Anda di halaman 1dari 11

EXTRANEOUS AND FALSE

LOAD

FLOW SOLUTIONS

B. K. Johnson

Power Technologies,Inc.
Schenectady, New York

ABSTRACT
M u l t i p l el o a df l o ws o l u t i o n sa r ep o s s i b l ef o rr e a l i s t i c systems w i t h r e a s o n a b l e i n i t i a l c o n d i t i o n s .
Two
d i f f e r e n t mechanisms whichproduce
m u l t i p l es o l u t i o n s
arediscussed
i nt h i s paper and several examples are
soincluded.Formostnetworks,falseorextraneous
lutionsareeasilydetected
because t h e r e s u l t s a r e u n reasonable.This
i s n o t alwaysthe case,however,
and
t h i s paper i d e n t i f i e s a c l a s s o f cases where extraneous
s o l u t i o n s y i e l d r e s u l t s whichlookreasonable.
INTRODUCTION
System planners must frequentlyperformloadflow
s t u d i e so fp o o r l yd e f i n e df u t u r e
systems.
One purpose
o f suchstudies
i st od e f i n er e q u i r e m e n t sf o sr h u n t
compensation, new transmissionlines,
and other system
a d d i t i o n sU. n t i l
such
system
additions
are
defined,
numerous nonconvergentloadflowsolutions
may be experienced.
However, asexplained
i nt h i s paper,convergence
does
notnecessarilyguarantee
a c o r r e c st o l u tion.
Erroneoussolutionsarealsopossiblewithwelld e f i n e d e x i s t i n g systems which have longheavilyloaded
lines or generators feeding capacitive loads.

power transmittedfrom

V,

i n terms
of

Yo,

X, B,

i s d e r i v e d i n Appendix I.

The followingobservations
t i o n( 1 ) :

i
l be used
Figure 1 shows a sample networkwhich w
i tnh feo l l o w i n ag n a l y s i s .
The symbols
used
i nt h e
f i g u r e a r e d e f i n e d as f o l l o w s :
= Per u n i t r e a l

SAMPLE NETWORK

The voltagemagnitude
and P , i s given by

Thisexpression

EXTRANEOUS
SOLUTIONS

FIGURE 1.

may
be
made

from equa

1) Ifthe two solutionsareclose,thentheexpression

v,"

bus 1 t o

under t h er a d i c a l

4(PX) 2 ( 1

must be

-y

small

, or

z 0

T h i si m p l i e st h a t

Vo = Per u n i t v o l t a g e m a g n i t u d e a t
V

= Per u n i tv o l t a g em a g n i t u d ea t

al = Voltageangle

at

buses 1 and 3

bus 2

P =

2X(1

bus 1 measured w i t h r e s p e c t

t o bus 3
a2 = Voltageangle

a t bus 2 measured

b u t as
f o rt h e
tem.

w i t hr e s p e c t

t o bus 3
X

XB
-7
)

Appendix I 1 shows, t h i si st h ef o r m u l a
maximum power transferacrossthesys-

The two solutions will, therefore, be close


when t h e power transferacrossthe
system i s
near i t s maximum.
2
2
as Vo, as
2) I f V i s the same orderofmagnitude

= Per u n i t 1 inereactance

B = Per u n i t l i n e c h a r g i n g

would be expected f o r a normalsystem,and


the
expressionundertheradical
i s small , thenthe
q u a n t i ti nyh e
denominator equation
of
1
should be c l o s et ou n i t y .
A reasonablerange
forthedenominatorof(1)
is

76 ~ 5 2 - 3 .

A paper reccmmded and approved by

I f t h el i n ec h a r g i n gi st h eo n l ys o u r c eo fs h u n t

the IEEE P w e r System Engineering Cumittee of the

admittance, then the quantity (1


7
XB)2 i s a f u n c t i o n
of the line length which is fairly constant for differentvoltageratings.Forhighvoltagelines,equation
( 3 )i ss a t i s f i e d
i f each l i n es e c t i o ni si nt h er a n g e

IEEE P m e r Engineering Society for presentation at


the IEEE PES Surrrer k e t i n g , Portland, OR, July 1823, 1976, M u s c r i p t submitted January 21,1976;
made available for printing May 5 , 1976.
524

of 300-500 miles.Therefore,if
both r o o t s of equation
(1) a r e t o appearreasonable,
the t o t a l systemseparationshould
be 600-1000 miles and t h el i n e s should be
heavily 1oaded .

SAMPLE CALCULATION FOR EXTRANEOUS SOLUTION


Table
Figure 1.
MVA 500 kV
loaded 500

P
V = .96

Operating
Point
for

I gives assumed parametersforthesystem


of
Theseparametersareinperunit
on a 100
base and correspondapproximatelytoheavily
kV l i n e s e c t i o n s of300 miles each.

gperatingPointfor

V = .572

VALUE

PARAMETER
X

.063
6.3

10.66
1.04

vO

TABLE I .

solu-

occurfor a u p t o 90". Therefore, bothloadflow


t i o n s would correspondtostableoperatingpointsif
constantvoltageis
held on bus 2.

I
I

90"

SAMPLE PARAMETERS FOR FIGURE 1

Two valuesforthevoltage
equation (1).

180'

2a2

V arecalculatedusing
FIGURE 2.

SINUSOIDAL POWER CHARACTERISTIC

V = .960 or V = .872
Using equation (1.1) from
Appendix
I
thevoltageangle
on
bus
1 corresponding
calculated voltage magnitudes.

PX
-

Sin(a)

FALSE SOLUTION

calculates

t o t h e two

vOv

f o r V = .96
10.66 x .063 = .673
1.04 x .96

The second mechanismwhich creates double load flow


solutionisdependent
on the way var c o n s t r a i n t sa r e
appliedinatraditional
Newton Raphson i t e r a t i o n . The
unexpected solution produced by t h i s mechanism will be
referred t o as a f a l s es o l u t i o n because i t i st e c h n i c a l l y n o t extraneous and because thisterminologydistinguishes i t from theextraneoussolutiondiscussedin
theprevioussections.
Newton Raphson load flow programs apply andrelease
generatorvarlimitsusinglogic
a t the end ofeach
iteration.Representativelogicisasfollows:

a2 = 42.27'

1) Ifgeneratorvoltageisfixed

and t h er e a c t i v e
power exceedsthe
maximum l i m i t , thentherea c t i v e power will be f i x e d a t t h e maximum l i m i t
and thevoltageconstraintwill
be released.

f o r V = .872
Sin(a2) =

47.75O

a2
The angleacrossthe

.7405

whole
system

the a n g l e a t bus 2 or

2 ) Ifthevoltageisfixed
and thereactive power
i s l e s s thanthe minimum l i m i t , thenthereact i v e power will be f i x e da ti t s
minimum l i m i t
and thevoltageconstraintwill
be released.

wil 1 be twice

3I)tfh e
maximum r e a c t i v e power l i m iites n forced and thevoltage magnitude i sl e s s
than
desired,thentheconstraintswillnotchange.

al = 2a2 = 84.54" f o r V = .96


a1 = 2a2 = 95.50'

f o r V = ,872

4)
If
the
minimum r e a c t i v e power l i m iites n forced and the
voltage
magnitude igs r e a t e r
not
than
desired,
then
the
constraints
will
change.

Equation(4)forthe
power t r a n s f e r P i s derived
in Appendix I1 and i s shown graphicallyin Fi gure 2 a s
a function of angle
n

5)If noneof
t h e s ec o n d i t i o n sa r es a t i s f i e d ,
then
thegeneratorvoltagewill
be held a t thedesiredvalue and i t s var constraintswill be r e leased.

(4)
This i s thefamiliar
sinusoidal power t r a n s f e r
c h a r a c t e r i s t fi oc r
which oDeratinq w i n t s w i t h 2a
greaterthan
90' areunstable(Appehix111)
when only
constant impedance s h u n t elementsarepresentatthe
center o f theline.Therefore,forthiscasethe
sol u t i o n of theloadflowwiththelowervoltagescorrewhich would not
sponds t o a n unstableoperatingpoint
appearinpractice.
However, i ft h e
same charging a t
bus 2 weresupplied
by a synchronouscondenserholding
constantvoltage,thenstableoperatingpointscould

The
more
sophisticated Newton loadflows may use
additionallogic t o keep generators from o s c i l l a t i n g on
and offor between varlimits, b u t t h i sa d d i t i o n a ll o g icdoesnotaffectthe
argumentgivenbelow.Similar
l o g i ci sa l s o
used in some nodal i t e r a t i v e programs.

The r a t i o n a l e f o r s t e p
3 and 4 i s t h a t a n increase
inthevar
o u t p u t of the generator is expected to raise
var output
theterminalvoltage,
and adecreaseinthe

525

o ft h eg e n e r a t o ri se x p e c t e dt ol o w e rv o l t a g e .T h i si s
t r u e when theapparentsystem
impedance i s inductive,
as shown i n equation 5.

wouldabsorbincreasingvarsasthevoltageregulator
i n c r e a s e se x c i t a t i o nt oi n c r e a s et h et e r m i n a lv o l t a g e .
Thus, the normal l o a df l o wl o g i c
models thegenerator
i n c o r r e c t l yi nt h i sc i r c u m s t a n c e
and a f a l s es o l u t i o n
wil be found.
I f t h es o l u t i o ni ss t a r t e df r o m
a differentinitialcondition,thegeneratormightnever
become pinnedduringtheiterativeprocess,
and t h u s f i n d
t h e" c o r r e c t "s o l u t i o n .

2
L

A l o a d flow user may t e s tt o


see i f he hasconverged t o a f a l s e s o l u t i o n o f t h i s t y p e
bychangingthe
v a r l i m i t s on theaffectedgenerator
and seeing i f t h e
terminalvoltage changes i n theexpecteddirection.

However, thevoltagemagnitude
may v a r yi n v e r s e l yw i t h
varoutputfor
a capacitive load asshownin equation
6.

EXAMPLES

-VB=Q

(6)

S i x cases areincluded
where
two
s o l u t i o n s were
obtained using standard load flow algorithms.
Case 1 i s f o r a 150 bus studysystemwhichrepresentsproposedadditions
to anactual system. The onel i n e diagrams shown i n Figures 3-5 o n l yr e p r e s e n tt h e
p a r t o f t h e system where theextraneoussolution
seems
t o o r i g i n a t e and does n o t show a l lo ft h ei n t e r c o n n e c t i o n so ra l ol ft h el o a d .T h i s
case
converges
t o an
extraneoussolution
when usingthe
Newton Raphson a l w
i
l starto
gorithm.
The nodal i t e r a t i v ea l g o r i t h m
convergeontheextraneoussolution,reducingthesysw
t o orders of magnitude, but eventually
temmismatchby
diverges if a s u f f i c i e n t l y s m a l l t o l e r a n c e i s
used. The
nodal i t e r a t i v e methodwould,
therefore, appear toconverge ifa l a r g e rt o l e r a n c e
was used t ot e s tc o n v e r gence.

Such capacitiveloads may o c c u rd u r i n gl i n ee n e r g i z i n g


o rl o a dr e j e c t i o n .
The voltagemagnitude may a l s o v a r y
inversely with var output for
systems w i t h l o n g h e a v i l y
IV).
loadedlines(Appendix
Suppose t h a t a generator becomes pinned a ti t s
maximum var 1m
i t
i d u r i n g one o f t h ee a r l yi t e r a t i o n s
l e a d i ntgtohseo l u t i oonf
a network.
Then, using
i
l remainpinned
a t i t s maxit e s t 3, thegenerator w
i
m
i
t aslongastheterminalvoltageremains
mum var l
lessthanthedesiredvalue
oneachsucceeding
iteration.But
a realgeneratorsupplyingcapacitiveload

TEST CASE FOR DulBLE WJTIONS

CASE 1

FIGURE 3.

EXPECTED SOLUTION

526

FIGURE 4.

FIGURE 5.

EXTRANEOUS
SOLUTIOP!

IMPEDANCES OF1 4 100 MVA BASE


527

FIGURE 6.

FIGURE 7.

FIGURE 8.

EXPECTEr) SOLUTIO!'

FIGURE 9.

FIGURE 10.

EXTRAHEOUS SOLUTION

IHPEDANCES O H A 100 HVA BASE

FIGURE 11.

EXTRANEOUS SOLUTION

IVPEDANCES ON A 100 tlVA BASE

Many 1 oad flow programs a l l ow the user to delay i m posinggeneratorvarlimitsuntilafterseveraliterat i o n s .T h i si s


done t o make convergence more c e r t a i n .
For t h i s system the Newton Raphson solutionconverges
tothecorrectsolutionifthegenerator
var l i m i t s a r e
not imposed u n t i l a f t e r t h e second i t e r a t i o n .

Case 3 i s the same a s Case 2 e x c e p tt h a tl i n er e s i s t a n c e is represented and a lighter load is used.

Case 4 was solved


using
several
solution
algorithms. In eachcasethesolution
was initializedwith
to zero,thevoltage
magnitudes
a1 1 voltage angles set
a at l gl e n e r a t o r s e at t h ed e s i r e dv a l u e
and the
voltage
magnitude
a t each
load
bus s e at 1t . 0 .
The
Newton Raphson algorithmconverged
t ot h e h i g h voltage
solution.
A decoupled Newton Raphson a l g o r i t h n and a
nodal iterativealgorithm
bothconverged
t o theother
solution.
With d i f f e r e nitn i t i asl t a r t i n gc o n d i t i o n s ,
t h e Newton Raphson a l g o r i t h n will converge t o e i t h e r
solution.

EXPECTED SOLUTION

This c a s e i s anexample of a f a l s e s o l u t i o n caused


by incorrectlyappliedgeneratorconstraints.
Observe
Figures 13 and 14which show t h ef a l s es o l u t i o n sw i t h
bus 708 s e t a t w
t o differentvalues.
t h ev a rl i m i t sa t
These f i g u r e s show t h a tt h e bus v o l t a g e sr e a c tt ot h e
changeinreactive
power limitintheopposite
way t o
that normally expected.

Case 2 usesthenetworkfromthesamplecalculationintheprevioussection.
The
Newton
Raphson a1 gorithmwillreadilyconverge
t o e i t h e r of two soluan i n i t i a lg u e s s midway between the two sot i o n s .I f
l u t i o n si su s e d ,
however, thealgorithm
shows a d e f i nitepreferencefortheexpectedsolution.
The nodal
iterativealgorithnwillonlyconverge
on theexpected
solution.

a 19-bus equivalent of planned


a
Case 4 i sf o r
South
American
system.
For c l a r i t yt h eo n e - l i n ed i a gram
shown
i n Appendix IV does n o t show a l lt h e
high
impedance connectionspresentintheequivalent.
Two
s o l u t i o n sf o rt h e
same operatingconditionsare
shown
f o rt h i s
system.
One of the solutions has
reasonable
has unreasonably h i g h
voltages and theothersolution
voltages.

hm

CONCLUSIONS
Multiple network s o l u t i o n s canoccur when r e a l i s t i c systemsaresolved
by t r a d i t i o n a l methods starting
from reasonable
initial
conditions.
Each solution
to
but
the networkcan representastableoperatingpoint,
i n many cases one solutionisextraneous
or f a l s e .
In
sanecasesextraneous
or f a l s e s o l u t i o n s may give reali s t i c lookingvoltages.Extraneoussolutionsaremost
1 ikely to occur when thesystembeingstudiedhaslong,
heavily
loaded,
high
voltage
lines.
False
solutions
may occurfora
systemhaving
generators which seea
capacitiveapparent systemimpedance.Severalexamples
inof m u l t i p l e s o l u t i o n s a r e documented in this paper,
cludingcases from w
t o actualstudysystems.

528

TEST CASE FOR DOUBLE SOLUTIONS


CASE #Y

FIGURE 12.

FIGURE

13.

EXPECTED
SOLUTION

FALSE SOLUTION
529

039NVH3 80L 3 N I H 3 W NO I
I
H
I
l
W h H1II.I N O I l f l l O S 3SlVJ

3tlnEII,

APPENDIX I

VOV
Sin(a2)
X

Network o f F i g u r e 1

VoltageSolutionfor

V,

S t a r t i n gw i t he q u a t i o n s (I.l),
(1.2) and(1.3)
for
t h e r e a l and r e a c t i v e power o u t o f bus 2, s o l v e f o r t h e
voltage magnitude V a t bus 2.

Cos(a2

- al)

+ V,

- al)

Since
Cos(a2

= P

Cos(a2)

(1.2)

2V + VXB = 0 (1.6)

must
equal
Cos(a2)
equation

(I
.6) may be r e w r i t t e n a s e q u a t i o n ( I I.1).
vOv
Sin(a2
X

- al)

= -P

v v

0
X Sin(a2)

v2 -

VV,

Cos(a2
X

- ul)

= P

V,

(1.3)

VXB
7
=

V +

(11.1)

PX

vo Sin(a2)

(1

XB
-7
)=

V.
(11.2)

Rearrangeequation(11.2).

(I.l),
(1.2) and (1.3).

Rearrangeequations

cos(a2)
= V2B

Substituteequation(1.2)toeliminatevariable

(1.2)

Yo Cos(a2)

v2 -

Yo Cos(a2)

Vo2 Cos(a2) Sin(a2)

PX(1

- 7XB)

= 0

(11.3)

Vo2 Cos(a2) Sin(a2)


P =
X(1
Vo Cos(a2

- al)

Vo Cos(a2)

2V + VXB = 0

Usingthetrigonometricidentity
substitute(1.4)and(1.5)into(1.6).
/

- $1B

(11.4)

Use Sin(28) = 2 Sin(B)Cos(e) to get equation

( I .6)

(11.5).

Vo2 Sin(2a2)

Cos(e)=h-Sin(8)'

P =
ZX(1

(11.5)

XB

i
l occur when the
The
maximum
power t r a n s f e r w
1 and will, therefore, be given by equation
Sin(2a2)
(11.6).

Rearrange equation (I. 7).

vo

J: -m2

\I

= V(1

-7
XB )

PMAX

2X(1

(11.6)

XB
-7
)

Startwithequation(1.5)
and (1.11) from Appendix
I and equation(11.2)
from Appendix B.

Rearrange equation (I. 9).

APPENDIX I11

Squarebothsides.

$2

"0

(Px)2

- v4

(1

-7
XB )

(1.10)

Using
the
quadratic
formula
to
solve
equation
( I . 10).

v2

= 0

* /v4 0 2(1

4(PX) 2 (1

- 7XB1 2

7
XB)2
(1.11)

XB
v

XB)2

2(1

(I.
11)

-7

Square equation

APPENDIX I 1

Power Equation for Network of Figure

Start withequation(I.l),(1.2)
Appendix I.

vv

0
Sin(a2
X

( I .5).
(111.1)

and (1.6)

fran
Substitute(1.11)intoequation(111.1).

- a1

= -P

(PX) 2 2 ( 1
Sin 2(a,) =

531

-7
XB)2

(111.2)

APPENDIXV
I

Rearrange equation (111.23.

Change i n GeneratorReactive
i n Yo1 tage

Power Outputwith

Change

Rearrange equation ( I I.2).

p2x2 (1

70

- 9)2
= Sin2 (a,)

Substituteequation

COS 2 (a,)

(111.4)

(111.4) i n t o e q u a t i o n (111.3).

2
= -Bvl
- + - - - v:
2,v:
Q,
2
x

Rearrange equation (111.5).

Cos(al)

(111.6)
I f P1 remainsconstantthen
Therefore,

la1 > 145"l f o r t h e r o o t c o r r e s p o n d i n g t o a

- sign.

la1 < 145"l f o r t h e r o o t c o r r e s p o n d i n g t o a

+ sign.

Equation (11.5) gives an e x p r e s s i o nf o rt h e


t r a n s f e r i n terms o f t h e a n g l e a.

v02

P =

~in(2a~)

vX 2 Sin(al)
dP1 = 0 = -

2V1 + '1'2X
Cos(al)

2al

or

power

(11.5)

Using t h i se q u a t i o n ,c a l c u l a t e
power w i t h r e s p e c t t o a n g l e .

"=riVative

' ~ I C

of

x-

2v1 v2
-aQl
=
-BY1 +
X Cos(al)
avl

2
V2 Sin (a,)
x Cos(al)

(111.7)
Note t h a t

t o be a
(111.7)
stableoperatingpoint.?Accordingtoequation
t h ed e r i v a t i v e w
i
l be p o s i t i v e i f 1.1
< 45" and negaT h e r e f o r e ,o n l yt h ep o s i t i v er o o t
t i v e f o r la1 > 45".
ofequation(1.11)representsastableoperatingpoint.
This derivative

must be p o s i t i v e f o r t h e r e

aQl tends

t o be n e g a t i v e f o r l o n g e r l i n e s

avl
and f o r h e a v i l y l o a d e d l i n e s w i t h a l a r g e

al

systemangle

REFERENCES

1) H. W. Domnel, W. F. TinneyandW. L. Powell,


"Further
Development I n Newton'sMethodFor
Power System Apat
IEEE
The a u t h ogr r a t e f u l l y
acknowledges t h he e l oppf l i c a t i o n s , "
Paper 70 CP 161-PWR, presented
L. N . Hannett o f Power Technologies,Inc.forprovidingWinter
Power Meeting, New York,January,
1970.
one o f t h e examplecases
and o f H. K. Clark and J. M.
U n d r i l lo ft h e
same firm forseveralsuggestionswhich
2) S. B. Crary, Power System S t a b i l i t y ,V o l .
I , pp. 75have c o n t r i b u t et hodweo r tohfpi sa p e r .
79, John Wiley 8 Sons, Inc., New York, 1955.
ACKNOWLEDGMENT

532

Discussion
H. E. Brown (North Carolina State University, Raleigh, North Carolina):
I believe the title of the paper has been poorly chosen and is actually
misleading. The solutions are neither extraneous nor false but are possible actual operating conditions of the system. Systems that have considerablelinecharginglike
the extensivecablesysteminChicago,
or
verylongtransmissionlineslikeinBrazil;
and operating with many
generators at or near their var limits can by improper operating prcceduresarrive at, and be maintainedindefinitely at one of the extraneous voltage solutions. The system will operate in this condition
all day with no chance to recover to the normal condition until the
load level is reduced in the daily load cycle.
In the morningpick-up, if the machinesare not committed on
schedule, the generators on the system will be driven against their var
limits without being able to keep the system voltages up to acceptable
values. The low system voltage will reduce the var contribution of the
line charging, which contributes to the low voltage difficulty. Recovery
to the correct solution is impossible.
I found such a case several yearsago in a severe emergencycase of
the Commonwealth Edison Company system. The van were so critical
that a reduction of 50 MVAR at a station near the center of the cable
area caused the system to collapse. Starting thesolution at 1.1 pu and at
0.9pugave two different solutions. These solutions would have been
duplicatedbythesystemitself
under differentoperatingprocedures
and would continue in this condition until theload was decreased or an
additional var supply was made available.
Has the author considered obtaining relief for such a system by
simultaneouslyover-loadingall generators by increasing their vars? Increasing the generator van would raise the voltage which would increasing the linechargingvars.
The systemwouldrecover
its normal
voltageand the generatorswouldautomatically be returned to their
rated loads. The only problem would be to convince the operators that
this is a workable method.
Manuscript received August 5,1976.

M. E. El-Hawary (Faculty of Engineering and Applied Science, Memorial


University of Newfoundland, St. Johns, Newfoundland, Canada): This
is an interesting paper in that it emphasizes the simple fact that systems
of simultaneous nonlinear equations will in general havemore than one
solution. The loadflowis just one of those problems. This discussor
wonders if the author has considered alternative models of transmission
lines in his investigation. Although the simple system used to illustrate
the point does so well, it opens more questions. It should be an easy
exercise to include the equivalent - pi model of the long line in the
present analysis. What effect would this have on the conclusions? I believe that in all probability a criterion for inclusion of adequate models
in parts of the network, may result. This, of course, is contingenton the
conclusion that only one practical solution is possible.
Another very interesting point that needs emphasis is that many a
power systems engineer assumes that a solution is obtained once the
specified tolerance on the error is met. The paper reports that this just
may not beenough.Falseconvergencecanoccur.Would
the author
indicate the values of safe tolerance for his example systems. This paper
indicates clearly the need for more studies of the LF problem. In particular more efficient algorithms are called for to handle such special
cases. This is a welcome addition to the literature on the LoadFlow
Problem.

to do this we would appreciate complete specifications including loads,


turns ratios, shunt admittances, acceleration factor, and swing bus. If
load-flow algorithms are usedto analyze steady-state stability,it may be
wise to consider sensitivity of loads to voltage changes that may OCCUT
in the dead time before transformer tapchangers are activated. Of the
two phenomenaMr. Johnson has illustrated, we feel that he has not
fully explored the occurrence and implications of multiple solutions.
WFERENCES
J. K. Delson, The Spurious Solution in Load Flow Calculations,
AIEE Conference Paper CP 62-277 presented at the AIEE Winter
General Meeting, New York, N. Y., January 28-Febmary 2, 1962.
2.A.J.Korsak,On
the QuestionofStableLoad-FlowSolutions,
IEEE Transactions on PowerApparatus and Systems, PAS9 1,
May-June, 1972, pp. 1093-1 100.
1.

Jerome Meisel (Wayne State University, Detroit, Michigan): This paper


is presenting data documenting the well known fact that the load-flow
problem, simply phrased in the vector form
f(x) = 0

(1)

can have numerous solutions here designated as (x1, x2,...).These solutions correspond to system
equilibrium
states which
are
either
asymptotically stable or not. Theoretically the system can operate at
any of these asymptotically stable states. However, the real power system may not be able to exist at certain of the asymptotically stable
equilibrium states obtained as solutions of(1) dueto component limitations not included in the problem formulation. The author is using adjectives such as erroneous, false, extraneous, expected, and correct to
describe these multiple solutions. In my view, these terms are confusing
and need further mathematical clarification. The adjective false appears
to be reserved for solutions obtained by changing equation (1) to a different set say
g(x) = 0

(2)

due, for example, to a var limit condition. Solutionsto (2), denoted as


as (El,3 , ...) are termed false if the switch from(1) to(2) were not really required due to the specific system under study.
In summary, I have the following two points to make concerning
this paper:
(1) Consideration of the asymptotic stability of system equilibrium
states needs to be included for the higher order systems if we want to
consider whether or not the system can indeed operate at an obtained
solution. Has the author checked the asymptotic stability of the nine
cases he has presented in the paper?
(2) Changing the load-flow equations from (1) to (2), due to such
matters as var limits, simply leads
to a solution of (2) even though a
solution of (1) may still be desired. The term false solution is, in my
view, confusing when used to describe such an instance. The author is
cautioning the users of standard load-flow programsthat simply examining output to see if results look reasonable is not always sufficient.
Manuscript received August 16, 1976,

Manuscript received August 5 , 1976.

J. Dekon (Ben Gurion University of the Negev, Beer-Sheva, Israel) and


E. Ofry (Tel Aviv University, Ramat Aviv, Israel): Mr. Johnson presents
two phenomena: multiple solutions of the load-flow problem and false
solutions. Multiple solutions were analyzed
for networks that include
resistance as well as reactance in Reference 1. (Mr. Johnsons analysis is
limited to reactivenetworks.) Mr. Johnson demonstrates howfalse
solutions may arise from a defect in the algorithmused to represent
constraints on the reactive output ofgenerators. In contrast to false
solutions, multiple solutions describe states that canphysically occur
ina network. To quote one example, Apparently power circulation
hasbeenobservedoccasionally
in the donut of the WesternUnited
States interconnected system ( 2 ) . We believe there are cases where a
power system operator must increase reactive power input in the early
morning hours or else be locked in to one of the extraneous, lowvoltage states throughout the day. From the computer-analysis point of
view, Mr. Johnsons example could be explored
more thoroughly, but

B.K. Johnson: The author thanks the discussers for their thoughtful
contributions.
Sincepresenting the paper the author hastalked to severalengineers who have experienced multiple load flow solutions.
It appears
that multiple solutions may be more common than generally supposed.
H. E. Browns description of his experience with multiple solutions is a
valuable addition to the paper. It appears to be similar to the expectations expressed by J. Delson.
The mechanism which caused the multiple solutions
reported by
Mr. Brown is apparently different from those described in the paper. In
his case the multiple solutions occurred when generator var limits were
applied as they should have been. The fust mechanism described in the
paper did not involve generator var limits, and the second mechanism
involved generator var limits which were incorrectly applied.
From H. E. Browns brief description,
it is not readily apparent
what mechanism was responsible for the multiple solution, but the following scenario seemsto be consistent with his description.
1. Several generators are supplying a central load and
are driven
against their var limitswithout being able to keep voltage up.

Manuscript received August 10, 1976.

Manuscript received September 30,1976.

533

2. If any one generator increases its voltage set point, it will cause
a slight decrease in the total var requirement because of the increased
line charging. However,it will increaseits own var output because it will
feed vars to neighboring machines (on regulator control). It will thus
exceed its var limit, and the operator will decrease the voltage set point
to its previous value.
3. If all of the generators could increase their voltage set points
simultaneously, they would decrease the total var requirement as well
as their individual requirements, and the system would reachthe desired
operating point.
This scenario seems possiblein real life, and it might be possibleto
converee
.~~~to both solutions with the traditional load flow. But neither
of the& solutions would be a false solution as described in my paper.
In a false solution the load flow incorrectly constrains the generator
at maximum reactive output because the derivative of reactive generation with respect to voltage is negative. In step two of the above scenario
the generator encounters an actual var limit, and the derivative of reactive power with respect to voltage is positive.

effective
range
var
requirements
of
load
of
transformer

~~

HighSide
Vol tage
varrequirements
oflinecharging

FIG. 3

DES

= maximum var
c a p b1i
it y
o f generator
VDES

desired voltage

FIG. 1

FIG. 2

Other mechanisms which produce multiple load flow solutions no


doubt exist. For instance, multiple solutions would exist if a generators
var requirements varied with voltage as shown in Figure 1.
Such a characteristic might result for the network in Figure 2,
which shows a line with substantial charging supplying a reactive load
through a tap changing transformer.
The reactive requirement of a load normally increasesas the square
or some higher power of voltage. The tap changing transformer, however, will hold the secondary voltage (and thus the load) relatively constant over some range of high side voltage
so the var requirement shown
in Figure 3 results.
As the discussers correctly point out,the presenceof multiple
solutions does not imply that one of the solutions is steady state unstable or extraneous. This fact was mentioned in the paper but perhaps
with insufficient emphasis. The threebus sample network was,of
course, not intended to represent all of the complexities of a real system.
For this simplifiedmodel with constant voltage sources, one of the
operating points is steady state stable (aP/aa is positive) and the other
operating point is unstable (aP/aa is negative). With different load and
generator exciter representations, both operating points might be stable
534

or bot1
; points might be unstable, DUT IT seems
reasonable to
expect less stability for the operating point with larger voltage angles
across the system.
The author cannot agree with J. Meisel that the term false solution
has been misapplied. The term false solution was reserved in the paper
for those solutions which result when the var limits of a generator are
incorrectly applied as J. Meisel states. To a mathematician this may not
be a false solution since the algorithm has solved the wrong equations
correctly. But the load flow program is responsible for both solving and
selecting the equations. To a powerengineer the answersare just as
wrong when the error isin selecting equations as they are when the
error is in solving them, so the term false solution seemed appropriate
to the intended audience.
The subject matter in J. K. Delsons paper and the present paper
are similarinsome respects, although he emphasized the solution algorithm while this paperemphasizespower
system characteristics.
Mr. Delsons paper proposes a nodal iterative algorithm, which will converge only to stable operating points. Reference 1 describes a similar
algorithm which uses Newtons method. The steady state stability of an
operating point depends upon many factors, which the traditional load
flow and Mr. Delsons algorithm do not represent. These include characteristics of generators, exciters and regulators, voltage and frequency
varying load characteristics, and complex shunt characteristics. I understand that J. Delson and E. Ofry are currently working on an improved
algorithm which will represent some of these characteristics. In the
meantime, when steady state stability is in doubt, it can be checked
using a dynamic simulation program or several other algorithms, which
are currently available and designedfor that purpose.
Unfortunately, because of space limitations the author is unable
to provide full documentation of the 15Bbus and a 19-bus load flow
case as requested by J. Delson. The documentation is complete for the
other cases described in the paper.
In response to M.E. El Hawary,the author believes that the results
would not changesignificantly if the transmissionline in the simple
system were represented by several equivalent n sections. In the 150-bus
case the transmission lines were broken up into short sections and did
not prevent the multiple solution from occurring.
A tolerance of .OOO1 p.u. (voltage change per iteration) proved to
be adequate to prevent false convergence as described for Case 1 in the
paper.
In conclusion, it appears that there are several mechanisms which
cancause multiple operating points for the real and/or a load flow
representation of the system. In some cases multiple solutions may occur in the load flow representation which do not occur for the real system. These occur because of deficiencies in the load flow representation
of load, generators, exciters, and voltage regulators.There may be other
cases where multiple operating points occur in the real system, which
cannot be represented in the traditional load flow.
REFERENCE
1.V.

A. Venikov, V. A. Stroev, V. I. Idelchick, V.I. Tarasov, Estimation of Electrical Power System Steady-State Stability In Load
Flow Calculations, IEEE Transactions On Power Apparatus And
Systems, PAS 94, May/June 1975, pp. 1034-1041.

Anda mungkin juga menyukai