Control of Electric Power Systems*

Olle 1. Elgerd

Introduction zyxwvuts
Department of Electrical Engineering
University of Florida
Gainesville, Florida

A powersystemmust be abletomeetreasonablepower
demands by large and
small
customers of domestic.
commercialand industrial type. It mustwithstand with
reasonable security the capricious forces of nature. In an
Tobus*12
1

///
Bus
Coding

1 2
9- zy
[I-
-
Generator

Transformer

Bus
Circuit breaker
z
age of high energy costs it is called upon to transform the
prime energy resources into electric form with an optimum
overallefficiency.Thecontrolfunctions areobviously
many and varied.
Some control and decision processes, exemplified by the
optimal utilization of the controlled flow of river systems
involvedynamicswithmonth-longtimeconstants.Other
phenomena,like thetransientsonthetransmissionlines
followinglightningstrikes,
milliseconds.
Theslowercontrolprocesses
run theircourse in afew

arenormallyhandled by
To bus*
/
4

7
z
zyxwvuts
Fig. 1. Power system
symbols.
neighboringsfstem

computer-assistedhumanoperators.The faster
control
functions are trusted to fully automatic control systems of 2. Load Characteristics and Generator Mix
either open or closed-loop nature.
The objective of this article is first to outline briefly the The loads vary widely over daily, weekly and seasonal
basic functional features of a power system and, secondly, cycles.Fig. 2 showsa typicaldaily fluctuation. Electric
power travels onthe lines at a speed closeto that of light. In

zyxwvutsrqp
describe some of the more important controls required for
its satisfactory operation. Finally some of the more relevant addition,the linesthemselves(unlikegastransmission
research and development areas are identifiedand discussed. lines)havenoenergystoragecapacity,andasaconse-
quence the electric power mustbe generated at the instant it
1. The Power Grid is being demanded by the loads. The installed generating
Fig. 1 shows a one-fine diagram of a section of a larger capacity of the system must thus equal the peak demand plus
system. The electric power is produced in the generators. a spinning reserve of 10-20 percent, the latter consisting of
partially loaded generators.

zyxwvu
transformedto an appropriatevoltage level in the trans-
formers and then via the buses sent out on the transmission
fines forfinal distribution to the loads. Via tie-lines the
A
system is connected to neighboring systems belonging to the Spinning R e s e w

same pool.
Fig. 1 does not show the low-voltage distribution portion
of thesystem,whichcontains themajority of the load
objects. For most important system studies it is sufficient to
use lumped or composite representations of the loads. The
load symbols in Fig. 1 are of the latter type.
The circuirbreakers permit
the
trippingof
faulty
componentsandalsosectionalizing of thesystem. High
voltage dc (HVDC) is being used in special cases. However,
thevastmajority of theworld’selectricpower is being
generated, transformed, transmitted and distributed as high

zyxwvutsrqp
voltage ac (HVAC) of the threephase variety. Collectively
thegenerators,transformers,buses,
stitute the power network or grid.
lines and loadscon-
6 12
I

18 24 6 12 18 -t
Noon Midnight Noon Hours

*Received January 8, 1981; in final form March 12. 1981. Accepted in Fig. 2. Daily load fluctuation.
final form by Associate Editor L. H. Fink.
0 2 7 2 - 1 7 0 8 / 8 1 / ~ $ W . 7 . 5 @ 1 9 8 1 IEEE
4
consisting of

laws
generators

that
driven

determine the
from

flow
zyxwvutsr
The power demand in Fig. 2 is met by a generator mix

zyxwvutsrq
1. Baseloaded units, running fulloaded on a 24 hour basis.

Reactor cores and huge boilers do

generators
short-time
are

zyxwvu
3. The Power Flow Equations (PFE)
common,

important topic. (For more details see ref. 3.)

zyxwvu
zyxwvutsrq
Fig. 3.
\ I
0
I \
v

ni outgoing lines
Bus voltage and current symbols.

Consider an N-bus system. For a typical

\

US power system
N > 100. For very large systems N > 1000. Fig. 3 details

zyxwvutsrq
z
Nuclear and large fossilfired units fall in this category.
not tolerate fast power
changes once thermal balance has been reached.
2 . Controllable units,consisting of hydrogenerators and
smaller fossil units. All such units have rate limits giving

but also
energy storage
facilities like pumped hydro, compressed gas or thermal
storage (ref's. 1 and 2 ) .

Central to all analysis of power systems are the physical

of the electric
power
throughout the system. We give a brief-presentation of this

bus # i of such a system, containing generation, load and tzr

outgoing lines. For typical systems

The bus is characterized by the bus voltage phasor V ,

measured between the reference phase and ground. V I has
magnitude and phase defined by

We choose bus # 1 as reference bus and assign to it the

voltage
All othervoltageandcurrent
relative to V I .
phasors are measured

The bus current. I ; , is defined as the difference between

the generator current IGland the load current I L i , i.e.

The bus potver consists of two components, the real and

the fastest rates (MW/sec) with which their loads can be reactive powers, P iand Q, respectively, defined as follows:
changed.

zyxwvutsrq
zyxwvutsrqp
3 Peakloaders, whichcan pick up load relativelyfast.
Gasturbine-driven

where Q i is the relative phase between V , and I ; , i.e.,

p i and Q , represent the generated minus the load power at

bus i, measured in MW and Mvar par phase resp.
It can be readily shown that P I and Q, satisfy the complex
equation

where

The N

the form
( )*

X
denotes "conjugate."
As the network is linear, electric circuit theory tells us
that the followinglinear relationship exists between theV i ' s
and l i ' s :

' b u s = 'busVbus

The N-dimensional vectors 'bus

N busadmittance matrix
and VbUSare defined by

is symmetric and contains the elements yG which are complex of

(9)

As a result of the nonequality ( l ) , Ybusis a sparse matrix,

i.e., most of itsoff-diagonalelementsarezero.

we obtain
(A non-
existing line between buses i and j means that yi, is zero.)
Upon substitution of the ith of equations (9) into eq. (8)

5
 zyxwvu
zyxwvutsrqp
Separation of thereal and imaginaryparts of eq. (13)
yields the 2N real equations:
where Xkl is the line reactance. As this power increases, the
line power angle, 6 k - Si, may reach 90” in which case we
have reached the static power limit

Any attempt to further increase the power would result in

zyxwvuts
theloss of synchronismandtransmissioncollapse. For
relatively short lines (less than 200 miles) the thermal limit

zyxwvutsrq
These arethe famous power flow equations (PFE). In typicallydeterminesthelineloading. For long linesthe
contrast to eq’s (9) they are highly nonlinear. Physically situation is reversed and the static power limitnow becomes
they expressreal and reactivepower balance at the N buses. the critical concern.
We can view the bus powers P i and Q i as control inputs.
By their manipulation we can affect or control the voltage 5. The State Transition Diagram
sfafe variablesI V i [and hi. In accepted control lingowe thus Dy Liacco (ref. 4) defined the various “states” in which a
have the state and control vectors power system may be found. Fink and Carlsen (ref. 5) went
further and suggested the state transition diagram shown in
Fig. 4. This diagram provides a good conceptual picture of
the overall control requirements of a power system.

E,I

zyxwvutsrq
and NORMAL

zyxwvuts I

Restarts
11
Preventive
I

zyxwvutsrq
‘
.‘M.
...
.”i
“Load pickup “Control
Thevectors x and u eachare of dimension 2N. The
subvectors 6 , lVl,P and Q each are of dimension N.
4. NormalOperating State RESTORATIVE
A power system operates in a normal state if the following
conditions are met:
I 1 I
Resynchron - Emergency
(1) All the load demands are met andthe
k oPFE’s
n t r oare
l satisfied. bation
( 2 ) The frequency, f, is constant (60 Hz in USA)
(3) The bus voltage magnitudes IVi I are within prescribed -- -
E,! E,I -
limits.
-
(4) No components are overloaded.
According to Fig. 2 the demands vary slowly with time.
Thus, in order to track the loads to meet requirement# 1 the
normal state is “drifting” as the hours wear on.
Frequency constancy is required for a number of reasons;
I
EXTREMIS

’E’ = Equality Constraint

A

’ EMERGENCY

’I‘ = Inequality Constraint

electric clocks must be accurate, steamturbines mustnot be ’-‘ = Negation
subject to blade resonance, motor speeds must be kept constant.
However, the most important reason for keeping f constant is Fig. 4. Statetransitiondiagram.
that its constancy indicates that total system powerbalance is
For more than 99 percent of the time we find the system in
maintained (Sect.9).
its normal state as defined in the previous section. In sect. 7
Voltage constancy isrequiredbecause all loadobjects
we shall discuss thenormal state controls required to keep the

zyxwvutsrq
from lightbulbs to giant motors are voltage rated.
system in this state. The symbol “E” refers to “equality”
Component overload must be avoided as it results in elevated
and means that the PFE‘s are satisfied and frequency and
temperatures andrisk for damage. For a transmission line
voltageconstancyobserved.Thesymbol “1” refersto
thermal damage isonly half the problem. It can be shown (ref.
“inequality” and means that we are operating within rated
3) that the real power, P,,,transmitted on the line connecting
component limits.
buses k and 1 follows the formula
Assume now that the system would suffer the sudden loss
of a generator or experience some other event that would
reduce the security level. The system would now enter the
alert state. The “E” and “I” would still be satisfied and,
6
withluck,
zyxwvutsrqpo
Withthesystemstill
zyxwvu
we couldoperate in this state indefinitely.
However, by preventive controls (for example start-up of

in thealertstatesomeadditional
The four Jacobian submatrices J , ... J4 have as elements
thepartialderivativescomputed
reserve generators) we would seek to return the systemto its matrices are sparse.
normal state. We can write eq (20) in “component” fornx

disturbance may occur, for example the tripping of a tie-line A P z J l A 6 + J , A I VI

in thenormalstate.The

or the loss of an additional generator. The resulting power

+ J , A l VAl Q z J 3 A 6 (2 1)
shift may then overload a line. The system remains intact,
i.e., “E” is still satisfied but “ I ” is negated. The system
now enters the emergency state. By means of emergency In a typical power networkthe impedance elements are
controls, we would now try to relieve the overload almost purely reactive which means that the angles yij of the
situation. For example, by lowering the bus voltages, one Ybus elements are close tof90”.In addition we seldom operate
would force a reduction in the loads (“brownouts”). with linepower angles, 6 j - t i i , above 30”. Underthese
Should the emergency controlsfail then the overloaded line circumstances it is easy to show that the submatrices J , and J4
dominate over 52 and J3 and in a first approximation eq’s (21)

z
must be tripped. We may then see a series of cascading events
which may lead to the extremis state. Typically, the system thus canbe written in simpler form:
would nowbreakup into islands, each ofwhichwould
operating attheir own frequencies. Both “E” and “I” are now
negated. Eachisland would typically be characterized by severe
power imbalance and“heroic” control measures like load
be

zyxwvu APzJlA6
AQ z J 4 A l V I (22)

zyxwvu
shedding or generator tripping would be tried to saveas much as
In words: A change in the reactive bus power components
possible of the system. On rare occasions, however, the efforts
will result in changes in the bus voltage magnitudes with
might fail and the system would end up in a total “blackout.”
onlyminor effects
onthebusvoltage
phase angles.
The restorativestate involvesgeneratorrestarts,re- Similarly, changes in the real bus power components bring
synchronizationandgradualloadpickup.Thisisaslow about changes in bus voltage phase angles with very minor
process and can in severe cases last for hours and days. effects on the voltage magnitudes.
6. Normal State Control-A Noninteraction We make these additional important observations:
Property
1 . Assumethat we manipulatethebuspower at one bus,
Maintaining a power system in its normal state is a high #k,only. As a resultthevoltages of all buses will
priority control function. The job is made relatively simple change.Thechanges will be largestatbus k and
by noninteraction
a property
characterizing all
power diminish with the distance from this bus. However, all
systems. We presently discuss this property. changes AIVi( and a S i will be of the same polarity as
Considerasystemoperating initsnormal state. Small AQk and A P k respectively. For example, if we increase
changes AP and AQ are now made in the control vectors p AQk thenthevoltagemagnitudes of a f f buses will
and Q. As a result the state vectors 6 and IVl will undergo increase with the largest increase measured at bus k .
small changes A6 and AlVl respectively. Eq’s(14) yield the 2 . The increments APk and AQk added at bus k will divide
followingrelationships
between
the
control
andstate and flow out on the nk lines terminating in bus k . At the
changes: ends of these lines there will be a further subdivision,
etc. Powerinjection at a bus thus dissipates as it flows
out in the network. This “dissipation effect” makes it
possible toanalyzephenomena in powersystems by
modelingonlythelocalportion of thenetwork.For
example, the Florida and California grids are electrically
interconnected, but anetworkdisturbance in Florida,
evenamajorone, will have no measurableeffects in
California.

zy
zyxwvutsrq
7. The Normal State Controllers
In sect. 4 we stated the four conditions that must be met
for the system to remain in its normal state. The obvious
way to maintain a perfect power balance at each bus would
for i = 1 .-, N
~
be to continuously keep the generated powers, PGi and QGi
Wecanwritetheselinearrelationships in thecompact in balance with the changing load powers PLiand QLi.This
vector-matrix form then would maintain allAPi and AQi at zero levels and thus
all busvoltages and line powers at constant values.
This, of course,isneitherdesirable nor possible. It is
undesirable because constant line powers would defeat the
real purpose of transmission lines which is to make possible
7
 zyxwvutsrqp
zyxwvutsrqp
zy
at each moment themosteconomicaltransfer

and reactive power sources.

of power.
Neither would it be possible since most buseslack both real

The most important power sources are the synchronous

generators. Typically, generators are available only at less
8. The Automatic Voltage Regulator Loop
Depending upon size and typeof the generator, excitation
systems come in several different models. Ref. 6 gives a
detailed classification. Here we brieflypointoutsome of
than five percent of the network buses. At these “generator the more important features, that are typical of most AVR
buses” both P , and Q, can be controlled. The real power is loops.
controlled via the turbine torque, the reactive power via the AS shown in Fig. 5 the busvoltage is measured via a
“exciter” and field winding. potential transformer (PT). After rectification and filtering
Generators obviously represent themost important means the output is compared with a reference. The resulting error
forsystemcontrol. Eachgeneratorisequipped with two voltage,afteramplificationserves as inputto an exciter
separate automatic feedback control loops both depicted in which feeds directly into the generator field. A drop in the
Fig. 5. TheAutomaticVoltageRegulator(AVR)loop terminal voltage causes a boost inthefield current. This
maintains
control of the bus voltage by means of increasesthereactivepoweroutput of themachinethus
manipulation of the reactive power output. The Automatic tending to offset the initiating voltage drop.
Load-Frequency Control (ALFC) loop maintains a constant Theanalysisoftheloopis not difficult(ref. 3 ) . The

zyxwvuts
frequency by manipulation of the real power output. Both amplifier, exciter and field circuit each represent separate
these loops are designed to operate around the normal state timeconstantsthelatterreachingvaluesashigh as 5-10
withsmallvariableexcursions.Thus the loops maybe sec’s.Thethreetimeconstantsadd thethreerealand
modeledwith
linear,
constant
coefficient
differential negative open-loop poles markeda, b and c in Fig. 6. There
equations and representedwith linear transfer functions. We will be threeclosed-looproot lociandthreeclosed-loop
look separately at the two loops.

zyxwv poles the latter marked A, B and C in the same figure.

r
Tietine powers

\L
Signal
zyxwvu -
ALFC LOODS
A
Secondary ALFC Loop
A

Primary ALFC Loop

\
\,
\ -
AVR Loop

voltage error
..

IviI

mixer

-f
Frequency
Sensor
zyxwvuts -
*PL
\;F&$$,;
Trans
Former

Local
Load
Burr’ i
Rectifier
& Filter

To
\ AP
I
Network
Fig. 5 . Basic generatorcontrol loops.

Normally, the closed loop should respond in less than 0.1 fix points” in the network. One normally needs additional
seconds. To overcome the slow field circuit and also for such fix points to assure an overall good voltage profile.
purposes of achieving sufficient
static
accuracy we This can be achieved by installing banks of shunt capacitors
obviously require very high loop gain. But this gain would at certain key buses. A capacitor delivers reactive power
pushtheconjugatecomplexpolepair (A, B) into the and constitutes thus a “Q-source.” If negative bus power
unstables-plane.The AVR loop is clearly in need of generation also is required banksof shunt reactors must also
effective stabilization if speed and accuracy requirements be installed. By controlling these capacitor and/or reactor

zyxwv
bothhave to be met.Ref. 6 describesvariouspractical banks from an error voltage similarto that of the AVR loop,
means for achieving stability compensation. automatic closed-loop voltage control canbe achieved. The
TheAVRloopmaintainsreactivepowerbalance at a control itselfcan be done in on-off fashion by means of
generator bus by indirectly maintaining a constant voltage circuit
breakers.
Modem installations
utilize
thyristor
level. The generator buses thus can be considered “voltage control which permits continuous variation of the reactive
power.

8
9. The Automatic Load-Frequency Control

“primary“ and “secondary“ in fig. 5. The purpose of both

theseloopsis to achieverealpowerbalance,
zyxwvut
decrease will accelerate the unit. In eithercase
generator frequency will undergo a change, Af, which thus
Actually there is not one ALFC loop but two, designated becomes an indicator of the existing power unbalance.
The ALFC loops are designed to maintain power balance
or “load by an appropriateadjustment of theturbinetorque.
the

By
tracking,” in the system.Just asthe AVR loopachieves means of the “primary” loop a relativelyfastbutcoarse
Q-balance by maintainingaconstantvoltage,the ALFC frequencycontrol is achieved.Theresponsetime of this

zyxwvutsr
loopsachieveP-balance by maintainingaconstantfre- loop is limited by the inherent speed of the turbine and is
quency. typically measured in seconds.

zyxwvutsrqp
Thereisanimportantdifference,however.The AVR The “secondary” ALFC loop works in a slow reset mode
loopisable tomaintainperfectQ-balanceonly at those to eliminate the small frequency errors which still remain
buses that are voltage controlled. The ALFC loops maintain after the actionsof the primary loop. This loop also controls
primarily P-balance at the generator buses but because rhe the power interchange between poolmembers.We now look
frequencyis the samethroughout the systemthey thus at these two loops in more detail.
collectively, achieve P-balance on a system wide-basis.
To understand the functioningof the ALFC loopswe need 10. The Primary ALFC Loop-

zyxwvuts
to brieflyreviewthemechanismwherebythegenerator
supplies power to the network. Consider thus the generator
in Fig. 5 to operate initsnormalstate.Itdeliversthe
constant electrical power, PC’ megawatts, to the network.
Through a rather intricate mechanism (ref. 3, chap. 4) the
generatorcurrentsand therotormagneticfieldcreatea
Mathematical Modeling
The purpose of this loop is to achieve the fastest possible
adjustment of the turbine power in response to a change in
frequency. To this endthe speed governor measures
continuously the frequency(or speed) and produces a power

zyxwvutsrqpon
command, P C ,of the linear form

TGO=
zyxwvutsrqp
zyxwvut
constant electro-mechanical decelerating torque TG’ which

zyxwvutsrq
is related to the generator power through the equation

7
PG O

wrn (Ref. 3 gives hardware details)

Pre. is areferencepowersetting.Theconstant Kg has
isthemechanicalrotationalspeed
amo dimension MW/Hz. Its inverse value (Hz/MW) is referred
of theturbine-

zyxwvuts
generator expressed in radtsec. to as“regulation” andinforms of the staticdrop in
The turbine delivers a constant accelerating torque, TTo frequency as caused by increased power output. In USA the
which if expressed in turbine power, amounts to regulation is typically set at 5% meaning that the frequency
would drop 5% ( = 3 Hz) for a change in power between zero
PT’ = w, TTo ’ and full load.
Thecommand, AP,., is fed into a hydraulicamplifier
The torques T G o and TTo(and the powers P G o and P r o ) which causes a position change, A P v , of the steam control
are in complete balance and the speed andfrequency,P, are valve (or control gate in the case of a hydrogenerator).
thus constant. The hydraulic amplifier typically has a transfer function
This equilibrium is suddenly upset by an electrical load
change, A P G , (Fig. 5 ) dueeither to achange, AP,,
in the local load or a change, AP,, in the line* power or
both. The load incrementAPL due, for example,to an added
motor will be referred to as “new” load, in contrast to where the time constant TH lies in the range 0.1-0.2 sec’s.
changes in already connectedor “old” loads (see also Sect. The change in valve (or gate) opening translates in the
11C). As a result of these load changes the generator power turbine into a power increment, A P T .
changes instantaneously with the amount of AP,. Electrical We can now readily assemble the block diagram shown in
power balance requires that
Fig. 7. (Disregard for the time being the dotted portion.) The
portions labeled “network” and “turbine“ require further
elaboration.
In the moments following this electrical loadchange no 11. Network Dynamic Representation
change takes place in the turbine torque, and the turbine-
generator thus experiences a slight torque or power im- The turbine power, A P T , will be used for four different
purposes:
balance.
If the electrical load change is positive signifying a load 1 . To supply the demanded “new” load APL.
increase theturbine-generatorwill
decelerate.
Aload 2. To accelerate the turbine-generator, thus increasing the
kinetic energy, Wkin, of the unit.
*For simplicity we assume only one outgoing line in Fig. 5 . 3. To increase the powers in outgoing lines, i.e., A P , .

9
 zyxwv
zyxwvutsrq
zyxwvutsr
zyxwv

zyxw
Fig. 6. Root-loci for uncompensated AVR loop.
Semndary ALF C loop
A
1 \

Primary ALFC loop

A
/ \

Network

A
Speed
governor II ApL

-
* PC1 -
A pv KN
- -
Af

zyxwvuts
l+rTH 1~ T N

Hydradim Turbine
1 2XT.- A f - Afj

7-1
'I t_

I;Af l2-J
Line

zyxwvutsr
A fi
Fig. 7. Modeling of ALFC loops. (from bus j)

Tothe
4. increase
meet in the "old" load. d
Wediscussbriefly the three last power
components. @kin =- (Wkin)
dt dt L
A . Kinetic Power Increment, AP,,,
Thekineticenergyserves as a buffer storage. For
example when a
motortothesystem
customer suddenlyconnects
a
it obviouslycannot
100 kW
be met by a
corresponding increase in the slowchanging turbine power.
Instead, the generator will supply itby "borrowing" from
thekineticenergy.Since the latter varies as the square of
%- zyxwv
dt
2Wkin0

fo
d
-
dt
(An
where Wkino representsthekineticenergyas
(28)

measured at
the speed this power component can be expressed as follows normal speed.

10
 zyxwvutsr
zyxw
zyxwvutsrqpzy
B . Line Power Increment, A P , 12. Turbine Representation
For simplicity we assume there is only one outgoing line Theturbinedynamicsis of centralimportance. It will
connecting our generator bus, # i , with an other generator vary widely &pending upon type of turbine used. We give
bus, # j . thetransferfunctionrepresentationforthreedifferent
From
eq (17)obtain
we
turbine
types.
(More
details canobtained
be from ref. 7.)

A5 A6 zyxwvutsrqpo
= 2.rrSAfdt we can write

zyxwvutsrqp
where the parameter, T u , the synchronizing coefficient of
the line, is defined by
Valve

turbine
Steam
chest
turbine
Reheater

-
APT

C . Frequency Dependency of ‘‘Old” Load I To Condenser

As the frequency increasesso will the speed of all motors
fedfromthebus.Addedspeed meansaddedtorque and
power.One may expressthis frequencydependency of
existing load by an empiric parameter, D, having the unit
MW/Hz. Thus the increase in the ”old” load equals DAf. ‘Head’ \\ ’
By Laplace transforming all the above power components
andadding themthe dynamic powerbalance at the bus
reads:

We can rearrange this equation as follows:

Fig. 8. Turbine types:(a)

A . Non-Reheat Steam Turbine

Thisturbinetype(Fig.8a)has
zyxw
Non-reheat steam,(b)
( c ) Hydro.
Reheat steam, and

a simpledesign. After
passing the control valve the high pressure steam enters the
turbine via the steamchest. The chest introduces a delay,
TcH, in the steamflow resulting in the transfer function

where

1
KNo - Typical values for TCHlie in the range 0.2-0.5 sec‘s.
D
(34) B . Re-heat Steam Turbine
This type of turbine has several turbine stages, between
which the steam is led via reheaters. The design increases
efficiency and is always used for large units.
Putting eq (33) in blockdiagramformyieldsthepart Assume that the two stagesin Fig. 8b are rated at half total
labelled “network” in Fig. 7. powereach. If we alsoassume that thereheatercan be

11
 zyxwvutsrq
represented by a timeconstant, T R H ,then the total turbine whole“area”which in practicetypicallycanembracea

The overall transfer function would be

zy
power equals (neglecting the delay in the chest):

zyxwvuts
whole power system.
If this “area” via tie-lines is connected to neighboring
“areas” then we talk about multi-area dynamics.
situations all the power commands are executed in unison
amongallgenerators that areundercontrol.
generator in the area has the samepercentage“regula-
In such

If each

tion’ ’ then each generator will participate in proportion to

its rating.
The secondary ALFC loops in multi-area systems contain
(37)
controlsignals, now referred to as “areacontrolerrors”
(ACE), which in additiontofrequencyerror, Af, also
TRH hastypicalvalues in therange 4-10 sec’sthus containtheerrors in thecontractedtie-linepowers.A
resulting in slow overall response times. typical such ACE would be of the form

zyxwvutsrqp
+ SAf

zyxwvutsrqp
C . Hydro Turbine ACE = (39)

zyxwvutsrq
Depending upon the magnitude of water “head” (fig. 8c)
this type of turbine is of varying design. Without proof (see 15. Optimal LQR Design
ref. 7) we give the following transfer function The AVR andALFCloopswerederived on the as-
sumption of total noninteraction. There
is,
however,
1 - ST,
GT=-

zyxwvu
1 + sTp

T p isthetime it takesforthewaterto
penstock.
Fig. 9 showsacomparisonbetweenthethreeturbine
crosscouplingbetween the channels,whichunder certain
circumstances willhavenoticeable
verydeleteriousnature(ref.
effects,sometimes of
3, chap. 9). Forexample,
pass through the sometimes whole areas will start to oscillate at frequencies
around 1 Hz. Theoscillationsshow up in frequencyand
tie-linepowersandcangrow
types in regards to the time response to a stepchange in the nization occurs.
valve position. It is interesting to note the momentary power
tolevelswhendesynchro-

In principle it is not difficult to expand the mathematical

decrease for a hydroturbine. (Electrical engineers refer to models toaccountforthesephenomena.However,the

zyxwvutsrqp
this type of behavior as “non-minimum-phase.”) models become of high dimensionality and classic control
design becomes difficult.
13. The Secondary ALFC Loop Insituations likethisoptimum LQR designbecomes
The primary ALFC loop would yield a frequency drop of attractive. Following initialattempts by Yu, Elgerdand
about 3 Hz between zero and full-loadof the generator. This Fosha (ref. 8, 9) many additional contributions have been
poor accuracy is entirely unsatisfactory. This is where the reported (ref. 10). Space does not permit a discussion here,
“secondary“loopenters the picture. It performs slow butthereaderisreferredtoref. 3 formoredetailed

zyxw
“reset”adjustments of thefrequency by changing the coverage.
reference power command Pref. 16. Emergency Controls
The dotted portion of Fig. 7 shows how this can be best
accomplished by a low-gain integrator loop. Allpreviousdiscussionshaveconcernedcontrol loops
Followingasudden load increase, AP,, theturbine whichareintended tomaintainthepowersystem in its
output, APT, is increased to a new value as rapidly as the normal state. In sect. 5 we indicated how cascading events,
primary ALFC loop will permit. Aswe noted, the turbine or multicontingencies, can bring the power system into an
response sets the pace.However, we areleft with a emergency state,which if not properlycontrolled can
considerable negative frequency error, which now causes a deteriorate into an extremis state. Due to the slow turbine
slowlygrowing positive integratoroutputand a cor- responsethe ALFCloopsareineffective in emergencies.
responding increase in power reference setting. Whereasthe The AVR loopsarefasterbut the limitedexciterpower
primary loop responseis over in seconds. the secondary fine renders even these loops essentially useless.
adjustment may take of the order of one minute andwill not Powersystemcomponents,especiallygenerators.are
stop until the frequency error is zero. veryexpensiveand theprimaryobjective of emergency
controls is topreventdamage totheequipment.Forthis
14. Extension to “Multi-Area’’ Systems
reason one always finds a first line of defence consisting of
The loop model in Fig. 7 is in strictest sense valid for a protective devices organizedinto unitprotectilme svstenzs.
singlegeneratoronly. We havenoted. that the frequency For example, generators, transformers.lines and buses have
dynamics is relatively slow.Thistends tomake a whole their own specialized fastacting protective devices. which
group of generatorsmove in unison.or coherently. thus mostly are
setto
operate i n preset or openloop mode.
permitting us to represent them all with the same Af. For Typicallya r e l q detects thefaultandinitiates a circuit
this reason it is common to let the model in Fig. 7 represent a breaker trip. It is important that tripping involves onlythose

12
thushave zyxwvutsrqpo
zyxwvutsrqpo
components that are subject to damage and the relays must
high selectivitx. Microprocessorsarefinding
increased use in modern relay design.
Thesecondobjective
perform
automaticre-energizing
oftheemergencycontrols
of components. For
is to
and 4 to open. At that instant the generator powers P C , and
PG2 change to match the local loads which will remain at
their prefault values*.Unit # 1 thus findsitself with a power
surplus of 300 MW causing an acceleration. Unit #2 will
haveapowerdeficiency of 300 MWand will thusde-

zyxwvu
example, following the tripping of a line in most instances celerate. The angles 6 , and 6 2 will thus move apart at an
the shortcircuit will “heal” in a fraction of a second and the accelerating rate. If line reclosure can be accomplished fast
lineafter reclosure canfunctionnormally.The reclosing enough the two machines will regain synchronization.
must take place fast and automatically to be successful.
The third objective of emergency control is to prevent the
system from desynchronization. i.e.. breaking up into parts.
A system is said to be rrunsientstable foraparticular
t!
i Non-reheat
disturbance if thegeneratorrotors, following the initial \

transient angular swings, tend to stick together. The system

may not in itself haveasufficientlystrongsynchronizing
“glue“ and it will then be necessary to insert on
temporary
stability.
basis
components designed to enhance the
a
-
4;
t, sec‘s

The fourth and final objective of emergency controls is to

save a deteriorating frequency.
17. Transient Stability Control-
Classical Approach Fig. 9. A comparison of turbine responses. zyx
zyxwvutsrqp
Under normal conditions the turbinepower. P,. and
generatorpower, P , . are in balance
and the turbine-

zyxwvu
generator unit is running at constant speed. During major
fault
disturbances P , remainsapproximately
constant
whereas P , undergoessudden and large changes.The

-
1 2
postfault difference power P , - P G . depending upon sign.
will
either
accelerate
or
decelerate the unit.
Theac-
celerations of thegenerator units follow from the sMsitzg

zyxwvut
300 MW 500 MV!
equariorzs

Fig. 10. A two-busexamplesystem

where S is the angular rotor coordinate andM = Wki,o/lrfo

is the rotor inertia constant.

zyxwvuts
18. Transient Stability Control-New Approaches
The name “swing equation” implies that power systems
are oscillatory in nature. When brought out of balance the Consider now the following alternative chain of events.
machine rotors tend to perform torsional oscillations which Atthemoment of linetrippingthecircuitbreaker #2
are
quite
undamped.As the electrical
restoring,
or inserts a 300 MW “brake” resistor R. At the same instant
synchronizing,powers of thelines(eq’s 17, 29) are 300 MW of the load at bus # 2 is disconnected in a “fast
nonlinearfunctions of theangularcoordinatesthe oscil- load-shed.” These breaker operations will in effect restore
lations become nonlinear of the limit-cycle Qpe. power balance at each rotating unit, thus avoidingsevere
A transient stability study involves the integration of the angularaccelerations. At the later moment when line
coupled swing equations(40). The coupling takes place via reclosure takes place the brake resistor is disconnected and
thenetwork. In sucha simulated study it is, of course, the load is restored at bus 2.
necessary to make contingency assumputions. The combined use of brakeresistor and “loadskipping”
Classically, transient stability controlhas centered on use obviously is a more active approach to preserving transient
of fast:acting circuit breakers. For example, consider the stability. It has lately received considerable attention (refs
two-machine system in Fig. 10. Initially the system is in its 1 1 , 12). For this type of control to be effective the control
normal state
characterized by the powerflows shown decisions must be arrived at fast. The controllers must also
(disregard the dotted portion). be robust as they must function under vastly different fault
A sudden short on the line causes the circuit breakers 1 conditions.

*The bus voltages will change (in spite of the best efforts by the AVR
loops) and as the voltages change so will the loads. We neglect this in our
discussion.
19. Frequency Control and Longterm Dynamics zyxwvu problems as
beingthe most relevant ones from the
operator's point of view. Other important control problems

zyxwvutsrqpo
Following a major fault the transient stabilityis typically related to resource optimization. security enhancement and
determined within oneorseveralseconds. Even if the environmental protection have been excluded due to space
system remains initially
synchronized,
faultinduced limitation. Due to the enormous capital
investments,

zyxwvutsr
problems often occurwhich make themselves known within changes in power systems technology tend to be
of
several seconds or possibily minutesandwhich will give evolutionary rather than revolutionary nature. We presently
rise to longterm frequency dynamics. identify some of the areas that probably will be subject to
Forexample,thelinetrip in theaboveexample may increased futureattention by control andsystems re-
cause voltage swings so severe that a feedwaterpumpmotor searchers.
will trip making it necessary to take G 2 out of service, thus The graph in Fig. 2 depictsone of the basic problems
creating a systemwide generation shortage of 200 MW. If facingthe power systems planner-disparity between
the system were partof a power pool, support power would maximumand minimum loading-which results.on the
immediately flow in overthetie-lines. If thesystem is average, in poor utilization of generating equipment. In an
operating alone the 200 megawatts will be taken from the earlierandmoreplentiful energy era the electricutilities
kinetic storage resulting in a rapid frequency deterioration. never attempted to interfere with the customer's power use
Power balance must be rapidlyrestoredand permanent habits.However.adding new generatingcapacity is ex-

zyxwvu
loadshedding will often be the last resort. It can be done ceedinglyexpensive in bothmonetaryandenvironmental
manually by the operator or automatically upon command terms. By shaving the demand peaks it is possible to either
from underfrequency relays. cancel or at least postpone for several yearsthe construction
Simulation of the frequency dynamicscould,theo- of new plants and/or the installation of expensive peaking
retically, be performed by extending the integration of the units. Controlling the power demand. or load management,
swing equations (40). now rewritten in the frequency form hasthus today becomea very high priority item on the
engineering agenda of most utility industries.
Load control can in principle be achieved in a number of
ways ranging from the totally voluntaryapproach to
compulsory shutoff (ref's. 13, 13). Using microprocessors
As some of the generators are performing relatively fast in combination withnovelrate schedules(ref. 15) offers
intermachineswingsthisintegration procedurerequires new possibilities. The industry is looking urgently for new
smallintegrationsteps (zO.01 s). Aconsiderably more and imaginative ideas in this field.
practical approachis to turn theattention to the average Areawide blackouts
although
rare have
dramatic
frequency of all the area generators. impacts
and
often
serious
consequences. Increased
We thus define a lumped area generator having lhe inertia operating security thus is high on the industry priority list.
Finding improved methodsforpredictingtransient and
frequency stability- and developing new emergency control
methods are tasks singularly suited to the control specialist.
and its rotor position ("center of inertia") defined by Determination of transient stability has classically been
performed indirectl! by integration of the swing equations.
Time-domain simulation is
computationally costly
and
places a constraint upon the number of contingensies that
canbestudied. Direct stabilityanalysismethodsseem to
offerbetter promise in thisregard.These methodsare
The frequency of this imaginary machine then equals known under
the
acronymTESA (Transient
Energy
StabilityAnalysis).Directstability assessmentmethods
exemplified by the famous "equal area criterion" (ref. 3)
have been used by powerengineersfor many years.The
TESA methods based as they all are upon Lyapunov theory
and can be integrated from a "lumped" swing equation are not new either.However,recentcontributors(ref. 16)
claim a "breakthrough" in the use of these methods as a

zyxwvuts
zyxwvu
zyxwvutsrq
As the frequency, f,,.is a fairly slowly changing variable
resultofphysicallybasedenergyfunctionsandamore
intuitiveinterpretation
goal in TESA researchwould
of computerresults. An ultimate
be to developamethod
whereby the system operator could assess in real time the
degree of stabilitycharacterizingthe system at acertain

z
the integration steps can now be chosen fairly large ( 2 I 5 ) . operating time and in a certain operating configuration.
In sect. 18 we brieflymentioned the new emergency
20. Future Trends
controlschemes the theory of which has been recently
In this brief exposition of power system control we have studied by ZaborszkyandMeisel (ref's. 1 1 . 12). Those
focused on the normal and
emergency
state
control methods represent
distinct
a departure from existing

14
practice of emergencycontrolsince they would require “Aggregation” is theonelarge-scalesystemtechnique
extensiveaddition of controlhardwareto the system at that maypossibly offer the best payoffs in futurepower
every generating bus. The proposed control schemes would systemscontrol. It is awell-knownfact that all power
function as follows: systemdynamics, both of thesmalland large magnitude
A few critical state variables would be locally monitored varieties, is characterized by “coherency,” i.e., groups of
at each control point. When the variables would indicate a generators tending to swing in unison. For example, it is

zyxwv
major disturbance a local microprocessor would determine intuitivelyfeltthatthe new emergencycontrolmethods

zyxwvuts
the best control strategy for applying the locally available discussedabovecould be greatlysimplified by taking
controlforces in an attemptto achieveoptimalpower advantage of this feature. Rather than installing the control
balance in the overall system. devicesateachgeneratorbus they may proveequally
Theauthorsdemonstrate impressivesuccess with their effective if lumped together into a few “switching
respective control strategies during fault situations in large centers.”
scale systems. Of course, at this early stage O E ~ Ysimulated
data is available.Extensive follow-up work is needed before
the utility industry will gain confidence in these methods.
Theauthors limit theirstudies to symmetrical network
faults.Extensive simulation seems required to check the
robustness for allthe various types of unbalanced faults that
canbefallpowersystems.Mostimportantly,hardware
studies-performed first at reducedpower levels-must
confirm beyond any doubts that the added equipment does
not in factintroducemoresecurity problems than it is
intendedtocure.At this stagethe workrepresents an References
exciting new development in power systems control which I . Brown Boveri Review, Aug. 1980, vol. 67, Baden/Switzerland.
2. Olle I . Elgerd, Basic Electric Power Engineering, Addison-Wesley

zyxwvutsrqpo
deserves the attention by more investigators. Publ. Co., Reading. Mass., USA. 1977.
Inrecentyearscontrol of “large-scale” systemshas 3. Olle I. Elgerd. Electric Energy Systems T h e o y , McGraw-Hill Book
become a popularresearch area of control theoreticians. It is C o . , New York, N.Y. 1971. (A second edition due in Jan. 82).
E. Dy Liacco, “The Adaptive Reliability Control System,” IEEE
not surprising that power systemsin particular have become 4. T.

zyxwvuts
Trans., Vol. PAS-86, May 1967.
the focus of much attention, since typical largescalesystem 5. L. Fink and K . Carlsen, “Operating under Stress and Strain,“ IEEE
topics like aggregation and decentralized control long have Specrrum, March 1978.
“ComputerRepresentationofExcitationSystems.”IEEECom-
beenofgreatconcern(underdifferent names) by power 6. mittee Report, IEEE Trans.. Vol. PAS-87. No. 6. June 1968.
systems engineers.The“dissipationeffect”
(sect. 6) 7. IEEE Committee Report, “Dynamic Models for Steam and Hydro
manifesting itself in localization of even major disturbances Turbines in Power System Studies.” IEEE Transactions on Power
Apparatus and Systems, Vol.PAS-92,No. 6 , pp.1904-1915,
haslong beenknown and been the basic reason why so-
November/December,1973.
called“areacontrol” has
historicallyemerged as an 8 . Y. Yu and H.A.M.Moussa.“OptimalStabilizationofMulti-
unquestioned natural feature of power systems control. The machine Systems,” IEEE Trans.Power Apparatusand Systems,
divisioninto“areas” wasoftenperformed on apolitical Vol. PAS-91, No. 3, May/June 1972.
9. 0. I. Elgerd and C. E. Fosha, “The Megawatt-Frequency Control
basis. Problem:A New ApproachViaOptimalControlTheory.’‘ IEEE
The intuitive use of decentralized control did not always Trans. Power Apparatus and Systems. Vol. PAS-89. No. 4, April
turn successful. The classical example was the emergence 1970.
IO. T. W . Reddoch,“Load-FrequencyControlPerformanceCriteria
of the “tie-line frequency-bias“ control law described by with Reference to the Use of Advanced Control Theory,” Systems
equation (39). Originally the attempt was made to delegate Engineering For Power:StatusandProspects,Henniker. New
regulation of thetie-linepowers to the individualpower Hampshire.Aug. 17-22. 1975.
1 1 . J. Zaborszky, etal.. “Monitoring, Evaluation and Control of Power
companies but let the frequency control be handled by one

zyxw
SystemEmergencies,”DOEConferenceonSystemsEngineering
(usually the largest) company. It turned out that this pool for Power, Davos. S e p t . 4 c t . 1979, Proceedings #CONF-790904-
memberexperienced unacceptablylargepower swings. PI.
12. J. Meisel,etal..“EmergencyOperatingStateControlon Bulk
Sharing the controlof both the frequency and tie-line power Interconnected Power Systems,” Ibid.
as implied by eq. (39) was suggested by Cohn (ref. 17) and 13. “Load Management,” EPRI Journal, pp. 6-1 I . May 1977.
proved a radical improvement. 14. G. Kaplan.“Two-WayCommunicationforLoadManagement,”
IEEE Spectrum, pp. 47-50, August 1977.
The powerful analysis toolsof modern linear control have 15. J.C. Mears.“Consumer-InteractiveReal-timePriceSignalling
laterconfirmedthesoundness of theempirically found Load Management Scheme,’’ Masters Thesis. Dept. of El.Eng.,
Univ. of Florida, 1978.

zyxw
formula (39). For example, it was pointed out in ref. 9 how
16. R. Athay et al., “A Practical Method for the Direct Analysis of
use of only local state variables in the optimal control of a Transient Stability,’. IEEE Transactions, Vol. PAS-98, No. 2, pp.
two-area system added only insignificantly to the optimum 573-584. March/April.1979.
integral index. This has been confirmed by later researchers 17. N. Cohn, Control of Generation and Power Flow on Interconnected‘
working with much larger systems andusing more extensive Sysrems, Wiley, New York, 1971.
18. M. Jamshidi, Large-Scale Systems-Modeling, Controland Applica-
models. A good discussion is given in ref. 18 which also lions, Elsevier North-Holland Book Co., New York, N.Y.,(Chapt.
contains a substantial bibliography on related topics. 9) to appear.

15
IEEE Professional Activities and the

zyxwvu
Control Systems Society*
Tim Johnsonf-

zyxwvutsrqp
Professional Activities Coordinator

zyxwvutsrqp zyxwvutsr
This is the first in aseries of short
articles about IEEE professional activi-
ties as they relate to the Control System
Societ?, member. In this article, we look
nized that engineering developments un-
avoidably have a profound effect on the
sociopoliticalsystemandonindividual
lifestylesandemployment.Fortheen-
come about within the decade, and some
of the details will be described in future
articles inthisseries.Ourpresentpur-
pose is to develop a framework in which

zyxwvutsrqp
at IEEEprofessionalactivitiesandex- gineer tomaintainhistraditionalde- to view IEEE professional activities and
amine their relationship to the primarily tachedandobjectiveviewpoint on such by which the relationship ofour technical
technical activities of theControlSys- matters would be irresponsible to society and professional activities as control and
terns Society. Future installmentswill and moreover suicidal to the profession. systemsengineers can beunderstood.
deal with organization and financing of A large number of laws, including social These considerations apply more or less
professionalactivities.programplans legislationand R&D funding,appeared equallyto U.S. and to international
andperformance assessment, specijic to be “dealing-out”
engineering and members of the Control System Society.
programareasrelevantto thecontrol science. Not enough attention was being The fact that IEEE professional activities
engineer, and future activitieswhich can paid to engineeringandscienceenroll- arecurrentlyfinancedby,andforthe
be pursued through the Control Systems mentsandmuchlessattentionandsup- benefit of, the U.S. membershipshould
Society. port was being voiced for education. In be regardedaslargelyapragmatic del
some instances,engineerswerevictim- velopment;ithappensthatthere is no

zyxwvutsrqp
ized by wage-squeezes,flawedpension other significantlylarge
professional
SignificantgrowthinIEEEprofes-

z
programs, or managementpressuresto organizationtorepresentelectricalen-
sional activities has occurred mostly over overlook public safety. Under these cir- gineers intheUnitedStates,andatthe
thelastdecadeandhasinvolvedpri- cumstances, theconsequences of inac- sametimethere are probably too few
marily U.S. MembersoftheInstitute. tivitywereclear:government,society, IEEE members in most other countries to
Some of thefactors whichled to the big business andeven
bigeducation warrant such a substantial investment in
institutionalization of IEEE professional would be free to push engineers around at professional activities through the IEEE.
activities-representedbytheestablish- will. Without any form of organization, Such developmentsareoccurringinde-
ment of the United States
Activities theefforts ofindividualstocounteract pendently, insomeinstances,through
Board (USAB)-arose fromthesocial such influences would be largely in vain. other appropriatenationalengineering
activismofthe 1960’s. Withthespace In spite of theengineer‘straditionally organizations abroad.
program and the rapid evolution of com- conservativeandindividualisticimage, Fitting enough, feedback is a concept
puter technology. many members recog- the cost of doing nothing had become so central to theunderstanding of IEEE
highthatorganization,planning,anda professional activities. Societymaybe
+ReceivedJan. 8. 1981. recommended by financial commitment had become ines- viewed as a complex interconnection of
D. H . Elliott, AssociateEditor-Features. capable. socioeconomicgroups,someofwhich
?Bolt Beranek and Newton Inc.. Cam- Suffice ittosaythatthesehaveall are institutionalized. Anystudentofin-
bridge, MA

16