame.

such that A$ = 0, eliminates the cross coupling betweeni$ an d A$. Note, however, that A$ f 0. With goodcurrent regulation, a correctly tuned rotor-flux-orientedsystem as shown in Fig. 2 exhibits decoupled control oftorque and rotor flux.Rath er than directly calculating the d-axis current com-mand in a feedforward appro ach, as is common in indirectfield orientation (IFO), a flux regulator CrC s typicallyused in DFO systems. One purpose of the regulator is tocompensate fo r potential ph ase lags in the current regula-tor. Current phase lags result in increased d-axis currentwhich attempts to build the flux beyond the commandedlevel. The flux regulator counteracts by decreasing thed-axis current command, effectively advancing the currentangle. As will be shown, the flux regulator can also reduceparam eter sensitivity.By calculating the torque current command i$*, usingthe estimated rotor flux (feedback) rather than the com-manded rotor flux, dynamic torque control is virtuallyindependent of the flux regulator bandwidth.Unfortunately, the rotor flux is not directly measurable,and thus significant erro rs can exist in the estimated roto rflux feedback signal, and hence also in the actual operat-ing flux and the actual developed torque.

111.A CLOSED-LOOPOTORFLUX BSERVERThe closed-loop rotor flux observer pro posed in [8]-[10]is illustrated in Fig. 3in block diagram form using com-plex vector variables. Unlike indirect field orientationwhere flux is estimated in the synchronous frame, thisapproa ch is implemented primarily in the stationary frame.The closed-loop observer is formed from two open-looprotor flux observers which are referred to as the current

and voltage models. The current model utilizes measuredsfatur current and rotor position to produce a flux esti-mate, while the voltage model relies on the measuredstator voltage and current. T he smooth transition betweencurrent and voltage model flux estimates is governed bythe closed-loop eigenvalues (observer bandwidth) deter-

..................r . _ _ _ _ _ _ _ _ _ _ _ . _ _ _ _ _

: rre M ...................-.-.u.-nt.odel_____: .. sa h1 . IFig. 3. Closed-loop rotor flux observer based upon the current andvoltage rotor flux models in complex vector notation.

ministically set by gains K , an d K , of the linear con-troller.Note that the current model is best implemented in therotor frame (i.e., physical rotor, not rotor flux frame) andthus, as depicted in Fig. 3, requires transformations be-tween the stationary and the rotor frames using the mea-sured rotor position. When implemented in the stationaryframe, the current model requires measured rotor velocityand has velocity-dependent cross coupling that can lead t onumerical instabilities at higher velocities when digitallyimplemented. (Note that this is the same cross couplingpresent in Fig. 1. ) Transformation to the rotor framecompletely eliminates th e undesirable cross coupling, andperm its the u se of rotor position which, unlike velocity, isgenerally directly available and sufficiently accurate.Under field-oriented operation, the slip is generallysmall, and r otor velocity correspo nds closely to excitationfrequency. This causes the observer estim ate to transitionbetween the current and voltage models very nearly inproportion to the rotor speed. The flux estimation fre-quency response function (FRF) from [9] and replotted inFig. 4relates the estimate d flux to the actu al rotor flux forrated slip operation using gains set to achieve 10 Hzclosed-loop eigenvalues. The flux observer sensitivity toparameter estimates corresponds to that of the currentmodel at velocities below the observer bandwidth, and tothe voltage m odel at velocities above the b andwidth.


3/9

JANSEN ef al.: OBSERVER-BASED DIRECT FIELD ORIENTATION 947

15IO

h$ 533 0

-10-1 50.01 0.1 1 2 3 4Rotor Velocity (P.u.)

Fig. 4. Closed-loop rotor flux estimation FRF, A% r / A , d r , illustratingparameter sensitivity of the closed-loop rotor flux o server at rated slipoperation (10 Hz eigenvalues (0.17 PA.), (A3) in Appendix).

IV . FLUX EGULATIONND CURRENTMODELPARAMETERENSITIVITYThe current model was shown in [9] to be the samemodel used for indirect field orientation (IFO). A D F Osystem based on the current model would thus be ex-pected to have the same sensitivity to the rotor timeconstant and the magnetizing inductance. However, fromthe current model flux estimation FR F phase plot in Fig.

5, one might incorrectly conclude that the parametersensitivity of a flux-regulated DFO system would be re-duced at high operating slip frequencies o, correspond-ing to the field-weakened operation). Despite the fluxestimation phase accuracy improvement, errors in therotor time constant will now be shown to cause errors viaan incorrect operating slip, similar to I F 0 slip frequencycalculation error.Fig. 6(a) shows the slip frequency to which the DFOsystem will converge, as a function of the effective DFOcommanded slip frequency w * ~ ,.e.,

for errors in rotor resistance estimation. As expected, theoperating slip frequency is the same as the commandedslip frequency of an I F 0 system. The corresponding oper-ating flux level is plotted in Fig. 6(b). Also plotted in Fig.6(b) is the estimated flux level, which fo r roto r resistanceerrors agrees with the commanded level, i.e, incorrectorientation due to flux estimate phase errors results influx level errors equivalent to the estimation error. Inother words, both flux phase and magnitude errors resultin equivalent operating errors in the DFO system.Given this inherent property, the flux regulator has noinfluence with respect to rotor resistance errors.Fig. 7(a) and (b) plot the corresponding operating slipfrequency and flux level for errors in magnetizing induc-tance estimation. Unlike the case with rotor resistanceerrors, the estimated flux level [also plotted in Fig. 7(b)l isnot in agreement with the commanded level, i.e., fluxphase and magnitude errors do not result in equivalentoperating errors in the D FO system.However, the DFO flux regulator will drive the esti-mated magnitude to the commanded level. The resulting

1.6A-rA, 1.43.z 1.22

1.0

0.8

. . . . . . . . . . . . . . ._..................' 2 = 1.5rr

n.... ....,....L,= 1.5 L,, ..............

0 1 2 3 4d w s r a t e d Ws /W B rated

Fig. 5 . Current model FRF, % d , C / A i d r , illustrating the parametersensitivity as a function of ope rating slip frequency [(A l) in Appendix)].

W s ratedFig. 6. Resulting (a) slip frequency and (b) flux of a DFO system basedupon the current model as a function of effective commanded slipfrequency for rotor resistance parameter errors.

. . . . .

.. .

..........

. . . . . . . . .

0 1 2 3 4w;/wsrated

1.6. 1.42* 1.2

1.00.80.6

0 1 2 3 4w f / w rated

Fig. 7. Resulting (a) slip frequency and (b) flux of a D FO system basedupon the current model as a function of effective commanded slipfrequency for magnetizing inductance parameter errors.

operating point will then correspond t o the level indicatedby the F R F magnitude plot in Fig. 5 .If the flux feedback regulator is replaced with thefeedforward approach used in I F 0 such that1

il;, = ~ - [ l trp1hT (2 )th e DFO system will converge on the same operatingpoint as IFO.Errors in the estimated rotor leakage inductance influ-ence the system in the same manner as the rotor resis-tance, although at a greatly reduced level. Thus, a major

L m


4/9

948

0.9

IEEE TRANSACTIONSON INDUSTRY APPLICATIONS, VOL. 30,NO. 4, JULY / AUGUST 1994

conclusion, although not surprising, is that under steady-state conditions, the flux regulator offers no advantageover the feedforward approach used in IF0 systems withrespect to parameter sensitivity of the current model inDF O systems. The flux regulator will be shown, however,to Significantly improve the DFO system when using thevoltage model.V. PARAMETER E NSITIVITYN THEFIELD-WEAKENINGEGION

At high speeds, especially in the field-weakening/con-stant-horsepower region, both the motor and the drivetend to be highly stressed. It is under these operatingconditions that the system is most sensitive to losses andvoltage limits, and hence to accurate field orientation.The accuracy of the flux estimate used for DFO whichoriginates from the voltage model at these high speeds isthus of great importance.Although the closed-loop rotor flux observer FRF plotin Fig. 4 indicates that the rotor flux estimate is onlymildly sensitive to the leakage inductance estimates atrated slip frequency, the sensitivity increases with operat-ing slip frequen cy. Fig. 8 plots the voltage model FR F as afunction of slip frequency with varying errors in the rotorleakage inductance.At high slip frequencies common to field-weakenedoperation, the magnitude and phase errors are substan-tial. With closed rotor slots, the rotor leakage inductancecan easily vary by a factor of 2 or more. Because thestator leakage variation is much smaller due to the openstator slots, the corresponding stator leakage plots ar e notincluded.The sensitivity to magnetizing inductance estimate isinsignificant relative to the leakage inductance sensitivity.Without flux regulation (i.e., via a feedforward ap-proach), the phase errors in Fig. 8 will cause the DFOsystem to converge on an incorrect operating slip andhence flux level, as plotted in Fig. 9. Note that the actualflux errors a re significantly greater th an the flux estimatemagnitude errors in Fig. 8.With rotor flux regulation, the DFO system will con-verge to a more correct flux level indicated by the FRFmagnitude plots in Fig. 8. Although a thorough stabilityanalysis is beyond the scope of this paper, system stabilitycan be shown to be greatly improved by the addition ofthe flux regulator.Even with flux regulation, however, the potential errorin the operating flux level due to errors in the rotorleakage inductance estimate can limit an d/ or significantlydegrade very high-speed operation. Because stator flux isobtained directly from the stator voltage, it is not suscep-tible to leakage inductance parameter errors, and thusorientation to the stator flux has attracted considerableattention.

VI . A CLOSEDLOOPTATOR LUX BSERVERA closed-loop stator flux observer structured similar tothe closed-loop rotor flux observer is shown in Fig. 10.

1.21.153 1.19 .osE1 o

0.95

......... I .........

2010

c3 02A 10-20-30

a

4 F A- - ... 3[ . . . Ir.O L.......................... 2 0 ...........lr OLE ~ ..................

.......... 1.5 .......... .....

0 1 2 3 4 0 1 2 3 4

flux regulation based upon the voltage model as a function of effectivecommanded slip (1 ) for rotor leakage inductance parameter errors.......................................................... ............. ................................ .Conhuller j ;p!

Fig. 10. A closed-loop stator flux observer of identical topology to therotor flux observer in Fig. 3.

Note that the leakage terms have simply moved from therotor flux voltage model to the stator flux current model,and thus the net computation and parameters required bythe closed-loop observer are unchanged.The F RF relating the estimated to the actual stator fluxfor rated slip operation is plotted in Fig. 11with gains setfo r 10Hz closed-loop eigenvalues. Th e para mete r sensitiv-ity is nearly identical to that of the rotor flux observer, themajor difference being the sensitivity to leakage induc-


5/9

949JANSEN et al.: OBSERVER-BASED DIRECT FIELD ORIENTATION

Rotor Velocity (P.u.)Fig. 1 1 . Closed-loop FRF, A,,,/A,,, , llustrating parameter sensitivityof the closed-loop stator flux observer at rated slip operation (10 Hzeigenvalues) [(A6) in Appendix].

Rotor Velocity (P .u.)

tance which has moved from the voltage model to thecurrent model. For less than rated slip frequency, charac-teristic of low-speed operation below rated torque, theincreased leakage sensitivity of the current model is smallrelative to the rotor resistance and magnetizing induc-tance sensitivities. The observer also functions at zerospeed.VII. A STATOR LUXD F O SYSTEMIn addition to the elimination of parameter sensitivityat high speeds, orientation and regulation of the statorflux can result in nearly optimal torqu e capability over theentir e field-weakening speed range [81.

A significant difference between stator and rotor fluxorientation is the level of cross coupling that occursbetween the d and q axes. When oriented to the rotorflux (i.e., A$ = 0), cross coupling of the stator q-aiscurrent to the d-axis rotor flux is completely eliminated,as was shown in Fig. 2. However, with stat or flux orien ta-tion (i.e., A$ = 0), the q-axis rotor flux is not zero, andthus a significant am oun t of cross coupling still exists. Thiscross coupling can be expressed in the form of a cross-coupling current i $ acting as a flux producing input inparallel with i;% as shown in Fig. 12.For dynamic flux control, a decoupler, also shown inFig. 12, can be implemented to eliminate the undesirablecross coupling [3]. Unfortunately, the decoupler is param-eter sensitive and requires additional computation.

A stator-flux-regulated stator-flux-oriented (SFR-SFO)DFO system com plete with decoupler is shown in Fig. 13.Since the torque command (q-axis) current is calculatedvia the estimated stator flux which is most accurate athigh speeds (assuming negligible voltage measurementand integration errors), the dynamics and accuracy of th edeveloped torque are dependent solely upon the currentregulation. T he flux regulator drives the stator flux to thedesired level. However, the decoupler can be removed,causing dynamic stator flux variations dependent on fluxregulator bandwidth.

VIII. A ROTOR-FLUX-ORIENTED STATOR LUXREGULATED FO SYSTEMThe attractive features of the SFR-SFO system in Fig.13are not unique to stator flux orientation, but rather to

I I; / I.... ___..... ____.... . .... . _ . _ _ _ _ _ _.. ..... ..... . .... . ..... . ____ ..... . .... ..... ..... ..... .Fig. 12 . A tuned stator-flux-orien ted machin e model illustrating crosscoupling and a control-based decoupler.

Fig. 13. A stator-flux-regulated stator-flux-oriented (SFR-SFO) DF Osystem with decoupler (C,< = stator flux regulator, e.g ., PI controller).

stator flux regulation. The stator-flux-regulated rotor-flux-oriented (SFR -RFO ) system illustrated in Fig. 14 isalso capable of achieving correct torque and flux control,even under detuned conditions.Compared to the previous rotor and stator-flux-ori-ented systems, the torque command current must now becalculated via the full torque equation in the rotor fluxframe , i.e.,(3)

In addition, the stator flux magnitude must also be calcu-lated, although the computationally expensive square rootfunction can be avoided by simply regulating t he s quar e ofthe stator flux magnitude.To gain perspective on the different possible DFOsystems, a rotor-flux-oriented machine detuned with re-spect to the stator transient inductance is illustrated inFig. 15.The model is derived in [7]. The cross couplingdue to detuning is of the same form as that always presentin the stator-flux-oriented system in Fig. 12.The control-based decoupler in Fig. 12 attempts toremove the cross coupling that is not present in the tunedrotor-flux-oriented system. Because the decoupler is sensi-tive to additional machine parameters such as the rotortime constant, correct decoupling in the stator-flux-ori-ented system is more difficult than in the rotor-flux-ori-ented system.The rotor flux orientation attempts to maintain con-stant rotor flux under torque command changes, while


6/9

950 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS,VOL. 30,NO. 4, JULY / AUGUST 1994..........................................................................................

Fig. 14. A tuned, stator-flux-regulated rotor-flux-oriented (SFR-RFO)DFO system.

rf : Detuned Rotor Flux Oriented Machine. . . q s j

Fig. 15. A rotor-flux-oriented machine with de tp ed stator transientinductance ( A a L , = aL , - CL,) .

all9wing the stator flux to change, i.e., A$ = 0, but Ai< =GLsii! # 0. The stator flux regulator counteracts by at-tempting to maintain constant stator flux. Thus, a tunedrotor-flux-oriented stator-flux-regulated system with lim-ited flux regulation bandwidth will experience a perturba-tion in the stator flux magnitude under torque commandchanges. The perturbation is not present in the stator-flux-oriented system with a correct decoupler, but is pre-sent at a potentially larger degree when either detuned orthe decoupler is eliminated.IX . IMPLEMENTATIONF OBSERVERASEDDIRECT

FIELD RIENTATIONThe three observer-based DFO systems designated

RFR -RFO , SFR-RFO , a nd SFR-SFO were experimen-tally evaluated via the implementations in Fig. 16. T heobservers and coordinate transformations were imple-mented in software on the M otorola DS P56001 with a 4kHz sample rate. A faster timed interrupt routine sam-pled and averaged the measured voltages and currents(after th e anti-alias filters), providing an effective integra-

Inductionachine

Velocity i er jObsewer jj DS P 56001L ..................................................................................

InductionMachine

n . Velocity i erObserverw :: ;j DSP 56001I................................................................................ >

Fig. 16. Experimental DFO systems using closed-loop flux observers.Upper diagram: stator-flux-regulated stator-f lux-oriented (SFR-SFO).Lower diagram: stator-flux-regulated rotor-flux-oriented (SFR-RFO).The rotor-flux-regulated rotor-flux-oriented (RFR-RFO) system imple-mentation was of the same form as the SFR-SFO system without thedecoupler. Al l three were implemented with a Motorola DSP56001.

tion step of 15.6 ps (64 kHz). The PWM VSI switched at3.4 kHz with a stationary frame PI current regulator with= 350 Hz bandwidth. D ue to signal noise, the flux regula-tor bandwidth was limited to real eigenvalues of = 15-20Hz. Th e flux observer two eigenvalues were set at 0.5 and5.0 Hz.Experimental results illustrating the response of th ethree systems to square-wave torque commands toggledbetween velocity limits are shown in Figs. 17-20. T he loadwas dominated by the rotor inertia with negligible damp-ing; thus, a triangular velocity waveform would be ex-pected for a tuned system.Under rated conditions and with a reasonably high fluxregulator bandwidth, all three systems behaved similarly,providing good torque and flux control, independent ofdetuning and/or the SFO decoupler.Zero- and low-speed operation of the SFR-SFO systemis demonstrated in Fig. 17 unde r tuned and det une d(F r = 2.0r,) conditions with a rated torque command. Asexpected (see Fig. 111, some sensitivity to Fr is presentfrom the stator flux current model.To demonstrate the differences among the three DFOapproaches, the systems were pushed to the limits, i.e.,wide-speed operation up to +2.7 P.u., field weakening to1/3 rated fl?, 1/3 rated torque commands, and severedetuning to L,, = 3.0L,, (see Figs. 18-20). The resulting


7/9

JANSEN et al.: OBSERVER-BASED DIRECT FIELD ORIENTATION 951

I I I I

0 :AWr 0.10

(P.U.) 0.050

1 : : : I I : : I0 0.05 0.1 0.15 0.2 0 0.05 0.1 0.15 0.2Time (sec.) Time (sec.)

(8.) hmed (b.) &anmi(&Z.O rr)

Fig. 17. Experimental low-speed results for estimated rotor and statorflux magnitudes and rotor velocity for the SFR-SFO system withoutdecoupler under tuned and detuned conditions.

. . . . . . :. . . . . . . . . . . . . . . . . . . . . . . .. ...... . ;... .

....... . . . . . . . . . . . . . . ..... . . . . . . . . . . . . . . . . . . . . . -13-1-2-3

Time (sec) Time (sec)(a.) hmed (b.) deI uned (k 3.0 L1 , )

Fig. 18. Experimental results for estimated rotor and stator flux androtor velocity for the RFR-RFO system under tuned and detunedconditions.

0 5i r 0 4

(P.U 1O0.20 10 5

h s 0 4( P U ) 0 3

0 20 1

3hr( P u ) o

1-2-70 5 10 15 20Time ( sec) 0 5 IO 15 20Time (sec )

(a.) hmed (h.) tlehinrtl(l+=3.0br)

-3 o 5 10 0 5 10 15 20Time (sec) Time (sec)(a.) tun4 drcwpler (b.)without d e c w p l a

Fig. 20. Experimental results illustrating estimated rotor and stator fluxmagnitudes and rotor velocity for the SFR-SFO system with (a) tuneddecoupler and (b) without decoupler.

high-slip operation was shown to increase the sensitivityto the leakage estimates. The flux regulator is forced tocompensate for errors attributable to poor parameterestimates, the chosen orientation reference, and to thestationary frame PI current regulator limits.In Fig. 18, the RF R-RF O system performed well whentuned and field weakened, but not detuned and fieldweakened. In Fig. 19, the SFR-RFO system performednearly as well as the RFR -FR O system when tuned andfield weakened, but was substantially better than theRF R-R FO in the detuned, field-weakened operation . InFig. 20, the SFR-SFO system with the tuned decouplerperformed excellently up to intermediate speeds = 1.7p.u. velocity. At speeds greater than this, the decoupledsystem gave evidence of instabilities. In Fig. 20, th eSFR-SFO system without the decou pler operated ad e-quately at higher speeds, although the cross coupling wassubstantially more than the flux regulator could easilycompensate for.

X. CONCLUSIONSThis paper presented an analysis and comparison ofdifferent methods of achieving observer-based direct fieldorientation (DFO ). DFO based upon rotor and stator fluxclosed-loop observers is an excellent means of achievingflux and torque control over very wide speed rangesincluding zero and field-weakened regions. The approachhas substantially improved parameter insensitivity com-

pared to existing methods.Three observer-based D FO approaches were evaluated.Below rated op eration, all showed good performance.Zero- and low-speed operation of stator-flux-ob-server, stator-flux-oriented control was dem onstrated .Fig. 19. Experimental results for estimated rotor and stator flux androtor velocity for the SFR-RFO system under tuned and detunedconditions. limits were pushed.The three DFo approaches differed when the system


8/9

952 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 30, NO . 4, JULY / A U G U S T 1994

The rotor-flux-regulated and -oriented system wasshown to be highly sensitive to leakage inductance esti-mates under high-slip operation.A stator-flux-regulated and -oriented system withoutdecoupler is parameter insensitive at high speeds, butsuffers from excessive cross coupling under high-slip oper-ation that must be handled by the flux regulator.A stator-flux-regulated and rotor flux-oriented systemappears to offer intermediate parameter sensitivity andcross coupling which is substantially easier t o remove thanwith stator flux regulation and orientation.

APPENDIXA. Current Model, Voltage Model, and Clos ed-L oop ObserverFRF Equations

Rotor Flux Observers:

where K = K , + K , / p (A3)Stator Flux Observers:A; L , 1 + r r w , 1 + jc??,w,l = L (i d s L , 1+J?,q( 1 + j m r w , ) = F R F s c (A41

B. Induc tion Machine ParametersWestinghouse T E E 11, Frame 215T, 10 hp, three phase,

460/230 V, 12.2/24.4 A, 1750 rpm, rated w, = 0.037 p.u.

ACKNOWLEDGMENTThe authors wish to acknowledge the motivation pro-vided by the Wisconsin Electric Machines and PowerElectronics Consortium (WEMPEC) of the University ofWisconsin-Madison.

*e$9 sf

REFERENCESD. W. Novotny and R. D. Lorenz, Eds., Introduction to FieldOrientation and High Performance AC Driues, tutorial book from1985 and 1986 IEEE-IAS Ann. Meet.K. B. Nordin, D. W. Novotny, and D. S. Zinger, The influence ofmotor parameter deviations in feedforward field orientation drivesystem, IEEE Trans. Ind. Appl. , vol. IA-21, pp. 1009 -1015, July/Aug. 1985.X. Xu , R . De Doncker, and D. W. Novotny, A stator flux orientedinduction machine drive, in Proc. I988 Power Electr on. SpecialistsConf., Kyoto, Japan, Apr. 1988. pp. 870-876.-, Stator flux orientation control of induction machines in thefield weakening region, in Proc. IEEE-US Annu. Meet. , Oct.X. Xu and D. W. Novotny, Implementation of direct stator fluxorientat ion control on a versatile DSP based system, in Proc.IEEE-L4S Annu. Meet., Oct. 1990.-, Selecting the flux reference for induction machine drives inthe field weakening region, in Proc . IEEE-US Annu. Meet. , Oct.1991.X. Xu , Stator flux orientation-A robust control technique forinduction machines, Ph.D. dissertation, Univ. Wisconsin, Madi-son, 1990.P. L. Jansen and R. D. Lorenz, A physically insightful approachto the design and accuracy assessment of flux observers for fieldoriented induction machine drives, in Proc. IEEE - U S Annu.Meet. , Oct. 1992, pp. 570-577.P. L. Jansen, C .0.Thompson, and R . D. Lorenz, Observer-baseddirect field orientation for both zero and very high speed opera-tion, in Proc. PCC, Yokohama, Japan, Apr. 1993, pp 432-437.P. L. Jansen and R . D. Lorenz, Accuracy limitations of velocityand flux estimation in direct field oriented induction machines, inProc. EPE Conf., Brighton, England, Sept. 1993.I. Takahashi and T. Noguchi, A new quick-response and high-ef-ficiency control strategy of an induction motor, IEEE Trans. Id.Appl. , vol. IA-22, pp. 820-827, Set./Oct. 1986.

1988,pp. 437-443.

Sfqdr, f q d sPTef

NOMENCLATURE

Estimated quantities.Commanded or reference quantities.Arbitrarily aligned synchronous frame quantity.Rotor and stator flux synchronous fram e qu an-tities, respectively.Stationary frame quantity.Complex rotor and stator quantities, respec-tively, i.e., f q d r = qr - fdr.Differential operator; p = w e at steady state.Normalized electromagnetic torque, T, =3/2 P/2Tb.Rot or an d s tator flux magnitudes, respectively.Stator transient inductance, L, - L L / L r .Rotor time constant; L r / r r .rotor position (elec. rad).Rotor and stator flux angles, respectively (elec.rad).rs = 0.20 fl L , , = 1. 5 m H L , = 32.3 m H . we Fundamental excitation frequency (rad/&

r, = 0.20 0 U,, os Rotor velocity and slip frequency, respectively(rad/$.,, = 1.5 m H


9/9

JANSEN et al.: OBSERVER-BASED DIRECT FIELD ORIENTATION 953Patrick L. Jansen 692) received the B.S., M.S.,and Ph.D. degrees in electrical engineering fromthe Universi ty of Wisconsin, Madison, in 1985,1987, and 1993, respectively.From 1987 to 1989 he was an experimentalScientist in the Commonwealth Scientific andIndustrial Research Organization (CSIRO), Di-vision of Applied Physics, New South Wales,Australia and a Development Engineer at Ma-chine Dynamics Pty. Ltd., Victoria, Australia.His dissertation research included the integra-tion of electric machine design and transducerless position and velocityestimation, observer-based direct field orientation, and linear inductionmachine design for high-speed material transport systems. He is cur-rently involved with the development of ac drive locomotives for theMorrison Knudsen Corporation, Boise, ID.

Robert D. Lorenz (SM91) received the B.S.,M.S., and Ph.D. degrees from the University ofWisconsin, Madison, in 1969, 1970, and 1984,respectively.Since 1984 he has been a member of theFaculty of the University of Wisconsin, Madi-son, where he is a Professor of MechanicalEngineering and of Electrical and ComputerEngineering. In this position, he acts a s Associ-ate Director of the Wisconsin Electric Machinesand Power Electronics Consortium and as Co-Director of the Advanced Automation and Robotics Consorhum. He wasa visiting Research Professor in the Electrical Drives Group of theCatholic University of Leuven, Leuven, Belgium and in the ElectricalDrives Institute of the Technical University of Aachen, Germany, in theSummer of 1989 and th e Summers of 1987 and 1991, respectively. In1969-1970 he did his Master thesis research at the Technical Universityof Aachen, Germany. From 1972 to 1982 he was a member of theResearch Staff at the Gleason Works, Rochester, NY. His currentresearch interests include sensor integrated electromagnetic actuator

technologies, real-time digital signal processing and estimation tech-niques, and ac drive and high-precision machine control technologies.Dr. Lorenz is a Chairman of the IE EE IAS Industrial Drives Commit-tee, and is a member of the Industrial Automation and Control Commit-tee, the Electrical Machines Committee, and the Industrial Power Con-verter Committee. He is an active consultant to many organizations andis a Registered Professional Engineer in the states of New York andWisconsin. He is a member of the ASME, the ISA, and the SPIE.

Donald W. Novotny (F87) received the B.S. andM.S. degrees in electrical engineering from theIllinois Institute of Technology, Chicago, in 1956and 1957, and the Ph.D. degree from the Uni-versity of Wisconsin, Madison, in 1961.Since 1961 he has been a member of theFaculty at the University of Wisconsin, Madison,where he is currently Grainger Professor ofElectric Machines and Power Electronics andCO-Director of the Wisconsin Electric Machinesand Power Electronics Consortium (WEMPEC).He served as Chairman of the Electrical and Computer EngineeringDepartment from 1976 to 1980 and as an Associate Director of theUniversity-Industry Research Program from 1972 to 1974 and from1980 to the present. He has been active as a consultant to manyorganizations including Marathon Electric Company, Borg Warner Cor-poration, Barber Coleman Company, Otis Elevator Corporation, AllenBradley Company, Eaton Corporation, and the Wisconsin Departmentof Natural Resources. He has also been a visiting Professor at MontanaState University, the Technical University of Eindhoven, Eindhoven, TheNetherlands, the Catholic Universi ty of Leuven, Leuven, Belgium, and aFulbright Lecturer at the University of Ghent, Ghent, Belgium. He haspublished over 90 technical articles on electric machines, variable fre-quency drives, and power electronic control of industrial systems, severalof which have received prize paper awards from the IEEE IndustryApplications Society.Dr. Novotny is a member of ASEE, Sigma Xi, Eta Kappa Nu, and TauBeta Pi and is a Registered Professional Engineer in Wisconsin.

Observer-Based Direct Field Orientation Analysis and Comparison of Alternative Methods

Documents