Sie sind auf Seite 1von 10



Industrial and Cogeneration Protection Problems Requiring Simulation

Abstract-The procedures outlined in relay instruction manuals for setting out-of-step, loss-of-excitation, reverse power, external fault, and similar protective relays cover typical situations, and may not be appropriate in all circumstances. Cogenerators, small power producers, and industrial generation can face long fault clearing times, high fault levels, lower fault levels, high voltage, and a range of other unusual conditions in the adjoining utility system. Relay setting procedures must address all possible operating conditions, and the dynamic response of the system if nuisance tripping is to be minimized and the operating process and equipment are to receive maximum protection from faults. Some critical relay applications and the simulations required to select the best settings are outlined.

INTRODUCTION HE ROUTINE protection needs of cogeneration and industrial plant generation have been well documented [ 11-[3]. However, the increasing number of nonutility owned generators being interconnected with utility systems reveals protection problems that were not recognized, or did not exist, in the past. Many of the problems are associated with the application of sizeable generators to utility distribution circuits and lower voltage subtransmission lines. Others are the result of very small generators being placed on relatively stiff subtransmission and distribution systems. The generator itself may impose special requirements on protective relays. For example, bus-fed excitation systems may allow fault current to drop below overcurrent relay pickup before time out can occur. Both reliability of service and protection of the generator and associated equipment require close attention to these problems. Utilities generally accept no responsibility for the protection of cogeneration and industrial equipment. It is therefore essential that owners of generation equipment, particularly plants above 5-10 MVA, evaluate protection requirements and plan for proper application and settings. OUTOF STEP PROTECTION Generators on distribution systems and 34.5-kV systems protected by overcurrent relays can be exposed to fault clearing times of 0.5-1 s or more. Multiphase faults of this
Paper ICPSD 87-31, approved by the Power Systems Protection Committee of the IEEE Industry Applications Society for presentation at the 1987 Industry Applications Society Annual Meeting, Atlanta, GA, October 18-23. Manuscript released for publication October 17, 1988. H. K. Clark is with Power Technologies, Inc., 1 Surrogate Plaza, Suite 340B, Roseville, CA 95678. J. W. Feltes is with Power Technologies, Inc., P.O. Box 1058, Schenectady, NY 12301-1058. IEEE Log Number 8927836.

duration are likely to cause instability, as are ground faults that last more than about 1 s. Instability results in large swings in generator power and potentially damaging shaft torque excursions. Also, if the fault level at the generator location is high or the generator is small, generator current during poleslipping can be well above the current the generator would deliver to a three-phase fault. Since the generator windings are braced only for three-phase fault levels, these high currents can loosen or abrade stator windings. Winding and shaft damage from pole-slipping is cumulative and may reach catastrophic levels only after several events. In addition, instability will usually cause a plant outage. Little can be done to reduce the risk of instability other than to reduce the fault clearing times. Reducing the fault clearing time will usually require costly protection changes in the utility system, and possibly in the industrial plant and neighboring plants as well (the utility relay time settings may be long simply to be selective with plant relay settings). Though relay improvements should be made if practical, the high cost of such changes is usually not justified by the reduction in the number of unstable events and reduced risk of turbine-generator damage. The alternative is to install relays that will protect the generator, insofar as practical, from damage when it becomes out-of-step. It is not practical with present relay technology to trip a generator before a pole slip occurs. Relays can, however, limit the exposure to one cycle of the pole-slip power swing. Out-of-step trip relays, ANSI device 78, respond to the change in apparent impedance at the terminals of the generator as pole slipping occurs. The relay effectively calculates an impedance from the ratio of voltage to current:

R + j X = V/ZLO
where 0 is the angle between the voltage and current. As the generator pulls out of step, the voltage falls and the current rises. The active power out of the generator also drops and the reactive power increases. The result is an impedance that falls to some minimum and then increases as the generator completes one slip cycle. Fig. 1 shows a typical apparent impedance plot for a small generator forced out of step by a long duration fault. In this case the electrical center, the point where the apparent impedance crosses the generatorsystem impedance line falls within the generator, indicating that generator current at this point in time will be well above generator three-phase fault current level. Loss-of-synchronism relays detect an out-of-step condition from the path of the apparent impedance and its rate of change.

0093-9994/89/0700-0766$01 .OO 0 1989 IEEE



Fig. 1 shows the most frequently used out-of-step relay characteristic. Each of the two angle impedance relay units closes or opens its contacts as the apparent impedance moves across the R-X diagram. If the two units change contact position in succession, the generator is tripped. If just one changes contact position, as would occur for a fault near the generator, no tripping occurs. One of the problems with loss of synchronism of small generators exposed to long fault clearing times is the high rotor speed that is reached. The generator may reach 62 or 63 Hz by the time the fault is removed. If the fault causes motor starter contactors to drop open, the loss of load can further accelerate the generator. The generator may thus slip poles at a rate of two or three per second. The out-of-step relay may miss the out-of-step condition if the pole slip rate is too high. Angle-impedance units settings other than those recommended by the manufacturer may be required. A mho unit is sometimes used to limit the reach of the outof-step relay so it will not trip for an out-of-step condition elsewhere in the network. The mho unit is shown by the dashed line, in Fig. 1. The out-of-step relay will trip only if the apparent impedance enters the mho unit before it crosses the two angle-impedance units. The apparent impedance trajectory is not always a smooth curve more or less perpendicular to the system impedance, as shown in Fig. 1. Nearby generators, synchronous motors, and induction motors can cause the apparent impedance to wander about the R-X diagram as loss of sychronism occurs. The mho unit may prevent operation in such cases. A modification of the normal mho blocking logic is necessary in such cases. Experience with simulation of a large number of small

generator situations has shown that out-of-step protection sometimes must be modified to be effective and can almost always by given specific settings that will improve protection reliability. Ensuring that a generator will be tripped on the first pole slip when instability occurs can significantly increase generator life. LOSS-OF-EXCITATION PROTECTION There are two main concerns for loss-of excitation protection. One is to ensure that the relay will trip the turbine generator upon an actual loss of field quickly enough to avoid machine damage or adverse effects on the system. The second is to ensure that the relay will not trip the turbine generator unnecessarily for stable system swings or transient disturbances which will not harm the unit. Loss-of-excitation protection is important from a system standpoint. Two things happen when loss of excitation occurs. The turbine generator will accelerate to a higher speed and operate as an induction generator at 40-60 percent of rated power, with resultant high amortisseur currents in the rotor iron. Secondly, exciting current (reactive power) will flow from the system into the generator. Also, though the machine will be running without excitation, it still has saliency in the rotor structure, and this will cause large power swings into and out of the generator, superimposed on the induction power being produced. A generator is not likely to be damaged by loss of excitation so long as the condition does not persist for more than 5-10 s. However, the heavy reactive load on the system can seriously reduce voltages, possibly causing collapse, instability, loss of load, tie tripping, and related problems. Because of these possible consequences, it is



-4 0

0 D

z D


Fig. 2 .

Loss-d-excitation relay characteristic and typical response under different load levels.

essential that a generator be tripped quickly when loss of excitation occurs. Loss-of-excitation protection is similar to out-of-step protection in several ways. Like out-of-step relays, loss-ofexcitation relays (ANSI device 40) are distance type and operate from the apparent impedance viewed from the terminals of the generator. However, the characteristic of these relays sits below the R axis to detect heavy reactive power flow into the generator as shown in Fig. 2. The trajectory of the apparent impedance upon loss of excitation is shown in this figure for two levels of generator loading, The apparent impedance trajectory is dependent on the generator characteristics and the electrical system the generator is tied to. The steady-state operating point of the generator will always be on the right side of the impedance plot (power out of the unit) and usually in the upper quadrant (reactive power out of the unit). Upon loss of field, the flux in the generator will decay and the reactive power out of the unit will decrease. The impedance will move into the lower right quadrant, entering the circle and tripping the unit. Curves A and B of Fig. 2 show the simulated response of a 90-MVA gas turbine for loss of excitation at two power levels. Curve A represents an initial operating point of full power and rated power factor. The impedance enters the circle in about 1.4 s. The loss of excitation at 25-percent load is shown by curve B. Due to the lower initial loading, the impedance starts off the plot in the upper right quadrant. It takes about 4.2 s for the impedance to enter the circle. The circle is representative of typical settings: an offset of one-half of the transient L ; and a diameter of the synchronous reactance Xd. reactance X

For very lightly loaded units (typically below 15-20 percent of rated), it is possible that the typical circle shown will not be entered upon loss of field. The impedance will curve in approaching the circle but stop outside the circle as a steadystate operating point is reached. At these low loadings, the generator can remain in synchronism and continues to run without the field as a synchronous reluctance generator. The generator will be absorbing a large amount of reactive power. The relay circle can be expanded to contain these operating points to minimize the possibility of end turn overheating if operation frequently occurs at low power. A large circle increases the chances of nuisance trips as discussed later. At very low power output, the loss-of-excitation relay will not operate, and the operator will be responsible for tripping the unit. Other relays can be used to cover these extreme cases. Improper trips of the loss-of-excitation relays can be caused by events that cause a rapid increase in network voltage and corresponding large change in the reactive power of the generators. If the generator reactive power goes to a large negative value (vars into the machine) the relay will interpret this as a loss of field. The usual causes for such a var swing are the switching of a large capacitor bank, loss of a large load, or the opening of the far end of an EHV line, leaving the line charging current flowing into a generating unit. To keep the generator from being tripped on transient swings of reactive power into the unit, the circle size can be reduced or a timer can be installed so that the impedance must remain in the circle for a set time period, usually 0.5-1 .O s. Reduction of the circle size will increase relay operating time and reduces relay sensitivity at low power loadings and to partial loss of field



such as due to shorted turns in the field winding. A larger circle with a timer and a smaller circle with an instantaneous trip can be used so that the trip occurs quickly for full load conditions, while tripping is delayed for less harmful light load conditions, and nuisance tripping on swings is avoided. Simulations provide the only means of selecting the timing delay and appropriate relay settings as the response of the voltage regulator must be modeled. Loss-of-excitation relays rarely will suffice for out-of-step protection. Out-of-step protection will usually appear as a swing from right to left on the impedance plot, similar in form to a loss of excitation. However, the swing can go through the area above the relay circles depending on where the electrical center of the system is. In this case, the out-of-step condition would not be seen, and the relay will not trip until the circle is entered on subsequent swings, possibly several slip cycles later. The use of a timer makes operation even more unlikely as the impedance may swing through the circle faster than the time setting. REVERSE POWER PROTECTION Reverse power relays (ANSI device 32) are used to trip turbine generators to prevent motoring of the generator. Generator motoring would occur if, for example, the turbine is tripped but the generator breaker is not. Generator motoring would cause damage to the prime mover such as turbine overheating in steam turbines and cavitation due to low flow in hydro turbines. Reverse power relays are sometimes used as backup to other relays in this situation, such as steam temperature in a steam turbine or exhaust hood temperature in a gas turbine. Gas turbines have a motoring power use that can reach 50120 percent of full load power. However there are mechanical components which may be damaged by reverse power flows as low as 5-10 percent. A relay setting of 3-5 percent of rated power is normally adequate. For steam turbines, reverse power relays must be set at a very low level due to the small amount of power drawn by the motoring generator (typically 2-3 percent). A typical setting may be one-half percent of rated power. To prevent operation for system disturbances, it is necessary to use a time delay as part of the relay. Simulation of system disturbances causing a negative excursion in generator power can help to determine the delay necessary, although a delay of a few seconds will normally be adequate. However, for industrial sites with more than one generator, reverse power settings can be critical in the case of separation from the utility. The ability of the plant to remain in service following an islanding condition depends on the response of the turbine governors, the amount of connected load, and the initial loadings of the turbines. In the case where generation exceeds load, the governors will respond to the overfrequency and decrease the output of the generators. If the governors have different droop settings, the generators have different initial loadings, or the turbines have different capabilities, some units may trip on reverse power. In the seconds following islanding, especially where a fault was the cause of the islanding, power oscillations can occur among generators and motors in the plant. However, as for

external disturbances, these short swings of power into the units will not usually cause unit trip with a properly selected time delay. Of more concern is the steady-state operating point reached after these initial oscillations diminish. If the governors will operate on a droop characteristic, the steady-state operating condition can be predicted graphically or by simple solution of simultaneous equations for the straight line droop curves ( Y = m X + b). Fig. 3(a) shows a typical droop curve. The ordinate is the system frequency or unit speed, while the abscissa is the generator power in MW. For a four-percent droop, the unit speed will proceed from unity (60 Hz) to 1.04 upon a drop from full load to no load. The droop characteristic from a second turbine generator can be plotted on the same graph (using the same scales) to compare the power output of the two at any given speed. For instance upon separation from the network, the two will be operating at the same speed or frequency, and will share available load as indicated by a horizontal straight line just above 1.0 and crossing the two droop curves where the sum of the power output of the two units equals the available load. Fig. 3(b) shows an example of this. A gas turbine curve will extend to the left of the origin since a gas turbine (GT) can absorb some power before combustion instability or other problems force a shutdown. Properly operating GT controls typically limit absorption to about 5-10 percent of rating, in which case the GT droop curve rises vertically at this point on the left of the speed axis. A steam turbine droop curve also rises vertically to the left of the speed axis, doing so at about two percent of unit rating. This absorbtion occurs with steam valves closed, and with windage and friction losses absorbing the power. The horizontal line that provides a sum of unit outputs equal to the available load may include a negative output on one or more of the units. In this case, the negative output units will be motoring and may be subject to possible trip by reverse power relays (i.e., depending on the threshold of the reverse power relay). The initial operating point of two or more turbines may be anywhere between zero or full load. Adjustment of the governor set point to select a certain unit power output when operating into a grid at 60 Hz is represented by shifting the droop characteristic vertically up or down until a horizontal line at 60 Hz crosses the droop characteristic at the specified power output. The initial output of each unit is set in the same manner. Upon loss of the connection to the grid, the units will share the available load at the frequency where the sum of their outputs would equal the load. Fig. 3(c) is an example of two units initially operating at other than full load. Fig. 3(c) demonstrates that two units with the same droop but different initial operating points (i.e., in percent of full load) will not share a reduced load in proportion to their fullload ratings or their initial outputs, if load is not sufficient to keep both units at positive output. One will motor and may be tripped from the system. The initial operating points are thus shown to be very important in determining whether motoring will cause a unit trip following loss of the connection to the grid. The adjustment of droop and the linearity of the droop characteristic are not nearly as precise as the lines on a plot of



1 . 0 4

( 6 0 HZI





LOAD f LINE (LOAD (I 1.0 P1 + P A2 11 )



p1 1 p1 INITIAL














droop curves might indicate. The condition of a unit may also affect its droop. Hence droop is likely to be quite variable from unit to unit, even though settings are intended to achieve similar droops. The initial operating point is also a major factor and will vary from time to time depending on unit condition (hours since last overhaul, fuel in use, etc.). If droop could be easily and precisely adjusted, the droop could be set to compensate for different loading levels on individual units so that all units would approach zero load at the same frequency (Fig. 3(d)). However, the droop adjustment range is limited, and the adjustment is not precise or easy. In cogeneration systems with both gas turbine and steam turbines (for example, steam extraction turbines with heat recovery steam generators using the gas turbine exhaust gases) coordination of the governor controls following islanding can be quite complex due to their different capabilities and their different operating characteristics. The first step in this coordination is to determine the amount of generatiodoad imbalance which may occur upon separation from the utility. Since separation will usually occur due to a fault, the loss of load due to motor contactor dropout must be taken into account. Typical generation dispatches must be studied. The worst case is where the steam turbines are at a low output level. As shown in Fig. 3(c), units at lower initial per unit loadings are more likely to be forced to a negative loading and the steam turbines have little negative capability. If analysis shows that unit trip on reverse power will occur, and the unit trip will cause disruption of the steam system, then corrective measures can be studied. These include isolation of some generation onto its auxiliaries to achieve acceptable load sharing, use of the gas turbine to control frequency back to 60 Hz (one GT with isochronous governor and the other GTs following proportionately), and the restoration of load in the plant by the use of circuitry to reclose motor contactors after a voltage dip (staggered to minimize voltage problems). Reverse power relays are also used to alarm or trip the utility tie when power flows into the utility system. This is useful, of course, only for cogenerators where the electrical load is normally much larger than the generated power. In these cases, power into the utility indicates an abnormal condition, such as the cogenerator having been islanded with utility load. OVERCURRENT PROTECTION The fault current supplied by a generator falls with time. Fig. 4 shows the drop in current for a fault in an industrial plant with a generator and a large amount of motor load. The generator is equipped with a rotating exciter. A static exciter with terminal current boost would be similar. The oscillatory variation in current beyond about 0.25 s is due to synchronous motors going out of step. These motors should trip quickly, leaving the utility and/or the industrial generator as the only sources. The first few cycles of the plot reflect contributions from induction motors as well as subtransient currents from synchronous motors. The torque in inverse time-overcurrent relays drops along with the current, making it difficult to determine operating





Fig. 3. (a) Typical droops curve. (b) Two units with same initial loading and same droop will share load change equally. (c) Two units with same droop but different initial operating points will not share load changes equally when large load occurs. (d) Adjustment of droop to get equal load sharing among units.


77 1




Fig. 5 .




Fault current sources include plant motors as well as utility and generator.

2- 10 cycles of fault current depending on the proximity of the fault to the bus. The synchronous motor fault current will decay to zero over a period of 1-2 s if exciters are static, or to a low level over that period if exciters are rotating type. The generator fault current will drop to 200-300 percent of generator rated current over 1-2 s if its exciter is a rotating type or is static with terminal current boost. In addition, dropout of contactors in motor starters will remove fault sources u.u 0.5 and load. Induction motors on circuit breakers may stall, T I X (SEUXTS) thereby drawing reactive power that will reduce fault current. Decay in fault current must be known to set time-overcurrent relays Fig. 4. Synchronous motors may pull out of step, causing oscillations properly. in fault current, or a drop in fault current as they are tripped by time for any given pickup and time lever setting. The problem power factor relays. Plant generators may also lose synchrois compounded in voltage-restrained overcurrent relays used nism with the utility or with each other, thus causing further for generator external fault backup protection since the variations in fault current, and possibly tripping or causing trip response of these relays is also a function of voltage which of the utility tie by out-of-step relays. drops with time along with the fault current. VoltageBecause of the above variables, the current that inverse time controlled relays used in the same application may also be overcurrent relays will see is not easily predicted by hand affected in that they do not begin operating until the voltage calculations. A simulation taking into account the electrical falls below a set level, usually 80 percent. and mechanical dynamics of each component can give Relay operating times for these conditions have historically excellent predictions of fault current. Contactor drop-out been predicted by application of a Simpsons rule integration characteristics, undervoltage relays on large motors, and procedure. The current is assumed to drop in steps, and the similar devices can also be included in the simulation model to relay disk rotation at each successive step added until the improve accuracy of the fault current prediction. contacts close. This is a very tedious procedure, and is an The simulation can be repeated for each of the several approximation at best. The actual operating environment is critical fault locations, and the fault current at each overcurquite complex, making prediction of the actual fault current rent device monitored and plotted for each case. It is also over the several second period difficult. practical to include the overcurrent devices and proposed Fig. 5 shows an example of a system with four fault current settings in the simulation model to demonstrate operating sources, and a number of requirements: times for various fault locations and operating condition scenarios. The generator relays must be selective with the tie relays Simulations in time-overcurrent relay coordination work are for faults in the utility system. particularly useful where static generator excitation systems The generator relays must be selective with feeder relays. without terminal current boost are to be used. Without The feeder and generator relays must be able to detect all terminal current boost, the generator excitation voltage falls faults within their zone of protection under all credible immediately to zero when a nearby multiphase fault occurs, or conditions. to a small fraction of its ceiling value when the fault is remote. The first-cycle current that instantaneous relays will see is The result is that generator fault current will also fall to zero quite predictable since all fault current sources will remain in within 1-2 s for close-in faults or will drop to a fraction of place, machine internal flux will not change appreciably, and rated generator current in the same time for remote faults. inertia will prevent any dynamic changes in the system. Ensuring that overcurrent relays will detect all faults in their However, fault current in the following several seconds is primary and backup zones may become the primary problem another matter. in this situation, with selectivity a secondary consideration. Fault current from the utility will be fairly uniform if utility That is, generator relays may have to be so fast and sensitive to generators are some distance away and there is no generation ensure successful operation that any hope of selectivity must in neighboring plants. The induction motors will deliver just be abandoned.



Fig. 7. Radial reclosing results in unacceptably high shaft torque.



reclosing can occur. Options to do this include: Quick and sensitive fault protection to trip the tie for faults that will cause trip and reclose in the utility; under/over frequency relaying; underlover voltage relaying; transfer trip from the utility substation.

(b) Fig. 6. (a) Reclosing of radial lines is extremely hazardous to plant generators and motors. (b) Reclosing on lines that uill not isolate plant can also be hazardous.

Several manufacturers now offer as standard equipment static excitation systems without terminal current boost. It is up to the industrial or cogenerator to recognize the problem this can cause and specify the terminal current boost option. SHAFT-TORQUES FROM UTILITY RECLOSING Two types of reclosing need to be recognized. The first might be called radial reclosing (Fig. 6(a)) and is widely recognized. A momentary opening of the utility tie will allow the generator to accelerate or decelerate and be out-of-phase with the utility when reclosing occurs. The second might be called passive reclosing. In this case the line opening and reclosing does not occur on a branch serving the plant directly, but on a nearby branch which can change power flows significantly on the branch serving the plant (Fig. 6(b)). This reclosing situation is not as dangerous as radial reclosing but can present significant hazards. Radial Reclosing

The first option is the least costly and is suitable if all faults that can cause trip and reclose can be detected at the plant tie. To be successful, fault protection at the plant must be quicker and more sensitive than the utility protection that initiates reclosing. To meet this requirement the tie protection may not be fully selective with utility protection, thus inviting occasional nuisance tripping. Underfrequency relays have long been used successfully to remove synchronous motors from the line during the dead time between trip and reclose. However, motors will always slow when the source of power is lost. Over and under frequency relays will do the job for a turbine-generator only if there is a measureable frequency excursion upon trip of the utility line. The frequency excursion will depend on: normal power flow across the utility tie, turbine-governor operating point and maximum capability turbine-governor response time, the load that will be dropped by the fault (in the plant and in neighboring customers), the reclosing delay (reclosing must not be instantaneous, and 1 or 2 s delay may be required, the voltage in the island formed by the utility line trip.

The frequency excursion is clearly difficult to predict. It is Winding and shaft damage can result from the currents and shaft torques caused by radial reclosing. Torques of several also clear that a frequency-based scheme cannot be 100times rated can occur on the turbine-generator shaft [4], [ 5 ] . percent reliable. If frequency relays are to be used, enough Fig. 7 shows the electrical and mechanical torque oscillations analysis must be done to show that the probability of relays caused by a single line-to-ground fault with successful reclose detecting the open condition is acceptable. Voltage relays are similar to frequency relays in that it is from [5]. The large torque oscillations and resulting shaft fatigue can lead to premature shaft failure. Radial reclosing difficult to ensure that reasonable voltage relay settings will simply cannot be tolerated. The potential for serious generator reliably detect the open condition. A combination of voltage damage is so high that major investment to avoid it is and frequency relays can be used to increase the probability of warranted. One solution is to open the plant utility tie before successful detection of an open line, but when the tie flow is



low, or the neighboring load is about equal to the normal tie 1 I I 1 I I I I I flow to the utility, both frequency and voltage are likely to change little when the utility line is opened. Transfer trip from the utility is costly but is probably the * _---__ VOLTAGE ( J V o 2 + V b 2 + V c Z ) ___ : most effective means of ensuring that the plant tie will be open I: *Py-Y I ,/'& P U A V when the line is reclosed. The transfer trip must be initiated " L from each location where a breaker operation can isolate the SLIP 001 p U ( S L I P ) ' plant from the main utility grid. I Dead-line-check voltage relays in the utility substation can be used to block reclosing until the plant tie has been opened (automatically or manually) or voltage in the island has otherwise collapsed. This approach can make the less reliable tie trip schemes outlined acceptable since it will block I / reclosing if the scheme fails to operate. Dead-line-check relays require a potential transformer on the line side of the breaker and this results in an excess facilities charge. Most utilities prefer to avoid "islanding" of customer generation and utility load. Utility lines energized by customer generation present a safety hazard to utility personnel and the ? * PHASE B CURRENT public and can delay restoration. The utility will thus also be interested in prompt trip of the tie when the line is opened for any reason and may insist that equipment to do so be in place. The utility requirements will usually not address the risks of out-of-phase reclosing, however. The turbine generator owner l p u I PEAK TO PEAK I 1 11 I I I I 1 I must recognize this problem, assess the risks. and make decisions as to what additional protection. if any. is warranted. 0o m 0 1500 0 2wo 0 3500 04500 TIME The prospect for successful operation of underiovervoltage and underloverfrequency relays can be assessed by simulation. Fig. 8. Motor transients following step change in supply voltage phase angle. A detailed stability model can be subjected to trip of the utility line or feeder with varying amounts of utilitj load left on the line and with varying in-plant conditions. Though the worst the shaft torsional oscillations initiated by the first event. Since case for islanding is trip of the utility breakers without a fault shaft damping is low, a series of events over a period of 10-15 on the line, most line trips will be caused b j faults. The effects s can build up severe torsional excursions. In fact, it is often of different fault types and durations can be assessed by the long exposure to the slowly decaying shaft torsional simulation. motion that initiates shaft cracking. It is not reasonable to ask a utility to disable all reclosing Passive Reclosing near a plant. However, it may be practical to reduce the Reclosing does not have to occur on a radial line to a plant to number of times reclosing is attempted or increase the time cause potentially damaging shaft torques in industrial genera- between reclosure attempts to allow shaft oscillations to decay. tors. Motor loads are similarly susceptible. Fig. 6(b) shows a Reclosures delayed by 20 s or more are unlikely to be a threat situation in which reclosing can cause a rapid and large change if each reclosure causes less than a 50-percent change in airin generator air-gap torque. Fig. 8 shows a step change in gap torque. If a system model is assembled for stability purposes, it torque of 350 percent of motor rated torque on a typical large motor hit by a 40" change in phase angle at a remote point in should be used to examine the potential impact of all reclosing the utility network. Trip of the generator breaker will cause no that will occur in the vicinity of the plant. Problems so more than a 100-percent change in generator air-gap torque, identified can be discussed with the utility in hopes of a but a generator or utility tie trip can impose a step change of negotiated change in utility reclosing practices. several hundred percent on in-plant motor loads. LOAD SHEDDING Many utilities use a rule of thumb that reclosing events Where ample low-priority load is available for shedding should not cause more than a 50-percent change in air-gap torque. Industrial turbine-generators and motors are probably when a generation shortage occurs and the frequency decay less susceptible to damage, but a similar rule of thumb has not rate is 1 Hz/s or less, underfrequency load shedding can be been developed for them. Such a rule of thumb is of only based on hand calculations [6]. However, where load shedding limited use because the shaft torque stress is not always due to must be more precise and voltage excursions are of concern, a single reclose event. A series of events such as fault-on, fault simulations are essential. Such simulations should utilize clearing, reclose into a fault, and clear the fault imposes four detailed models of turbine-generators, governors, excitation separate step changes in generator air-gap torque. If the timing systems, and, especially,plant motors. The usual stability of each event is "right," each will increase the magnitude of analysis model with some equipment detail and underfre-$





quency relay models added will be quite adequate for this purpose. Hand calculations, by necessity, must neglect spinning reserve that may be available in plant generators. When it is essential to protect critical loads with spinning reserve, the amount of reserve required, and the coordinated underfrequency relay settings to avoid unnecessary trip, will depend largely on turbine governor response characteristics. These characteristics can be taken into account only in simulations. Low voltage during faults can open contactors, thereby removing low-priority load from the system (high-priority load is presumably protected from loss during voltage dips) and may result in further voltage and frequency excursions as motors slowed by the fault reaccelerate during the loadshedding period. Loss of load caused by utility faults that immediately precede loss of the utility tie can significantly affect the load-shedding system design and can also only be properly considered in simulations. Where feeder or line trip in the utility system can leave neighboring customer load on the plant, it is essential that the tie be opened quickly to protect in-plant load. The highest set underfrequency relay can be used for this purpose. though it may lead to nuisance tripping of the tie when the utility frequency drops or a power swing in the utility network causes a short duration frequency excursion. An underfrequency relay supervised by a directional power relay can improve security of such a scheme. It is useful to include simulation cases to confirm correct operation and freedom from nuisance tripping. Where very precise load-shedding is required to protect large critical loads such as paper machines or refining units. a microprocessor based load-shedding system can be applied [ 7 ] . Such systems allow rapid precise load-shedding that takes into account spinning reserve while limiting frequency excursions to 1 Hz or less. SELF-EXCITATIOS-OVERVOLTAGES




69 kV 138 kV 230 kV 345 kV

26 kvar/mile 105 k?ar/mile 380 kvadmile 860 kvar/mile


1.0 TIME, S



Fig. 9. Self-excitation can occur when generation in islanded with capacitor banks on high-voltage lines.

Overvoltage due to self-excitation is a problem closely associated with the islanding problem discussed earlier. The utility industry has faced this problem for many years in the planning of bulk transmission systems, and while not a new problem to industrials, the increase in cogeneration has brought more attention to it. The two culprits in the problem are line charging and capacitor banks. If a generator becomes isolated with a capacitor bank that supplies more reactive power than the generator can absorb, voltages will rise. In the case of a synchronous generator there may be temporary control of voltage by the field winding in the direct axis. However, if the capacitive reactance and the generator quadrature axis reactance are at or near resonance (equal), voltage can grow without bound. Induction generators can experience the same problem, but without a field winding CONCLUSION there is not even a temporary control of voltage. The problem can arise when no capacitor banks are in the There are many tasks in engineering a safe and reliable island if the capacitance of utility overhead lines and cables is utility interconnection. Many of the protective devices can be sufficient. Typical charging kvar of utility overhead lines are selected and set according to manufacturers instructions and listed in Table I. If the charging or capacitor bank kvar information avialable in the literature. However, the proper exceeds the reactive load in the island, at least modest setting of some protective equipment is heavily dependent on

overvoltages are likely to occur upon trip of the breaker that causes the islanding. Fig. 9 shows the voltage excursion for two scenarios. Curve A shows an initial voltage excursion as charging current flows into the generator from the system, followed by restoration to a lower voltage as generator terminal voltage is returned to normal by the voltage regulator. Curve B shows an immediate and dramatic rise due to near resonance between the capacitance and the quadrature axis reactance. Selfexcitation must be prevented and voltages more than about 15 percent above normal should be avoided. Higher voltages cause transformer saturation that will generate harmonics. These harmonics can fail surge arresters and prevent proper operation of relays. Since load on the generator will significantly affect the overvoltages that occur upon formation of an island, predicting overvoltages is difficult. Simulations are the most reliable means of assessing the potential problem and determining solutions. The basic stability model with appropriate modeling detail is sufficient. The effects of transformer saturation need not be modeled since the final simulations with a solution in place must show fundamental frequency voltages that do not exceed 120 percent of normal. Solutions include moving capacitor banks, transfer trip of the plant tie ahead of the utility line trip and shunt reactors to absorb the reactive power that the generator cannot safely handle. Operating constraints may also be feasible.



the local plant and utility dynamics. For these, the only practical method available to determine these settings is dynamic simulation. Some critical relay applications and the simulations required to select the best settings were discussed.

IEEE Recommended Practice for Protection and Coordination of Industrial and Commercial Power Systems, ANSUIEEE Standard 242, 1986. Applied Protective Relrlying, Westinghouse Electric Corp., 1979. G. Potochney and L. Powell, Application of protective relays on a large industrial-Utility tie with industrial cogeneration, IEEE Trans. Ind. Appl., vol. IA-19, no. 3, pp. 461-470, May/June 1983. J. M. Undrill and L. N. Hannett, Turbine-generator impact torques in routine and fault operations, IEEE Trans. Power App. Syst., vol. PAS-98, no. 2, pp. 618-628, Mar./Apr. 1979. J. S. Joyce, T. Kulig, and D. Lambrecht, Torsional fatigue of turbinegenerator shafts caused by different electrical system faults and switching operations, IEEE Trans. Power App. Syst., vol. PAS-97, no. 5, pp. 1965-1973, Sept./Oct. 1978. H.Clark, Load shedding for industrial plants, presented at the 8th Annu. Meeting IEEE Industrial Application Society, October 8-1 I , 1973, Paper ICP-WED-PM2 725. , Microprocessor based load shedding for industrial plants, presented at the IEEE Industrial Application Society ICPS Conference, Cleveland, OH, Mar. 5-8, 1986.

Harrison K. Clark (M69-SM86) was born in Alameda, CA, on November 11, 1941 He received the B.S.E.E. In power from California State Polytechnic University in 1966. He joined General Electric as an application engineer responsible for system planning and design studies on large industrial power systems. He joined PTI as an andytical engineer in 1970 and has since accumulated experience in a wide variety of industrial power system problems, major utility transrmssion planning projects, criteria development, equipment failure analysis and blackout investigations. Mr. Clark was promoted to Senior Engineer in 1974: Manager, Utility System Performance in 1984: and Manager, Western Office in 1987.
James W. Feltes (S74-M79) was born in Belle-

vue, IA, on February 15, 1956. He received the B.S. degree in electrical engineering (with honors) from Iowa State University, Ames, in 1979. Upon graduation, he joined Power Technologies, Inc. as an Analytical Engineer in the Systems Performance, Design and Operations Department He was promoted to Senior Engineer in 1985. He is currently a member of the Utility System Performance section. At PTI he works on large-scale power system studies involving load flow and transient and dynamic stability, cogeneration studies including transient stability, short circuit, relaying, and equipment tuning, as well as many other study areas.