You are on page 1of 6

IEEE Transactions on Power Systems, Vol. 10, No.

3, August 1995

1467

Load Shedding on an Isolated System


Charles Concordia Consultant Lester H. Fink Boza Avramovic ECC, Inc. George Poullikkas Electricity Authority of Cyprus

Abstract
Underfrequenc load shedding OJFLS), in previous generations shunned gy the industry as an admission of poor planning and operation, has become a enerally accepted practice. There is by now a considerab7e body of literature on the subect, dating back to midcentury. That literature, however, deals with the subject in the context of large interconnected systems. Smaller, isolated systems, which are more vulnerable to serious disturbances, have o erating characteristics that require somewhat different gui&lines, due to their lower inertia, limited reserves, and lack of access to offsystem assistance. This paper reviews the objectives and principles of UFLS, and reports their application, to a small island power system.

links. The sheddin schedule should not require frequent modifications to fo low the continual load and system changes, as a well-designed and well distributed scheme will be sufficiently robust to accommodate such changes.

Introduction
Underfrequency load shedding FFLS) is a practice used throughout the power industry. ower systems are designed to withstand the effects of an array of credibleworst disturbances. The function of UFLS is to serve as back-up protection for the system in cases which might occur that were not covered by the design process. Fortunately, even in the emergency state, the power system is usually robust enough to enable carefully designed heuristic load shedding strategies to be effective. It is important however, that these strategies be designed on the basis of mature understanding of the charactenstics of the system involved, including system topology and dynamic characteristics of its generation and its load. A poorly designed load shedding program may be ineffective, or worse may exacerbate stresses on the transmission network leading to its cascading disruption. Over the years, however, utility experience and extensive studies on a number of systems have resulted in dependable guidelines for the design of effective load shedding programs. When plans designed on the basis of these guidelines are validated by carefully modeled simulations, they provide reliable tools for preservation of the system under severe conditions.

The major parameters that must be considered in devising an UFLS pro ram are the level and distribution of spinning reserve and the specific load sheddin arameters: the frequency thresholds, the total amount of Lad to be shed, the MW step sizes, and intentional time delays. An important characteristic of any schedule is the amount of undershedding and/or overshedding in which it is like1 to result over the range of cases that mi ht be encountere$. The response of system frequency to a oad shedding schedule will be characterized by an average value within a band of actual values pertaining to individual machines and determined by the interdifferences across the netmachine oscillations. work can result in randomness o load shed.

Frequencr

Amount o load to be shed. Since underfrequency load shedf ding is a last resort that must work if the system is to be saved from colla se, it does not pay to be timid. If the s s tem goes down &cause a particular load was not shed, tza; load oes down anyway, carrying the rest of the system with it. 8or an isolated (non-interconnected) system, maximum load subject to and scheduled for shedding should, in principle, include all load. As a practical matter, perhaps 80% should be included, some with appreciable time delay. Frequency threshold. The first step threshold should not be too close to normal frequency, to avoid tripping on severe but non-emergency frequency swings. However, frequency thresholds must be coordinated with machine protective relaying (and vice versa). Step size & number o steps. A significant consideration in f determining step sizes is the potential for excess load shedding. Frequenc steps must be far enough apart to avoid overlap of sheding due to (intentional or inherent) time delays. Althou h average excess load shed might theoretically be reducefby increasin the number (and reducing the MW size) of steps, the n d for rapid, decisive action is more important. Moreover, there are other important considerations weighing against more than a minimal number of steps. Step frequency thresholds too close together may in themselves cause overshedding: a second step at one bus may be triggered before the first step at another, due to the transient differences in bus frequencies. Again, a second step before system inertia responds to the previ, since load shed at each ste should be dise entire system, many sma 1 steps requires very many more relays.

In the following section, rinciples and guidelines for load shedding are reviewed. &e succeeding section deals with ap lication of these guidelines in development and testing of a 8FLS plan for a small island system.

Principles and Guidelines


Load Shedding
Underfrequency load shedding is a coarse tool for use in an extreme situation. As a last-resort ex edient, it should be simple, rapid, and decisive. To be erfective, it should be automatic, be distributed uniformly across the system at each ste to avoid aggravating line overloading, be locally controlle(Pin response to local frequenc to be independent of system splitting, and should not be &pendent on communication

Time delay. Some delay is needed in order both to ride out short time transient frequenc excursions, and to accommodate the res onse time of txe system to each step of load d w e v e r , time delays should be introduced cautious sheddin? y, for many of the reasons just cited for separation of frequency thresholds. Priorities and Distribution. There are two basic riorities for planning the shedding of articular blocks of road: the importance of that load, and et benefit of the action to the ! l system. The importance of each load block establishes priorities, but these must be subject to the requirement that each step must be distributed as equally as ossible across the system, since the location and extent of a Sisturbance cannot be known in advance.

95 WM 140-4 PWRS

A paper recommended and approved

by the IEEE Power System Engineering Committee of the IEEE Power Engineering Society for presentation at the 1995 IEEEfPES Winter Meeting, January 29, to February 2, 1995, New York, NY. Manuscript submitted July 14, 1994; made available for printing January 5 , 1995.

0885-8950/95/$04.00 0 1995 IEEE

1468

Other considerations. It might seem that, to be optimally effective, UFLS should be based on an accurate estimate of the amount and location of the generation that caused the system upset, and control means should shed that amount of load in that area. Both of these objectives are unattainable. While system frequency in the steady-state is a sensitive indicator of any small mismatch between load and generation within a closed system, it is not an unambiguous immediate indicator of the magnitude of a sudden change in the loadgeneration balance, and no indicator at all of the location. Even. the ipitial rate of change of fr uency is ambi uous, since individual generators react i3ividually accor%in to their own inertia, and to their electrical distance from t e disturbance and from other machines. A third problem, related to the first, is inability to determine, in the available time frame, the exact topology and d namic state of the s stem. Finall the amount of load a c d y shed by relays wib de end on t k actual load demand at the time of shedding, wkch is not likely to be equal to the potential (peak connected) load on those feeders.

lished by load priorities, sub'ect to the operator's jud ement that no security constramts (e.g. line loadings) will violated. Thus, although in rinci le it might seem that the higher priority loads shoulf be sted last and restored first, t h s guideline is subject to differing constraints in shedding and restoration.

/%

The arameters to be considered for an automatic restoration sch ule are the frequency threshold for restoration, the size of the steps, and time delays.

Integration into SCADA/EMS


For maximum effectiveness, UFLS should function in full coordination with SCADMEMS facilities. Basic capabilities to be considered include monitonng of relay status, and monitoring and archiving of load shedding actions. Beyond this, three modes of coordination may be considered: (i) suplemental manual load shedding, (ii) reclosure of selected Ireakers ("restoration") following either manual or automatic sheddm , and (iii) remote setting of.UFLS relays. The first two of gese are conventional functions of modem SCADA s stemsare conventionally known as "Load Zurtai1ient!Reh e third function is unconventional, and to date does not appear to have been implemented on any system. Each of these modes will be discussed m turn.

Rate of change of frequency, while it is in princi le an earlier indicator of an instantaneous change in loa generation balance than absolute frequency, is even more subject to distortion b local dynamics than is the latter. This is another as ect wherein, because of the emergency conditions under wgich UFLS is invoked, simplici and robustness are to be preferred to more refined, careful y tuned designs. eral, a reliable UFLS schedule can be developed wi out resorting to rate of change conditioning. If for some reason it seems desirable to resort to the rate of change, measurements must be regressed in order to avoid erratic performance, which may mtroduce unacceptable delays.

dP

=r-

In effect, UFLS supplements spinning reserve. For small load-generation mismatch and without load shedding, maximum deviation from normal frequenc increases sharply as the magnitude of the mismatch approacies the amount of spinning reserve. Conversely, as the amount of spinning reserye approaches several times the disturbance magmtude, the maximum deviation a proaches the associated steady state frequenc offset. {pinning reserve also serves to reduce expFt.dexcess shedding for any given schedule, but the effect dimnishes as size of disturbance increases [ 1,4]. (It should be noted that the ma itude of a frequency swing following a disturbance is ve %pendent on the size of the disturbance ((amount of actuaYgeneration, not generatin ca aci e that is effeclost)) relative to the concurrent level of the oa .) tiveness of spinning reserve is strong1 affected by its distribution over the equivalent electricalrarea of the network; and the number o , and its distribution among, responsive generators [4],

f B

Load Restoration
Second in importance to a reliable load sheddin program is a correspondmg program for safe restoration of interrupted load as rapidly as may be feasible. This is a desirable element in load shedding plans for most systems, and a necessary element for many. A robust UFLS, in order to be effective, may often have to shed load in excess of generation that has been lost, and (slower) s inning reserve will usually be available to pick up some or ay1 of that load. It is usually the case that load shedding is effected at unattended sites, and consequently local manual restoration will be time consuming and protracted. The same basic priorities affect the restoration of load as do the sheddin of load: the importance of the load, andathe benefit of t i e action to the system. However, the implications of these two considerations are not the same for sheddin and restoring circumstances. For planning the shedding of foad, importance must be subject to the r uirement that each step must be distributed as equally as p o s s l e across the system. For restoring load, however, the condition of the system should be known, and in this case the sequence of restoration may hew more closely to the optimal sequence estab-

Supplemntal Load Shedding. Ca ability should be provided for manual shedding of selected bgcks of load by the system operator under circumstances other than those that trigger automatic UFLS. Such circumstances will include may not be limited to) individual line or circuit overloa s, unacceptably low bus volta es s stem-wide cajacity shorta es, and any situations in w k l h Jequency decline is arrestefby automatic UFLS, but fr uency does not fully recover to its normal level. Operator%rected shedding should be limited to steps no reater than 5 % of total s stem load. The location of shed in for line overloads or ow bus voltages necessari1 will be fictated by the incidence of those conditions. S h e d k g for s stem-wide capacity shortages should follow the rinci le oty distribution across the system recommended F for % A which may be achieved by utilizing the. first step(s) of the UFLS schedule. However, thi? policy will require modification for rotating "brownouts, if such are required.

Put

Reclosure o Breakers. Following UFLS, a capability for perf missive automatic load restoration (where "permissive automatic" means automatic followin operator approval and acknowledgement) is highly desireabk. Reclosure on loads, subject to securi constraints, should be in order of priority (higher priority oads first) subject to security constralnts. Accordingly, once frequency has recovered, and available generation exceeds that required for necessary spinning reserve, the o erator should be provided with the reverse order list of loa& to be restored, m blocks not to exceed 5% of total system load prior to the initiation of shedding. Each associated breaker should be reclosed only given a satisfactory level of bus voltage and following operator acknowledgement.

Remote Setting o UFLS Relays. Utmost caution is advisable f before implementation of any scheme for remote setting and enabling of relays. Properly designed load shedding, as already emphasized, is calculated to respond to unforeseen disturbances in a modulated but sufficiently drastic manner to save the system when design criteria have been exceeded. Attempts at elaborate fme-tunin , and frequent arming and disarming of UFLS relays, coul leave the system vulnerable

1469

at the time when drastic UFLS is most needed. As an instance, the 1977 New York blackout was due in part to a partial disabling of load shedding activation.

MW as turbine units, and in 1997 a new Station C will be addecfcontaining three 60 MW steam units. EAC transmission includes 390 kilometers of 132kV circuits. and 850 km of 66kV circuits. The system incorporates some protection a ainst underfrequency conditions which shed load at predefmJ levels between 49.2Hz and 48.4Hz. No intentional time dela s are included in the settings. Following tripping, manua restoration of load is required.

Development of a System Specific UFLS Plan


Small Isolated System Considerations

General principles. In applying general principles to small, isolated systems, the distinguishmg charactenstics of such s stems should be kept in view. Small isolated systems lack t i e supporting inertia of an embedding interconnection; m other words, they lack the sup ort of fr uency provided by inertial power flows on ties. hence, U % schedules must accommodate a much wider band of normal frequenc deviations, but at the same time, more drastic action must ge taken to avoid unnacceptable frequency excursions. There is smaller margin for error. Fewer and larger steps will be reuired than for larger interconnected system, time dela s %odd be minimized, and a greater percentage of load suxject to shedding will be required. Reserves. For systems embedded in large interconnections, spinning reserve is intended to replace lost generation within several minutes; immediate covera e for a local loadpractical1 infiniteeneration deficiency is provided by f u s inertial reserve of the interconnection (within t e pick-up capacity of the tie lines). This is not true for small, isolated systems, and the effective inertial reserve ("responsive 4. reserve'' [ 2 ] )must be considered [ ]

In 1991, a trip of one of the two 132kV circuits out of Station B while the other circuit was on maintenance
precipitated a splitting of the system and a total systein shutdown. This and other considerations resulted in EAC commissioning a thourough study of its protective system.

Modeling
Development and validation was done in three stages, using successively more detailed models: Stage 1, a screening stage to explore a wide variety of UFLS schedules; Stage 2 , testing of one or a few selected schedules; Stage 3, a fm verification of the recommended schedule. Stage 1 used a lumped model - a single equivalent ienerator, and a uivalent load. Sta e 2 representate the individual units, in ividual bus loads, a 1 connected to appropriate buses of a network model. Stage 3 used a full transient-stability model.

credibilfity

be

This means that it is even more important on small isolated systems than in general, that reserve margins be spread across as many units as feasible, rather than assi ed to a few larger, or to smaller inexpensive, units. InertiaYresponse of all units not at their limits will decrease frequency excursions [%and thus minimize exposure of the system to activation of U LS,as well as to the amount of load shedding that will be incurred. Obviously, the effective amount of inertial reserye of a unit is much less than its unloaded margin (its nominal s inning reserve) [2]. These considerations will influence tie setting of the frequency threshold for the first step.

The Sta e 2 model consisted of dynamic (AGC level) models of all pfants, cou led with a load flow re resentation of the system. System Synamics were integrate at one-second intervals, and plant outputs and line flows recorded. The system was sub ected to a series of increasingly severe generation losses, deginning with one unit, and extending to entire loss of each of the plants (two plants for 1993 and 1997, three plants for 2000).

CQ

Ln view of the possibility of system splitting durin a disturbance, reserves (active and reactive), as well as oad shedding sites, should be spread as uniformly as possible across the system: active reserves on all units, other reactive reserves as close as possible to loads.

The Stage 3 model was used for a final test of system dynamic behavior during the course of the load shedding schedule. This test was mtended for very limited purposes, since the bus schedule that was assumed was somewhat arbitra . The Stage 3 test, accordingly, was intended only as (i) a%al verification of the validity of the Stage 1 and Stage 2 models b comparison with a recognized, fully detailed transient stagility program, and (ii) to indicate the stability of the system during the course of the load shedding.

Restoration. For restoration on an isolated system, activation should be well above normal frequency. Ste size(s) in eneral should be smaller than those used in loalsheddmg. h e y should be well within s inning reserve capability of the units on line; stated conversi , spuming reserve dunng restoration should be kept well atove the magnitude of restoration step sizes. Time delays should be provided to avoid reaction to short duration frequency excursions (such as might result from overshedding), and to accommodate the response capability of spinning reserve generation.
The EAC System
The Electricity Authority of Cyprus (EAC) is responsible for the generation, trimsmssion, and distnbution of electric power throughout the island of C IUS (although the Turkish occu ied ortion of the island (?&!OA) is inaccesible to the EAC!staf$. With the exce tion of 25 MW of eneration located within the TOA, A C F ! provides all o[ the island's ener y requirements. The resent system peak is in excess of 5 0 0 k W , and is expectecf to reach 880 MW by the year 2000. The EAC system includes two oil-fired power stations, Station A having six 30 MW steam units and two 37.5 MW gas turbine units, and Station B havin six 60 MW steam umts. Station A will be expanded in 1955 by two additional 37.5

Development o a Load Shedding Plan f Criteria. In order to provide a screening (Stage 1) of plausible UFLS plans, three load and eneration levels on the existing 1993 s stem were chosen as stown below. For each candidate U F d lan, the system was subjected to the loss of one, two, three, four, and in the case of the medium and high load levels, five 60 MW units.
Load Level Light Medium Heavy Generating Lost Generator Capacity % Loading 200 360 503 Spinning Reserve
130 MW 50 MW 27 MW

330 410 530

60 100 100

Early decisions, in accordance with the objectives and ciples, were i) to make a total of 80% of the system loacf$$bect to load s edding, dependin on the severit of the disturkame, (ii) to separate the steps %yat least 0.2 Az.

In order to provide an objective analysis that would reflect the relative performance of the alternative schedules, four statistics were calculated for each schedule:
" (i) "RMS Hz. the root-mean-square (rms) value of the maximum deviation of frequency over the entire scenario set (i.e. the cases that were simulated),

1470

(ii) "StD. Hz. " the sample standard deviation of the minimum frequency, (iii) "S.S. .Hz. the sample standard deviation of the postsheddin stead state frequency, and "Max.fkc. " the maximum frequency excursion (i.e. the minimum of the minimum frequencies).

J. z

Of these four statistics, the first provides an indication of the performance of the schedule in minimizing the frequency excursion over the range of test cases; the second provides an indication of the consistency of performance over the scenario set, the third provides an indication of the performance relative to overlunder-sheddin and the fourth provides a worst-case performance over t e scenario set.

0 O.P.40.60.8 1 1.2l.41.61.8 2 2.22.42.62.8 3

Time (sec)

Ultimately, sixteen candidate UFLS schedules were tested in a process involving several hundred tests on an evolving sequence of schedules developed to minimize the selected statistical measures.

A: With 1991 UFLS Schedule


51.0

- ._- ........
.~

Results. Improvements in performance were achieved by retaining the gradual separation of thresholds, increasing the amount of load shed for the first three steps, and increasing the cumulative amount shed over the first three steps (while kee in the total load shed to 80%). In this wa , the stand a r i feviation of minimum frequency was r uced from 0.823 to 0.684, while the sample standard deviation of ostshedding steady-state frequency was held to 0.189, xown from 0.476 for the existing schedule. This latter statistic is deemed less critical than the former, since it can, and always will, be corrected by either automatic or manual generation control. The worst frequency excursion for the preferred schedule was 3.18 Hz., down from 3.58 for the existing schedule.

.......

0 O.P.40.W.8 1 1.21.41.61.8 2 222.42.62.8 3

lime (sec)

B: With 1993 UFLS Schedule

Review of the results indicated that one schedule provided the best over-all performance:

E3 Shed, % a Delay, sec.

Th. Hz. 49.0 48.8 48.4 47.8 47.0

15 20 25 10 10 0.20 0.20 0.35 0.35 0.35 Cum. % Shed 15 35 60 70 80

Several experiments were made with extended time delays for later stages of shedding. No overall improvement in erformance was achieved b this means. For instance, in tge case of the preferred Schdule, there were two cases in which all five steps were shed: light loading with a loss of 240 MW, and medium loading with a loss of 300 MW. In the former case, the effect of a 1 second time delay on the last ste was to avoid the shedding of the last ste , resulting in 4 d W of undershedding, rather than 16 M d o f overshedding; there was no effect on the minimum frequency. In the latter case, however, shedding of the last step was not avoided, and the effect of the 1 second time dela was to depress the minimum On balance, the latter effrequency an additional 0.41 fect seems more detrimental than the slight overshedding in the former case.

0 0.a3.40.60.8 1.21.41.61.8 2 222.42.62.8 3 1 Time (sec)

C: With Preferred UFLS Schedule

Simulation of Incident of November 1991


Figure 1
November 1991 Incident Simulation. The Sta e 1 model, which had been used to develop the UFLS Sckdule, was used further to study the res onse of the system, under three schedules, to the event of November 1991, when area 3 was split from the remainder of the system. The event was simulated with three UFLS schedules: the one in effect at the time of the incident, the one currently in effect, and the one that is now proposed. The assumptions, matching system conditions at the time of the event, were that area 3 included four full loaded units (200 MW) and 22% of the total system l o d For the first two cases, ste s 1, 2, and 3 of the UFLS schedule consisted onl of area f; load, which was already lost in the 22%. In SUC a case, from the standpoint of the remaining portion of the system, 200 MW of generation would have been lost alon with 22% of the system !oad; the remanun steps would be b l l y effective. For the thud case, with loafshedding spread across the system, only the area 3 ortion (22%) of the first three steps would have been lost, gut available load shedding on the remaining portion of the system would be discounted at each step to 78% of its nominal value. The results are shown in Figures 1 A, B, and

dz.

Evaluation o the Load Shedding Plan f


Once the screening process had generated a preferred plan, attention was turned to rigorous evaluation of that plan. The first sta e of this evaluation, still utilizing the Stage 1 model, tested t i e plan against the circumstances that led to the 1991 system blackout, and against both existing and forecast system conditions. The next sta e was evaluation on the Sta e 2 model, which required speciication of bus specific shed ing plans. The final sta e was varification of representative results using the Stage transient stability model.

li

In what follows, we will refer to a division of the system into three, somewhat arbitrary, areas by two boundaries at which the system might conceivably split under certain extreme conditions. Two of these areas contained one of the two existing power lants, the third could remain with one or the other plant, Bepending on which of the two boundaries opened.

1471

C. This example shows the importance o dktributin each f f step o the load shedding across the system, and injicates clearly that, had the proposed UFLS schedule been in use, areas 1 and 2 of the system, with an unshed 24% of the system load, would not have been lost.
N

52 51 50

Bus-s ecijic schedules. The next step was to provide buss eci& schedules for load shedding that satisfied reasonably tle recommended schedule. No attempt was made at thls sta e to provide a practical schedule meeting all the multitu inous constraints of the actual system. Instead, an attem t was made to schedule load sheddmg at the proper level f& each step within each of the three areas alluded to above, and for the system as a whole. Minimum adjustments were made in the schedule for the two later ears to accommodate system rowth. The resulting schedure for the year 1997, showing &e percent load in each of the three areas, the percent of area load to be shed at each step in area, and the resulting percent of system load to be shed at each step in the system as a whole, is shown as typical:

I49
48

cf

47

46

Loss of 360 MW on Year 1997 System


Figure 3
52

1997 System: Load Shedding Schedule by Area


Area:
% Load 1

2 3 4 5

2 3 System 39.2 41.3 19.5 100 Cumulative percent shed at each step 13.5 18.0 10.0 14.7 34.2 32.6 43.5 35.4 61.0 57.1 62.2 59.6 73.0 66.4 73.4 70.4 82.5 74.9 81.3 79.1
1

51

. - . . - . . . - . -

50
N

_._ -.

I49
48

--

- . ... -.

..

... . .

..- .....

- .- . .- .

Stage 2 model evaluation. The UFLS plan was evaluated a g m s t all combinations of unit losses for the system models provided for the years 1993, 1997, and 2000. Fi re 2 presents the results of simulation of the loss of six 6 r M W units (two of which had been loaded at 45 MW), with a s stem load of 500 MW, and a capacity on line of 530 M k Four of the five sta es of shedding were invoked, for a totai of 355 MW shed. %he minimum frequency was 48 Hz,the z maximum was 51 Hz;after ten seconds it settled at 50.4 H .
Figure 3 resents the results of the loss of six 60 MW units (one of whch had been overloaded to 70 MW) with a system load of 718 MW and a capacity on line of 708 MW.
In this case, three of the five stages of shedding were invoked, for a total of 431 MW shed. The minimum frequency was 48.2 Hz, the maximum was 51.1 Hz, and after ten seconds, it settled at 50.6 Hz.

47 46
0

10

15

20

Time (sec)

Loss of 540 M W on Year 2000 System


Figure 4
Sta e 3 model evaluation. The Stage 3 model was used for a fin$ evaluation of system d namic behavior during the course of the load shedding scledule. These studies, as exlained, were intended for a very limited purpose. One knitation was encountered; the stability program that was used has a limitation in the number of underfrequency relays such that only a few relays at level 3 could be represented. Accordingly, attention was limited to the most severe of the cases where only 2 levels of shedding were encountered. It was ossible to represent a few step 3 relays, to determine whetger or not step 3 would have been initiated.
Figures 5A, B, and C present the results of one scenario for the 2000 system, the loss of a new Station C , using the Sta e 1, Stage 2, and Stage 3 models res tively. In this case, t e load was 882 MW, and 180 M f i e r e lost out of available generation of 891 MW. Two of the five stages of shedding were involved, for a total of 309 MW shed. The minimum fr uencies indicated b the three models were 48.77 Hz, 4 8 3 0 H ,and 48.75 z, res ectively; the frequencies after z ten seconds were 50.45 Hz,&.40 H and 50.39 Hz respecz tively.

Figure 4 presents the results of the loss of nine 60 MW units at two stations, with a s stem load of 882 MW and a capacity this case, four of the five steps of on line of 891 MW. shedding were invoked, for a total of 626 MW shed. Frez quency dipped to 47.6 H ,overshot to 51.8 Hz,and settled z out after ten seconds at 50.8 H .

52

It should be noted that, due to their being produced b dif ferent pro rams, the plots in the three cases have di&eren; scales. d s t importantly, the time scales have a duration of 3 seconds in Figure 5A, a duration of 20 seconds in Figure 5B,and a duration of 10 seconds in Figure 5C.

46 4
0

10 Time (sec)

15

20

E ect o S innin Reserve Level. It will be noted that only f t e first o the t ee scenarios rovided for testing, that of

Loss of 360 M W on year 1993 System


Figure 2

the light load level, met the f a r est sin le unit rule-ofthumb for spinning reserve; and %is is tie only case that never resulted in an load sheddin The medium load level rovided 50 of the esirable 60 M b reserve, while the heavy road level scenario provided only 27. A clear indication of the benefits [4] of increased levels of spinning reserve is

F L

1472 shown by comparin three cases with 0, 60 MW and 120 MW of reserve. RJuctions in the RMS of the minimum frequency excursion were 12% and 20% for the two levels, and in the maximum frequency excursion were 15% and 25% respectively. These reductions are significant.

Table: E f c s of Spinning Reserve Level fet


spinning mum Reserve

RMS
Exc. 2.49 2.19 1.98

MaxiExc. Hz. 4.20 3.58 3.14

Hz.

+ 60MW + 120 MW

None

.
48.0
47.0
...........

........

"

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

46.0
0 0.23.40.60.8 1.21.41.61.8 2 2.22.42.62.8 3 1 Time (sec)

A significant factor, illustrated b the entire range of cases studied, is that, in general, the efktiveness of load shedding increases with increasing load. This is, of course, because with increasin load, a reater amount of load is shed at each step: at a l e v i of 200 W, 60 MW re resents 30 percent of the load and, under the su gested schdule, such loss would require the sheddm of at feast two steps, whereas at a level of 500 MW, 60 hfW re resent only 12% of the load and such loss would r uire tfe shedding of only one step. This effect can be seen y c o m p m g the dramatic difference in erformance at the light load level to the severe loss of 144 !hW with 60 MW and 130 MW of reserve. It may also be seen, more realistically, b comparing the erformance of the UFLS at medium and h i g i load levels wit: either 60 MW or 120 MW reserve: load sheddmg is more effective (requires fewer ste s to be shed) at the higher load level for each number of 6 8 M W units lost. This effect is fortunate because it is easier to provide a larger margin of spinning reserve at light loads than at heavy loads.

"b

Conclusions An effective, robust UFLS program is a necessa contribution to the reliability of eve power system, an it is most important for small, i s o l a t 3 s stems. Properly designed, implemented, and maintained, &LS can enable such systems to survive many disturbances that otherwise would result. in total blackouts. This paper has developed and a plied criteria for UFLS lans that may differ in some aegree (although not i kin ) from those appropriate to large, intern connected systems.

48 ...........................................................................
47

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

The authors are not unaware of work on optimal load shedding rograms [e. . 31, but remain convinced that such lans may fail to be ekectwe when most needed, especial$ on smaller, more vulnerable systems.

Vertical divisions represent 1 Hz. Horizontal scale w e n d s to 20 seconds B: Stage 2 Model

References
1. R.M.Maliszewski et al.: Fr uency Actuated Load Shedding and Restoration; IEEg Trans. v.PAS-90 n.4, 1971, pp. 1452-59. R.J.Kafka: Load and Generation Coordination; presented at ECC Seminar on Power System Restoration, A ril 1992, Arlington, Vir inia, and October 1992, d m t r i c h t , Netherlands, E C E Inc. D.Kottick, M.Blau, Y.Halevi: Evaluation of an UnderFrequencx Load Shedding Optimization Algorithm; Proc. 11 PSCC, Avignon, France, Aug./Sept. 1993, pp.437-41. C.Concordia: Performance of hterconnected S stems Following Disturbances; IEEE Spectrum, v.2 n . l June 1965, pp.68-80.

2.
IO.

30

IO. 20

in. I O

3.

4.

Biographies
Vertical divisions represent 0.5 Hz. Horizontal scale w e n d s to 10 seconds C: Stage 3 Model Charles Concordia, D.Sc., LF, independent consultant. Lester H.Fink, LF, Exec. V . P . R&D Mgr., ECC, Inc. Boza Avramovic, Ph.D., SM, Lead Consultant ECC, Inc. GeorgePoullikkas, Asst. T&D Mgr., EAC.

Loss of 180 M W at Station C, 2000 System Figure 5

You might also like