Sie sind auf Seite 1von 17

I

1106 Transactions on Power Systems. Vol. 7, No. 3, August 1992

A report of the AGC Task Force of the IEEE/PES/PSE/System Control Subcommittee

UNDERSTANDING
AUTOMATIC GENERATION CONTROL
Nasser Jaleeli 1 (TF Chairman) Donald N. Ewart 3 Lester H.Fink 4
Louis S. VanSlyck 2 Power Technologies, Inc. Arthur G. Hoffmann 2
American Electric Power Service Corporation Schenectadv, New York ECC, Inc.
Columbus, Ohio Fairfax, Virginia
1 Member, IEEE z Senior Member, IEEE 3 Fellow. IEEE 4 Life Fellow, IEEE

Abstruct - This paper describes what automatic generation SYSTEM NATURAL PERFORMANCE
control (AGC) might be expected to do, and what may not be Power system loads and losses are sensitive to frequency.
possible or expedient for it to do. The purposes and objectives Data captured right after frequency disturbances indicate that
of AGC are limited by physical elements involved in the pro- their aggregate initial change is in the same direction as the
cess and, hence, the relevant characteristics of these elements frequency change.
are described. For reasons given in this paper, it is desired that o n c e a generating unit is tripped or a block of load is
AGC act slowly and deliberately over tens of seconds or a few added to the system, the power mismatch is initially compen-
minutes. From a perspective of utility operations, there is no sated by an extraction of kinetic energy from system inertial
particular economic or control purpose served by speeding up storage which causes a declining system frequency. As the fre-
the AGC action. quency decreases, the power taken by loads decreases.
By this Task Force paper, the System Control Subcom- Equilibrium for large systems is often obtained when the fre-
mittee is providing a resource to the power engineering com- quency sensitive reduction of loads bal'wces the output power
munity to help guide research into topics related to AGC. of the tripped unit or that delivered to the added block of load
at the resulting (new) frequency. If this effect halts the
Key Words- Area Control Error, Automatic Generntion kquency decline it usually does so in less than 2 seconds.
Control, Economic Dispatch, Frequency Response, Govemor If the mismatch is large enough to cause the frequency to
Action, Inadvertent Energy, Load Frequency Control, Power deviate beyond the govemor deadbarid of generating units, thcir
System Control, Power System Operation, Speed Droop. output will be increased by govemor action. For such mis-
matches, an equilibrium is obtained when the reduction in the
INTRODUCTION
power taken by loads plus the increased geeneration due to gov-
Present schemes of automatic generation control (ACC), emor action compenqates for the mismatch. Such equilibrium
have evolved over the past six decades and are in use on inter- is normally obtained within a dozen seconds after the tripping
connected systems as large as one-fourth the North American of a unit or connection of the additional load.[ I ]
continent. Continued enhancement of these schemes is Many govemor deadbands are beyond 35 mHz. This
expected via new application9 and yet to be developed logic amount of frequency deviation requires the upset of more than
algorithms and prwcss con~rollechnology. However, certain 1000 M W in the eastem interconnection of the US.Thus, in
concepts, objectives, and simulation models that too often this interconnection many govemors may be called upon as
have been assumed for research effofis in AGC are not applica- unit speed stabilizers (21only a few times per month.
ble for present day power systems. This paper (prepared under Typical speed droops for active govemors are in the range
the auspices of the System Control Subcommittee) attempts
of about 5%. (Govemor droop is the percent change in fre-
to describe basics that are applicable to today's power systenis
quency which would cause the unit's generation to change by
and AGC in order to assist interested research parties in direct-
Io()%of its capability.) This level of sensitivity to frequency
ing their work at nieanirigful objectives.
allows many isolated systems, which are not necessarily small
The way loads and unit govemors respond to various
in capacity, to perform satisfactorily without AGC. Thus, at
upsets of electric power mismatch in the system is presented. the expense of some tiequency deviation, generation adjust-
This is followed by a brief description of types of generating nient by govemors provides aniple opportunity lor a follow up
units, arid constraints on their range and rate of response to manual control of units. The objectives of the follow up con-
AGC signals. Within these constraints, the objectives of AGC trol, especially under normal changes of load, are to retum the
in isolated, and then in multi-conlrol areas are presented. For frequency to the schedule, to minimize productioti cost, and to
the latter, the philosophy of tie-line bias control is reviewed. operate the system at an adequate level of security.
Additional fundamental considerations affecting AGC, together The purpose of AGC is to replace portions of the above
wilh some concluding rematks, m presented at the end
mentioned manual control. As it automatically responds to
normal load changes, AGC reduces the response time to a
91 WM 229-5 PWRS A paper recommended and approved minute or two, more or less. Mainly due to delays associated
by the IEEE Power System Engineering Committee of with physically limited response rates of energy coiivcrsion,
the IEEE Power Engineering Society for presentation further reductioti in the respoiisc tliiie ol AGC i s ucillier pos-
at the IEEE/PES 1991 Winter Meeting, New York,
New York, February 3-7, 1991. Manuscript submitted sible nor desired.
September 4, 1990; made available for printing NeiIher Iiw follow up nianrml cnntrol nor' AGC. i.\ irhle
January 3, 1991. or' expected to pluy urry role in linrititrrg tire ntugrri~rrtle(d rlw

0885-8950/92$03.00 Q 1992 IEEE

I
1107

first frequency swing which occurs within seconds ajier the Nuclear Units
loss of a block of generation or loud in the system. In fact, in Nearly all nuclear plants have either boiling-water (ESWR)
the U.S., the procedure in most control areas requires AGC to or pressurized-water (PWR) steam generators.[8,9] Most are
be suspended when the frequency deviates 200 mHz or more. not currently controlled by AGC, but there are exceptions.
For conditions where change of generation due to govemor BWR units operated under AGC, typically can respond at 3%
action and change of load due to its sensitivity to frequency are per minute for 10 minutes or so within their regulating range.
not enough to intercept the run-away frequency, over- and To move outside the range requires malcing changes manually
under-frequency relays are among the last resorts for shedding in the control-rod pattem, a more lengthy process.
loads to prevent system collapse, or tripping generating units Power control in P W R units is accomplished by
to prevent their damage. adjusting control rods in the reactor core, and for larger
excumions at slower rates, by changing the concentration of
GENERATING UNIT CHARACTERISTICS boric acid in the primary loop. These units are capable of
A G C realizes generation changes in the system by making 20% excursions at rates of nearly 3% per minute.
sending signals to units under its control. The design and
performance of an AGC system is very dependent on how fi
units respond to such signals.[ 3,4,5] Unit response Combustion turbines, particularly simple-cycle gas tur-
characteristics vary widely and are dependent on many factors bines, are capable of the fa..test response of m y units on util-
such as: ity systems, but since typically used as peaking units they are
Type of generating unit; for example fossil-fired, combustion seldom equipped for operation under AGC. Combined-cycle
tuhine, combined cycle, nuclear, or hydro. units, where the gas-turbine exhaust gas is used to generate
Type of he1 being used; i.e. coal, oil, gas, or uranium. steam to power a steam turbine, are more likely to be equipped
Generic plant type; i.e. drum-type or once-through boiler, for AGC control if operated by a utility. The response of a
boiling- or pressurized-water nuclear steam supply, high- or combined-cycle plant is not quite as rapid as a simple gas tur-
low-head hydro plant. bine. Combined-cycle units operated by independent power
Type of plant control; i.e. boiler-follow, turbine-follow, or producers and cogenerators may constitute a significant
coordinated. Also, whether the unit is operated in a sliding fraction of generation in some areas, but are seldom operated
pressure or fixed presswe mode. under AGC.
Operating point; frequently the ability of a unit to respond is
different at one load point than at another. E.g., operation near Hydro Units
a valve point will be different than operation between valve Low-head hydro units, such as used in run-of-river plants,
points. Also, a generating unit operating with valves wide have excellent response capabilities. Many can be cycled over
open cannot respond to a signal to generate more power. their entire operating range in under a minute. High-head units
Operator actions; unit operators may take a unit off AGC must have their response rates somewhat curtailed to prevent
control for various reasons. Problems with auxiliaries such as water-hammer damage in long penstocks. Even with such a
boiler feed pumps, and coal mius are two examples. limitation, the units can respond with very large excursions, if
A brief description of the characteristics of several types desired. However, hydro plant loading maneuvers may require
of generating units as they affect the response to AGC signals careful coordination with other hydro units upstream or down-
Ls given below.[6] stream on the same river system.[lO]

-
Fossil-Fired Stem-Turbine Units AGC OBJECTIVES
Many existing drum-type units are still controlled in a The objectives traditionally defined for AGC appear to be
boiler-following or turbine-following mode. Controlling the vague and incomplete. However, any attempt in this paper to
steam flow at the turbine inlet allows fast initial unit power precisely detine them may introduce a constraint on future
response by altering the rate of boiler energy conversion. opportunities. Hence, we only compare attributes of AGC
Boiler-following controllers tend to he fairly responsive to strategies from different aspects. For each attribute, the pre-
AGC signals, on the order of 3% per minute for a 30% excur- ferred strategy is indicated. Yet, we leave the selection (or
sion, particularly if fueled by oil or gaq. The AC‘ signal usu- weighting) of attribute importance and the exact determination
ally drives the speed-load setpoint adjuster on the speed-gover- of m overall score for individual consideration.
nor control which, in turn, cauSes turbine valve movement. For the comparison of the strategies, a power system hav-
The boiler controls sense the changes in steam pressure to ing some units under AGC and others manually controlled, is
adjust flows of air, fuel, etc. considered. Various attributes are to be assessed over a selected
Once-through steam generators are frequently supplied time window (duration of their comparison.) This comparison
with coordinated control systems, meaning that the AGC sig- is lint made when the system operates as a single (isolated)
nal is processed by a master controller that “coordinates” the control area. The concepts developed for the single control area
fuel, air, temperature, and turbine valve controls so as to limit caye are then extended to that of an interconnection comprising
undesirable stresses on the plant components. Many of the several control areas.
newer drum-type steam generators are &“oiled this way too.
Depending upon how these controls are designed and adjusled, Shide (isolated) Control Area
unit response cm v i l q widely.l71 A wcll-adjustcd unit of this ‘The following nspects for which the strategies are coni-
type be o f Iliaking a 20% in 10 pared ant by no niems intended to be coinplcte nor to k appli-
minutes. cable for all systems. We also acknowledge that niucli is lelt
1108

for improving the definition of each aspect and the qualifying restoration becomes very limited. Almost all units can provide
words used in them. For example, how should the acceptabil- a higber rate and range of control from the unit control room
ity of a frequency trend be defined, or measured? This issue than via AGC signals from a control center. Following a large
will be addressed again later. frequency deviation, the mode of operation of a number of
Of AGC strategies, the one which yields a geaeration trend units is normally changed from automatic to manual and unit
acceptably matching the trend required to serve the varying operators are usually asked to adjust the generation manually.
load at scheduled frequency over tbe selected t h e window gets Therefore, the time required for restoration of frequency to the
a high score, The more acceptable the generation trend, the schedule mainly depends on how quickly manual changes can
more acceptable becomes the kquency head, and vice versa. berealized.
Therefore, a main input to AGC for a single control area is the While frequency is maintained within the acceptable
system frequency. range, electric clocks may still gain or lose time. To maintain
The strategy which accumulates lower fuel cost over the accurate time, the scheduled frequency is offset from nonlinal
time window is preferred. Such an AGC may also be expected whenever the time error exceeds a threshold.
to recognize undesirable unit generation ranges (including When the frequency is lower than nominal, the system
steam valve points ot hydro rough spots) and avoid un- loses a small component of sewed load. When it is higher, the
necessary sustained operation in such ranges. system imposes a small additional component onto the con-
The strategy should maintain a sufficient level of reserved nected load.[ll] As long as the time e m r is forced to cross
control range and a sufficient level of control rate. System load zero frequently, the difference between tbese lost and gained
can potentially have more sudden sustained changes than those load components remaios negligible. (Depending on frequency
which occurred in the selected time window. Prior to such amplitude excursion, “fkquently” may mean merely “several
changes the operating point of each unit would be preferred to times per week.”)
be in the middle of its controllable range. Also at any time,
the number of speed-govemor type units in a wound-up state Multi-Area Control
should be minimized A ”multi-area interconnection” is comprised of regions, or
The strategy should operate the system with a better security “areas”, that afv interconnected by tie-lines. Tie-lines have the
margin as specified by the system’s management. benefit of providing inter-area support for abnormal conditions
The strategy which accumulates lower cost associated With as well as transmission paths for contractual energy exchanges
the wear and tear of regulation for all units combined is pre- between the ateas. The area boundaries are determined by tie-
ferred. This strategy, therefore, is expected to avoid unneces- line metering for AGC and contractual billing purposes. Both
sary rapid maneuvering of unit generation (or the chasing of power and energy flows are metered. Energy metering is usu-
high frequency components of demand change.) ally on an hourly basis and the data values used for accounting
The strategy which requires less effort fiom system operators purposes must be identical (after auditing) for each corporate
gets a higher score. party sharing the tie-line. Accounting, and auditing, are
The strategy which provides timely recommendations for described in reference 12.
changing tbe output of units that are manually controlled gets The mnd of fiquency measured in any area is an indicator
a higher score. of the trend of mismatch power in the interconnection and not
The strategy should provide timely recommendations for in the area alone. Any area, of a multi-area system, that
changing the automatic regulation band (whether set at the attempts to adjust its generation to restore the frequency to
plant ot at the control center) for units controUed by AGC. schedule, requires a block of controllable generation large
The strategy should provide meaningful alarms, e.g. for enough to respond to the mismatch power in the interconnec-
units not responding to either system operators’ requests or to tion. Furthermore, the interconnection requires tie-lines which
the AGC signals. This may include displays in the control can carry such amounts of power between this area and the
center and/or at the plant for prior deviation from desired Others.
generation level and anticipated trends of future desired In an interconnection where AGC in more than one area
generation. This could also include the automatic observation is driven solely by a frequency signal, there will be large
of security constraints on the generation level of units. power oscillations between controlling areas unless regulating
The strategy which requires less computing power and other actions taken by all areas can be realized simultaneodsly.
hardware may deserve a higher score. This would recognize Further, the operation of such an interconnection would face a
that “simpler” is easier to understand and maintain, and is more severe problem if the areas attempting to control
usually more reliable. frequency had measurement error. An area that measured the
We now address the issue of frequency trend acceptability. frequency at a value higher than others would reduce its
Under normal variation of load, the frequency must be main- generation, while others raised, both attempting lo force
tained sufficiently within a band where under- or over- frequency (as they each measured it) to the scheduled value.
frequency relays could not be actuated by the next credible O n the surface, an alternative to frequency based control
contingency, e.g., a unit trip, or loss of a block of load. In may seem to be that each area generate enough power to serve
most systems, AGC can accomplish this objective even if it its intemal loads and losses, plus the total scheduled power
yields a generation trend which laps by several minutes that interchange ‘Ts” with other areas. However, if every area oper-
wliich would colrtirlually serve the load at the scheduled ated wilh this objective, then for reiso~lsdescribed below U t e
fquency. interconnectiorltiright not be able to operite salisfactorily.
Oacr the ficquency goes out of the above band for any ‘ h e tliflerence between an area’s pelwrition and the power
reason, eitllcr suddenly or slowly, the role of AGC UI its quick takeo internally by loads and losses is the suln of power llows

1
1109

“Ta” on all tie-lines between this area and others. The scheme has an “acceptablematch” with the sum of the area’s own load
whereby each area controls its Ta to match its Ts is called con- and losses, Ts, and the area portion of 10 p (Fa - Fs).
stant net interchange (or flat tie-line) control. Consider a sys- The system natural response coefficient, J3, is not a con-
tem comprising two areas, 1 and 2, and suppose area 1 mea- stant, neither is it accurately obtainable nor predictable. It
sures its interchange several M W below the actual, and area 2 depends on the current status and governor response character-
measures its own correctly. Then both areas will continuously istics of the presently on-line units [15] and the sensitivity of
raise their generation and no equilibrium can ever be reached. loads. Depending on the magnitude of upset from the prevail-
Altemately, assume area 1 uses an erroneous interchange ing pre-disturbance frequency, the variable number of gover-
schedule whose magnitude is higher than that used by area 2 nors coming out of deadband causes J3 to be highly sensitive to
for its purchase. The same instability occurs. upset size.[ 1,121Moreover, the observation or measurement of
Even in the absence of the above errors, this method of natural response can be obscured by normal system activities.
control is unable to provide satisfactory operation because it E.g., generating units may be actively responding to prior
m u l l s in severely depressed frequency when either area is not control signals and, of course, individual system loads are con-
able to produce its share of generation. As area 2 will attempt stantly and arbitrarily changing.
to maintain its interchange at the schedule, the system may The North American Electric Reliability Council (NERC)
collapse if a large unit is tripped in area 1. Guidelines specify using 10 B (Fa - Fs) to represent the area
Just as it was found impossible, for reasons that are now portion of the frequency response term in area’s AGC process.
obvious, for generators to operate stably in parallel without The Guidelines allow the use of a constant for B which i to
providing a suitable droop to their govemor characteristics, so be estimated annually based on the average of apparent area p
it was found impossible for systems to operate stably in p a d - values as observed for disturbances that occurred during on-
le1 under either constant frequency or constant net interchange peak hours. (The sum of all area J3 values is system J3.)
control. Ther@ore,a quick return of eitherfrequency or inter- Guidelines altemately allow using a variable B so as to repre-
change to the schedule must not be used as a basis for advocat- sent an estimate of the area’s p variation. However, they fur-
ing an AGC algorithm. ther require that the monthly average of IBI should not be
numerically smaller than 1% of the control area’s estiniated
Tie-Line Bias Control yearly native peak load demand (IBI is the magnitude of B.)
Multi-area control philosophy 1131 constitutes assistance Hevery area used an underestimated value for IBI, opera-
from area 2 to area 1 beyond the scheduled interchange when tion of the interconnection would tend to show characteristics
area 1 is not, for whatever reasons, satisfying its obligations. similar to those associated with constant net interchange con-
All areas of an interconnected system respond relatively trol. On the other hand, indiscriminate use of over-estimated
quickly, under load and govemor action, to changes in demand values for IBI, would tend to yield inter-area generation oscilla-
as signalled by changes in frequency. The concept of tie-line tions.
bias control, developed in the mid Ihirtias, permits areas, over Using B as proxy for an area’s J3 restates one of the major
the longer run, to readjust generation (to follow changes in objectives of AGC - it is desired that ACK in each control
their intemal load) without compromising the area’s share! of area provide a generation trend which has an acceptable matdi
natural response for frequency support. witli that representing the sum of the area’s own load and
To understand the basis of the above mentioned control losses, 10 B (Fa - Fs),and Ts. Generation in this area equals
philosophy, we examine a power system in two cases, first the sum of load and losses, and Ta. Therefore, the difference,
operated as one control area, and then as two control areas. The called the area control e m r (ACE), between the trend of gener-
system is assumed to be operating at scheduled frequency, Fs, ation and that with which an acceptable match is desired, is:
when it loses a block of load or generation.
However operated, system natural response (aka.regula- ACE=(Ta-Ts)-lOB(Fa-Fs)
tion, or primary response) based on design criteria, is expected
(for credible contingencies) to stabilize the system at a new Even if it were possible, it would not be desirable to
frequency, Fa. With Fs and Fa in Hz,this response is conven- maintain ACE at zero because this would require unnecessarily
tionally expressed as 10 J3 (Fa - Fs). The coefficient, J3, is neg- rapid maneuvering of units. The definition of a desirable gen-
ative and traditionally is given in M w / O . 1 Hz. It represents eration trend remains an open issue.
the combination of both load and govemor sensitivity 1141 to As long as each area manages to keep ACE and its
frequency. As system frequency is retumed to the schedule, pri- (several minute) average bounded, the interconnection will
mary response diminishes. continue to operate successfully despite reasonable errors in
When the system operates as one control area, it benefits metering of frequency and tie-line tlows, and/or the use of
from primary response of the complete system until central imprecise fmquency and power interchange schedules. Meeting
(supplementary or secondary) control actions can fully com- these conditions, each area will be supported by other mas
pensate for the power upset. Assume the same system operates when it faces a large change of load, or the tripping of a unit,
as two control areas and the location of the power upset is in that it otherwise could not alone handle satisfactorily.
area 1. It is desired, of course, to obtain a smooth restoration Furthermore, huge economic savings are realii~dby utility
of frequency in this case roo. Therefore, area 2 should, as interconnections versus the altemative of each operating a$an
before, assist area I with iis natural response - a power coni- isolated entily.
parable to its share or IO J3 (Fa - Fs) - aid ihis natural Power nihmatch in the intcrconncction causes frequency
rcsponse should riot he unduly compromised by AGC action. to change, and sustained frequency dcviation lroni Uie noniind
Thus, the AGC iri area 2 should yield a generation tretid wliidi value causes t h e error. Siniilarly, sustained deviation lroni
I

1110

tbe interchange power schedule results in inadvertent energy a dominate role in providing an acceptable generation trend
exchange between area$.Even if an area were able to maintain under nonnal conditions.
its ACE at zero all the time, sustained frequency deviation Such a tend should have an acceptable match with that of
from the scheduled value (whether offset from nominal or not) the area demand [i.e. the sum of the area's own load and losses,
would cause inadvertent accumulation. However, such inadver- -
10 B (Fa Fs), and Ts.] Area demand unpredictably varies
tent exchange benefits those areas requiring support and they around an average mnd. With the existing limits on the rate
accept respomibility for its subsequent retum.lo multi-area and range of generation change and the fact that steam units
operation, over the very long term, the energy generated in take a few to several dozen seconds to fully respond,
each area must he made to match the area's native load and maneuvering generation to match fast varying components of
losses, and the area's interchange energy schedule. There is no area demand is impossible. Furthermore, maneuvering
equivalent energy balance consmint for isolated operation generation based on these components will not necessarily
because Ta and Ts are each identically zero. produce a moR desirable ACE brend[l7]
"bere are procedures by which time error is corrected and Even if generation control rate and range were not as lim-
inadvertent energy is paid back. Methods m described in =fer- ited as they are and units could respond much faster, it may
ences 12 and 16. still be undesirable to maneuver generation attempting to
To define the objectives of AGC when the area is a part match fast varying components of the m a demand. Such oper-
of an interconnection we again refer to the comparison of ation would increase equipment wear and tear.
AGC strategies. However, for a multi-area interconnection the For normal operation, NERC guidelines encourage each
first attribute listed for a single controt area should be replaced area to control so that ACE crosses zero at least every ten
by: minutes and has a ten-minute average below an area specific
Of AGC strategies, the one that yields a more acceptable value. Lack of generation reaction to the components of area
ACE trend gets a higher score. demand whose periods are smaller than several dozen seconds
All other aspects for which the AGC strategies were com- should not reduce the level of compliance with these guide-
pared for a single control area are also applicable for multi-area lineS.
operation. In addition, the following ataibute is included for
multi-area systems.
The strategy which maintains a mote acceptable range of Almost all systems filter tie-line power flow measure-
inadvertent energy and repays its accumulation in a timely ments to compute ACE, and then process ACE to arrive at an
manner is preferred. Strategies that exploit unilateral inadver- AGC decision in each cycle. Any form of signal processing is
tent corrections, which simultaneously reduce time error, may in fact a filtering action: the sampling process in itself is a fil-
be consided to deserve a higher score.[12,161 ter; therefore filtering is inescapable. Filtering of any type
Timely and accurate calibration of power metering on introduces delays and distortion, and design should attempt to
each tie-line versus energy metering over recent hours can con- minimize them.Moreover, unless digital sampling is preceded
tribute to the veracity of the AGC process. by coordinated analog filtering, disruptive aliasing will be
introduced into the process.
T N Typical filters on tie-line metering introduce delays of up
It is typical for only a subset of units in any area to be to a few seconds. ACE processing logic introduces further
equipped with remote control hardware. For a variety of rea- delays. Given these delays and the response characteristics of
sons, a fraction of those equipped m not operated under AGC units, most systems use 2 or 4 seconds for the data acquisition
at any given time. Of those that are, some do not always re- *and decision cycles of the AGC system. Reduction of this
spond to control signals, e.g. because of backlash and wind-up cycle to a fraction of a second would substantially increaqe the
in speed-govemormotom. Thus, the rate and range of output percent of overall CPU time required for AGC with little or no
M W change that AGC can realiie is limited. improvement in system perfomaoce; in the presence of alias-
With this limitation, and notwithstanding the details of ing, performance could even be degmded.
AGC algorithm designs, control areas are Undbk to compen-
sate quickly for an abrupt large power mismatch caused, for SIMULATIONMODELS F0R AGC STUDIES
example, by a unit trip. NERC Guidelines define disturbance In various literature one can find analyses related to elec-
conditions for such events. The guidelines require that ACE in tro-mechanical transients and intermachine oscillations which
the disturbed area should be returned to zero within ten follow power balance upsets. Where these analyses utilize
minutes following the start of the disturbance. To satisfy this such variables as rate of change of power angles, or local
requirement for large power upsets, system operators typically frequency, etc., the phenomena being studied are really not the
need to manually intervene, to alter interchange schedules, to slow and deliberate process of AGC, but the dynamics
bring some hydro and combustion units on line, and/or to themselves. The requirement of anti-aliasing filters for the
manually q u e s t generation cbanges Erom some units whether typical time scales of AGC measurements prevents the
or not they are on AGC. In non-disturbed areas, depending on observation of such dynamics. The studied systems may come
the match between B and area p, AGC may be quite in- to some final off-nominal frequency, or other imbalance from
sensitive to the power upsel - thereby avoiding undo scheduled conditions which AGC is then responsible to
comproniise of naturdl wsponse. correct on its own time scale of action.
Disturbance conditions it] an area are not frequent evcnts. As U cpntrol control process AGC is n P i t k p r able nor t n n
They typically occur only a few tinics pcr nioritli. I n contrast be e.yper*ted to pluy any role in dcmipinR elecuw-mPclruiiic,~I
with its limited role under disturbancc cordictilions, AGC plays transients iricluding intermachine o.wi1kution.v.111 systems pos-
1111

sessing coordinated control steam units, it takes a few to sev- A. M. DiCaprio, A. A. Fouad, R. K. Green, Jr, K. Hill, R.
eral dozen seconds for AGC actions to be reali74. Therefore, K. McCrea, T. L. Overly, E. H. Preston, III, J. E. Price, R.
system models developed for AGC studies need not repment P. Schulte, G. B. Sheble, R. R. Shoults, L. H. Siddiqi, J.
phenomena having time coostants shorter than a few seconds. Singh, J. E. Troutman, S. V e m d , a d s . Virmani.
AGC studies, the momentary difference between the
frequency of different areas can be ignored. For all AGC pur- REFERENCES
poses, the frequency used for one area to compute ACE should [l] D. N. Ewart, “Automatic Generation Control -
be the same as used in other areas so long as they remain Performance Under Normal Conditions,” Systems
interconnected. In AGC practice, rapidly varying components Engineering for Power: Status and Prospects, U.S.
of frequency are almost unobservable due to filters involved in Government Document COW-750867.1975, pp. 1-14.
the process. [2] C. Concordia, “Power System Objectives’ Side Effects:
Good and Rad,” IEEE Power Engineering Review,
OVFRVIFW AND CONCLUSIONS September 1990, pp. 12-13.
A major factor that must be comidered in the design of a [3] R. D. Dunlop, and D. N. Ewm, “System Requirements
control system is the nature of the utility plant to be con- for Dynamic Performance and Response of Generating
tmlled. If it already exists, or its design has been completed, Units,” IEEE PAS, May/June 1975, pp 838-849.
the control system must accommodate the characteristics of [4] C. Concordia, F. P. de Mello, L. K. Kirchmayer, and R.
that design. Even if the control system is designed as part of P. Schulz, “Effect of Prime-Mover Response and
the overall design, so that insofar as possible the plant and the Governing Characteristics on System Dynamic
Control system designs are coordinated with each other, physi- Performance,” Pmxedings of American Power Conference,
cal and thermodynamic limits am non-negotiable, and often 1966, pp 1074-1085.
constrain control objectives. [5] M. R. Stambach, and D. N. Ewart, “Dynamics of
In the case of automatic generation control of electric Interconnected Power Systems: A Tutorial for System
power systems, the utility plant, i.e. those individual com- Dispatchers ami Plant Operators,” Electric Power Research
pany systems, pools or interconnections that comprise In.titute, Report EL6360-L, Section 9, May 1989.
subcontinental or continental systems, already exists and [6] E E E WG on Power Plant Response to Load Changes,
represents a massive investment that can only be changed “ M W Response of Fossil Fueled Steam Units,” IEEE
slowly over time in an evolutionary manner. Accordingly, any PAS, March/ApriI 1973, pp 455-463.
proposed contribution to improved control of such systems [7] D. N. Ewart, “Who‘s Watching Frequency These Days,”
must accommodate existing physical realities if they am to be Power Technology, Issue 53, April 1988.
meaningful. [8] D. G. Carroll, R. G. Serenka, and H. R. Propst, “BWR
Most important of these physical realities are Newton’s Maneuvering Capability,” Proceedings of American Power
laws of motion, and the effects of inertia that are entailed Conference, 1979, pp 73-78.
thereby. These effects are governing, not only at the level of [9] N. P. Mueller, “Response of Pressurized Water Reactors to
the entire system, with its massive inertial effects, but ako at Network Power Generation Demands,” IEEE PAS, October
the level of the individual boiler-turbine-generator units. An 1982, ~~3943-3950.
inescapable consequence is the impossibility of matching load [IO] A. Klopfenstein, uReSponse of Steam and Hydroelectric
and generation other than in a time-average manner. The degree Generating Plants to Generation Control Tests,” AlEE
of mirmatch that must be accepted is dictated by (i) the opera- PAS, December 1959, pp 1371-1381.
tional constraints of the utility plant, (ii) the limits of control [ l l ] P. D. Henderson, er al, “Cost Aspects of AGC,
effom that are available and accepted, and (iii) the signal/noise Inadvertent Energy and Time Error,” IEEE TPS,February
ratio in the instantaneous estimate of generation demand. 1990, pplll-118.
Control realization is limited by how much stored energy [12] L. S . VanSlyck, N. Jaleeli, and W. R. Kelley,
is available in generating units, and how rapidly its conversion “Implications of Frequency Bias Settings on Interconnected
rate can be changed. The signal/noise ratio is limited by ele- System Operation and Inadvertent Energy Accounting,”
ments of the component signals that do not represent true IEEE TPS,May 1989, pp712-723.
changes in load, by instrumentation noise and error (possibly E131 N. Cohn, “Discussion of : The Megawatt-Frequency
including heavy aliasing if not purposely avoided by filtering), Control Problem: a New Approach Via Optimal Control
by non-synchronicity of signals, etc. Additional, and some- Theory,” IEEE PAS, April 1970, ~ 5 7 4 - 5 7 6 .
times purposeful, constraints are imposed by deadbands in [14] N. Cohn “Control of Generation and Power Flow on
mechanical linkages, by tnmport delays, by catch-up speeds Interconnected Systems,” John Wiley and Sons, Inc. New
of control motors, and other inescapable nonlinearities. York 1961.
In view of the considerations described in this paper, [15] T. Kennedy, S . M. Hoyt, and C. F. Abell, “Variable,
attempts to reduce the root-mean-square magnitude of a sys- Non-Linear Tie-Line Frequency Bias for Interconnected
tem’s ACE below some threshold, or to achieve any reduction Systems Control,” IEEE TPS, August 1988, pp1244-
in a portion of its spectrum beyond some frequency, are 1253.
inevitably futile and counterproductive. 1161 K. P. Schulte, W.L. McReynolds, and D. E. Badley,
“Modified Automatic Time Error Control and Inadvertent
Interchange Reduction for thc WSCC Inlcrconneded Power
ACKNOWLEDGEMENTS
Systems,” Paper 90SM 304-7 PWRS, IVesenled at I99U
T h e Task Force (Authors) wisll to acknowledge
1EWpES Sunmer Meeting, m K a y o l i s , July 16-19.
contributioiis from: C. F. Abcll, J. R. Beachman, A. Bose,

I
I

1112

[I71 L. S . VanSlyck, N. Jaleeli, and W. R. Kelley, “A Scbeoectady, NY,where be conducted studies in bulk power
Comprehensive Shakedown of an Automatic Generation system stability, boiler and control simulations, excitation and
C00tr0l Pr-s~,” IEEE TPS,May 1989, ~ ~ 7 7 1 - 7 8 1 . turbine control analyses, automatic generation control synthe-
sis, and subsynchronous resonance. Mr. Ewart was responsible
BIOGRAPHIES for the formu1;ltioOof GE’s digital implementation of AGC and
Nasser Jaleeli (M’79) graduated from the College of following through with field installation. Mr.Ewart was: named
Engineering, University of Tehran in 1%7. He earned the PhD Manager, System Dynamics and Control in 1969, and
in Electrical Engineering from tbe Imperial College of Science Manager, System Performance Engineering in 1977. During
and Technology, of the University of London in 1975. He this period he made major contributions to define the role of
joined the teaching staff of Arya-Mehr University of large interconnected power systems.
Technology (AMUT) in 1967. While serving AMUT as an In 1980, Don changed assignments, becoming Manager,
Transmission and Distribution, with respomibilities for studies
Assistant Professor from 1975 to 1979, be actively served the
local industry and, in particular, as the head of Electrical and product application engineering for ac and HVDC
Services Division of Bonyan Consulting Engineers. Transmission, Distribution Systems, and for EPRI’s High
Dr. Jaleeli joined Ohio University in 1979, attaining the Voltage Transmission Research Center in Lenox, MA. Systeni
rank of associate professor before joining AEP Service engineering for several large General Electric HVDC projects
Corporation in 1983. At AEP he has been principal or co- was carried out under his direction.
principal investigator for several system control center pro- Mr.Ewart joined Power Technologies, TOC. in 1987 where
he is Manager, Consulting Services. Don i s a Fellow of the
jects. Automatic generation control has been his main respon-
sibility in recent years. He is a Principal Engineer in LEEE, and an active member of the Power Engineering Society
Operations Control Systems. and the Fellows Committee. He is the author of over 40 tech-
Dr. Jaleeli is lhe author or coauthor of several research nical articles and papers, and a registered Professional Engineer
papers in the Electric Power Systems field. He is a registered in the states of New York and Massachusetts.
Professional Engineer in the State of Ohio, and a member of
Eta Kappa Nu. Lester H. Fink (M’51, SM’S8, F’73, LF’90) received the
BS and MS degrees f” the University of Pennsylvania in
Louis S. VaoSIyck (M’SS, SM’63) has over 35 yean expe- 1950 and 1960. He has forty years experience in’eledric utility
rience in the elecaic industry, primarily in computer applica- systems engineering and re.search, including twenty-four years
tions for power system operation and engineering. He earned with tbe Philadelphia Electric Company. At Philadelphia, he
the BSEE and MSEE degrees from North Dakota State conducted researcb for and drafted the functional specifications
University, and tbe PhD from Illinois Institute of Technology, for the indusby’s first digitaUy directed automatic generation
all in Electrical Engineering. control system and, a decade later, its second-generation
Joining the American Electric Power Service Corporation replacement. Les has authored or coauthod some chi- papers,
in New York in 1968, Dr. VanSlyck participated in research, and holds two U.S. patents related to power system control.
development, and implementation of the h t large power sys-
tem network state estimator in the world. He is presently an Arthur C. Hoffmann (M’62, SM’75) graduated fmm the
AEP Senior Staff Engineer in Engineering Computer University of Pittsburgh with a BSEE degree. After graduation
Applications at Columbus, Ohio. be joined the EMS Department of Westinghouse Electric wbere
lo the WEE, Dr. VanSlyck has been chairman, 1966, of he beld various positions from Project Manager through EMS
the Red River Valley Subsection. He was an original member Depmment Manager. He is currently Chairman of ECC, Inc.,
of the Power Engineering Education Committee (PEEC) and which he co-founded in 1978. Mr.Hoffmann is also Chairman
served drat committee from 1963 to 1975, including four years of the System Control Subcommittee of the Power System
as its secretary. He was responsible for publishing the surveys Engineering Committee. He is a Registered Professional
of Electric Power Engineering E<lucatiooal Resources in U.S. Engineer in several states and a member of Eta Kappa Nu.
accredited schools for 1970, 1972, and 1974. I n 1980 he
nxeived the IEEE award for recognition of dislinguished seMce
to the Power Engineering Society.
Dr. VanSlyck is a meniber of Tau Beta Pi, Eta Kappa Nu,
and the Society of the Sigma Xi for “Dedication to Research in
Science.” He has been 3 registered Professional Engineer in
North Dakota since 1959, and in the state of Ohio siiice 1972.
He is the author or coauthor of several research papers in the
Electric Power Systems lieid.

Donald N. Ewart (M’SS, SM’75, F’78) is a native of


Buffalo, New York. He holds a BSEE degree from Come11
University and an MS degree i n Engineering from Union
College.
After gaining experience in the U S . Air Force and the
Generrrl Eleclric Transli)rlner Depdrtlllenl he joined GE’s
Electric Utility Systeais Engineering Department in
I

1113

DIBCUBSION Assume a three area system which is to operate under AGC. for
each area the formation of an ACE according the formula in the
JACX Y. YILLER, Cajun Electric Power Coopera- paper is assumed.
tive, Baton Rouge, LA. The authors are to be
complimented for presenting the technical de- ACE, = Ta, - Ts, - IOB, (Fa - Fs,), i = 1,2,3
scription of AGC so clearly. The industry
has needed such a fundamental review. It In addition the power balance must hold:
would be interesting if the authors would ex-
pand the subject somewhat and discuss the
Tal + Ta, + Ta, = 0
economics of AGC as well. Specifically, what Each AGC controller should function according to a control law which
is the affect of generating unit load- in the final steady state reduces the ACE to a small value which is
frequency control on production cost? Does zero if the law is integrating. If the ideal case is assumed (ACE = 0)
system load-frequency control increase or the following system of equations can be set up
reduce the system production cost?
Tal Ta, Ta, Fa
a1 - 10 B, = Ts, - 10 B, Fs,
1 - 10 B, = Ts, - 10 B, Fs,
1 - 10 B, = Ts, - 10 B,- Fs,.
1 1 1 =o
All the setpoints governing the exchange of power and the frequency
are on the right-hand side. All the variables are on the left side.
H. Glavitsch, (Swiss Federal Institute of Technology, CH8092 Zurich, The system has a solution if the matrix is non-singular which can be
Switzerland): This is a very much needed and timely paper which achieved if the sum of the bias values does not disappear. A first
clarifies a number of subjects of automatic generation control. The conclusion from this requirement is that at least one area has to do
authors and the task force are to be congratulated on their efforts and frequency control (one B must be non-zero). The converse is also true.
care in explaining the concepts, limitations and facts. Unfortunately, None of the areas or several together can do flat frequency control.
there have been several misconceptions of AGC earlier which have led This is equivalent to avoiding that one of the diagonal one’s becomes
to unrealistic investigations out of which papers have been published zero which would lead to singularity. As far as errors in measuring
whose results are quite impractical. The present paper puts almost all tie-line power or frequency is concerned which can be transformed to
of them in the right perspective and says what is possible and reason- errors in the setpoints it can be stated that the system has a stable
able in this field. solution irrespective of inaccurate setpoints. The right-hand side must
One of the main items which quite often has been the starting point not be consistent or must not have accurate values for Fsi, for
of misled investigations is the turbine-generator model which was example. The AGC system will generate a set of tie-line powers and
expanded to a two-area model for AGC. The parameters of this model one well defined frequency. They will deviate from the setpoints but
give the impression that the output of the turbine can be changed in a the system is stable which in general leads to inadvertent interchange
fast way. Studies which do not take into account the crossover ele- and to a time error.
ments in a thermal unit or the behavior of the penstocks in a hydraulic Thus, there is another indication that the AGC system is quite
installation nor the real sampling interval of the data acquisition robust allowing non-consistent setpoints without jeopardizing the sys-
system come to results where frequency or tie-line power could be tem.
corrected within one second. Irrespective of the fact that this is not It is hoped that the findings of this paper will find its due recogni-
possible there is no incentive to control the system in such a way as tion among system analysts and researchers.
the paper makes it clear from various directions. If frequency devia-
tions (several hundreds of mHz or over one Hz) are to be expected Manuscript received March 3, 1991.
other means like load shedding have to be employed.
In Europe there is an extended experience with AGC in the
UCPTE system. One feature is that area controllers with different
sampling rates (decision cycles) and different control laws (propor-
tional, proportional-integral) having different parameters cooperate
without difficulties. In at least one country AGC is implemented on
two levels, one receiving setpoints from the other. So the basic concept
of AGC is very robust.
It might be useful to consider the following few additional remarks Charles Concordia, Consulting Engineer,
which could be helpful in understanding the material in the paper. Venice, Florida. I would like to offer the
Under “Tie-Line Bias Control” a verbal explanation is given as to following comments:
the adjustment of the bias B, e.g. by the term “acceptable match.” In
the years of active development of AGC the term “non-interactive As a general comment, transmission in a single
control” has been introduced which means that no control actions company not only serves to go from generator
should be initiated, for example in area 2 when the power upset is in to load, but also serves to permit sharing of
area 1 (as assumed in the paper). This can be achieved by setting B load changes among all generators. This is
equal to p. The paper explains that p is not constant and difficult to well known, and within a company advantage is
measure or to obtain which is certainly correct. As a concept, however, taken of this to avoid large generation
it is worthwhile to mention it. It is to be kept in mind, that no “control changes on any particular generator. It would
actions” means no “AGC control” (ACE = 0) whereas primary con- be ridiculous to have tie-line control exten-
trol will contribute significantly to the support \of the other area. sively within a company. From a physical
At another point, namely in the conclusion of the chapter on point of view the-- benefits should be
multi-area control (written in italics), it is found by this discusser that obtained from interconnections between com-
the foregoing arguments are not logically leading to this statement. panies, Thus it is almost too obvious that
The points mentioned before that statement are subjects of the rigidly controlling “tie line” loading is -
control structure, observability and controllability. It is agreed that a silly.
quick return of the frequency is no basis for advocating an AGC
algorithm but the reasons for not aiming at a quick return lie in the The only negative comment I have is that, even
technical constraints of the turbine-generator set and not in the though I agree with most of the paper, it is
control structure. So, an explanation of this point may be useful. always dangerous to say that something cannot
For the purposes of discussing the effects of incorrect measure- be done in the future, and the somewhat nega-
ments and of various modes of AGC operation the following analysis tive tone of the paper may “turn some people
is offered which may complement the explanations in the paper. off“, who ought to read it.
1114

THOMAS KENNEDY (System Operations Consultant, The NERC interconnections have a multi-
St. Louis,Missouri) The authors and the System control area configuration and therefore must
Control Subcommittee are to be complemented rely on tie line bias control. However, some
for setting the stage for further discussion interconnections of the world, such as the
of the very complex topic of automatic British grid and the Chinese provincial
generation control (AGC). It is the opinion of systems, are centrally controlled as a single
this discusser that the conclusions reached in area. Single area control requires that
the paper (in italics) are correct. The North frequency be controlled by AGC but, of course,
American Electric Reliability Council (NERC) does not require direct AGC control of tie
Operating Committee is certainly very much lines. While the multi-area configuration must
concerned with developing an acceptable closely control frequency and interchange,
"definition of a desirable generation trend". because they are so closely interlocked (l),
The NERC Operating Committee promotes the centrally controlled system can use
the reliable and coordinated operation of the frequency as a regulating tool if their
North American interconnections through "biggest contingency to system size" ratio is
operating policies and procedures that are small.
contained in the Reliability Criteria for Areas of AGC technology that the
Interconnected Systems Operation (Criteria) discusser feels need further study include
and the Operating Guides (Guides). The HVDC interties characteristics, proper
Criteria and Guides (rather than guidelines) application of expert systems, and automatic
deal with all aspects of interconnection MVAR control between control areas.
contro1,including AGC, and are designed to
promote an orderly sharing of interconnection References
resources and burdens. While AGC is certainly [I] L.A.Mollman,and T-Kennedy.
an important part of interconnection control, "Interrelationship of Time Error, Frequency
it is just one facet of the overall Deviation, and Inadvertent Flow on an
interconnected systems operation. Interconnected System", Paper 31 TP 67-136
The paper should be read with the presented at the IEEE Winter Power Meeting,
perspective that the system operator has New York, N.Y. January 29-February 3 , 1967.
several operating tools and procedures at his
disposal, including AGC, that aid him in Manuscript received February 19, 1991.
operating his system in a reliable and
economic manner. Those same tools also allow
him to contribute to the reliable and economic
operation of his interconnection by following
the NERC Criteria and Guides. Present day
control room facilities coupled with present
day operating techniques and procedures are ROBERT P. SCHULTE (Consulting Engineer,
quite adequate for the system operator to Portland, Oregon): This paper does indeed
fulfill both of these obligations. The authors provide a good viev of many AGC topics that
are quite correct in stating that a number of are of interest. The folloving discussion is
the requirements and recommendations of the about the topics of acceptable ACE trend and
NERC Operating Guides, regarding system vear on regulation units.
regulation, cannot and should not be
accomplished through AGC. However, it must be In the paper the highest rated attribute of
remembered that the NERC Criteria and Guides AGC strategies for a multi-area inter-
can all be followed by the proper application connection is one that yields a more
of the operating tools and procedures at the acceptable ACE trend. Much further dovn the
system operator's disposal. AGC capability or list of strategies is an attribute that
intended use should not be confused with the accumulates lover cost associated vith vear
control area's obligation to follow the NERC and tear of regulation on all units. These
Criteria and Guides. strategies, vhich are given in preferred
The paper is correct in its description order, are not exclusive. To a limit, as is
of the "flat tie line " unstable control pointed out in the paper, an improved ACE
action when there is a tie line measurement trend does require greater response by
error on the part of one of two control areas. regulating units to AGC signals.
However, where the system response is
recognized through a frequency bias setting, Thus a more acceptable ACE means more strict
a stable frequency point will always load and schedule folloving by energy sources
theoretically be reached. Also, both the low in the area. In assessing the present state
and high bias setting cases, described on page of AGC the authors point out in their
4 of the paper, will result in a frequency and conclusion the futility of attempts to reduce
interchange movement similar to that shown in ACE performance belov some threshold. But
Figure 1 of reference 15. It should be noted vhat vi11 happen in the future as pover
that tie line bias control was developed to system demands grow? Increasingly additional
eliminate some of the disadvantages of "flat loads are being met vith scheduled imports
tie line" or "zero bias" control. and alternative source or non-utility
The strategy that discusses the cost of generators vithout AGC Capability. Also, as
"wear and tear" of regulating generating units is mentioned, nuclear plants are not usually
should also mention the obligation of each used for AGC. Will ve then be obliged to put
control area to carry its share of the even greater AGC duty on existing regulation
regulation burden and not cause undue burden units, vi11 ACE control performance threshold
to its interconnection neighbors. It also suffer or vi11 we reach a time when ve must
should consider the differential cost of an employ special energy conversion devices of
uneconomic combination of regulating units some kind designed for area load folloving?
that could result from insufficient regulating
capability of some generating units. Manuscript received March 4 , 1991.
I

1115

Carson W. Taylor, Bonneville Power Administration, make very large hydro generation changes within one
Portland, Oregon: The purpose of this paper seems to be minute.
to lay to rest the "mischief" started by Elgerd and Fosha The Puget Sound (Seattle) area voltage stability
[A,B] over twenty years ago. Although many of the problems can be improved by dropping hydro generation
basics in the present paper were given in the discussions in eastern Washington state and ramping generation to
of references A and B, many subseqent academic papers the north (B. C. Hydro) and to the south. Activation of
have made unwarranted assumptions. reactive reserve at generators or SVCs could be used to
AGC is a slow (tens of seconds) follow-up control to trigger the generation changes.
return frequency and interchange to scheduled values
following disturbances. For system dynamic perfor-
mance, attention should be paid mainly to the primary You say you are providing a resource to guide research
powerplant controls (excitation, prime mover, and en- into topics related to AGC. What research do you think is
ergy supply controls). These controls are often poorly needed?
tuned. Data is generally unavailable for energy supply I think better characterization of both prime mover con-
system models. Area control cannot be responsive if the trol deadbands (andor backlash) and natural regulation
power plant generation controls are either sluggish or may be useful. The statement is made that many gover-
excessively oscillatory. nor deadbands are beyond 0.035 Hz (0.06%). Reference F
Besides helping the bookkeepers and keeping clocks and its discussions address this subject; see also refer-
accurate, there are other reasons for good AGC perfor- ence G.We should realize that a t any point in time, gov-
mance. Part of the difficulties in connecting the eastern ernors may be positioned randomly within deadbands
and western North American interconnection in the late and for large systems a statistical characterization is
1960s was probably due to poor AGC performance such useful. Many hydraulic governors have vibration motors
a s undergeneration during morning pickup. The diffi- t o reduce deadband by dither modulation; process noise
culties led to an expensive solution-back-to-back dc may also provide dither modulation. Do you advocate in-
links. My study indicated, however, that the generation tentional deadbands for electronic prime mover control?
and transmission additions in Montana, Wyoming, Incidentally, IEEE Std 122-1985 IEEE Recommended
Nebraska, and the Dakotas of the 1970s would have Practice for Functional and Performance Char-
allowed successful ac interconnection. Partly because of acteristics of Control Systems for Steam Turbine-
concerns related to the poor AGC performance of the Generator Units depreciates the term governor % ad-
eastern interconnection the study was rejected by west- vance the understanding that control systems need not
ern utilities. Special generation control at selected power be limited to rotating flyweights but include mechanical,
plants and improved AGC logic might have improved hydraulic, and electronic components."
performance of the interconnection. Improved AGC Do you have evidence regarding the statements that
logic might have been reduced frequency bias coeffi- relatively small generation-load imbalances are arrested
cients for western utilities during morning pickup mainly by load response?
hours of the eastern interconnection-. The bias coeffi-
cients should be closer to the early morning western You say that "the system natural response coefficient, p,
natural response (p). Such logic is feasible with digital is not a constant, neither is it accurately obtainable nor
AGC . predictable." Do you favor research of on-line estimation
of p with a view toward adaptive control (setting B = p)?
Related to the above paragraph, recall that a motivation From time-scale separation or psuedo steady-state
for the Elgerd and Fosha work was the east-west inter- notions, "perfect" calculation of ACE results with B = p. I
connection problems. have always felt this attacks unnecessary control and
The paper focuses on the potential for automatic genera- oscillation problems at a fundamental level.
tion control to improve system dynamic performance. I commend you for a paper that advances the under-
standing of automatic generation control.
Nowadays, slowly occurring voltage instability is a
major industry concern. One solution is fast emergency A . 0. I. Elgerd and C. E. Fosha, Jr., "Optimum Megawatt-
automatic generation control. There is a race between Frequency Control of Multiarea Electric Energy Systems,'
generation increase to improve stability and generator IEEE Transactiom on Power Systems,vol. 89, no. 4, pp. 556-563,
current limiting and transformer tap changing. Several April 1970.
examples can be cited to show the importance of genera- B . C. E. Fosha, Jr. and 0. I. Elgerd, "The Megawatt-Frequency
tion control: Control Problem: A New Approach via Optimal Control
Theory," IEEE Transactwns on Power Systems, vol. 89, no. 4,
A voltage collapse in southern England on May 20, pp. 563477, April 1970.
1986 was averted by emergency start of 1000 Mw of gas C M. G. Dwek, Study Group 38 discussion, Proceedings of
turbines within five minutes [Cl. 33rd CIGRE Session, vol. 11, 1988.
Voltage stability along the Pacific AC Intertie [Dl D . W. Mittelstadt, C. Taylor, M. Minger, J. LuiN, J. McCalley,
for disturbances such as loss of the 3100 MW Pacific DC and J. Mechenbier, 'Voltage Instability Modeling and Solutions
as Applied to the Pacific Intertie," CIGRE 38-230,1990.
Intertie could be improved by fast generation changes in
E . D. N. Scott, R. L. Cresap et al., 'Closed Loop Digital Automatic
the Pacific Southwest. The Hoover hydro power plant Generation Controller,' IEEfPES C73 518-8.
and several pumped storage plants are candidates for
F . C . W. Taylor, K. Y. Lee, and D. P. Dave, 'Automatic
large power increase. AGC in the Southwest can detect Generation Control with Governor Deadband Effects," I E E E
the loss of the major tie-lie (DC Intertie) and command Transactions on Power Apparatus and Systems, vol. PAS-99, pp.
the generation change. Controls are available [E] to 2030-2036, Novembermecember 1979.
I

1116

G. C. W. Taylor, discussion of "Inertial, Governor, and A w l nitude constraint:


Economic Dispatch Load Flow Simulations of Loss of Gen-
eration Contingencies," IEEE Transactions on Power
Apparatus and Systems, vol. PAS-104,no. 11, pp. 3020-3028,
+
( X ( t 1) - X ( t ) ( I&,e
November 1985.
e) The rate of change of state variables must be zero at t = N;
i.e., the system must reach a smooth level at the final time
of control (i.e., X(t,) = 0)

X ( t - 1) - X ( t ) = 0
A. Kryhrni (The Ohio State University, Columbus, OH): The
authors have performed an invaluable nervice to the utility in- The problem is to find a control sequence U =
dustry by presenting the practical limitations of AGC which are (Uo, U , , . . . Ut-l) and a corresponding trajectory S =
constrained by the dynamic response of the process involved. (Xo,X I , . . . Xt) determined by the above equations and condi-
The authors have defined a number of strategies that clearly tions while minimizing the performance index J:
define what AGC is expected to perform when it is controlled
M a process. In these strategies, limited knowledge of process
dynamics is utilized. However, a strategy which uses the gross t=o
dynamics of a boiler for the dynamic control of generation and
load may also be considered. In the following, such a strategy where D ( t ) is a weighting vector on control actions.
will be presented, and the authors' comments concerning the For such a control, the gross boiler model developed by
proposed strategy are welcomed. Astrom [I] can be used. In this model, the control variables
Let the gross dynamics of a boiler be represented by the fol- are fuel flow, valve setting and feed water flow, and the state
lowing linear, vector difference equations. variable is the drum pressure.
The above control strategy may be possible, if the models of
X(t + 1) = X ( t ) + B U ( t ) power plant components are developed for AGC control. AU-
thors' comments concerning the above are welcomed.
where
[I] Astram, K.J. and Eklund, K., "A Simplified Non-Linear
Model of a Drum Boiler-Turbine U i t , " Int. Journal of Con-
trol, Vol. 16, No. 1, 1972.
and P ( t ) is the MW generated at time t .
Also, ansume the energy delivered to the system is con-
strained to be equal to the energy consumed by the load over
the next few minutes; that is
T.S.Bhatti,D.P.Kothari,and J - S a t i s h (Indian
I n s t i t u t e o f Technology D e l h i , I n d i a ) : We w i s h
t o commend t h e a u t h o r s f o r their extremely
valuable and t i m e l y c o n t r i b u t i o n i n the
where &(to) is generated M W at time to;ACEis the area control important area o f AGC. T h i s Task Force paper
would go a l o n g way i n p r o v i d i n g a g r e a t
+
error, Pf is the forecseted load a t time to 2'; and T is the resource t o r e s e a r c h e r s , p r a c t i s i n g engineers
final time of control. T is appropriately chosen based on boiler and s t u d e n t s t o understand and t o c a r r y Out
response. The discrete version of the above equation can be further r e s e a r c h i n t o AGC related topics.
written BB However, we would l i k e t o o f f e r some comments
and seek the author's c l a r i f i c a t i o n on some
N-1
points.
pc(t) t A C E t PfT 1. I t i s a p r a c t i c e by u t i l i t i e s t o use a
P(t)At = small value o f speed droops f o r active
1=0 2
generators i n the range o f 3% t o 5%.The
percentage droop o f each u n i t determines the
The constraints of the control problem can be expressed as
s h a r i n g o f a d d i t i o n a l l o a d amongest the var-
i o u s generating u n i t s . & higher value o f speed
a) Final time of control in fixed droop ensures more s t a b l e p a r a l l e l o p e r a t i o n
whereas a l o w e r v a l u e o f speed droop p r o v i d e s
b) Initial and terminal boundary values are given better l o a d frequency c o n t r o l . T h u s a b a l a n c e
has t o be s t r u c k between t h e two t o a r r i v e a t
X ( 0 ) = So optimum v a l u e . ~n o u r computer simulation
X ( N )= X(T) s t u d i e s f o r AGC we have f o u n d t h a t l o w e r v a l u e
o f speed droop(4%) and h i g h e r v a l u e o f speed
droop (10%) g i v e more or l e s s same dynamic
c) The control variables and the state variables must satisfy responses(frequency,tie-power and ACE d e v i a t -
a magnitude constraint: i o n s ) f o r reheat t h e r m a l u n i t s i n presence o f
generation r a t e constraint(GRC) of 3%/min.
T h e r e f o r e , one wonders i f i t i s possible t o
Xnin I X ( t ) I Xln, p r e f e r a higher value of speed droop for
k i n I U ( t ) 5 Um.x t h e r m a l u n i t s ? . I f t h e answer i s yes, then,
w i l l t h e governor be e c o n o m i c a l l y c h e a p e r ? . I f
d) The rate of change of state variables must satisfy a mag- t h e answer i s no, t h e n f o r what reasons?
-

1117

However. f o r hydro u n i t s g e n e r a t i o n r a t e cons- addition to ACE processing logic


traint i s v e r y h i g h (270%/min f o r raising introducing further dealys. I n v i e w of
g e n e r a t i o n and 360%/min f o r l o w e r i n g g e n e r a t - t h i s how f o r i t i s j u s t i f i e d t o u s e 2 - 4
ion) and i n o u r s i m u l a t i o n s t u d i e s we have seconds for a d a t a a c q u i s i t i o n and
found t h a t a l o w e r v a l u e o f speed droop(4%) decision cycles of the AGC s y s t e m s ?
g i v e s much b e t t e r frequency s t a b i l i z a t i o n as Shculd a h i g h e r v ~ l u eo f s a m p l i n g p e r i o d
compared t o t h e h i g h e r v a l u e o f speed droop b e e x p l o r e d t o p r o v i d e more o r less b e s t
and GRC l i m i t was never v i o l a t e d . I n f a c t . i n t h e
transient response and reduces
Northern g r i d o f I n d i a , speed droops o f many simultaneously the sampllng eiforr?
h y d r o u n i t s a r e i n t h e range o f 3% t o 4%. Carpentier[l] i n h i s s t a t e of t h e a r t
2. We f u l l y agree t h a t for t h e t h e r m a l r e v i e w p e r t a i n i n g t o AGC h a s s t a t e d t h a t
u n i t s , g o v e r n o r t i m e c o n s t a n t ( T g ) a n d t h e steam a sampling period v a r y i n g f r o m 2-10
c h e s t t i m e c o n s t a n t ( T t ) a r e much s m a l l e r than
seconds is used by u t i l i t i e s . B o s e a n d
Atiyyah [ 2 ] h a v e u s e d sampling periods
t h e o t h e r t i m e c o n s t a n t s used f o r simulation varying f r o m 2-30 s e c o n d s . would
o f AGC m o d e l [ l ] . We
t h e r e f o r e l i k e t h e a u t h o r s t o comment o n
3.The l o a d changes occur always i n random t h e o p t i m u m s e l e c t i o n of t h e sampling
f a s h i o n i n a r e a l system which r e s u l t random period.
o p e r a t i o n of t h e governor a c t i o n and produce
s t o c h a s t i c v a r i a t i o n s i n t h e frequency and t h e 3. I n t h e objectives of AGC t h e a u t h o r s
tie-line power nominal values[2].How t h e ACE state the strategies that exploit
signal will be m o n i t o r e d i n this case u n i l a t e r a l i n a d v e r t e n t c o r r e c t i o n s , which
f o r d e r i v i n g t h e supplementary c o n t r o l l e r ? . simultaneously reduce time e r r o r m a y b e
4. We have a l s o found [21 t h a t t h e e f f e c t of considered to deserve higher score.
governor deadband o n t h e dynamic response i s T h e s e are a c h i e v e d by a n AGC s t r a t e g y
n o t s i g n i f i c a n t b u t i t has a tendency t o f o r c e based on n e w a r e a c o n t r o l e r r o r (ACEN)
t h e dynamic response t o o s c i l l a t e for long 1 3 1 . T h e ACEN i s based on t i e power
p e r i o d around i t s steady s t a t e v a l u e . deviation, frequency deviation, time
Once a g a i n we would l i k e t o heartily e r r o r and inadvertent interchange. The
c o n g r a t u l a t e the authors f o r an e x c e l l e e n t AGC s t r a t e g y b a s e d o n ACEN g u a r a n G e e s
piece of work and l o o k f o r w a r d t o their zero steady s t a t e e r r o r of i n a d v e r t e n t
f u r t h e r i n v e s t i g a t i o n s i n the f i e l d . i n t e r c h a n g e a n d time e r r o r a c c u m u l a t i o n s
REFERENCES i n a d d i t i o n t o r e g u l a t i n g ACES t o z e r o
[l]. 1.J.Nagrath and D . P . K o t h a r i " Modern following load perturbations. W e would
Power System A n a l y s i s ",I1 edn, Tata McGraw- a p p r e c i a t e t h e a u t h o r s comment o n t h e u s e
Hill,NewDelhi,l989. o f AGC s t r a t e g y b a s e d o n ACEN.
[ 2 ] : S.C.Tripathy, T . S . B h a t t i , C.S.Jha, O.P.
M a l i k and G.S.Hope, " Sampled d a t a Automatic 4. I n a n AGC s y s t e m t h e n ~ t h o r sm e n t i o n t h a t
G e n e r a t i o n c o n t r o l A n a l y s i s w i t h Reheat Steam it is not d e s i r b a l c L O m a i n t a i n ACE a t
T u r b i n e s and Governor Dead-Band E f f e c t s " , I E E E z e r o as t h i s would r e q i l i r e u n n e c e s a r i l y
Trans.on P A S . v o l . P A S - 1 0 3 , N o . 5 , M a y 1984,pp.1045 r a p i d maneuvering of units. Could t h e
-1051. authors throw some l i g h t on a p p r o p r i a t e
selection o f m i n i m u m l e v e l o f ACE f o r
AGC?

REFERENCES
J.NANDA, M.L.KOTHAR1, L. H A R I , G .G . B H I S E
( D e p a r t m e n t of E l e c t r i c a l E n g i n e e r i n g , I . I . T .
Delhi, I n d i a ) : We w o u l d l i k e t o c o n g r a t u l a t e 1. J. CARPENTIER " S t a t e o f t h e a r t r e v i e w ,
t h e a u t h o r s f o r a useful. and i n t e r e s t i n g To b e or n o t t o b e modern" t h a t i s t h e
paper . providing indepth understanding of q u e s t i o n for Automatic Generation Control
a u t o m a t i c g e n e r a t i o n c o n t r o l . We s h a l l v e r y (point of v i e w of a u t i l i - . y e n g i n e e r ) " ,
much a p p r e c i a t e t o h a v e t h e commeiits o f t h e Int. J. E l e c t r i c a l Power a n d Energy
authors on the following points : S y t e m s , Vo1.7, A p r i l 1 9 8 5 , pp.81-91.
1. The a u t h o r s mention s p e e d d r o o p s for 2. A.BOSE a n d 1.ATIYYAH " R e g u l a t i o n E r r o r i n
active governors t o be i n t h e range of Load F r e q u e n c y C o n t r o l " , IEEE T r a n s o n
5%. A p p a r e n t l y l i t t l e a t t e n t i o n h a s b e e n Power A p p a r a t u s a n d S y s t e m s , V o l . PAS-99,
p a i d t o t h e optimum s e l e c t i o n o f g o v e r n o r M a r c h / A p r i l 1980, p p . 6 5 0 - 5 5 7 .
regulation parameter R which d e c i d e s t h e
g o v e r n o r droop. Our i n v e s t i g a t i o n s for a 3. M.L.KOTHAR1, J.NANDA, D.P.KOTHAR1, and
two e q u a l area s y s t e m show t h a t i n t h e D.DAS. "Discrete Mode Automatic
presence of t h e supplementary control Generation C o n t r o l o f a Two Area R e h e a t
(i.e. AGC) with a generation rate Thermal System w i t h New Area C o n t r o l
c o n s t r a i n t (GRC) o f J X l m i n , a m u c h h i g h e r Error", I E E E T r a n s On P o w e r S y s t e m s V o l .
value of percentage droop even t o t h e 4 , May 1 9 8 9 , p p . 7 3 0 - 7 3 8 .
t u n e of a r o u n d 1 5 % may b e p r e f e r r e d
that provides b e t t e r transient response.
We b e l i e v e t h a t a h i g h e r v a l u e o f R m a k e s
the realization of t h e governor s i m p l e r
and r e d u c e s its c o s t . Could t h e a u t h o r s
from t h e i r e x p e r i e n c e s u g g e s t g u i d e l i n e s J. Z. PONDER, G. A. CUCCHI, (Pennsylvania-New Jersey-
for optimum s e l e c t i o n o f R f o r g o v e r n o r Maryland (PJM) Interconnection, Nomstown, PA. 19403):
design?
This paper provides an excellent tutorial summary of
2. The a u t h o r s r i g h t l y mention about t y p i c a l Automatic Generation Control. However, it is a bit frustrating
f i l t e r s on tie-line metering introducing for the reader to see the stage set so well and then be left
delays of up t o a few seconds in without seeing the play! While the general notion of what
1118

constitutes "good control" has been discussed at length within Traditional AGC schemes do not allow for this. Shouldn't this
the industry, no one has yet defined the specifics. be one of the aspects for which strategies are compared to
As it stands today, some control areas do an excellent job rank AGC schemes? At PJM, we find that we must operate
of control as evidenced by their NERC A1 and A2 control to transfer limit restrictions about 60% of the time. This
performance criteria. Some control areas approach 100% "split cost" operation of the system requires us to use manual
compliance with the NERC criteria. Other control areas control.
perform very poorly, as defined by existing standards, with A1 Another subject of interest is the NERC disturbance
and A2 compliances below 50%. Probably adequate recovery criteria. The control areas in the MAAC and NPCC
performance lies somewhere in between these two extremes. regions of NERC have instituted a joint Shared Reserve
Can the authors comment on the reasonability of the existing procedure in which neighboring control areas will assist a
NERC criteria for measuring control performance and/or control area which has experienced a large disturbance. The
suggest other specific measures of performance which would assisting areas help the disturbed area to return ACE to zero
be better? within ten minutes. Are the authors proposing that this is
It may also be desirable to maintain "good control" for unnecessary? It would seem that there is indeed some value
other than reliability reasons. For instance, it is desirable to in quickly recovering from a system disturbance, since
maintain a reasonable match between scheduled and actual additional disturbances, if they should occur before recovery
interchange for accurate accounting purposes. If a control of the initial disturbance, would cause a further drop in
area does a poor job of generation control, then its actual frequency, possibly resulting in load shedding.
interchange may have little relationship to its scheduled There are many specific issues yet to discuss and resolve
interchange, resulting in large inadvertent interchange concerning automatic generation control. We hope that the
accumulations and gross inequities for the parties involved Task Force and the industry will be able to develop more
with the transactions. Inadvertent interchange is often paid specific recommendations for automatic generation control.
back at a more favorable time than it was accumulated. The The specific answers to the question 'What is good generation
requirements for "good control" to prevent such inequities control?" still remain unresolved.
may exceed that required for reliability. Has the task force
examined this and if so, what were the conclusions? Manuscript received March 4 , 1991.
On the subject of generating unit characteristics, we have
seen instances where the modem coordinated control systems
have resulted in poorer initial response from units.
Apparently, the coordination of the turbine output with the
boiler response restricts the unit from using stored energy in
the boiler. Older systems allowed wider swings of boiler M. K. Enns, (Electrocon International, Ann Arbor, Michigan): A few
pressure and temperature as the stored energy was drawn years ago I had occasion to examine the AGC behavior of a control
area in connection with a dispute between a utility and a steel-making
from the boiler. Coordinated controls block the steam valves customer. The arc furnaces operated by the .customer caused large
from opening if it would mean drawing down the boiler swings in demand that made it impossible for the utility, which
pressure or temperature. This is of course good from a boiler constituted a small control area, to meet the NERC Control Perfor-
control standpoint, but results in less aggressive initial mance Criteria. The utility experienced rapid and erratic response of
governor and AGC response from the unit. their units under AGC without any apparent benefit in the form of
good control. They were, in the language of the paper, "chasing high
The paper discusses kinetic energy and load frequency frequency components of demand change." They eventually took most
response from a very traditional viewpoint. Can the authors if not all of their units off AGC and lived, with no cost and little
comment on the effects of the newer dynamic controlled inconvenience, with a somewhat large and nonconforming ACE.
loads, such as load commutating inverters, which digitally This solution was really quite satisfactory. The ACE was large
sense frequency and may draw more power to maintain the because the control area was so small. Combining with neighboring
utilities to form a control area ten times larger would reduce the
local loading level? Such modem air conditioning controls effective ACE by a factor of ten, which would easily conform to the
had an impact on the July 23, 1987 Tokyo voltage collapse. NERC criteria. What appeared to be excessive interarea flows would
Can traditional AGC techniques cope with these dynamic types simply be internal flows, unmeasured and unnoticed.
of local load control?
On a more general level, the coordination of AGC with
frequency, stability and voltage controls needs to be
addressed. While the authors point out that many of these
items have a much shorter time constant than AGC and H. H. Thompson, (Chetty Mamandur, Energy Services, Inc., Pine
Bluff, AR): We congratulate the authors for clearly summarizing the
therefor AGC cannot be expected to react to them, there have importance for tie-line bias control it he interconnected power system
been instances where such problems have developed over operations. We agree with the comments and conclusions of the paper
periods of time consistent with AGC response capability. Lack on impacts and penalties of overachieving AGC control. In systems
of coordination of AGC with these problems can aggravate with significant nuclear units, which are not currently controlled by
rather than help the situation. It is quite conceivable that AGC, fewer remaining units are expected to provide all the control,
hence requiring increased maintenance. Moreover, to accommodate
AGC may take action counter to logic programmed for special the nuclear base load generation during minimum load periods,
relay schemes, under voltage relay load shedding and thyristor fossil-fired system-turbine units are being converted to sliding pressure
controlled devices which act independently from the AGC. type control to achieve lower minimum run levels. However, this mode
Shouldn't the AGC strategy which best coordinates with these of operation seems to result in reduced controlling capabilities for the
other independent controls be given some bonus points in the units hence resulting in increased control demands from other units.
Thus, the systems are forced to regulate with fewer and .fewer units
ranking? that can be trusted to control. Hence, it is becoming very important
Many control areas today must control to specific power that significant efforts be devoted to identify any unit problems and
transfer interface limits in addition to net interchange. get as many units as possible, to participate in AGC control.
1119

Recently, there seems to be discussions for including integrated


ACE in addition to ACE to determine unit desired generations.
Would the authors kindly comment on the implications and impacts of
this control.
Manuscript received March 4, 1991.
Economic Tnnd

Tm
Unit 1
Jaleeli, VanSlyck, Ewart, Fink, and Hoffmann: We would like A

1.
to thank all the discussers for their interest in the paper as well
as their contributions. Our closure addresses points raised i n
the discussions and provides additional comments.
Mr. Miller’s questions are very much appreciated a s tliey
provide an opportunity to emphatically state that: “of-course,
operating units with a well-designed AGC procedure rctliicrs 7
the total production cost in the area.” AGC systenis allow I b
m. Tlnv
much more frequent adjustment of the output power of cdcli Unit 2
unit as compared to a manual system where the adjustnicrit
Figure 3: Unit 1 la verbally maneuveredand,
of each unit requires verbal communication b e t ~ e e i ia central hance, can reallze generationchanges only
control coordinator and the unit operator. Therefore, output in large sleps. As a result, to match total
power adjustments of units under AGC can be performed with generationwith load, a step change in target,
and otherwise unnecessary movement, is
much finer steps than are possible with verbal control. Hence, unit2 ma imposed on automaticallycontrolledunit 2.
. .-
the unit output power trends can be held much closer to eco- Figure 2; BMh unita are aulMMticplly Both units are consequently operatsdaway
nomical trends with AGC than wlieri all units are unclrr verbal Conbolled. from their economic trends.

control, or a mixture of AGC and verbal C O l l h Y J l .


A review of Figures 1, 2 and 3 supports this claim. Figure
under AGC, can realize a savings of $350,000 per year in fuel
1 shows the load trend over a period of several minutes lor
costs.
an area with two units. The optimal economic trend for each
unit, as shown in Figures 2 and 3, are those that will minimize In reality, load variation is not a smooth line as shown
the area production cost over the given time window while in Figure 1; it fluctuates irregularly about its trend. While
their sum equals the area load trend. As shown in Figure 2 present NERC Control Criteria do not demand a tiglit match
where both units are under AGC, the output power of each between generation and load, many control areas cause their
unit is adjusted frequently at fine increments and trcnds are generation to follow these fluctuations to some extent. Moving
therefore maintained close to those of optimal rcononiiis. “lie generation up and down will cause units to operate off eco-
small deviation of the output power trend of each iinit fioin nomic base points some portion of the time. This can increase
optimal is due to irreducible imperfections of tlie AGC process production cost and, in some literature, it is called rrgulation
including inaccuracies in the prediction of unit response to the cost. Various authors go still further afield and call this the
control signals sent. cost associated with putting units on LFC or even, unfortu-
nately, on AGC.
In the scenario depicted in Figure 3, unit 1 is verl)ally coli-
trolled and unit 2 is under AGC. Limitation on tlie frequency Matching generation with area load fluctuations results in
of verbal communications between a control center coordinator an unavoidable increase in cost over that of constant power
and the unit operator constrain the output powcr of unit 1 to output. Whether this matching is accomplished by verbal coni-
be adjusted just once in the given time window, and with a munications or automatically by AGC, the cost of doiiig so is-
necessarily large magnitude as shown conceptually in the up- an inherent requirement of the business, not of AGC, (it is sig-
per portion of Figure 3. Therefore the deviation of the trcntl nificantly less, we might add, than would be required without
of unit 1 output power from that of its optimal economic trrntl the benefil of interconnected operation.) Rut, in fact, ihr cost
is much larger than that of the Figure 2 scenario. This devin- of matching load with generation, when the matching is ac-
tion also impacts operation of unit 2. As the sum of the pover complished by a well-designed AGC, can be much lower than
generated by the two units must equal the load, the generalion when units are verbally controlled. Operating an increased
trend required from unit 2 must also differ substantially from number of units under a well-designed AGC realizes reduction
its optimal economic trend. in system production cost.
The difference between area production cost undcr tlie scc- The first two paragraphs of Professor Glavitsch’s discussion
nario shown in Figure 2 and that in 3 depends on the incremen- which emphatically state the main messages of the paper are
tal cost characteristics of the units. Let XI and XL respecti\ely very much appreciated. We agree with him that, if R = 8,
be the area incremental cost at the start and at the end of the ACE would remain insensitive to external disturbances, but \{-e
shown time window. The greater the difference betweell dl have not advocated adjustment of 13 in an attempt to obtaiii
and Xp,the more savings result from operating the units uiidrr an “acceptable match.” Our words “acceptable match” were
AGC. Of-course, if A , and X2 are equal, the prodiiction cost between generation trend and demand, not I? and 8. In line
for the scenarios shown in Figures 2 and 3 would he tlic same. with the discusser, our message for normal operation is: “it
However, tests conducted in 1989 on the AEP system (with in- is impractical and unnecessary to attempt a precise matrli of
stalled generation capacity of 23500 M W ) have indicated that, area generation with the trend of an area’s own load and losses,
on average, operation of an additional 1000 MW of capability T,, and the area portion of lO,B(F, - F?).”The iriterconnrctioii

F
11%

will continue successful operation as long as these trends arc the area can and should avoid following fast varying cornpo-
“acceptably matched.” nents of load. Such attempts are without benefit to the area
On the paper’s statement of “a quick return of eit her fre- or to the interconnection, and often can amplify ACE and in-
quency or interchange to the schedule must not be used as crease the cost of wear and tear.
a basis for advocating an AGC algorithm,” and in agrernieiit, The belief stated above is also relevant to some aspects of
with what can be inferred from the discusser’s comments, we Mr. Schulte’s comments, to which we would like to add the
mean that the interconnection is hurt if every area should at- following. As far as response to fast varying compoiicnts of
tempt to quickly (or even slowly): load is concerned, the typical frequency trend in tbe Eastein
Return its interchange to the schedule with no consider- Interconnection indicates that the present level of AGC regitla-
ation of system frequency. tion is satisfactory. To continue this level of performance, the
or generation adjustment capability, as related to the r a t r ol load
Return frequency to the schedule with no consideration change during pick-up and drop-off, must be maintained near
of its interchange. its existing level. Relative to the “ACE control performance
threshold” point of the discusser, in our opinion, no energy
The very astute equations presented by Professor Glavitscli
conversion device of short h c d output will be necessary in f h e
for analysis of a three area system support the previous point
AGC domain, notwithstanding potential applications of such
and complement the explanations in the paper on the subject.
devices for suppression of undesirable transients in a fnturr,
The AGC process has some characteristics which are intrinsic
and some which are a manifestation of inaccuracies and un- but as yet undefined and unconventional manner.
avoidable imprecision. An attribute of Dr. Glavitscli’s equa- When ACE, in compliancr with performance criteria, frr-
tions is that all error effects, offsets, etc. (of which there are quently crosses zero and has a small average, it essentially mea-
many in the AGC process) are constrained to the right Iiand sures fast area load dynamics that ought iiot be a concern of
side while everything on the left hand side is exact. AGC, and acceptable control does not require any reaction to
From the left hand side of the equations, Dr. Clavitsch them. The present performance criteria for normal conditions
made the observation that at least one area must do frequency have a ten minute time scale. While AGC should adjust area
control. It can be noted too, that “flat” frequency control can generation to yield an acceptable ACE trend, it should not at-
be done in one area, but not more than one. By the cun- tempt to react to components of load whose oscillation periods
ning construction of Professor Glavitsch’s equations these ob- are comparable to area generation response time. Attempting
to react to such components does not produce an iniprovcd
servations can be seen to be intrinsic to the ACC process and
ACE trend and requires unnecessary nianeuveriIig of units.
independent of its inaccuracies.
In line with Mr. Taylor’s comments, we believe in utiliza-
Considering collective AGC action in the interconiircf i o n .
tion of AGC for enhancing any aspect of area operation lwfor-
we may add:
mance, including voltage stability, when it can be accomplished
C A C E = -1OB,(F,, - F % ) by adjusting unit generation within the constraints of their as-
where sociated characteristics. AGC data captiiring and plotting p ~ < ’ -
B , = B1 + BP -t B ~ land
, grams developed at AEP (171 have provided the opportunity of
observing the response of about sixty major AEP units to small
+
Fa = ( B I F ~ I + BJ’,,)/B,. and large (60 mHz) frequency upsets. The measured frequency
Therefore, for stability, B , must be negative as wrll as non- in these plots is the one second average of local frequellcy, atid
the measured unit power output is observed through an rxpo-
zero. Clearly, if B , is positive, any deviation from F, will be
nential filter having a time constant of -3 seconds. Presented
amplified until the system collapses. Beyond this point how-
below is a summary of these observations.
ever, it is not sufficient that l?,be negative, because tlicre may
be occasions when a number of areas are unable to match thrir While many units present a discernible response to frr-
obligations. For successful operation of the interconnection on quency upsets exceeding 40 mHz, only a few respond wlien
such occasions, it is necessary for the sum of the L? in the set the upset is -15 mHz. Wherever “dead-band” is referenced
of remaining areas to be negative. in the paper or closure, it indicates the effective dead band
The first paragraph of Mr. Concordia’s discussion nicely in the governor loop. Inspection of the frequency trend and
helps to promote one o f t e main messages of our paper. We the power output of units, even for tlie latter few units, lias
feel so strongly about th j that we want to reempliasize his shown no correlation when upsets are below -10 m l L . Since
point: “It ts obvtow th .t over-control of t i e - h e loading ES AEP’s primary generation response to frequency upsets, espr-
sally.’’ It is certainly not intended to discourage research in cially those exceeding -25 m h , is very close to that of tliv
the AGC area. We have stated today’s practical limitatioiis Eastern Interconnection on a per hIW basis, units with govrr-
and theoretical basis to help increase the usefulness of rrscarch nor dead-bands less than 10 to 15 mHz appear to be rare.
efforts on this subject. We believe tlie state of governor control in the Eastern In-
Mr. Kennedy’s observation concerning“NERC Guides” vcsr- terconnection is adequate for non-islanding scenarios. (Tlir
sus “Guidelines” is correct, of-course, and we find his com- largest frequency upset in recent years has been less than 100
ments to be compatible with our views. Uy broadening tlie mHZ.) While we do not advocate tlie introduction of intert-
understanding of AGC, we hope this paper will inspire the tional dead-band in governors, we feel it would be cliflicult to
continued evolution and development of tools and proccdiirrs justify any large expenditure for reducing dead-baricls brlow
to assist system operators, beyond tlie present level, so that the existing level. This assumes that the percentage of units
improved system operation and control can be acliirvd. We with smaller dead-bands would be maintained near the present
believe that while an area is satisfying its obligation to the ratio.
interconnection, even as measured by today’s NEHC Criteria, As mentioned in the paper, ACC shoulcl not i i n r l i i l , ~com-
1121

promise an area’s primary response. This does not inean t h a t parameters that affect selection of governor droops in order 10
B has to nearly equal p. There are many other reasons for arrest the frequency at acceptable level. AGC is a slow antl
oscillations, including the nature of load in many arras. that deliberate process relative to governor action. I’ropcr tlrsigu of
substantially limits perceived benefits from setting 13 = @. AGC parameters prevents interaction hctween AGC aiid gor-
Professor Keyhani’s concept of matching energy demand ernor response. Due to limited rate and range of control, AGC
with minimum unit maneuvering is commendable. Ile presents of the disturbed area takes minutes to return ACE to zero. Tlir
a model for obtaining a solutioii that attempts to control uiiits governor droops only affect the level at which tlie frequency is
with this objective. This model requires a forecast of future initially arrested. These droops have no significaiit effect on
load Pj and could be enhanced to iiiclude minimization or fucl how the interconnection frequency returns from its post u p w t
cost over the forecast period T. Characteristics of units and level to the scheduled value.
loads in many areas, however, substantially limit the usefulness This paragraph responds to the third renlark from hlrssrs.
of the proposed algorithm. Bhatti, Kothari, and Satish. The ACE signal is computed aiitl
Many units can fully realize a requested generation change plotted in most control areas. In many of them the coordinator
within one minute after AGC has stopped ramping tlieni at can even monitor it on his CRT. Most of the AGC algorithms
the allowed rate and essentially all units will have complclrd in service, use a processed ACE to decide if generation nrrtl
this task within two minutes. To realize the full bciiefit of a n be changed and by how much at each AGC cycle. Tlicsr pro-
extended boiler model, the period T should therefore be about cesses normally filter out fast varying components of ACE antl,
two minutes. Although load can satisfactorily be forrcastetl hence, to some extent prevent AGC froiii responding to rail-
for the next hour, methods to reliably forecast the load for the dom changes in the load.
next minute or two are not available, and may be infeasible in The following three bullets respond to the second to fourth
our opinion. comments from Messrs. Nanda, Kothari, Hari, and Hhisr.
If T is taken to be several minutes, allowing flie forecasled In many areas, the upper limit of the spectrum of collrc-
load Pj to be more reliable, then a detailed prediction of R I L tive achievable unit response is roughly 0.015 Hz,correspond-
output power trajectory of each unit over horizons of up to two ing to about 60 seconds/cycle. With such generation response
minutes does not play a significant role. Despite these consid characteristics, it would not only be futile, but corinterpro-
erations, some areas have implemented ideas along the lilies ductive to attempt to follow load fluctuation components with
suggested by the discusser. AEP uses a 90 second anlicipa- periods shorter than 60 seconds.
tion of interchange schedule changes and initiates generation
AGC must be on the constant look-out for disturbance con-
maneuvering to avoid ACE. Some areas also use a vrry simple
ditions which normally require operator alert or emergelicy ac-
model to compute what power change is yet to be aiiticipaietl
tion. This requirement must also be considered in the selection
from a unit if the ramping were to be stopped at the presciit
of the data acquisition sampling rate and generation response
cycle.
immediately after the occurrence of a disturbance in the area.
We would like to offer the following comriicrits on Ihe first The NERC B2 criterion somewhat addresses this issiie by re-
and fourth remark from Messrs. Hhatti, Kotliari, a i i d Satish quiring response within one minute.
and on the first remark from Messrs. Nanda. Kothari, llari,
Characteristics of the necessary analog filters in a data ac-
and Bhise.
Under normal conditions, the US Eastern Interconnection
’ quisition system must also he compatible with the coiitrol ry-
cle. The output signal from any such filters should at least l)e
frequency rarely experiences a sudden deviation of more t tian sampled at the rate of 1/4th the period of the highest frequency
20 mHz from its one or two minute average. Even for the upper component present in the signal. Conversely, if the sampling
range of such deviations, only a small number of units partic- cycle is chosen, say 2 seconds, one has to make sure that all fre-
ipate in arresting frequency excursions. Since the majority quencies beyond 1/8 Hz are removed by analog filters in order
of generating unit governors apparently remain in dead-baud, to avoid aliasing.
it appears governors have a minimal influenrr on operation
under normal conditions in this large interconiiectiou. Conse- Many areas use an integral term of ACE in tlicir AGC
quently, supplementary control (wlidher providcd inaiiually 01 algorithms. Fortunately, their algorithms bencfit from some
by AGC) and the performance resulting from it undei iiornial sophistication that limits or disregards the integral term whcrt-
conditions, is not sensitive to the droop of governors. Such ever its use may potentially deteriorate AGC performance.
droop only comes into consideration when a sudden deviation Any significant power mismatch in the interconnection is vcry
from prevailing frequency takes a substantial number of gover- undesirable. Adverse impact of such a power mismalch by an
nors out of dead-band. integral of ACE offset should be avoided.
For such interconnections, governors (and their droops) bc- No algorithm can guarantee zero inadvertent and time er-
come of great importance when units find themselves in an ror. There are many periods over which an area or several
island. For an island to be able to continue its operation, it areas are unable, for whatever reason including unknown erlor
initidly must arrive at a stable frequency. For this, tlie com- and mistakes, to match their generation with their obligal ion.
bined contribution of governors, components of load sensitive During such periods, frequency may stay off schedule. TOs t o p
to frequency, and (when necessary) load slieddiiig or ovrrspretl further degradation of frequency, other areas are requzrrd to
tripping have to overcome the initial power mismatch in tlic provide their share of system frequency support. For this, thcy
island. accumulate inadvertent.
The ratio of the maximum probable mismatch o r r r thc sine Unilateral inadvertent payback with a limited rate is al-
of the island (or interconnection when islanding is n o t of coii- lowed by NERC when it helps iiiterconnrction perforrnaiicc.
cern,) the sensitivity of the island load to frequency dcvia tioii, Many areas repay their inadvertent unilaterally if it liclps to
and the available automatic load shedding are among {hose reduce time error and if the frequency is within a proper range.
1122

0 The ACE level acceptable by the current NERC Guidt-s, We don’t believe taking units off AGC is the proper re-
is stated in the last paragraph under “ACCEPTABLE ACE sponse to the performance criteria compliance problrnl cited
TREND” of the paper. The NERC definition of an acccptable by Dr. Enns. An area with this problem should, ntost impor-
ACE trend may, however, be changed in the future. tantly, control to meet its short term (less than an hour) area
As Messrs. Ponder and Cucchi are aware, NERC is present- energy demand. The more units on control, the less actioii
ly considering other proposals for measuring contrrd 1 1 r r h - needed from any one of them for good area performance. It
mance. Some of the Task Force members are involved in t liis rf- may be worthwhile to consider telemetry from some customers
fort and hope to soon publish suggestions. Improvetl intercon- that could provide an anticipatory signal of load change to use
nection reliability and economic equity as well as increased i n - in AGC.
terconnected operation benefits to member areas are the basrs On the comments of Messrs. Thompson and h4amandur:
of proposals being examined. we would like to include remarks from Mr. R. A. Bulley, Man-
Messrs. Ponder and Cucchi state that coordinated unit con- ager of Power System Supply, Commonwealth Edison C o n -
trols could result “in less aggressive initial governor and AGC pany. All of Commonwealth Edison’s nuclear units, both BM‘H
response from a unit.” We do agree that response of such units and PWR, are equipped for AGC regulation and they have u p
to changes in target MW (via AGC or hand-auto control) arc to five years operating experience with some of them. Unit re-
delayed by several tens of seconds for coordination with atl- sponse rates are up to 0.5% per minute. Depending on system
justments of air, fuel, etc. We do not believe delays of such conditions, most on-line nuclear units may be operated in a
durations are significant to the function of AGC. regulating mode.

Coordinated unit controls do not necessarily introduce a There are various ways of implementing sliding pressure
time delay for aggressive and immediate governor response. operation for generating units. These result in corresponcl-
However, to avoid a follow up compromise of the initial gover- ingly various unit regulating capabilities, some of nliicli may
nor response in such units, a boiler-loop bias logic should SCIW be rather minimal as the discussers suggest. AEP has iinple-
off-nominal frequency and allow a sustained net RI\V (droop mented a hybrid sliding pressure mode 011 some of its supercrit-
like) deviation from target value.[7] If AGC survives, o r is rc- ical units and could maintain a turbine control valve reserve
turned to service after an upset, it will subsequently Iradjust for AGC action. This has not been found necessary as control
unit output, regardless of any offset from target due to a boilrr obtained via the pressure control valve has been adequate.
loop bias, so as to satisfy interchange, frequency and economic The discussers also ask about including “integrated A C T ”
objectives. as an area control error compoIient. integrated ACE includes
The discussers mention the Tokyo voltage collapse. Cer- both an inadvertent component and a time error component.
tainly the stabilizing effect of load sensitivity to frequency will Over a term of days, energy metering provides an inadvrrteiit
be compromised if modern load controls, such as those involvrtl accumulation to be controlled within bounds. Heduring the
in the Tokyo incident, become extensively utilized. This is time scale to hours, one must be concerned with energy audit-
an issue of primary control and not that of AGC; one line ing uncertainty. Use of integrated ACE is simply shortening
of defense could involve larger governor response, if a n d when the time scale still further. But, if done judiciously, we believe
the existing aggregate load-frequency characteristics slioulti be- this can be beneficial. One can think of the current inad\rrtent
come drastically changed. as having three components: 1.The audited value, norniallj- as
of midnight yesterday, 2.The sum of the hourly valtirs since
AGC is automatic generatton control. It changes genera- midnight (from energy metering if available), a d 3. The inte-
tion by adjusting the output of each unit. Within the con- gral of the tie-line component of ACE since hourly data was
straints of unit response delay and other characteristics. we last obtained. Separate accounts must be maintained for on-
encourage any AGC strategy that can enhance any nsprr t o f peak and off-peak periods. Depending on one’s confidence iii
operation performance. This includes, as statrtl in i lw paprr. the accuracy of these three components their weiglited s i ~ i i i
the strategy that operates the system with a bettrr security could be an estimate of the up-to-the minute inadrertrnt. A
margin. Under this umbrella, are considerations of trarisfcr unilateral inadvertent correction might be invoked, following
interface limits and coordination with relay schemes. NERC rules, of course, and including proper consideration o f
The discussers refer to a jointly shared reserve proccdiire any change in time error.
of MAAC and NPCC. We understand this as a scheme of in- In conclusion we would like again to tliank all the discussers
voking an interchange transaction to split the ACE of a largc for the time and effort they have contributed. Tlieir questions
disturbance between areas, getting them both into a sensed dis- and interest have added clarification to many of the points of
turbance state so that more units are available for emiw?prg the paper and have improved its qualtity and usefull~~ess as a
action. We wonder which portion of the paper has given tllc reference for AGC practice axid development.
discussers the impression that such a scheiiie could he undr-
sirable, indeed it may be a means of realizing further economy
of interconnected operation. It may also be a consitleratiou,
if implemented automatically, for an area having a problem
situation, such as that described by Dr. Enus. Manuscript received A p r i l 5, 1991.

Das könnte Ihnen auch gefallen