Loss-of-excitation Protectionfor Synchronous Generators

GER-3183

LOSS OF EXCITATION PROTECTION FORMODERN SYNCHRONOUS GENERATORS

John BerdyGeneral Electric CompanySchenectady, New York

ABSTRACT

This paper presents the results of a study intothe application and performance of the offsetmho distance relay for the loss of excitation pro-tection of synchronous generators. Included isinformation on the loss of excitation characteris-tics of modern generators, on relay performanceduring transient swings and low frequency dis-turbances and on generator protection.

INTRODUCTION

In 1949,1 a single phase offset mho relay wasintroduced for the high speed detection of loss ofexcitation in synchronous generators. This distancerelay approach was developed to provide improvedselectivity between loss of excitation and othernormal or abnormal operating conditions and toprovide the operating times necessary for optimumprotection of both the generator and the system.

Over the years, the offset mho relay has beenwidely accepted for loss of excitation protectionand experience with the relay has been excellent.The relay has demonstrated its capability ofdetecting a variety of excitation system failuresand to discriminate between such failures andother operating conditions. The relatively fewcases of incorrect operation that have occurred canbe attributed to incorrect relay connections (majorcause), and blown potential transformer fuses.

In spite of this excellent experience, there hasbeen some user apprehension about the perform-ance of distance type of relaying for loss of excita-tion protection. In particular, there has beenconcern over possible incorrect operation of therelay when operating the generator in the under-excited region, during stable transient swings andduring major system disturbances that cause under-frequency conditions.

In view of this continuing concern over relayperformance and in view of the fact that machineparameters have changed appreciably during thepast twenty years, a general study was initiated toreview the application and the performance of theoffset mho loss of excitation relay for a variety of

system conditions. This paper discusses the resultsof this study and provides guidance on the applica-tion of loss of excitation protection.

REVIEW OF RELAY CHARACTERISTICSAND SETTINGS

The offset mho loss of excitation relay is a singlephase, single element distance relay which isappliedto the generator terminals and connected and setto look into the machine. On the R-X diagram (seeFig. 1) the relay characteristic is an offset circlewhich has an angle of maximum torque that fallson the (-X) ordinate. As viewed from the machineterminals the relay will operate for any imped-ance phasor that terminates inside the circularcharacteristic.

When the relay was introduced in 1949, it wasrecommended the offset be set equal to one-halfof the direct axis transient reactance (X’d/2) andthe diameter of the circle set equal to the directaxis synchronous reactance (Xd). It was shown1

that with the machine reactances that existed atthat time, these settings would detect a loss ofexcitation from any machine loading and thatthere would be optimum selectivity against oper-ation during stable power swings. Machine directaxis synchronous reactances were in the range of1.1 to 1.2 per unit.

Fig. 1. Operating characteristic of loss-of-excitation relay.

3

In more recent machine designs, these synchro-nous reactances have increased to a 1.5 to 2.0 perunit range. With the advent of these higher imped-ance machines there has been reluctance by someutilities to use relay settings proportional to syn-chronous reactance, mainly because of a fear thatthe resulting large circle diameters might infringeon the underexcited operating capability of themachine. Therefore where this possibility was aconcern, it has been recommended that the relayreach be limited to an assumed synchronousreactance of 1.0 per unit. When this recommenda-tion was made it was recognized that this reducedsetting would detect a loss of excitation with highmachine loadings (the most severe condition forboth the machine and the system) but would notprovide coverage if the machine was lightly loaded.While this limited coverage was acceptable to theconcerned user, there has been some question as tothe extent of protection being provided. Therefore,one of the purposes of the study was to determinequantitatively the protective limits of a reducedsetting.

probable mode of failure. For the less likely case ofan open field, the loss of excitation characteristicswill differ to some extent from those presentedhere but the final impedances as viewed from thegenerator terminal will be essentially the same asfor a short-circuited field.

While the discussion will be limited to steammachines with specific parameters, the results andphenomena described also apply to hydro-gener-ators and machines with other parameters.

As a point of interest, it should be noted thatthe loss of excitation characteristics and pheno-mena presented here do not differ appreciablyfrom those reported by Concordia,3 Temoshokand Mason some twenty years ago.

Loss of Excitation - TandemCompound Generators

GENERATOR LOSS OF EXCITATIONCHARACTERISTICS

This section presents and discusses in somedepth the loss of excitation characteristics ofmodern tandem and cross compound generators.As noted in reference 1, the loss of excitationcharacteristic refers to the locus of the apparentimpedances as viewed from the generator termi-nals during a loss of excitation condition. Thesecharacteristics were determined for typical ma-chine designs in a digital computer study using acomprehensive dynamic model2 of a turbinegenerator.

Figure 2 shows the loss of excitation character-istics for a typical large tandem compound gener-ator that is connected to a system through a step-up transformer having a .15 per unit impedance onthe machine base. These characteristics are shownas a function of both initial machine loading andsystem impedance.

The following discussion will consider the effectof initial generator loading and system impedanceon the impedance locus, on the generator terminalvoltage and on machine loading during a loss ofexcitation condition. The discussion will also con-sider the effect of voltage regulators on cross com-pound generators. In all cases the loss of excitationcharacteristics will be plotted with respect to tworelay settings: one setting will have a circle diam-eter of 1.0 per unit, the other will have a circlediameter equal to machine synchronous reactance.The offset, in both cases, will be equal to X’d/2.

0 . 5

-I3-

1

2

1 2 3 4

PER UNIT IMPEDANCE

CURVE INITIAL LOADING (per unit) SYSTEM IMPEDANCE

0.93 M V A 0.92 PF Lagging 0.4 PU0.98 M V A 0.98 PF Lagging 0.2 PU

0.92 M V A 0.90 PF Lagging 00.31 M V A 0.95 PF Leading 0.4 PU0.30 M V A 1.00 PF 0.2 PU

In all cases, it was assumed the loss of excitation Fig. 2. Loss-of-excitation characteristics for a tandem-

was caused by a short-circuited field, the most compound generator.

4

As noted in the diagram, curves (a), (b) and (c)show the impedance locii as a function of systemimpedance with the machine operating initially ator near full load. Curves (d) and (e) show the lociiat two values of system impedance with the ma-chine initially at about 30% load.

For the case of the machine operating at fullload, all of the impedance locii terminate in an areato the right of the (-X) ordinate and will approachimpedance values, which at the final steady-stateslip, will be somewhat higher than the average ofthe direct and quadrature axis subtransient impe-dances of the generator. The final impedances willalways be greater than the offset setting (X’d/2)and therefore will always fall inside the relay cha-racteristics as shown in Fig. 2.

For system impedances of zero and 0.2 per unit,the impedance locii (b, c) go directly to this areawhile the impedance locus for a .4 system spiralsinto the area as indicated by curve (a). The traversetime from the initial load point to the relay cha-racteristic of the impedance locii will be between 2to 7 seconds. The .4 system locus travels the fastest(2 seconds). It should be noted that when the im-pedances reach the area to the right of (-X) ordi-nate, the machine will be operating as an inductiongenerator at a speed of 2 to 5% above normal. Itwill be supplying some reduced power to the sys-tem and will be receiving its excitation (VARS)from the system. The machine slip and the poweroutput will be a function of the machine slip-torquecharacteristic (which in turn is a function of ma-chine and system impedances) and governor cha-racteristic. High system impedances produce a highslip and a low power output.

For the case of the machine operating initially at30% load, the impedance swing is more gradual andonly goes as far as point (A) just inside the 1.0 perunit circle before it reverses. The swing will oscil-late in the region between points (A) and (B). Thetraverse time from the initial point to point B isaround 7 to 9 seconds while the time to traversethe distance B-A can be up around 10 to 15 secondsor higher. For this initial loading, the machinespeed will only be 0.1 to .2% above normal and asbefore it will be operating as an induction generator.

For initial machine loadings between .3 and 1.0per unit, the impedance locii will terminate insidethe 1.0 per unit circle in the region above point A.For loadings below .3 per unit the locii will ter-

minate below point A and will only appear in thelarge circle (diameter = Xd). For a loss of excita-tion from no load, the relay will see an impedancewhich in the limit will vary between the direct andquadrature axis synchronous impedances (Xd’ Xq).

Machine Loading and Terminal Voltage: Figure 3shows the effect of loss of excitation on terminalvoltage, power output and reactive power for a 0.1per unit system impedance and for a machine oper-ating initially at full load. The abscissa is given inseconds while ordinates specify per unit volts,power and VARS. It should be noted that negativeVARS signify VARS into the machine.

I I I I I I I

Fig. 3. Variation in terminal voltage, power, vars for loss

of excitation on tandem-compound generator.

As noted in this diagram, the voltage decreasesand oscillates around an average of 0.5 per unit,the power output decreases and averages about 0.3per unit and the VARS go negative and averagearound-O.93 per unit.

For the case of a lightly loaded machine, thevariation in loading and terminal voltage will beconsiderably less when excitation is lost. For exam-ple, consider the case of a generator connected to a.2 system and with an initial loading of P = .3, Q =-. 156, VT = 1.0 per unit. Thirty (30) seconds afterlosing excitation, the lowest voltage reached was.78 per unit, the power dropped only to .275 perunit and the VARS reached -.6 per unit.

There are several points to note from theseresults. First, when a lightly loaded machine loses

5

excitation, the final MVA loading will probablynot be damaging to the machine but the VARdrain may be detrimental to the system. In the casediscussed the final machine MVA loading is .66 perunit and the stator current only reaches .85 perunit. When the machine is initially operating at fullload, a loss of excitation can be damaging to boththe machine and the system. While the final load-ing in terms of MVA is not excessive, the machinein Fig. 3 will have stator currents in excess of 2.0per unit. The high current is due to the fact thatthe resulting machine loading is at a substantiallyreduced terminal voltage. Of course, the VARdrain from the system can depress system voltagesand thereby affect the performance of other gener-ators in the same station or elsewhere on a system.In addition, the increased reactive flow across thesystem can cause tripping of transmission lines andthereby adversely affect system stability. Forexample, in 1951 a utility reported4 that loss ofexcitation on a 50 MW generator caused systemwide instability, the tripping of interconnectionsand tie lines and over 100 breaker operationsbefore the disturbance subsided. In this case, itwas evident that other generators and interconnec-tions could not stand the additional reactive loadimposed on the system. The possible effects onother generation will be discussed in a later section.

Loss of Excitation - Cross Compound Generators

The cross compound generator studied was atypical 900 MVA conductor-cooled machine. Itwas assumed, the high and low pressure units werebussed at generator voltage and connected to ahigh voltage system via a .15 per unit transformer.

As perhaps might be expected, the loss of excita-tion characteristics for cross-compound units followmuch the same pattern as those for a tandem gen-erator. With a loss of excitation on either the highpressure (HP) unit or the low pressure (LP) unit,the impedance locii as a function of system imped-ance and initial loading are similar in all respects tothose for tandem units. The behavior of the soundunit in the cross-compound arrangement (that is,the unit which still has excitation) will be a func-tion of system impedance and of whether or not avoltage regulator is in service. To illustrate thesimilarity in characteristics, Fig. 4 shows the lossof excitation characteristics for a cross-compoundgenerator connected to a .2 system and with initialloadings of .95 MVA and .3 per unit MVA. In bothcases, it was assumed the low pressure unit lost

6

excitation and the effect on the high pressure unitwas determined with and without regulator.

1

CURVES-a,b,e: LOADING = 0.95.MVA

\ CURVES d,e,f: LOADING = 0.3 MVA2 ’ I I 1 I I I I I I

-X 1 2 3

PER UNIT IMPEDANCE

LP-LOW PRESSURE UNITHP-HIGH PRESSURE UNITNR-NO VOLTAGE

REGULATORWR-WITH VOLTAGE

REGULATOR

Fig. 4. Loss-of-excitation characteristics for a cross-com-

pound generator.

As before, for a .95 MVA initial loading, the im-pedance locus (curve a) terminates to the right ofthe (-X) ordinate. The locus with an initial loadingof .3 per unit (curve d) again just reaches inside the1 per unit circle.

As noted by curves (b, c, e, f) in the diagram,the impedance locii of the sound unit (high pres-sure unit) will vary appreciably depending onwhether or not the voltage regulator is in service.On the other hand, the loss of excitation imped-ance locus for the low pressure unit is essentiallythe same with or without a regulator in service onthe high pressure unit.

In this case and for lower system impedances,the high pressure unit will not lose synchronism.However, with a 0.4 system impedance and withthe voltage regulator out of service, the high pres-sure unit will pull out of step with respect to thesystem. With the voltage regulator in service, thehigh pressure unit will remain in synchronism.

A loss of excitation on the high pressure unitproduces impedance locii which are almost iden-tical to those shown in this figure.

Machine Loading and Terminal Voltage: The lossof excitation on a unit in a cross-compound gener-ator imposes a more severe duty on the generatorthan in the case of a tandem machine. The mostsevere duty occurs when the generator is initiallyat futl load, when the system impedance is .2 andbelow, and when the voltage regulator is in serviceon the sound unit. With these conditions, the unitthat has lost excitation can have a peak MVA load-ing over 2.0 per unit and peak currents in excess of2.5 per unit. The sound unit can have peak MVAloadings above 1.5 per unit and peak currentsapproaching 2.0 per unit.

Even with the voltage regulator out of service,the unit that has lost excitation can have peakMVA loadings that approach 2.0 per unit and thepeak currents that approach 2.5 per unit. Thelarge variations in power and VARS that can occurare illustrated in Fig. 5 for the cross-compoundunit connected to a .2 system. Figure 5 shows thevariation in terminal voltage, power and VARSwhen the low pressure unit loses excitation and thevoltage regulator is in service on the high pressureunit.

system can exceed .5 per unit. With or without avoltage regulator in service, the terminal voltage isstill above .9 per unit after 60 seconds and the slipis negligible.

Effect on Generators in the Same Station

To study the effect on a “sound” machine afteranother machine in the same station loses excita-tion, it was assumed two similar tandem compoundgenerators were connected to a high voltage systemthrough separate step-up transformers as shown inFig. 6. It was also assumed both machines wereinitially fully loaded and that machine (A) lost ex-citation. The effect on machine (B) wasdeterminedfor three values of system impedance (Zs = .05, .2,.4), with and without a voltage regulator in serviceon machine (B).

Fig. 5. Variation in terminal voltage, power, vars for loss

of excitation on low-pressure unit for a cross-compound

generator.

When the generator is initially operating at .3per unit power the loading and current do not ex-ceed 1.0 per unit but the VARS taken from the

Fig. 6. Two tandem-compound generators in same station.

This study showed that when the system imped-ance is low (Zsys = .05), the loss of excitation onmachine (A) will have little effect on machine (B).Without a voltage regulator, the terminal voltageon (B) will drop about 5% but the power and VARoutput will remain essentially constant. With avoltage regulator, the VAR output from (B) in-creased slightly going from Q = +.09 to Q = +.3.

When the system impedance is increased toZ sys = .2, there is a greater effect on the perform-ance of machine (B). Without a regulator, theterminal voltage and machine output will varyappreciably. During a 5 second time interval, theterminal voltage dropped 20%, the power outputdecreased to .75 per unit and VAR output increasedto .3. However machine speed only increased .2%and the maximum angular swing was 20”. With avoltage regulator in service, there was a 5% varia-tion in terminal voltage and a negligible effect on

7

machine speed. However there was an appreciableincrease in machine (B) loading, mainly due to anincreased VAR output. During a 10 second interval,the machine MVA loading reached a peak of 1.4per unit and remained in the range of 1.1 to 1.3per unit for several seconds. The voltage regulatorremained at ceiling for 7.0 seconds.

With a system impedance of .4 there is a pro-nounced effect on the performance of machine (B).Without a voltage regulator, machine (B) will losesynchronism in about 1.0 second after machine(A) slips a pole. The loss of excitation and loss ofsynchronism characteristics for both machines areshown in Fig. 7. With a voltage regulator, machine(B) maintains synchronism but the MVA loadingreached and remained around 1.4 per unit for atleast 5 seconds.

-x I 1PER UNIT IMPEDANCE

Fig. 7. Loss of excitation - Machine A

Loss of synchronism - Machine B

While it would appear from the above resultsthat the loss of excitation on one machine willonly affect a nearby machine when the system im-pedance is unusually high, it should be recognizedthat this was a limited study which did not considerall possible machine characteristics, system config-urations and the interaction effects of other gener-ators. There has been at least one case reportedwhere a loss of excitation on a machine caused a

generator in a nearby station to lose synchronismand the equivalent system impedance was .2 orless. The study does indicate however the effective-ness of the voltage regulator in maintaining ma-chine stability during these conditions.

PERFORMANCE DURING TRANSIENT SWINGS

From time to time, there have been reports thatthe loss of excitation relay had operated during astable transient swing after the clearing of a nearbyexternal fault. An investigation of each case re-vealed that either the connections to the relay wereincorrect or that an incorrect voltage was appliedto the relay. In effect, the relay was “looking” outinto the system and not into the generator. In sev-eral cases, the loss of excitation relay had actuallydetected a loss of synchronism which was causedby prolonged fault clearing times.

In spite of the excellent performance of the lossof excitation relay in this regard, the concern thatthe relay might operate incorrectly during stableswings has persisted. In view of this concern, aninvestigation was made to determine the proximityof stable swings to the relay characteristic.

The impedance swing characteristic as viewedfrom the generator terminals was determined forstable transient swings after the clearing of a threephase fault on the high voltage side of the step-uptransformer. Both tandem and cross-compoundgenerators were considered in the study. For variousmachine parameters the impedance swing wasdetermined as a function of fault clearing time,system reactance and whether or not a voltageregulator was in service. After a number of com-puter runs, it soon became evident that the “worst”swings occurred when:

1. The voltage regulator was out of service.

2. The system impedance was low.

3. The fault clearing times were equal to the cri-tical switching times. (That is, the maximumswitching time for which the machine is juststable.)

4. The machine was initially operating at a leadingpower factor.

8

In this instance, “worst” swing refers to the im-pedance locus which comes closest to the relaycharacteristic.

To illustrate the extent of the swings, Fig. 8shows the impedance swing locii as viewed fromthe terminals of the tandem compound generatorused in the previous discussion.

I I .-XI 1 2 3 4

PER UNIT IMPEDANCE

Fig. 8. Stable transient swings - tandem-compound

generator.

This figure gives the impedance locii for threevalues of system impedance and for two machineloadings: full load - unity power factor; full load- .95 leading power factor. In all cases, the voltageregulator was out of service and critical switchingtimes were used. The point L indicates the initialload impedance; point S the short circuit impedance(S = XT in this case), and point R the apparent im-pedance the instant the fault is cleared. The changefrom L to S and from S to R is instantaneous.

Curves A and 8 show the impedance locii for thecase of the machine operating at full load - unitypower factor. For the .2 system, the impedancelocus swings up and away from the relay character-istic. The swing makes several oscillations beforesettling down to the initial load point. For the .05system, the impedance locus swings closer to therelay characteristic and actually makes a more ex-tensive excursion than indicated in the diagram. Inthis case the impedance locus will cross the (-X)axis at 6.0 per unit and swing into the -R regionbefore returning to the initial load point.

Curve C is for the case where the machine isoperating at a .95 leading power factor. As shownin this diagram, it is possible for a stable swing toenter the relay characteristic. In this instance, theimpedance locus enters the large relay setting andstays inside the relay characteristic for 0.3 seconds.It should be emphasized that this swing was due toclearing a fault at the critical switching time whichwas 0.18 seconds in this case. For faster clearingtimes, less leading power factors and for unity orlagging power factors, the transient swings remainedoutside the relay characteristic.

Another point for consideration in these swingcurves is point R, the apparent impedance after thefault is cleared. As shown on thisdiagram, the lowerthe system impedance, the closer this point comesto the relay characteristic. While this point cancome close to the relay characteristic, it did notenter the relay circle in any of the cases studied.

With no regulator in service or with a slow re-sponse regulator, the point R will invariably appearbelow the R axis. This is due to the fact that whenthe fault is cleared, the generator will be operatingat a higher angle on the power-angle curve andtherefore the power output will be above 1.0 perunit. For loading conditions around unity powerfactor the machine internal voltage will be lessthan 1 .O per unit and therefore this power transferwill be accompanied by a VAR transfer from thesystem into the machine. For example, for the .05system, the power, VARS and voltage at the ma-chine terminals at the instant the fault is cleared isP = 1.6 per unit, Q = .8 per unit, Vt = .71 per unit.

The use of a fast responsevoltage regulator wouldbe beneficial since it tends to drive the point Rand the impedance locus away from the relaycharacteristic.

It should be noted that while the above discus-sion was limited to tandem generators, cross-compound units will have almost identical imped-ance swing characteristics.

Unstable Swings

With the increasing concern about possible gen-erator loss of synchronism, a number of utilitieshave proposed to use the loss of excitation relay todetect this condition. As an adjunct to the preced-ing study, a number of cases were run to determineif the loss of excitation relay would indeed detect a

9

loss of synchronism under all conditions. Both tan-dem and cross-compound generators were consid-ered and the impedance locus was determined as afunction of system impedance. The effect of volt-age regulators was also considered.

The results of the study are summarized in Figs.9 and 10. Figure 9 shows the impedance locii for atandem machine connected to a 0.2 and 0.4 system.Curve A gives the impedance locus for a machinewithout voltage regulator connected to a .2 system.As shown, the impedance locus will enter the relaysetting which is proportional to synchronous react-ance and the relay may trip for this condition.However, with a voltage regulator in service, theimpedance locus, Curve B, increases in diameterand just barely enters the relay characteristic.Actually, on subsequent swings the locus increasesin diameter and remains outside the relay character-istic. These curves also apply for a cross-compoundgenerator connected to .05 and .1 systems.

- /-X 1 2

PER UNIT IMPEDANCE

Fig. 9. Loss-of-synchronism characteristics for a tandem-

compound generator.

Curve C shows the impedance locus, withoutregulator, for a tandem machine connected to a .4system. This locus remains outside the circle andgets larger when a voltage regulator is used.

Figure 10 shows the impedance locii for a cross-compound generator connected to either a .2 or a.4 system. It is apparent that neither the low pres-sure or the high pressure units would detect thisswing.

10

Fig. 10. Loss-of-synchronism characteristics for a cross.

compound generator.

The obvious conclusion istion relay can not be reliedsynchronism and thereforerelaying should be used.

that the loss of excita-on to detect a loss ofsome other form of

PERFORMANCE DURING LOWFREQUENCY DISTURBANCES

During the major disturbances in the Northeastof a few years ago, a number of generators weretripped from the systems by the loss of excitationrelay. At the time, it was not possible to pinpointthe cause of these tripouts because of the lack ofrecorded data on system and generator conditions.However, a post-mortem investigation revealedthe following:

1. The tripouts occurred many minutes after thedisturbance started.

2. System frequency was low.

3. The generators were either initially on manualcontrol or had been switched to manual controlafter the voltage regulators had exceeded thetime limit at ceiling operation.

4. All of the generators had excitation systemswhose output was a function of frequency (forexample, shaft driven exciters).

On the basis of this information, it was possibleto show that the relays had not operated incorrectlyor unnecessarily, but in effect had detected either aloss of excitation or a loss of synchronism. A qua-litative analysis indicated that the exciter charac-teristics as a function of frequency (speed) couldcause a loss of synchronism or practically a com-plete loss of excitation. To illustrate this pointconsider the exciter output characteristic shown inFig. 11.

This figure shows voltage output as a function ofspeed (or frequency). At normal frequency and onmanual control, the exciter will be operating onthe 1.0 per unit curve and at the point where therheostat line intersects this curve. As the speed(and frequency) decreases, the output of theexciter will decrease and at some speed the excitersaturation curve will become tangent to the rheo-stat line which remains fixed in position. The pointof tangency will be at zero armature volts which ofcourse would mean a collapse of the generator fieldvoltage and a complete loss of excitation. Even ifthe field voltage does not collapse immediately, thegradual decrease in exciter voltage could cause themachine to pull out of step.

t?i37 600

i 5002

6 00>4

2> 3 0 0

tg 200a

2 1 0 0I -

iw 0

10 20 30 40 50 60 70 80 90

EXCITER FIELD CURRENT -AMPERES

Fig. 11. Typical saturation curve for a 500~volt, shaft-

driven exciter.

To verify the conclusions reached in the qualita-tive analysis, a limited study was made to deter-mine quantitatively the performance of a generator

during this type of disturbance. For purposes ofthis study it was assumed that a 475 MVA tandemcompound generator was connected to a systemten times larger, 4750 MVA. It was further assumedthat the machine was initially fully loaded, it wason manual control and that the exciter output wasa function of speed (frequency). The disturbancewas initiated by sudden increase in load and theimpedance locus as viewed at the generator term-inals was determined as a function of system im-pedance. The results of this study are shown inFig. 12 for two values of system impedance Zsys =.2 and .4. Curve A gives the locus for a .2 systemwhile curve B is for the .4 system. For system im-pedances below .2, the impedance locus followsmuch the same pattern as for the .2 system.

PER UNIT IMPEDANCE

Fig. 12. Impedance locii during an underfrequency disturb-ance-tandem-compound generator.

The impedance locii shown are actually loss ofsynchronism characteristics which would causerelay operation. For the exciter characteristic used,a 10 to 15% reduction in excitation voltage causedthe generators to pull out of step. With the .4 sys-tem the generator lost synchronism at approxi-mately 58 Hz while with the .2 system, synchro-nism was lost at approximately 57 Hz.

It should be noted that at reduced frequencythe relay characteristic will shift into the thirdquadrant and the relay reach and offset will beslightly reduced. At 57 Hz, the angle of maximum

11

torque is at -105o, the offset is reduced 5% andcircle diameter is reduced 10%. However, evenwith this reduction and shift in characteristic, therelay would still operate for the impedance lociishown.

The above results would indicate that the phe-nomenon involved is essentially one of instability.The fact that the loss of excitation relay can detectan unstable condition was and still is considered adesirable operation. In the case of the major North-east disturbances none of the generators involvedhad loss of synchronism protection and trippingby the loss of excitation relay in all probabilityprevented machine damage. Moreover, it should benoted that the 0.2 impedance locus (curve A inFig. 12) covers a limited area and therefore maynot be detected by some conventional loss of syn-chronism relays.

While this has been a limited study, it might benoted these results substantiate some of the post-mortem data that indicated the loss of excitationrelay had tripped machines at or near 57 Hz duringthe major disturbances.

GENERATOR PROTECTIONCONSIDERATIONS

While the preceding data is based on a studythat, of necessity, has considered a limited numberof generator and system parameters, several conclu-sions can be drawn with regard to loss of excitationprotection:

1. It is readily apparent that a loss of excitationcan be damaging to the generator as well as de-trimental to the overall operation of the system.Therefore, loss of excitation protection shouldbe provided on all types of generators.

2. To detect a loss of excitation with any machineloading, the relay characteristic should be setwith a circle diameter equal to direct axis syn-chronous reactance (Xd) of the generator.

3. The offset mho loss of excitation relay can de-tect a generator loss of synchronism for somesystem conditions. However, the relay will notdetect a loss of synchronism under all systemconditions and therefore separate loss of syn-chronism relaying should be provided to protectthe generator.

12

4. Consideration should be given to the effect ofstable swings on relay performance.

With regard to the last point, it should be notedthat whether or not a stable swing will enter therelay characteristic is a function of generator load-ing (magnitude and power factor), generator andvoltage regulator characteristics, and system im-pedance. The effect of these parameters on relayperformance should be evaluated by the study ofa specific generator and system.

Another factor of concern to some users is theperformance of the voltage regulator when operat-ing on the underexcited limit. There is apprehen-sion that the regulator will “undershoot” whiletrying to maintain the limit and thereby cause amomentary excursion of the apparent impedanceinto the relay characteristic. While this has notbeen a widespread problem, users have reportedthat this possibility exists for some types ofregulators.

The selection and application of loss of excita-tion protection requires the consideration of twofactors:

1. Effect of stable swings,

2. Voltage regulator performance.

If an evaluation of these factors indicates that un-desired operations will not occur, then a single off-set mho characteristic should suffice to provideprotection. The relay would be set with an offsetequal to one-half the direct axis transient reactance(X’d/2) and the circle diameter equal to direct axissynchronous reactance (Xd) as shown in Fig. 1.This setting will detect a loss of excitation due toan open or shorted field circuit from any initialgenerator loading. Aside from the small time delayincorporated in the relay, no additional externaltime delay should be used because of the possibleadverse effects on the machine and/or system. Itshould be noted that the time to damage for largeconductor-cooled machines is considerably lessthan that for conventionally cooled machines.

On the other hand, if stable swings or voltageregulator performance are a concern, undesiredtripping can be avoided by the use of two relaycharacteristics set as shown in Fig. 13. One relaywould be set with a diameter equal to 1.0 per unitimpedance on the machine base and this relay

should be permitted to trip without any addedinternal time delay. This unit will provide fast pro-tection for a loss of excitation with high initial ma-chine loadings, the more severe condition in termsof possible machine damage and adverse effects onthe system. The 1.0 per unit impedance is an arbi-trary value that establishes a circle that will provideprotection for machine loadings in the range of 30to 100 percent.

43(I

+ OFFSET =

. DIAMETER =

-x 1I

I

Fig. 13. Generator protection using two loss-of-excitation

relays.

The second relay should be set with a diameterequal to direct axis synchronous reactance (Xd)and some external time delay should be used toride over the transient conditions that might causeundesirable operation. This setting will detect aloss of excitation when a generator is lightly loaded,a less severe condition. Both relays would be setwith an offset equal to one-half direct axis tran-sient reactance (X’d/2).

This combination of relays will detect a loss ofexcitation due to an open or shorted field circuitfrom any initial generator loading and providesmaximum security against undesired operations.

The amount of time delay used with the largesetting should be the minimum time required toride over transient conditions. A time delay of 0.5or 0.6 seconds appears to be sufficient to ride overstable transient swings. While there is no data avail-able on the transient performance of voltage regu-

lators, it would appear that a 1 to 3 second externaltime delay should prevent undesired operation dueto voltage regulator undershoot.

In either case, when selecting a time delay theuser should determine the effect of the time delayon possible generator damage and on the overalloperation of the system. It should be noted thateven in the case of a lightly loaded generator, aloss of excitation can cause a considerable VARdrain from the system (up to .5 or 0.6 per unit onmachine MVA base). A prolonged VAR drain maycause the tripping of transmission lines and generalsystem instability.

In conclusion, it should be emphasized thatthere is need for users to study the effects of gen-erator loss of excitation on system operation andto evaluate the performance of the loss of excita-tion protection for each generator. In the gener-alized study presented here it was not possible toconsider the effects of all combinations of gener-ator designs, voltage regulator characteristics, sys-tem parameters or the interaction effects of theother generators. These effects can only be com-pletely determined by the study of a generatorconnected to a specific system.

ACKNOWLEDGEMENTS

Grateful acknowledgement is extended to C.Concordia for his guidance and to E. H. Lenfestfor his computer study efforts.

REFERENCES

(1) C. R. Mason, “A New Loss of ExcitationRelay for Synchronous Generators,” AIEETrans., vol. 68, pp. 1240-l 245, 1949.

(2) R. P. Schulz, W. D. Jones, and D. N. Ewart,“Dynamic Models of Turbine GeneratorsDerived from Solid Rotor Equivalent Cir-cuits,” paper T 72 515-5, presented at the1972 IEEE Summer Power Meeting, SanFrancisco, Cal., July 9-14, 1972.

(3) C. Concordia and M. Temoshok, “Resyn-chronizing of Generators,” AIEE Trans.,vol. 66, pp. 1512-1518, 1947.

(4) A. W. Walton, “Loss of Excitation Expe-rience on Oklahoma Gas and Electric Co.System,” DP paper presented at AIEE South-west District Meeting, April 1952.

13

GE Protectionand Control

M e t e r e n d C o n t r o l B u s i n e s s D e p a r t m e n tG e n e r e l E l e c t r i c C o m p a n y205 Great Val ley ParkwayMalvern , PA 1935512/89 (2000)