IEEE Transactions on Power Systems, Vol. PWRS-2, No.

3, August 1987

652

LOW BUS VOLTAGE AND ILL-CONDITIONED NETWORK

SITUATIONS IN A COMPOSITE SYSTEM ADEQUACY EVALUATION
S. Kumar, Student Member IEEE
R. Billinton, Fellow IEEE
Power System Research Group
University of Saskatchewan
Canada
Abstract - The adequacy assessment of bulk power
systems generally involves solution of a network
configuration under different outage events. Various
solution techniques, depending upon the adequacy
criteria employed and the intent behind the studies
are available in order to analyze the adequacy of a
power system.
Transportation models [1,2] D.C. load
flow and A.C. load flow solutions [3,14,5] are the most
commonly used techniques.
The transportation model
approach and the D.C. load flow techniques do not take
into account bus voltage violations and generating
unit VAr limit violations in a network.
In order to
include
these
considerations,
it
is
therefore
necessary to employ an A.C. load flow technique to
evaluate the adequacy of the system.
A.C. load flow
techniques are computationally more expensive and
require larger storage than approximate techniques. A
drawback of an A.C. load flow technique is the
possible non-convergent situation for a network.
Under these circumstances, it is not possible to
calculate adequacy indices for the system. This paper
clearly illustrates these non-convergent situations in
two test systems on which adequacy studies were
conducted.
A heuristic approximate technique to
handle these non-convergent or ill-conditioned network
situations is also presented and illustrated by
application to the two test systems.
1.0 INTRODUCTION
Electrical energy plays a vital role in the
economic and social development of a region, a
This responsibility
state/province and a country.
places
considerable pressure on electric power
utilities to provide an uninterrupted adequate power
supply of acceptable quality to its customers. It is
not only the continuity of the power supply but the
quality that is of considerable concern to power
It is, therefore,
system engineers and managers.
necessary to examine the voltage levels at each major
load centre and the VAr limits of each generating unit
while considering the effect of outages of different
components, viz, generating units, transmission lines
and transformers.
Considering a power network as a
transportation model or using approximate simple load
flow techniques such as D.C. load flow etc. does not
provide any estimate of the bus voltages and the
reactive power limits of the generating units. If the
quality of power supply is an important adequacy
criterion, then more accurate A.C. load flow methods
[3,14] such as Gauss-Seidel, Newton-Raphson and second
order load flow techniques must be employed to assess
These techniques are computationally
the adequacy.
more expensive than approximate methods and have large
It is therefore mandatory in
storage requirements.
large system studies to use computationally faster

A paper recommended and approved

86 SM 310-7
by the IEEE Power System Engineering Committee of
the IEEE Power Engineering Society for presentation
at the IEEE/PES 1986 Stummer Meeting, Mexico City,
Mexico, July 20 - 25, 1986. Manuscript submitted
January 31, 1986; made available for printing
kpril 23, 1986.

Printed in the U.S.A.

A.C. load flow techniques. One such technique is the

fast decoupled A.C. load flow developed by Stott and
Alsac [5]. A digital computer algorithm [6,7] using a
fast decoupled A.C. load flow technique has been
developed at the University of Saskatchewan.
The
program solves the power network for each outage
contingency considered in the calculation of adequacy
indices.
The program can consider simultaneous
independent outages up to four generating units and
two lines/transformers.
The terms line and element
outages have been used interchangeably throughout this
paper.
They are intended to include both line and
transf ormer outages.
Using an A.C. load flow approach to solve a power
network sometimes, however, results in non-convergence
for the outage condition.
In this case, the
quantitative assessment of the situation becomes
extremely difficult. These non-convergence situations
are frequently encountered while
considering the
outages of transmission lines/transformers.
The
simplest solution to this situation is to allow the
system buses to stay at the low voltage and treat the
outage event as a system failure due to the bus
voltage violation(s)
This approach, however, does
not provide a quantitative measure of the voltage
violation problem
and also does not give due
consideration to the severity of the outage event.
These events are treated as failure events regardless
of the voltage magnitude at the system buses.
A major objection to this assumption is whether
low voltage really constitutes a system/bus failure.
Many power utilities use D.C. load flow for adequacy
studies because they do not view low bus voltage as a
failure but only as a minor problem which is normally
rectified
transformer
by
tap-settings
and/or
phase-shifter adjustments.
This treatment, however,
gives an optimistic assessment because correction of
voltage violations may not be possible for all voltage
violation contingencies. It also does not permit a
quantitative evaluation of the outage contingencies
The actual
using voltage as an adequacy criterion.
situation lies somewhere between the two viewpoints.
It is desirable to use voltage as an adequacy
criterion but suitable corrective action should be
taken when encountering any voltage problem.
A basic voltage correction model suitable for use
with the Newton-Raphson load flow is developed in
Section 3.0.
In addition to low voltage at one or more system
buses, another problem which is also experienced while
using an A.C. load flow is a non-convergent load flow
situation.
These situations pose a major obstacle in
system adequacy evaluation as it is difficult to
quantify the adequacy indices in the event of
Most
of
the
non-convergence.
non-convergent
situations result due to high values of the mismatch
in reactive power beyond the permissible tolerance
limit. Very few situations result due to high values
A third possibility
of the mismatch in active power.
is that a load flow may not converge although a
This non-convergence
solution, in fact, does exist.
could occur due to numerical problems with the fast
decoupled algorithm and/or the characteristics of the
.

numerical formulations used.

In order to avoid these

non-convergent situations when a solution does exist,

an
additional algorithm has been included in the
digital computer program. This algorithm is discussed
in Section 2.0.
If a non-convergence

0885-8950/87/0800-0652$01.00O1987

IEEE

situation

still

persists,

653
after checking for the non-cornvergence that may result
because of the numerical formulation, it is presumed
that it is due to the fact that under given operating
conditions of the network, the A.C. load flow does not
have a solution. These non-convergence situations as
well as the low bus voltage situations described
earlier are illustrated using the following two power
systems:

(1)

(2)

the IEEE Reliability test system (RTS) [8]

and
a 45 bus model of the Manitoba assisted
Saskatchewan Power Corporation (SPC) system

[7].

A brief description of these systems is as

follows.
This
(1) The IEEE Reliability Test System (RTS):
system was established by an IEEE task force in 1979
as a reference network to test and/or compare methods
for system reliability analyses.
The single line
The
diagram of this system is shown in Fig. 1.
generating units, bus and line data are given in
The system has 10
Tables 7 to 13 of Reference [6].
generator
(PV) buses, 10 load (PQ) buses, 33
transmission lines and 5 transformers.
The total
number of generating units is 32. The minimum and the
maximum rating of the generating units are 12 MW and
The voltage limits for the
400 MW respectively.
system buses are assumed to be 1.05 and 0.95 p.u.

Reston and one fictitious bus, are included to

represent equivalent assistance from the Manitoba
Hydro System. A power import of 300 MW from Manitoba
is represented by 3 units of 100 MW each at the
fictitious bus, which is connected to three buses, The
Reston
and
Roblin.
The
Pas,
lines
and
interconnections between the four buses are assumed to
be an equivalent power network of the Manitoba system
for these adequacy studies.
The system has 8
generator
(PV) buses, 37 load (PQ) buses, 71
transmission lines/transformers and 29 generating
units.
The minimum and the maximum rating of the
generating units are 15 MW and 280 MW respectively.
The voltage limits for the system buses are assumed to
be 1.05 and 0.95 p.u.
This paper has been divided into five sections.
Section 2 presents a technique for preventing a load
flow from diverging.
In Section 3, heuristic
algorithms for the correction of voltage violations
and non-convergent situations are given.
This is
followed by a discussion of the findings on the two
test systems and finally conclusions.

2.0 TECHNIQUE FOR PREVENTING A LOAD FLOW

DIVERGING DUE TO NUMERICAL FORMULATION

FROM

When solving an outage contingency using the

Newton-Raphson load flow technique, at any point in
iteration process the voltage/angle increment computed
on the basis of mismatch powers from the previous
iteration may project the voltage/angle solution
outside the local neighborhood where the solution
exists. The solution range is never encountered and
the iterative process diverges. Many approaches are
to
available
prevent
the
divergence
of
the
Newton-Raphson load flow solution under those
situations when a solution does exist. Powel [9] has
suggested that the convergence property of the
Newton-Raphson load flow can be improved by scaling
the solution projection calculated by the load flow
algorithm without changing the direction of the
projection.
A set of generalized equations for the
load flow can be written as:

AY(x) = J(x).Ax
old

new

where,
x

= solution

AY(x) =
J(x) =

Ax
a

Fig. 1.

(2) The

Single line diagram of the IEEE Reliability

Test System (RTS) .

Assisted
Saskatchewan
Power
The single line diagram of
the Manitoba assisted Saskatchewan Power Corporation
(SPC) system is shown in Fig. 2. The system has 45
buses in total, of which 4 buses, The Pas, Roblin,
Manitoba

Corporation (SPC) System:

=
=

vector, voltage magnitude

phase angle
mismatches
Jacobian matrix
solution vector correction
scaling factor or acceleration factor.

or

In an ordinary Newton-Raphson load flow, the

Powel [9] and recently
scaling factor a = 1.0.
Iwamoto and Tamura [10] have developed analytical
expressions for the scaling factor, such that the sum
of the squares of power mismatches is minimized. In a
recent EPRI [11] report on "Transmission System
Reliability", a heuristic technique was suggested to
adjust the scaling factor by monitoring the sum of the
squares of the power mismatches before and after each
voltage magnitude and phase angle correction.
This
technique is readily adaptable to the fast decoupled
load flow approach.
Under normal load flow
situations, the scaling factor is taken as 1.0, but in
the case when the sum of the squares of the power
mismatches for the new iteration exceeds the value
calculated from the previous iteration, the scaling
factor is decreased from its initial value of 1.0 by a
factor.
The value of the factor is arbitrary and
could lie between 0.0 and 1.0.
During one complete
load flow cycle, the value of the scaling factor is

654
Bus 38

60

70

'66
711

us42

Bus 45

Hypoth-

~tical
167

9I
Reston
Bus 44

--

Bus 41

Bus 30

Represents
ut01
Hydro Assistance

'aIi

Single line diagram of tf-Le Manitoba Assisted

Saskatchewan Power Corporation (SPC) System.
(3) If the load flow has converged, adequacy
decreased whenever the sum of the squares of the power
indices [7,121 are calculated and the next
mismatches exceeds its previous value. A flow diagram
In those cases
contingency is considered.
The main
of the algorithm is shown in Fig. 3.
when the load flow does not converge,
features of the algorithm are as follows:
appropriate corrective actions are employed
to handle the non-convergent situation.
(1) The load flow solution progresses in an
These corrective actions are discussed in
unmodified way until either the sum of the
Section 3.0.
squares of the real power mismatches SP or
the sum of the squares of the reactive power
After integrating the algorithm into the digital
mismatches SQ shows an increase rather than a
computer program, it was tested on two test systems
decrease when the P-8 or Q-V portions of the
described in this paper. Table I gives a list of the
load flow solution is executed respectively.
non-convergent outage contingencies examined before
(2) At any point during the iteration process, if
applying the above technique. It was found that even
SP increases from its previous value, the
after using the technique, none of the previous
scaling factor a for the P-8 portion of the
non-convergent situations realized convergence. This
load flow is deBreased to half of its old
implies that for these contingencies, there is no
New values of the phase angles are
value.
solution possible for the operating conditions under
This also indicates
calculated and again SP is calculated. If SP
which the load flow is solved.
is still larger than what it was when a was
that the solution for these contingencies does not
PThis
is further halved.
unity, then a
diverge because of numerical problems and/or the
characteristics of the mathematical formulation used.
process continSes until SP is smaller than
its value when a was 1 .0 or until such time
It is therefore necessary, in order to solve these
that a becomes 4:maller than a cut-off value.
non-convergent contingencies, to modify the operating
The sine procedure is repeated for the Q-V
conditions so that the load flow converges within the
A heuristic
portion of the load flow. The cut-off values
MW and MVAr mismatch tolerance limits.
of bot4 scale factors, a and a was chosen
algorithm was developed and successfully incorporated
as 10
; a is the scalin' factorqfor the Q-V
into the digital computer program. This algorithm is
portion of the load flow.
described in Section 3.0.

Fig. 2.

655
3.0 A

HEURISTIC

ALGORITHM

FOR

THE

CORRECTION

VOLTAGE; VIOLATIONS AND NON-CONVERGENT SITUATIONS

OF

The outage of one or more lines may result in low

voltage at some of the system buses and the A.C. load
flow may not converge. The outage of the line(s) in a
net power exporting area may result in slightly higher
values of the voltages at a few of the system buses i n
the area and significant increase in the bus angles at
the generation buses. This is caused by the fact that
power cannot be transferred from this area to other
areas of the system due to the outage of the line(s).
Conversely, the outage of a line(s) in a net power
importing area may cause lower values of the voltages
and a decrease in the bus angles at some of the buses,
particularly load buses, in the region.
Outages of lines 23 and 29 and of lines 24 and 28
for the IEEE RTS (Fig. 1) result in a significant
increase in the bus angles at generator buses 18 and
22.
These buses are situated in the north (230 Ky)
On the
region which is a net power exporting area.
other hand, outages of elements 6 and 7 and that of
lines 6 and 27 in the net power importing area result
in low voltage at bus 3 and a decrease in the bus
The outage of
angles at generation buses 1 and 2.
lines 11 and 12 and that of lines 11 and 13 result in
a low voltage at bus 8.
These non-convergent and low voltage situations
were solved by taking the following corrective
actions:

(i) Rescheduling of the generating units.

(ii) Injection of reactive power at the voltage
violating buses.

While

F ig. 3. Flow diagram for the application of a scaling

factor to prevent divergence of the load flow.

TABLE I
A list of non-convergent outage contingencies
for the two test systems

IEEE RTS
Elements 6 and 27 Out
Elements 6 and 7 Out
Elements 23 and 29 Out
Elements 11 and 13 Out
Elements 24 and 28 Out
1
1
10
10
10
11

11
11

11
17

23
24
25
26
26
tlements 28
El ements 29

and
and
and
and
and
and
and
and
and
and
and
and
and
and
and
and
and
and
and
and

SPC System

10 Out
57 Out
19 Out
56 Out
66 Out
14 Out

34 Out

57 Out
Out

66
26
24
26
56
46

Out

Out
Out

Out
Out

70 Out

36 Out
36 Out

Elements
Elements
Elements
Elements
Elements
Elements
Elements
Elements
Elements
Elements
Elements
Elements
Elements
Elements
Elements
Elements
Elements

the

A.C.

load

flow,

if

the

its initial estimated value, the generating units are

If the
rescheduled as explained in Section 3.1.
magnitude of the voltage at any bus during the load
flow iterations is not within the permissible limits
even though the load flow has converged, reactive
power is injected at the buses as described in Section

3.2.

3.1 Rescheduling of the Generating Units

Line/Transformer
Outage Contingencies

El ements
Elements
El ements
Elements
El ements
Elements
El ements
Elements
El ements
Elements
El ements
Elements
El ements
Elements
El ements

iterating

absolute value of any bus angle deviates beyond twice

adequacy indices

1 and 11 Out

5 and 26
10 and 34
10 and 57
11 and 13
11 and 18

11 and 56
11 and 60
16 and 26
18 and 57
23 and 26
25 and 34
26 and 37
26 and 51
28 and 29
28 and 41
29 and 41
Elements 38 and 65
Elements 58 and 59
Elements 64 and 65

41 Out
Elements 36
65 Out
El ements 54
63 Out
Elements 58
Outage of element 43 and any element outage
combination involving element 43.

Out

Out
Out

Out

Out
Out
Out
Out
Out
Out
Out

The bus angles are significantly influenced by the

real power generation pattern in a network. Therefore
outages of lines which are responsible for power
transfer from the net power exporting area(s) to the
net power, importing area(s) cause an increase in the
bus angles in the net power exporting region and a
decrease in the bus angles in the net power receiving
area. This can be seen in the IEEE RTS when lines 23
and 29, and lines 24 and 28 are out of service.
A heuristic algorithm which monitors the buis
angles and adjusts the generation schedule accordirng
to deviations in the bus angles was developed and
integrated within the digital computer program. The
algorithm resolves many of those situations which
A
otherwise would result in non-convergent cases.
flow diagram for the algorithm is shown in Fig. 4.
The main features of the algorithm are as follows:

Out
Out
Out
Out
Out
Out
Out

10l

(1)

If the absolute value of the bus angle

at
any bus increases beyond twice its initial
estimated value, the angle deviation Ae at
the generator buses having reserve capacity
available are calculated. The average value
of these angle deviations is also
AO

(2)

The

Out

deUgmined.

scheduled generation at the generator

buses, which have an angle deviation lower

than the average value is increased to the

total generation capacity available at those
buses. However, a restriction on the reserve

656

(3)

(4)

available for rescheduling * an also be

imposed, if required, by slightly modifying
the algorithm.
The scheduled generation at the generator
buses, which experience angle deviations
is
average
value,
the
larger
than
proportionately decreased by an amount which
is now available because of augmenting the
generation as described in (2).
After rescheduling the generating units, an
A.C. load flow is again carried out starting
with the initial estimates of the load flow
parameters.

(2)

(3)

that the load f low has converged btut the bus

voltages are lower than the permissible
voltage limit, the power factor of the load
is improved by 10% of its original value.
The A.C. load flow is again carried out with
the original estimates of voltage magnitude
V and bus angle 90 as the starting values.
I? the voltage at al0 buses is within the
specified voltage limits and the load flow
converges, the algorithm proceeds to step
(3), otherwise to step (2).
If the modified power factor of the load is
less than 0.9, it is further improved by 10%,
otherwise the net reactive power (local VAr
generation + load VAr) at a bus is increased
to a value which is 20% of the active load.
The A.C. load flow is again carried out and
If the
the voltage limits are checked.
voltage at the system buses is still not
further
is
power
reactive
acceptable,
increased by 20% of the active load and so
The process of injection of reactive
on.
power can be stopped after a predetermined
In
value of reactive power is supplied.
these studies, the maximum reactive power
that can be injected was taken as 140% of the
At this stage,
real load demand at a bus.
load flow is again carried out. If the load
flow diverges, a D.C. load flow is performed
are
calculated
indices
and
adequacy
accordingly.
The line overloads and the generator VAr
limits are then checked. After taking proper
corrective action [13] for these problems, if
are
the
indices
adequacy
necessary,

calculated.

Increase the

scheduled generation
proportionately

Fig. 4.

Decrease the
scheduled generation

proportionately

~ _-

Flow diagram for rescheduling the generating

units for the line outages.

3.2 Injection of Reactive

Violating Buses

Power

at

the

Voltage

The magnitude of the bus voltages are primarily

influenced by the reactive power generation pattern in
the network. Therefore, injection of a proper amount
of VAr at the voltage violating buses helps in
alleviating the voltage violation problem for most
Power factor improvement
outage contingencies.
devices such as synchronous condensers or static
capacitor banks could be installed at the appropriate
buses. Besides supplying the VAr support under normal
circumstances, these devices improve the quality of
A
the power supply under line outage events.
heuristic algorithm was developed to calculate the
reactive power required in order to correct the
voltage violations at the system buses for an outage
A flow diagram of the algorithm is shown in
event.
Fig. 5. The main characteristics of the algorithm are
as follows:

(1)

If the voltage at any load bus decreases to

50% of its initial estimated value (generally

1.0 p.u.) while solving the A.C. load flow or

iPig. 5.

Flow diagram for correcting the voltage

limits by injecting reactive power.

657
4.0

DISCUSSION OF THE RESULTS

Table II shows the total number of voltage

violation contingencies for the line outages in the
two test systems before employing any corrective
actions as described in Section 3. The three voltage
levels chosen are 0.95 p.u., 0.90 p.u. and 0.85 p.u.
for each bus in each system.
A brief description of the major voltage violation
contingencies for each system is as follows:

(2)

(3)

TABLE II
Number of voltage violation contingencies for three
voltage levels

System description

IEEE RTS
SPC system

Permissible minimum voltage

limit in p.u. at each bus

0.85

0.90

0.95

39
178

334

39

144

(4)

(5)

715

voltage at bus 5. Outages of lines 1 and 10,

10 and 18, 10 and 34 (transformer), 10 and
56, 10 and 57 and of 10 and 66 result in low
voltage at four system buses, namely bus 5,
6, 7 and 8.
The outage of element 11
(transformer)
together with the outage of any other
line/transformer result in low voltage at
buses 3 and 5.
The outage of lines 58 and 59 results in low
voltage at f ive system buses 12, 13, 114, 15
and 16. Buses 12, 13 and 14 also experience
low voltage due to the outage of elements 25
and 34.
The outage of line 36 and the outage
combination of any other line/transformer
with line 36 result in low voltage at bus 19
and bus 20.
Buses 22 and 24 experience low voltage due to
the outage of element 43 (transformer) and
any other outage combination with the element
43. Outages of transformers 43 and 45 also
result in low voltage at buses 23 and 25.

The IEEE RTS (Fig. 1.):

The number of voltage
violation contingencies increases significantly if the
voltage limit is raised from 0.90 p.u. to 0.95 p.u.
The number of voltage violation contingencies remains
unchanged if the voltage limit is further relaxed to
0.85 p.u.
A summary of major voltage violation
contingencies is as follows:

(b) Voltage limit = 0.90 p.u.

In addition to the above voltage violation
contingencies, the outage of line 40 and any other
outage combination involving line 40 result in low
The outage of line 41 and any
voltage at bus 19.
other outage combination with line 41 also result in
low voltage at bus 20.

(a) voltage limit = 0.85 p.u. or 0.90 p.u.

(1) The outage of line 10 and all other outage
combinations of line 10 with any other
element result in low voltage at bus 6.
(2) The outage of elements 2 and 7, 2 and 27, 6
and 7 and of 6 and 27 results in low voltage
at bus 3.
(3) The outage of lines 11 and 12 and of 11 and
13 causes low voltage at bus 8.

(c) Voltage limit = 0.95 p.u.

If the bus voltage is increased to 0.95 p.u. from
0.90 p.u., the following additional line outages
result in voltage violation contingencies.

(b) voltage limit = 0.95 p.u.

If the bus voltage is increased to 0.95 p.u. from
0.90 p.u. the following additional line outages result
in voltage violation contingencies.

(1)

(2)
(3)
(4)
(5)

The outage of element 7 (transformer) and all

other outage combinations of the transformer
with any other line/transformer result in low
voltage at bus 3.
The outage of line 11 together with the
outage of any other line/transformer result
in low voltage at bus 8.
The outage of line 27 and any other outage
combinations involving line 27 result in low
voltage at bus 3.
The outage of elements 4 and 7, 4 and 11, 4
and 14, 4 and 15 and of 4 and 27 results in
low voltage at bus 4.
The outage of elements 11 and 13 and of 14
and 15 results in low voltage at bus 9.

The SPC System (Fig. 2)

This system model is prone to frequent voltage

violation problems.
The total number of line
contingencies tested was 2556.
Voltage violation
contingencies represent 7%, 13% and 28% of the total
line contingencies at three voltage levels of 0.85
p.u., 0.90 p.u. and 0.95 p.u. respectively. A summary
of the major voltage violation contingencies is as
follows:

(a) Voltage limit

0.85 p.u.
(1) The outage of line 10 and any other outage
combination with this line result in low

(1)
(2)
(3)

(4)

(5)
(6)

The outage of line 25 and any other outage

combination with this line result in low
voltage at bus 13.
Many outage combinations of line 26, 27 and
of 35 result in low voltage at bus 18.
The outage of line 29 and any other outage
combination with line 29 result in low
voltage at bus 20.
Both buses 8 and 12 experience low voltage
due to the outage of, either element 34
(transformer) or line 56 and any other outage
combination involving either of them.
Many outage combinations of line 28 result in
low voltage at bus 20.
The outage of lines 58 and 59 results in low
voltage at bus 38.

After taking suitable corrective actions as

described earlier, the voltage violation problem no
longer exists in any of the systems.
Only 2
contingencies i.e., (i) the outage of transformers 23
and 24 and (ii) the outage of lines 28 and 29 for the
The non-convergence of
SPC system do not converge.
these two contingencies is due to the fact that the
the
iteration limit exceeds
maximum
number
of
iterations which was chosen as 30 in this case. It is
worthy of note that prior to employing the corrective
the
of
total
number
actions,
non-convergent
contingencies for the SPC system was 112. The maximum
VAr rating of the power factor improvement devices
required for test systems are given in Table III and
Table IV.
The three permissible voltage limits
considered are 0.95, 0.90 and 0.85 p.u.
The most
severe
outage contingency requiring the maximum
improvement is also given in each case.

5.0 CONCLUSIONS

A quantitative evaluation of the voltage violation

contingencies and non-convergent contingencies in

658
conjunction with the appropriate corrective <actions
Two simplified
has been presented i n this paper.
heuristic algorithms, that can be easily integrated
into the fast decoupled A.C. load flow have been
suggested f'or rescheduling the generating units and
for calculating the VAr rating of a power factor
violation
each
voltage
for
device
improvernent
has
techniques
The
incorporated
contingency.
successfully alleviated many of the voltage problems
cause
which
situations
non-convergent
and
the
discontinuities in the quantitative evaluation of
system adequacy indices. It should, however, be noted
that the effectiveness of the heuristic algorithms
presented in Section 3.0 depend on the availability of
real and reactive power at the -appropriate buses. In
this study it was assumed that the total installed
capacity at any bus is available for rescheduling.
Restriction can, however, be imposed on any generating
bus after making minor modifications in the algorithm
The availability of
presented in Section 3.1.
reactive power can be made possible by installing VAr
supplying devices at appropriate buses in a power
network.
This may, however, involve extra capitol
Decision regarding the trade-off
expenditures.
between the quality of the power supply and the
additional investment should be made judiciously by
system planning engineers and system managers.

6.0 REFERENCES
1.

3.
4.

5.
6.

7.
8.

TABLE III
MVAr rating of a power factor improvement device
for the IEEE RTS

9.

Bus Voltage in p.u.

0.85
0.90
0.95
Line
Line
Line
Bus MVAr Contingency MVAr Contingency MVAr Contingency

11.

3
4
6
8
9

108
15
42
203
35

6
4
9
11
14

& 7 out
& 7 out
& 10 out
& 13 out
& 15 out

72

25

68

7 out

72

9 & 10 out
11 & 13 out

25

6 &

68

10.

7 out

6 &

9 & 10 out
11 & 13 out

TABLE IV

MVAr rating of

imnprovernent device
for the SPC system
.,- -

--_--

0.85
0.95
Line
Line
Line
Bus MVAr Contingency MVAr Contingency MVAr Contingency
3

13

14
15
16
18

19

69
76
30
23
45
31
12
12
39

12

20
22
23
24

109
47
17

25
38

25

9
5

11
10
10
10
10
58
25
58
58
58
27
20
28
1
43
1
43
58

&
&
&
&

&

&

&

&

&

&

&

57
57
57
57
57
59
34
59
59
59
35

13.

a power factor

Bus Voltage in p.u.

0.90

5
6
7
8
12

12.

Ford, D. R. Ful kersonr, "Flows i n Networks",

(book), Princeton University Press, Princeton,
N.J., 1962.
H.J. Koglin, E. Roos, W.M. Wellsrow, "Application
of Reliability Calculation Methods to Planning of
Third
Networks",
Distribution
Voltage
High
International Conference on Reliability of Power
Supply Systems, The Power Div. of the TEE,
Conference Pub. No. 225, London, U.K., Sept.,
1 983.
G.W. Stagg, A.H. El-Abiad, "Computer Methods in
Power System Analysis", McGraw-Hill Book Company,
New York, 1968.
B. Stott, "Review of Load Flow Calculation
Methods", Proc. IEEE, Vol. 62, July 1974, pp.
91O0-929.
B. Stott, 0. Alsac, "Fast Decoupled Load Flow",
IEEE Trans. PAS-93, 1974, pp. 859-869.
T.K.P. Medicherla, "Reliability Evaluation of
Composite Generation and Transmission Systems",
Ph.D. Dissertation, lJniversity of Saskatchewan,
Saskatoon, Canada, Dec. 1978.
S. Kumar, "Adequacy Evaluation of Composite Power
of
Ulniversity
Thesis,
M.Sc.
Systems",
Saskatchewan, Saskatoon, Canada, July 1984.
IEEE Committee Report, "IEEE Reliability Test
System", IEEE Trans. PAS-98, 1979, pp. 2047-1054.
Unconstrained
to
"Problems
Powel,
M.J.D.
Optimization", New York Academic Press, 1972.
S. Iwamoto, Y. Tamura, "A Load Flow Calculation
Method for Ill-conditioned Power Systems", IEEE
Trans. PAS-100, 1981, pp. 1736-1743.
EPRI Report, "Transmission System Reliability
Methods - Mathematical Models, Computing Methods
and Results", Technical report EPRI EL-2126, Power
Technologies Ine., New York, July 1982.
IEEE Task Force, "Reliability Indices for Use in
Bulk Power System Adequacy Evaluation", I EEE
Trans., PA--97, 1978, pp. 1097-1103.
T.K.P. Medicherla, R. Billinton and M.S. Sachdev,
"Generation Resheduling and Load Shedding to
Alleviate Line Overloads - Analysis", IEEE'Trans.
PAS-98, 1979, pp. 1876-1884.
L. R.

out
out
out
out
out

out
out
out

out
out
out

& 40 out
& 36 out
& 43 out
& 45 out

& 43 out
& 45 out
& 59 out

4
69
76
24
23
45
20

12

20

11
10
10
10
10
58
25
58
58
58

&
&
&
&
&
&

&

&
&
&

57
5'7
57
57
57
59
34
59
59
59

out
out
out

out
out
out
out
out
out
out

17

20 & 40 out
28 & 36 out
& 43 out
43 & 45 out

13

43 & 45 out

12
82
47

t & 43 out

52
76
24
18

45

11
10
10
10
10
58

&
&
&
&
&
&

out
out
out

out
out
out
out

20
9
9
20

25
58
58
58

12
82

28 & 36 out
28 & 36 out
1 & 43 out
43 & 45 out
1 & 43 out
43 & 45 out

47
17
9
13

&
&
&
&

57
57
57
57
57
59
34
59
59
59

out
out
out

Discussion
M. G. Lauby (MAPP CC, Minneapolis, MN): The authors are to be
complimented on their paper presenting techniques to produce reasonable
analysis techniques for voltage adequacy assessment. I have a few
questions.
The discusser has many times wondered if the sole application of the
fast decoupled load flow (FDLF) for adequacy assessment would result
in a number of unsolved cases, though there is a solution, which many
times occurs when contingency enumeration is being performed. Namely,
though the FDLF is tolerant of the starting voltage vector, it can subsequently diverge because the starting poitit is not adequate enough for
solution. Many times a Gauss-Seidel technique is used to gain a better
starting vector in these cases. The authors have presented an application of Iwamoto et al. which they indicate will resolve this problem.
Unfortunately, this could not be demonstrated in the paper, as all conditions which diverged actually had no reasonable solution. Have the
authors continued their investigations with this technique and could they
demonstrate that an unsolved case using the FDLF is subsequently solved
using the FDLF with the Iwamota et al. technique?
I can appreciate the heuristic algorithms implemented for with adequacy
assessment where only one reasonable solution is required and minimization of generation shift or losses is not needed. Can it be demonstrated
that these algorithms are quicker, based on CPU requirements, than