Sie sind auf Seite 1von 26

A Guide to Steady-State Voltage Stability Analysis

Anthony B. Morton
November 2007
Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2 DC Power Flow Across Series Resistance . . . . . . . . . . . . . . . . . . . . . . . . . . 1
3 AC Power Flow Across Series Impedance: Two-Bus Load Flow . . . . . . . . . . . . . 3
4 Approximate Solution for Voltage in terms of P, Q . . . . . . . . . . . . . . . . . . . . 6
5 Sensitivity of Voltage to Reactive Power . . . . . . . . . . . . . . . . . . . . . . . . . . 7
6 Graphical Representation of Stability Limit . . . . . . . . . . . . . . . . . . . . . . . . 7
7 Relation to Power-Angle Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
8 Operation at Constant Power Factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
9 Remote Generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
10 Eect of Long Lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
11 Eect of Shunt Admittance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
12 Shunt Susceptance, Limits and Voltage Sensitivity . . . . . . . . . . . . . . . . . . . . 19
13 Summary and Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Appendix: Derivation of the Two-Bus Load Flow Solution . . . . . . . . . . . . . . . . 23
1 Introduction
Voltage stability, broadly speaking, refers to the ability of a power system to sustain appropriate
voltage levels through large and small disturbances. Kundur [2] discusses many aspects of voltage
stability and techniques for stability analysis.
The purpose of this paper is to explain in some detail one aspect of voltage stability most
commonly encountered in discussions with utilities. This is termed steady-state voltage stability
because it is assessed using steady-state load ow solutions of the network, rather than dynamical
considerations. (Despite this, it does have important implications for power system dynamics.)
Steady-state voltage stability is concerned with limits on the existence of steady-state operating
points for the network. The nonlinearity of the load ow equations for an AC network places
constraints on the maximum power ow that can occur on network branches before the equations
come to possess no solution and the network enters voltage collapse. These power ow constraints
are additional to the more familiar static constraints based on thermal limits and voltage drop, and
in transmission networks are often more stringent.
This paper has been written to provide some background to the voltage stability problem, and
to document exact solutions in the simplest cases. While these exact solutions are not technically
dicult to obtain, they appear to be omitted from most discussions of steady-state voltage stability.
2 DC Power Flow Across Series Resistance
A simple picture of the basic voltage stability problem is provided by considering power transfer in
a DC circuit, where there are no complex numbers to complicate matters. In Figure 1, power is
1

+
V
S
R
?
Load,
power P
+

V
Figure 1: Transfer of power in a DC circuit
delivered from a DC voltage source to a DC load via series resistance R. If V
S
denotes the source
voltage and V the load voltage, one can immediately write the current in the circuit as
I =
V
S
V
R
(1)
and the power P transferred to the load as
P = V I =
V (V
S
V )
R
. (2)
It is convenient to express this in terms of the ratio V/V
S
of load voltage to source voltage magnitude,
abbreviated to throughout this paper. The power equation that results is
P =
V
2
S
R
(1 ), =
V
V
S
. (3)
Figure 2 plots the power ow P against the voltage ratio . The left hand plot illustrates the
fundamental relationship (3) in the way it is usually taught to rst-year engineering studentsas
the maximum power transfer law. This way of viewing it emphasises the principle of optimising the
power transfer and only the power transferas for example when transmitting a signal down a wire
in an audio or telecommunications circuit. It is then true that one maximises the signal power by
matching the input impedance of the receiver to the impedance of the line, thereby making = 1/2
and attaining the peak of the power curve.
(Usually when this circuit is presented to students, the load is taken to be resistive, so that the
load voltage is given by a simple voltage-divider rule. Thus the impression is sometimes gained that
(3) is simply the maximum power that can be developed in a resistor. This misses the important
point that the relation (3) holds regardless of the nature of the load: the only assumption we used in
deriving it via (1) and (2) was that the load draws real power P. So (3) is a fundamental maximising
principle with broad application.)
In the power system context, however, it is more instructive to view the relationship as on
the right hand side of Figure 2, with power on the horizontal axis. This is because the technical
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
0.05
0.1
0.15
0.2
0.25
Voltage ratio
P
o
w
e
r

P
,

a
s

p
r
o
p
o
r
t
io
n

o
f

V
S 2

/

R
0 0.05 0.1 0.15 0.2 0.25
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
V
o
lt
a
g
e

r
a
t
io

Power P, as proportion of V
S
2
/ R
Figure 2: Power-voltage relationships in a DC circuit
2

V
S
R

jX
?
P, Q
+

V
Figure 3: Transfer of power in an AC circuit
objective in a power system is to satisfy a given demand for power while keeping both losses and
voltage uctuations to a minimum. Viewed this way, what equation (3) does is establish a hard limit
of V
2
S
/4R on the amount of power that can be delivered across the resistance R. But it is also clear
that operating at or near this transmission limit is undesirable, because:
1. Its too inecient. When = 1/2 the voltage drop across R is V
S
/2, and the I
2
R loss in the
line is equal to the power actually delivered to the load.
2. There is too much voltage uctuation. This is particularly clear from the right hand side of
Figure 2: we see that the sensitivity /P of load voltage to power becomes arbitrarily large
as P approaches the limit. In practice load power always uctuates, but large uctuations in
voltage give rise to icker in lamps and other undesirable eects.
3. The system is liable to collapse into an undesirable operating condition. Again, this is evident
from the right hand side of Figure 2. If P is less than the limiting value, there are two possible
operating points for , and these approach each other rapidly as P tends toward the limit. If
the load continues to draw the same power at any voltage level, operation close to the limit
increases the likelihood of dynamic uctuations causing a transition from a desirable high-
voltage, low-current operating point to an undesirable low-voltage, high-current operating
point.
These three considerationsparticularly the thirdsum up the reasons why power system engineers
concern themselves with steady-state voltage stability. Notice again that all three stem from the
nature of the steady-state operating conditions (Figure 2 is a steady-state characteristic), and not
from the dynamical characteristics of connected plant. Plant dynamics may exacerbate the condi-
tions that drive a system into instability, but the instability is still considered to originate in the
sensitivity of steady-state voltage to steady-state power ow.
3 AC Power Flow Across Series Impedance: Two-Bus Load
Flow
Voltage stability in conventional AC power systems depends on the vagaries of AC power ow, where
we must switch to phasor analysis and consider reactive power in addition to real power.
Figure 3 presents the AC analogue of Figure 1: an AC voltage source feeding an AC load (with
real power P and reactive power Q) across an AC series impedance R +jX.
In power system terms, this is a 2-bus load ow problem. Load ow equations are highly nonlinear
and except in the very simplest cases it is not practical to seek an analytical solution. Even in the
2-bus case the exact solution is rarely discussed, except in the special case where R = 0, when it is
stated as
P =
V
S
V
X
sin (4)
with V
S
and V the magnitudes of the terminal voltages, and the phase dierence between them
(the power angle). Equation (4) immediately gives the limit V
S
V/X for the maximum power transfer
in terms of the terminal voltage magnitudes. However, it suers from the drawback that it does
not provide any direct information on the way the load voltage V varies with P, in the way that
equation (3) does in the DC case. And, of course, it fails to apply when the resistance R is a
signicant fraction of the reactance X, as is the case with many real AC lines.
3
0 10 20 30 40 50 60 70 80 90 100
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
1.2
Real power P (MW)
V
o
l
t
a
g
e

r
a
t
i
o

30 MVAr
20 MVAr
10 MVAr
No Q
-10 MVAr
-20 MVAr
-30 MVAr
Figure 4: Power-voltage relationships in an AC circuit (r = 0.001, x = 0.005)
It appears to be little known that the two-bus load ow, for resistive as well as lossless lines, has
a fairly straightforward analytical solution. It takes the form of a quadratic equation for V
2
, where
V is the magnitude of the load voltage:
V
4
+
_
2(RP +XQ) V
2
S
_
V
2
+ (R
2
+X
2
)(P
2
+Q
2
) = 0. (5)
(See the Appendix for a derivation.)
As in the DC case, this can be written more conveniently in terms of , the ratio V/V
S
of load
to source voltage magnitude. For convenience we adopt the notational shorthand
r =
R
V
2
S
, x =
X
V
2
S
. (6)
It is best to regard this simply as a notational device rather than as a per-unit representation, since
V
S
will not always be equal to the nominal system voltage. A convenient scale of units in most power
system studies is to express V
S
and V in kV, R and X in ohms, P in MW and Q in MVAr. The
units of r and x are then inverse MVA, so that when multiplied by P or Q they give dimensionless
quantities.
In terms of , r and x, and the load power P and Q, equation (5) reads

4
+ (2(rP +xQ) 1)
2
+ (r
2
+x
2
)(P
2
+Q
2
) = 0. (7)
This can be solved as an ordinary quadratic equation for
2
. The solution is most conveniently
stated by dening the two dimensionless variables
= rP +xQ (8)
= xP rQ (9)
and the solution then reads

2
=
1
2
_
1 2
_
1 4( +
2
)
_
. (10)
Figure 4 illustrates the solution for in terms of P, for the values r = 0.001, x = 0.005 and several
values of Q.
Figure 4 for AC power transfer is broadly similar to Figure 2 for DC, with an additional eect
due to the reactive power Q. As one may expect, a positive Q (VAr import) at the load tends to
depress the voltage V relative to V
S
, while a negative Q (VAr export) raises the voltage. But at any
xed value of Q there are again two branches to the solution for the voltage in terms of P (the
4
PV curve), and these branches meet at a limiting value of P (the voltage collapse point), just
as in the DC case.
Note that the maximum power transfer for stable operation is highly sensitive to the reactive
power Q: in the example of Figure 4, a 30MVAr export at the load more than doubles the transfer
capacity compared with a 30MVAr import. But while injecting reactive power boosts the transfer
capacity, it also increases the critical voltage at the turning point of the PV curve. For a 30MVAr
import the critical voltage is around 55% of V
S
, while for a 30MVAr export the critical voltage is
around 75% of V
S
.
(Important: For simplicity we assume for the time being that the reactive power is associated
with a constant-Q source, such as would be provided by a Statcom. Section 8 deals with the case of
a load having a xed power factor, while Sections 11 and 12 deal with reactive compensation using
passive shunts such as capacitor banks.)
Clearly it is important to be able to identify the voltage collapse point on the PV curve, where
P reaches a maximum (P
max
, say) and the steady-state solution breaks down. Equation (10) allows
this to be done fairly easily, since this is just the point where the quantity under the square root
falls to zero. In other words, where
4( +
2
) = 1. (11)
Substituting the denitions of and and expanding, this results in a quadratic equation for P
max
:
4x
2
P
2
max
+ 4r(1 2xQ)P
max
+ 4r
2
Q
2
+ 4xQ1 = 0 (12)
with the single positive solution
P
max
=
_
(r
2
+x
2
)(1 4xQ) r(1 2xQ)
2x
2
. (13)
Equation (13) is the exact formula for the real power ow limit in an AC circuit that absorbs
constant reactive power Q at the receiving end. The formula has a number of useful approximations
and special cases. For example, if xQ 1 one has the approximate formula
P
max

z r
2x
2
(1 2xQ) (14)
where z =

r
2
+x
2
is the line impedance divided by V
2
S
. (See the following section for the square-
root approximation used.) Thus, the power transfer limit is a roughly linear function of the reactive
power, with VAr absorption reducing the limit and VAr injection increasing it.
If the circuit is lossless (r = 0) the formula (13) takes the simpler form
P
max
=

1 4xQ
2x
. (15)
This can also be well approximated when xQ 1 as
P
max

1
2x
Q, xQ 1. (16)
In other words, in a lossless circuit there is an eective limit on the sum of the real and reactive power
given by 1/2x, or equivalently V
2
S
/2X. So each 1MVAr of reactive power injected at the receiving
end increases the power transfer capacity by around 1MW. This however is an approximation and
becomes invalid when Q is large.
The other quantity important for steady-state stability is the critical voltage V
C
, or the voltage
at the point P = P
max
. Again, this is the point where the quantity under the square root in equation
(10) becomes zero, so the voltage ratio is
=
V
C
V
S
=
_
1
2
=
_
1
2
rP
max
xQ
=
_
z
2
(1 2xQ) rz

1 4xQ
2x
2
. (17)
If xQ 1, an approximation similar to that in (14) gives the approximate formula
V
C
V
S

_
z(z r)
2x
2
(1 2xQ). (18)
5
-50 -40 -30 -20 -10 0 10 20 30 40 50
0
20
40
60
80
100
120
Reactive power Q absorbed at load (MVAr)
P
o
w
e
r

t
r
a
n
s
f
e
r

(
M
W
)
,

V
o
l
t
a
g
e

(
%

o
f

V
S
)
Power transfer limit P
max
Critical voltage
Figure 5: Transfer limit and critical voltage as function of Q (r = 0.001, x = 0.005)
This conrms analytically that the critical voltage decreases with increasing VAr absorption at the
receiving end, and increases with increasing VAr injection. This imposes a limit on the strategy of
increasing circuit capacity by injecting large amounts of reactive power, because this injection also
raises the critical voltage until it becomes similar in magnitude to V
S
. It is undesirable to operate
the receiving end close to the critical voltage because this carries a greater risk of voltage collapse;
on the other hand, it is not practical to run the receiving end at a voltage V signicantly greater
than V
S
.
Figure 5 depicts the power transfer limit and the critical voltage as a function of Q, for a circuit
with the same parameters as in Figure 4.
4 Approximate Solution for Voltage in terms of P, Q
While it is not particularly relevant to the topic of voltage stability, it is worth noting that many com-
mon rule of thumb approximations to the voltage in a two-bus load ow can be easily derived from
the exact solution (10). The technique involved is the so-called Taylor or binomial approximation
for the square root:

1 x 1
x
2
, x 1. (19)
This can be applied to the square root in (10) in the case where and are small: that is, where
the line impedances (r, x) are small, or the load powers (P, Q) are small, or both.
Taking just the conventional positive solution in (10) and using the approximation (19) gives

1
2
_
1 2 + (1 2 2
2
)
_
= 1 2
2
. (20)
Then, taking one more square root and using the approximation (19) a second time gives
1

2
2
= 1 rP xQ
(xP rQ)
2
2
. (21)
For a rule-of-thumb approximation one usually ignores the
2
/2 term entirely (as it is an order of
magnitude smaller than ). Written in terms of the original voltages, the resulting approximation
reads
V V
S

RP
V
S

XQ
V
S
. (22)
6
Equation (22) provides a simple rst approximation for small voltage drops on an AC circuit.
When the circuit is lossless or nearly lossless (R 0), and per-unit quantities are used with the
source voltage close to nominal (V
S
1) the voltage drop is given approximately by the product
XQ. For resistive circuits the product RP also contributes to the voltage drop.
The validity of this approximation relies on the quantities RP/V
2
S
, XQ/V
2
S
, XP/V
2
S
and RQ/V
2
S
all being much smaller than 1. As is evident from Figure 4, it is a very poor guide to voltage drop
when this condition fails.
5 Sensitivity of Voltage to Reactive Power
An alternative criterion for steady-state voltage stability is the sensitivity V/Q of voltage to
reactive power, for a xed level of real power ow. According to this criterion, the circuit is considered
stable if an incremental reactive power draw reduces the voltage, and unstable if it increases the
voltage.
In Figure 4 it can be seen that at a xed value of P (that is, on a vertical line), the voltage will
always reduce with increasing VAr demand provided the circuit is operating on the desirable upper
half of the PV curve. However, if the circuit operates on the undesirable lower half of Figure
4, this relationship is inverted and an increasing VAr demand will increase the voltage. (Strictly
speaking, this is not true in some cases near the bottom left of the diagram, but this operating region
is of much less practical importance than the region near the turning points of the PV curves.)
This sensitivity criterion has a fairly clear practical rationale. Voltage control in transmission
and distribution networks is usually accomplished using reactive shunt devices that adjust their
reactive power, on the assumption that positive Q draw (more inductive) will suppress the voltage
and that negative Q draw (more capacitive) will boost it. If the system enters an unstable operating
condition where the reverse is the case, the voltage controller enters a positive-feedback situation,
where voltage is continually adjusted in the wrong direction until a protective trip of network
elements results, leading to the possibility of cascading failure.
The sensitivity of voltage to reactive power can be analysed with the help of equation (10).
Provided V
S
is held xed, incremental changes in V are equivalent to incremental changes in =
V/V
S
, and so the criterion can be evaluated using the partial derivative /Q. Taking the positive
solution in (10) gives

Q
=
1
2

Q
_

2
_
=
1
2
_
x +
x + 2r
2
Q2rxP
_
1 4 ( +
2
)
_
. (23)
Now, we are interested primarily in the sign of this quantity. So the multiplicative factor 1/2 need
not greatly concern us, since is always positive. Likewise, the quantity 1 4( +
2
) under the
square root is always positive provided P is less than P
max
.
It is seen that given our underlying assumptions (the circuit is a plain series impedance, source
voltage is held xed, receiving end power is xed at P and Q, and the circuit operates in the
positive branch with P < P
max
), the stability criterion V/Q < 0 can be violated only in circuits
with non-negligible resistance. Furthermore, instability requires either that Q be very large and
negative (of order 1/r) or that P be very large and positive (again of order 1/r). In the latter case
P is likely to exceed P
max
in any event.
When the eect of shunt admittance must be considered, the picture is more complicated, al-
though the overall conclusion is the same. See Section 12 below for further discussion.
6 Graphical Representation of Stability Limit
In Section 3 a quadratic equation (12) was derived for the power transfer limit P
max
in terms of the
reactive power Q and the impedance parameters r and x. Since quadratic equations can be graphed
relatively easily, this raises the possibility of illustrating the stability limit graphically, giving a
pictorial summary of the variation of P
max
with Q, r and x.
A convenient visualisation starts from equation (11), rewritten as
=
1
4

2
. (24)
This is the equation of a vertical parabola in the (, ) plane, shown on the left hand side of Figure
6. The limit curve crosses the axis at = 1/4 and is symmetric about this axis. It crosses the
7
-0.5 0 0.5
0
0.25
Stable region
Unstable region

-150 -100 -50 0 50 100 150


-100
-50
0
50
100
Stable region
Unstable region
- / z
/ z
P
Q

V
S
2
/ 4Z
V
S
2
/ 2Z
Figure 6: Graphical representation of stability limit equation (12)
axis at = 1/2 and = 1/2. Since the distance from the origin to these intercepts is twice the
distance to the apex, the origin is the focus of the parabola, speaking geometrically. The region of
stability is the interior of the parabola, where +
2
< 1/4; the contrary holds for the exterior
region.
To make this meaningful in terms of the power P and Q, notice that the denition of and
equations (8) and (9)can be written in the matrix form
_

_
=
_
r x
x r
_ _
P
Q
_
. (25)
The matrix appearing here is the matrix of a rotation in the plane, albeit slightly disguised. It
becomes more explicit upon dividing through by z =

r
2
+x
2
and reversing the sign of :
_
/z
/z
_
=
_
cos sin
sin cos
_ _
P
Q
_
, cos =
r
z
, sin =
x
z
, z =
_
r
2
+x
2
. (26)
Equation (26) says that if one takes the axes in a PQ chart and rotates them anticlockwise through
the angle , one obtains the and axes as in the left hand side of Figure 6only with a scaling
by z and a reection in the axis.
The result is shown on the right hand side of Figure 6. The stability limit curve is a parabola,
having its focus at the origin of the PQ plane and its axis of symmetry inclined at an angle
= arctan(X/R) to the P axis. The apex of the parabola is located a distance V
2
S
/4Z along this
axis, where Z =

R
2
+X
2
is the circuit impedance in the original units. The parabola passes
through two other points located symmetrically at a distance of V
2
S
/2Z along an axis perpendicular
to the axis of symmetry (the latus rectum, in technical terms).
The importance of this graphical visualisation is not as a method to construct the limit curves
(since we have computer software to do this), but rather as a tool to help build engineering intuition
about the limits of operation for AC circuits. By picturing how the parabola of Figure 6 moves
through PQ space as parameters are varied and how P
max
varies along the curve as Q changes,
initial conclusions about the capacity of circuits can be drawn which can then be rigorously tested
in computer simulations.
7 Relation to Power-Angle Analysis
Lets now return to equation (4), which is the way the voltage constraint on power ow is most
commonly expressed:
P =
V
S
V
X
sin.
The usefulness of this equation (at least in the lossless case) is that it establishes a relationship
between the power ow on the one hand, and the power angle on the otherand this depends only
weakly on other system variables when the system operates as intended. Since voltages in a power
system are usually controlled within a narrow range90% to 110% in most transmission systems
the quantity V
S
V/X varies only slightly from its nominal value V
2
b
/X, where V
b
is the nominal system
8
ZI
V
V
I I
S

Figure 7: Angle relationships for two-bus AC circuit


voltage. Accordingly, changes in power ow P are intimately connected with changes in the power
angle . Furthermore, since sin can never exceed 1, this equation establishes an absolute upper
limit on power ow in lossless circuits, which holds regardless of how much reactive compensation
is available.
How does our analysis in terms of and variables relate to this more familiar power-angle
analysis, and how might it be extended beyond the lossless case? Figure 7 helps answer this question:
it shows the phasor relationships between the voltages V
S
and V , the load current I, and the voltage
drop ZI = (R +jX)I. Aside from these magnitudes, there are four angles appearing in Figure 7:
The power angle between V
S
and V .
The power factor angle , being the angle by which the load current I lags the load voltage
V . (If the current leads the voltage, is dened to be negative.) The load powers P and Q
have their familiar expressions in terms of :
P = V I cos , Q = V I sin. (27)
The circuit impedance angle , which is just the phase of the complex number Z = R + jX
(or the normalised z = r +jx).
The associated power angle . This can be thought of as the amount by which the voltage
drop diers in phase from the ideal value /2, which occurs when the load has unity power
factor and the circuit is lossless. The value of can be derived by inspection of Figure 7 as
=

2
+ . (28)
Equations for the power angle are found by resolving the phasor V
S
into components in phase and
in quadrature with V . From Figure 7 and equation (28), the component in quadrature with V has
magnitude
V
S
sin = ZI cos = ZI sin( ). (29)
Multiply both sides by V and use (27), and there results
V V
S
sin = ZV I(sin cos cos sin)
= XP RQ. (30)
Lastly, normalise both sides by V
2
S
to obtain
sin = . (31)
The component of V
S
in phase with V is, from Figure 7
V
S
cos = V +ZI sin = V +ZI cos( ). (32)
Multiply both sides by V to obtain
V V
S
cos = V
2
+ZV I(cos cos + sin sin)
= V
2
+RP +XQ (33)
9
and then normalise by V
2
S
to obtain
cos =
2
+. (34)
So, one sees that knowing , and the ratio = V/V
S
immediately gives the power angle using
cos = +

, sin =

. (35)
This establishes the relationship between power-angle analysis using and the above analysis using
and .
The importance generally ascribed to power-angle analysis stems from the use of as a proxy
for the voltage stability margin. By (4), the point of maximum power ow occurs when reaches
/2: accordingly, the proximity of to /2 is taken as an approximate margin of stability. Thus
many transmission networks have operational criteria stipulating that for any transmission link
not exceed some threshold value (say, 30 degrees).
We can, however, use the preceding analysisspecically, equation (10)to derive a more precise
stability margin and relate this to the angle . Consider again the quantity under the square root
sign in (10):
=
_
1 4 ( +
2
). (36)
This makes an excellent candidate for a margin of stability: its square root is equal to the dierence
between the two solutions for
2
; it is equal to 1 when the network is quiescent (P = Q = 0); and it
is by denition equal to 0 at the point where voltage collapse occurs. Now, use (35) to write and
in terms of and substitute in (36):
=
_
1 4(cos ) 4
2
sin
2

=
_
1 4 cos + 4
2
(1 sin
2
)
=
_
1 4 cos + 4( cos )
2
= 2 cos 1. (37)
In particular, consider the case when = 0. Then is equal to the critical voltage ratio
C
= V
C
/V
S
,
and is equal to the critical angle
C
. From (37) the two are related by the identity

C
cos
C
=
1
2
. (38)
This relationship between
C
and
C
holds for all AC circuits that can be represented as in Figure
3: that is, with a series impedance and power (P, Q) independent of voltage. Just as importantly, it
demonstrates that the voltage ratio and the angle are of equal importance as stability indicators.
If the angle criterion for stability gains more attention, it is only because one tends to operate within
a narrow range of values, so that it is more common to hold steady and allow to increase with
increased loadings than it is to hold steady and allow to drift.
Of course, equation (38) can also be used to convert a critical voltage formula into a critical
angle formula. For example, taking equation (17) for critical voltage leads to the following formula
for the critical angle in terms of Q and the circuit impedances:

C
= arccos
_
x
_
2z
2
(1 2xQ) 2rz

1 4xQ
_
. (39)
The critical angle is plotted in Figure 8, for the same example as used in Figures 4 and 5. It is
seen that the critical angle declines steeply as reactive power is absorbed at the load, falling below
30 degrees when Q equals roughly 20MVAr. Of course, equation (38) guarantees that the critical
voltage and the critical angle rise and fall together, and that there is a trade-o in maintaining a
high margin between operating voltage and critical voltage on the one hand, and maintaining a high
angle margin on the other.
For reference, Figure 9 depicts the identity relation (38) between critical voltage and angle. It
can be seen that
C
declines precipitously as the critical voltage ratio falls below 75 per cent or so.
10
-50 -40 -30 -20 -10 0 10 20 30 40 50
0
10
20
30
40
50
60
Reactive power Q absorbed at load (MVAr)
C
r
i
t
i
c
a
l

a
n
g
l
e


(
d
e
g
r
e
e
s
)
Figure 8: Critical power angle as function of Q (r = 0.001, x = 0.005)
0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
0
10
20
30
40
50
60
70
80
Critical voltage ratio
C
= V
C
/ V
S
C
r
i
t
i
c
a
l

a
n
g
l
e

C

(
d
e
g
r
e
e
s
)
Figure 9: Relationship between critical voltage and critical angle
11
0 20 40 60 80 100 120
0
0.2
0.4
0.6
0.8
1
Real power P (MW)
V
o
l
t
a
g
e

r
a
t
i
o

0.90 lead, = -0.48


0.95 lead, = -0.33
0.98 lead, = -0.2
unity pf, = 0
0.98 lag, = +0.2
0.95 lag, = +0.33
0.90 lag, = +0.48
Figure 10: Power-voltage relationships for xed power factor (r = 0.001, x = 0.005)
8 Operation at Constant Power Factor
Textbook discussions of voltage stability, such as those of Kundur [2] or Gonen [1], present the basic
PV curves somewhat dierently to Figure 4. This is because they derive these curves not by
varying P with the reactive power held constant, but instead by varying P with the power factor
held constant.
This latter assumption is appropriate for transmission lines supplying a load centre, in the case
where the reactive power comes primarily from natural load variation rather than from auxiliary
compensation equipment. This was a valid assumption in the classical power systems of the early
to mid 20th century, but is increasingly inaccurate given the prevalence of capacitor banks, SVCs
and other compensating equipment in modern transmission systems. The presence of compensation
equipment means that reactive power is controlled more or less independently of real power, so that
curves such as those in Figure 4 may be more relevant.
Nonetheless, the equations for two-bus load ow can be solved just as well on the assumption
of xed power factor as on the assumption of xed reactive power. The equation used (7) and its
solution (10) are the same, only now we make the substitution
Q = P, =

1
pf
2
1. (40)
(The parameter is dened for convenience: it is just another way of expressing the power factor.
Note however that unlike the power factor, is given the same sign as Q: positive for VAr absorption
and negative for VAr injection. Thus for a load, 0.95 lagging power factor is equivalent to 0.33,
while 0.95 leading power factor gives 0.33.)
With the substitution (40), the solution (10) is still used, only with the parameters and
calculated as
= (r +x)P k
1
P (41)
= (x r)P k
2
P. (42)
Then the solution from (10) reads

2
=
1
2
_
1 2k
1
P
_
1 4k
1
P 4k
2
2
P
2
_
(43)
and is plotted in Figure 10 for various dierent power factors. A notable eect of operation at a
constant power factor is seen at some leading power factors, where the voltage can at rst rise
slightly with increasing power ow, before falling again as power approaches the stability limit. By
12
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
20
30
40
50
60
70
80
90
100
110
120
Power factor coefficient
P
o
w
e
r

t
r
a
n
s
f
e
r

(
M
W
)
,

V
o
l
t
a
g
e

(
%

o
f

V
S
)
,

A
n
g
l
e

(
d
e
g
r
e
e
s
)
Power transfer limit P
max
Critical voltage
Critical power angle
Figure 11: Transfer limit, critical voltage and angle at xed power factor (r = 0.001, x = 0.005)
comparison, as Figure 4 shows, when a load operates with constant reactive power irrespective of
real power, the voltage falls with increasing power ow in virtually all practical cases.
As before, the power limit occurs when the quantity under the square root becomes zero, that is
when
4k
2
2
P
2
+ 4k
1
P 1 = 0. (44)
The solution is
P
max
=
_
k
2
1
+k
2
2
k
1
2k
2
2
=
_
(1 +
2
)(r
2
+x
2
) r x
2(x r)
2
(45)
and the corresponding critical voltage V
C
and critical angle
C
are given by
V
C
V
S
=
_
1
2
(r +x)P
max
, cos
C
=
1
2(V
C
/V
S
)
. (46)
The transfer limit, critical voltage and angle for our example circuit (r = 0.001, x = 0.005) are given
in Figure 11. This is the analogue of Figure 5 for operation with constant power factor at the load.
9 Remote Generation
So far the discussion has focussed on the case shown in Figure 3, where real power ows from
the source V
S
to an AC load. Just as important is the case where real power ows in the reverse
direction, from a remote generator toward the source (or innite bus).
It is of course possible to apply all the previous formulae, such as equation (10) for voltage and
equation (12) for limiting power ow, and simply insert negative values of P to represent generation.
But generation normally follows a sign convention where generation quantities are positive, and it
is convenient to recast the formulae to reect this.
For a generator, dene the generated real power P
gen
to be positive for net export (thus equivalent
to P for a load). Likewise, dene Q
inj
to be positive for reactive power export (equivalent to Q
for loads), as is usual practice for a generator. Then for the voltage solution, we replace the variables
and with their sign-reversed equivalents (subscripted with g to avoid confusion with the original
variables):

g
= rP
gen
+xQ
inj
(47)

g
= xP
gen
rQ
inj
(48)
13
0 20 40 60 80 100 120 140 160
0
0.2
0.4
0.6
0.8
1
1.2
Generated power P
gen
(MW)
V
o
l
t
a
g
e

r
a
t
i
o

30 MVAr
20 MVAr
10 MVAr
No Q
-10 MVAr
-20 MVAr
-30 MVAr
Figure 12: Power-voltage relationships for AC generator (r = 0.001, x = 0.005)
and equation (10) now reads

2
=
1
2
_
1 + 2
g

_
1 + 4(
g

2
g
)
_
. (49)
Figure 12 illustrates the solution for remote generation, again using the values r = 0.001 and
x = 0.005. This is simply the negative half of Figure 4: it can be reected in the vertical axis and
adjoined to the left of Figure 4, whereupon the voltage traces become closed curves. Note again
that we are using the opposite sign convention for Q as well as P, so the 30MVAr curve for a load
is equivalent to the -30MVAr curve for a generator.
The power limit P
max
for generation occurs when the quantity under the square root in (49)
vanishes:
1 + 4
g
4
2
g
= 0, (50)
that is, when
4x
2
P
2
gen
4r(1 + 2xQ
inj
)P
gen
+ 4r
2
Q
2
inj
4xQ
inj
1 = 0. (51)
The solution, P
max
, is given by
P
max
=
_
(r
2
+x
2
)(1 + 4xQ
inj
) +r(1 + 2xQ
inj
)
2x
2
. (52)
Comparing (52) with (13) and noting that Q
inj
= Q, it is seen that the only dierence is the sign of
the term r(1 +2xQ
inj
). This reects the fact that the P
gen
gure for a remote generator is inclusive
of the circuit losses, but the P gure for a load excludes losses. For a lossless circuit (r = 0), the
power transfer limit P
max
is the same for generated power as for consumed power.
(This asymmetry in the transfer limits for load and for generation can also be seen in the
graphical visualisation of Figure 6. When the circuit is lossless, the angle in the construction of
the diagram is 90 degrees, and the limit curve is symmetrical as between positive and negative P.
The apparently greater capability in the negative P region in Figure 6, which depicts a relatively
lossy circuit, reects the fact that at points of large P and/or large Q, much of the generated power
is consumed in circuit losses.)
The critical voltage V
C
is obtained from (49) when the quantity under the square root is zero:
V
C
V
S
=
_
1
2
+rP
max
+xQ
inj
=

z
2
(1 + 2xQ
inj
) +rz
_
1 + 4xQ
inj
2x
2
. (53)
14
-50 -40 -30 -20 -10 0 10 20 30 40 50
0
20
40
60
80
100
120
140
160
180
Reactive power injecton Q
inj
(MVAr)
P
o
w
e
r

t
r
a
n
s
f
e
r

(
M
W
)
,

V
o
l
t
a
g
e

(
%

o
f

V
S
)
,

A
n
g
l
e

(
d
e
g
r
e
e
s
)
Power transfer limit P
max
Critical voltage
Critical power angle
Figure 13: Transfer limit, critical voltage and angle for remote generator (r = 0.001, x = 0.005)
Comparing (53) with (17), it is seen that the two formulae dier only in the sign of the second
term in the numerator. As with the formulae for P
max
, this dierence arises from the accounting for
circuit losses, and is absent in lossless circuits.
Figure 13 depicts the power transfer limit, critical voltage and critical angle as a function of
Q
inj
for the example circuit used above. (The relation (38) between
C
and
C
is unchanged for
generation: the key dierence is that V now leads V
S
in phase.) Figure 13 may be compared with
Figure 5 for an equivalent load. The power transfer limit for the generator is signicantly higher,
and again, this is due to the fact that the circuit losses are supplied from the generator, whereas in
the load case the losses are supplied from the source and not counted in the transferred power. The
critical voltage is also higher, and for large reactive power injections approaches close to V
S
itself.
This reects the fact that large reactive power injections will signicantly boost the local voltage at
a generator, and only when the stability limit is approached will this boosting eect not be seen. It
also shows, however, that there is a limit to the ability to boost the transfer capacity by injecting
reactive power at a generator: eventually it becomes impossible to sustain nominal voltage at the
generator without risking instability.
Lastly, it will be seen that the only dierence in the sensitivity of voltage to reactive power
(see Section 5) when moving from remote load to remote generation arises from the change in sign
convention:

Q
inj
=
1
2
_
_
x +
x + 2rxP
gen
2r
2
Q
inj
_
1 + 4
_

2
g
_
_
_
. (54)
Just as with a remote load, an injection of reactive power causes a rise in voltage under virtually
all circumstances, provided the network operates on the normal upper branch of the voltage curve
in Figure 12. This condition is violated only for a very large VAr injection of order 1/r, which is
generally too large to be practical.
The dierences between the network solutions for load and generation can be summed up as
follows:
1. Equation (10) gives the voltage ratio for a remote load fed through a series impedance (e.g.
Figure 4), while equation (49) gives the voltage ratio for a remote generator (e.g. Figure 12).
The only dierence is in the sign convention used for real and reactive power.
2. Equations (13) and (17) give the power transfer limit and critical voltage, respectively, for a
remote load. For a generator it is necessary only to change the sign of the term proportional
to r, and to pay attention to the sign convention used for reactive power. This results in
equations (52) and (53) respectively. For lossless circuits there is no dierence between the
transfer limit and critical voltage for loads and generators.
15
10 Eect of Long Lines
There are two more important generalisations to consider, which are dealt with in this and the
following section. All our analysis to this point has been for the simple circuit model of Figure 3,
where the transmission circuit is a simple lumped impedance, and the remote load or generator has
xed real power P, and either xed reactive power or xed power factor.
Real transmission lines and cables do not behave as lumped impedances. Cables in particular
have a relatively high capacitance per unit length and must be treated as distributed circuits if they
are more than a few kilometres long. Most overhead lines up to around 80km long can be adequately
modelled as series impedances, but longer lines must include the eect of capacitance. For medium
length overhead lines up to about 300km it usually suces to aggregate the shunt capacitance at
each end, but above this length the full distributed circuit representation must be used.
In any case, it does not greatly change any of the calculations above to use the full long-line
formulae in place of a series impedance. As the Appendix shows, for long lines there is an exact
analogue to the two-bus load ow equation (7), where r and x are replaced with coecients that
depend on the line parameters. The parameters for an exact representation of transmission lines
are:
The line length l;
The resistance R and reactance X per unit length;
The shunt susceptance B and conductance G per unit length;
The surge impedance Z
C
=
_
(R +jX)/(G+jB), split into real and imaginary components
as Z
C
= R
C
+jX
C
; and
The propagation constant =
_
(R +jX)(G+jB), split into real and imaginary components
as = +j.
From these parameters, dene three coecients as follows:
r

=
R
C
sinh2l X
C
sin2l
2V
2
S
, (55)
x

=
R
C
sin2l +X
C
sinh2l
2V
2
S
, (56)
a

=
cosh2l + cos 2l
2
. (57)
Aside from the normalisation of r

and x

by V
2
S
(as with r and x in (6)), these coecients depend
only on the line constants and not on the source or load conditions.
Having dened these coecients, let the variables
l
and
l
be dened (for a remote load)
analogously to (8) and (9):

l
= r

P +x

Q (58)

l
= x

P r

Q. (59)
Then the voltage formula is only slightly modied from (10):

2
=
1
2a

_
1 2
l

_
1 4(
l
+
2
l
)
_
. (60)
See the Appendix for a derivation of this formula. The main change in the form of (60) is the factor
1/2a

in place of 1/2: this is the Ferranti eect whereby lightly-loaded lines and cables can have a
higher voltage at the remote end than at the source, due to the voltage boosting eect of the line
capacitance.
The condition for a short line is that l 1 and l 1. Under this condition it can be shown
that r

and x

become approximately equal to the equivalent lumped impedances r = (Rl)/V


2
S
and x = (Xl)/V
2
S
, and that a

is very close to 1. Equation (60) then reduces to equation (10), as


expected.
Figure 14 illustrates the long-line eect. It shows the power-voltage relationships for a load at
unity power factor, where the line length is varied but the conductor impedance varied inversely
with length, such that the total line impedance in the absence of long-line eects is always equal to
16
0 10 20 30 40 50 60 70 80 90 100
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
1.2
Real power P (MW)
V
o
l
t
a
g
e

r
a
t
i
o

10 km
1000 km
2000 km
3000 km
4000 km
5000 km
Figure 14: Power-voltage relationships with a long AC line
that used in Figure 4. In this way, the eect of the long-line behaviour is isolated from that due
to multiplication of impedance as circuit length increases. The capacitance is as appropriate for
an overhead line, giving a wavelength in the order of several thousand kilometres: accordingly, the
long-line eect only becomes signicant at 1000km or greater in this example. Below this length
the short-line approximation holds, and the curve is then essentially identical to the Q = 0 curve in
Figure 4.
One important conclusion from Figure 14 is that the short-line approximation is conservative:
even where it is not strictly valid, the transfer limit given by the short-line approximation is less
than the actual transfer limit. However, the use of the approximation may still underestimate the
critical voltage.
For a remote generator, one uses the formulae of Section 9, with r

and x

in place of r and x:

lg
= r

P
gen
+x

Q
inj
, (61)

lg
= x

P
gen
r

Q
inj
, (62)

2
=
1
2a

_
1 + 2
lg

_
1 + 4(
lg

2
lg
)
_
. (63)
With either remote load or remote generation, or for operation at constant power factor, the
transfer limit formulae are identical to those in Section 3, Section 9 or Section 8 respectively, with
r

and x

in place of r and x. The critical voltage formulae, and the sensitivity of voltage to reactive
power, are also the same with one proviso: the factor a

now multiplies the formulae. This means


the critical voltage V
C
is scaled by the factor 1/

, and the sensitivity d/dQ by the factor 1/a

.
Since a

is typically less than 1, this means that V


C
can be slightly higher, and the sensitivity to Q
slightly greater, for long lines than for short lines of equivalent series impedance.
With long lines, however, one loses the neat phasor diagram of Figure 7 and with it the convenient
link between the power angle and the and variables given by (35). Similarly, the relation
(38) between the critical voltage and critical angle no longer holds, except approximately. It is,
nonetheless possible to dene a dierent angle

by analogy with (35):


cos

= +

l

, sin

=

l

, (64)
and this angle

does indeed attain a critical value

C
satisfying the relation

C
cos

C
=
1
2
. (65)
The angle

does bear an approximate relation to the true angle between V and V


S
, but there is
no longer an exact correspondence with the other stability variables.
17
11 Eect of Shunt Admittance
A similar generalisation to that in the previous section covers the case where there is shunt impedance
connected at the remote load or generation site: for example, where reactive compensation takes
the form of a passive capacitor bank rather than a constant-Q device. A capacitor bank injects
less reactive power under low voltage conditions than under nominal voltage conditions, and so will
result in a steady state voltage dierent from that indicated for a constant Q source in equation (10)
and Figure 4. This in turn will aect the power transfer limit.
Assume than that in addition to the load with xed real power P and reactive power Q, there
are shunts connected with shunt conductance G and susceptance B (either of which may be zero).
Dene the nominal shunt power as the real and reactive power that would be drawn by the shunt
impedances if the voltage V were equal to the source voltage V
S
, that is,
P
sh
= GV
2
S
, Q
sh
= BV
2
S
. (66)
(Notice the minus sign for Q
sh
: capacitors for example have positive susceptance B but negative
VAr absorption by the load sign convention.)
Then to take account of the shunts, the load ow equation (7) with normalised circuit impedances
must be expanded to the following:
a
sh

4
+ (2(r
sh
P +x
sh
Q) 1)
2
+ (r
2
+x
2
)(P
2
+Q
2
) = 0 (67)
where
a
sh
= 1 + 2(rP
sh
+xQ
sh
) + (r
2
+x
2
)(P
2
sh
+Q
2
sh
), (68)
r
sh
= r + (r
2
+x
2
)P
sh
, (69)
x
sh
= x + (r
2
+x
2
)Q
sh
. (70)
(See the Appendix for details.) The solution for again has almost the same form as previously:
the solution variables and are now dened as

sh
= r
sh
P +x
sh
Q (71)

sh
= x
sh
P r
sh
Q (72)
and = V/V
S
is found from the formula

2
=
1
2a
sh
_
1 2
sh

_
1 4(
sh
+
2
sh
)
_
. (73)
The important thing to note here is that the P and Q occurring in these formulae do not include the
shunt components. The shunt power instead enters through the denition of a
sh
, r
sh
and x
sh
. Notice
that the eect of a shunt is very similar (aside from a constant scale factor) to that of increasing the
line resistance parameter by an amount (r
2
+x
2
)P
sh
, and the line reactance by (r
2
+x
2
)Q
sh
. (The
eective reactance is reduced when the shunt is capacitive.)
Figure 15 illustrates the eect of compensating with a capacitor bank, and how this diers from
a constant VAr source. The solid blue line shows the uncompensated case, which is identical to that
in Figure 4. The blue lines show the eect of adding 30MVAr of constant Q (dashed) or capacitors
(solid). (The dashed blue line is also identical to the 30MVAr curve in Figure 4.) Notice that the
solid and dashed lines intersect at the point where V = V
S
( = 1), since this is the point where the
two forms of compensation provide the same actual reactive power. Also shown as a green dashed
line is the PV curve when no compensation is used but the line reactance is set equal to x
sh
; this
gives the same behaviour as the solid red line apart from a scaling by the factor 1/a
sh
. Accordingly,
it will be noticed that the capacitive compensation scenario (red solid line) and the reduced line
impedance scenario (dashed green line) give the same power transfer limit. This limit is lower than
for the constant Q compensation scenario (red dashed line).
Power-angle analysis is less helpful in the presence of shunt admittances than with pure constant-
power load. By substituting the voltage-dependent formulae for load power in (30) and (33), one
can show that
cos = (1 +rP
sh
+xQ
sh
) +
1

(rP +xQ), sin = (xP


sh
rQ
sh
) +
1

(xP rQ), (74)


but these expressions are dicult to relate to
sh
and
sh
. The critical-angle relation (38) is not
valid in general for voltage-dependent load.
18
0 10 20 30 40 50 60 70 80 90 100
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
1.2
Real power P (MW)
V
o
l
t
a
g
e

r
a
t
i
o

No VArs
-30MVAr: constant Q
-30MVAr: shunt cap
No VArs, equiv X
Figure 15: Power-voltage relationships: capacitor bank versus constant-Q source
In certain situations, one must take into account both long-line eects and constant-impedance
load. The solution is then found by combining elements of the above formulae with those in Section
10. As the Appendix shows, the parameters analogous to (68) through (70) are
a

sh
= a

+ 2(r

P
sh
+x

Q
sh
) +
1
a

_
(r

)
2
+ (x

)
2
_
(P
2
sh
+Q
2
sh
), (75)
r

sh
= r

+
1
a

_
(r

)
2
+ (x

)
2
_
P
sh
, (76)
x

sh
= x

+
1
a

_
(r

)
2
+ (x

)
2
_
Q
sh
, (77)
where r

, x

and a

are as dened above for long lines. The resemblance to equations (68) through
(70) is plain: one just uses the long-line equivalent parameters r

and x

in place of r and x, and


inserts a factor a

or 1/a

where indicated to account for the Ferranti eect.


With these denitions the solution formulae for = V/V
S
are again analogous to those for the
plain series impedance. The intermediate solution variables are

sh
= r

sh
P +x

sh
Q (78)

sh
= x

sh
P r

sh
Q (79)
and the solution for is

2
=
1
2a

sh
_
1 2

sh

_
1 4 (

sh
+ (

sh
)
2
)
_
. (80)
With remote generation, the reverse sign convention gives

shg
= r

sh
P
gen
+x

sh
Q
inj
(81)

shg
= x

sh
P
gen
r

sh
Q
inj
(82)
with the solution formula

2
=
1
2a

sh
_
1 + 2

shg

_
1 + 4
_

shg
(

shg
)
2
_
_
. (83)
12 Shunt Susceptance, Limits and Voltage Sensitivity
The equations of the previous section show that with shunts, as with long lines, there is no need to
radically revise the formulae for power transfer limit, critical voltage or sensitivity to reactive power,
19
-50 -40 -30 -20 -10 0 10 20 30 40 50
0
20
40
60
80
100
120
Reactive power Q absorbed at load (MVAr)
P
o
w
e
r

t
r
a
n
s
f
e
r

(
M
W
)
,

V
o
l
t
a
g
e

(
%

o
f

V
S
)
Transfer limit, constant Q
Transfer limit, shunt ind/cap
Crit. voltage, constant Q
Crit. voltage, shunt ind/cap
Figure 16: Transfer limit and critical voltage for passive versus constant-Q compensation
provided the shunt powers P
sh
and Q
sh
are considered to be excluded from the transferred power
and from the reactive power variable in the sensitivity analysis. In this case, one can simply bundle
the shunts into the equivalent impedance parameters r
sh
and x
sh
(or r

sh
and x

sh
) in the formulae
and scale by a
sh
(or a

sh
) where appropriate, just as for long lines.
There are, on the other hand, cases where the shunt powers P
sh
and (in particular) Q
sh
must
be treated as variables in their own right and not bundled in with other quantities. The most
important such case is where switched capacitor banks or shunt reactors are used to provide reactive
compensation. This is particularly common where large amounts of reactive power are required,
either to boost (capacitively) the transfer capability of high-impedance lines, or to compensate
(inductively) charging currents on long lines or cables. It can then be important to analyse the
sensitivity of transfer limits or voltage to the shunt reactive power Q
sh
, rather than to the intrinsic
reactive power Q of the load or generator. (Once again, Q
sh
is really just an alias for the amount of
capacitance connected: the amount of actual reactive power from a given capacitance will vary.)
In this situation, the transfer limit with a load is the usual equation (13) but with x
sh
in place
of x:
P
max
=
_
(r
2
+x
2
sh
)(1 4x
sh
Q) r(1 2x
sh
Q)
2x
2
sh
. (84)
Here P and Q are the xed real and reactive power associated with the load, and we assume
P
sh
= 0. If for simplicity it is assumed the load itself operates at unity power factor, then Q = 0
and substituting formula (70) for x
sh
gives (after some algebra)
P
max
=
_
(r
2
+x
2
)(1 + 2xQ
sh
+ (r
2
+x
2
)Q
2
sh
) r
2 (x
2
+ 2x(r
2
+x
2
)Q
sh
+ (r
2
+x
2
)
2
Q
2
sh
)
. (85)
Figure 16 depicts the limit given by (85) in comparison with the limit for constant-Q compensation
depicted in Figure 5. There is a clear dierence: in general a given amount of passive inductive or
capacitive VArs changes the transfer limit by less than the same amount of VArs from a constant-Q
source. This is not clear from equation (85), but if one assumes xQ
sh
to be small one can apply the
approximations
_
1 + +O(
2
) 1 +

2
,
1
1 + +O(
2
)
1 , 1 (86)
and after some algebra, there results the following approximate form of (85) (with z =

r
2
+x
2
):
P
max

z r
2x
2
_
1
_
1
r
z r
+
2r
2
x
2
_
xQ
sh
_
. (87)
20
Comparing this with the corresponding approximate formula (14), the dierence in sensitivity to
Q
sh
is apparent. In particular, when r/x is small, a small amount of reactive power Q
sh
causes P
max
to change by about half as much as an equivalent small amount of xed reactive power Q.
Figure 16 also shows the variation of critical voltage with reactive power. Surprisingly, it is
found that even though P
max
is signicantly aected by the change from constant-Q to passive
shunt compensation, the critical voltage in this example is virtually unchanged except for large
capacitive values. To see why, note that the formula for critical voltage with a shunt is
V
C
V
S
=
_
1 2r
sh
P
max
2x
sh
Q
2a
sh
. (88)
In this example we assume that r
sh
= r and Q = 0. Now notice that equation (85) for P
max
can be
written in terms of a
sh
and z as
P
max
=
z

a
sh
r
2(z
2
a
sh
r
2
)
. (89)
Substituting this into (88) now gives
V
C
V
S
=

1
2a
sh
_
1 r
z

a
sh
r
z
2
a
sh
r
2
_
=

1
2a
sh

z
2
a
sh
rz

a
sh
z
2
a
sh
r
2
=

1
2a
sh

z
2
z(r/

a
sh
)
z
2
(r/

a
sh
)
2
.
So far, all the working is exact. We now approximate, setting r/

a
sh
r, giving
V
C
V
S

1
2a
sh

z
2
zr
z
2
r
2
=

z(z r)
2a
sh
x
2

_
z(z r)
2x
2
(1 2xQ
sh
). (90)
Comparing (90) with the approximation (18) for critical voltage, the similarity is evident. The
approximation works best for low-resistance circuits, since if r is much smaller than z, the quantities
(z r) and (z r/

a
sh
) will be approximately the same over many values of a
sh
.
The above formulae for P
max
and V
C
are, again, not greatly changed if one places a generator
instead of a load at the end of the circuit. As already discussed in Section 9, certain terms will
change sign, either through the use of an opposing sign convention or as a result of the generator
now having to supply the circuit losses. Again, one may use the formulae in Section 9 and substitute
r
sh
and x
sh
for all occurrences of r and x.
The analysis of sensitivity to reactive power requires some attention, as the behaviour with
respect to Q
sh
looks very dierent to that with respect to Q. Looking again at Figure 15 for a load,
however, reveals nothing unexpected: as long as the circuit operates on the upper half of the PV
curve, the injection of capacitive VArs always raises the voltage.
One can conrm this analytically by dierentiating equation (73) with respect to Q
sh
. The
algebra is somewhat involved but eventually results in:

Q
sh
=
1
2

Q
sh
_

2
_
=
z
2
2a
sh

_
Q+
Q(1 2r
sh
P) + 2x
sh
P
2
_
1 4(
sh
+
2
sh
)
_


a
sh
_
x +z
2
Q
sh
_
. (91)
Consider rst the case of a unity-power-factor load (Q = 0). Provided Q
sh
is not too large (that
is, smaller than V
2
S
/X), it is not hard to see that the sign of /Q
sh
is always negative and the
expected behaviour results: absorption of VArs reduces the voltage and injection of VArs raises it.
The same will be true if the load is inductive (Q > 0): the quantity in the large parentheses
remains positive and the overall sensitivity is negative. On the other hand, if the load is capacitive
(Q < 0) it is theoretically possible for the quantity in parentheses to change sign, even with Q
substantially less than V
2
S
/X. Nonetheless, in most practical cases the second term, on the far right
of (91), dominates the rst term and keeps the overall sensitivity negative. This is because the rst
term is of order z
2
while the second term is of order x and so is generally larger by a factor of 1/z.
21
The equivalent sensitivity equation for a generator is found by reversing the sign convention for
P and Q, as follows:

Q
sh
=
z
2
2a
sh

_
_
Q
inj
+
Q
inj
(1 + 2r
sh
P
gen
) 2x
sh
P
2
gen
_
1 + 4(
shg

2
shg
)
_
_


a
sh
_
x +z
2
Q
sh
_
. (92)
Note that although we adhere to the generator sign convention for P
gen
and Q
inj
, we are using the
same load sign convention for Q
sh
as before. This is a common convention for shunts, whether
they appear in conjunction with loads or generators, and it is important not to be confused by this.
Accordingly, we expect that for a stable system /Q
sh
will be negative, as for a load.
The change in sign convention does not greatly aect the overall sensitivity result. If Q
inj
is zero,
then (92) is easily seen to be negative as long as Q
sh
is not impractically large (of order V
2
S
/X).
The same is true if Q
inj
is negative (the generator absorbs VArs). Only if Q
inj
is positive (generator
injects VArs) is it possible for the rst term in (92) to change sign and become positive. Yet in most
practical situations, the second term dominates and keeps the overall sensitivity negative.
13 Summary and Conclusion
Steady-state voltage stability refers to the constraint imposed on power ow in electrical circuits
by the nonlinear load-ow equations. Given any circuit conditions, there is a limit on the amount of
real power that can ow from one point to another in a circuit. In DC circuits this limit is given by
the maximum power transfer rule, while in AC circuits the analysis is more complicated and the
limit also depends on the reactive power ow. As this limit is approached, the likelihood of voltage
collapse increases due to the convergence of the dual solutions for voltage and the high sensitivity
of voltage to power ow.
Although the equations for general AC circuits have no useful analytical solution, usable formulae
are available for the case of power ow between two buses, given the voltage magnitude at one bus
(the source) and the power P and Q at the other (the load). Studies of larger power systems
often focus on one critical transmission link which can be studied in this manner.
The exact formula for voltage at the load has the same general form whether the circuit is mod-
elled as a simple series impedance or as a long transmission line. One rst denes two intermediate
variables and as
= rP +xQ
= xP rQ
where P and Q are the xed power components of the load, and r and x are impedance parameters
for the circuit. In terms of and the voltage formula is
_
V
V
S
_
2
=
1
2a
_
1 2
_
1 4 ( +
2
)
_
where V
S
is the source voltage and a is a parameter dependent on the circuit. Because the formula
is derived from a quadratic equation, there are two branches to the voltage solution corresponding
to the sign. The positive branch is held to be the normal or stable operating regime, while
the negative branch is held to be unstable as it results in abnormally low voltages.
If the load is actually a generator, an alternative sign convention is generally used for the power:
instead of the power consumption P and Q one has the generated power P
gen
and injected reactive
power Q
inj
which are sensed positive in the opposite direction, back toward the source. The above
equations are then modied to

g
= rP
gen
+xQ
inj

g
= xP
gen
rQ
inj
_
V
V
S
_
2
=
1
2a
_
1 + 2
g

_
1 + 4
_

2
g
_
_
.
The denition of the circuit parameters r, x and a depends on the nature of the circuit, and of
any shunts connected at the remote end. If the circuit can be modelled as a plain series impedance,
then a = 1 and r and x are just the impedance normalised by the square of the source voltage:
22
r = R/V
2
S
, x = X/V
2
S
. If the circuit must be modelled as a long line, then equations (55), (56) and
(57) are used for r, x and a. These impedance parameters are also normalised by V
2
S
, but otherwise
depend only on the characteristics of the line.
If there are shunt admittances connected at the remote end, so that changes in voltage aect
the power at the remote end, this is reected through the parameters r, x and a as indicated in
equations (68), (69) and (70). The eect of a shunt at the remote end is much the same as that of
increasing the line resistance and reactance by (r
2
+x
2
)P
sh
and (r
2
+x
2
)Q
sh
respectively, where P
sh
and Q
sh
are the powers consumed by the shunt at V = V
S
. (Note that if the shunt is capacitive, the
eect is to reduce the eective line reactance, in accordance with intuition.)
The steady-state stability margin is related to the size of the quantity under the square root in
the voltage equation. As the real power transfer P increases, this quantity falls to zero and the load
ow equation ceases to have a solution: one says that voltage collapse has occurred. Depending on
the sign convention used (load or generation), the equation for the stability limit is = 1/4
2
or

g
=
2
g
1/4, and has a convenient graphical visualisation.
The form of the limit equation is the same for long lines as for short lines, and also in the presence
of shunts, provided the appropriate impedance parameters r and x are used. However, the eect on
the limit of a xed reactive power Q is quite dierent to that of a nominal reactive power Q
sh
from a
passive shunt, because of the way the shunt power varies with voltage. This dierence is illustrated
in Figure 16.
The critical voltage is the voltage at the point where the solution collapses. It is therefore found
simply by omitting the square-root term in the voltage equation, since it corresponds to the point
where this term vanishes. Generally, the injection of VArs will increase both the transfer limit and
the critical voltage.
One may also relate the analysis based on and to the more traditional type of analysis based
on power angle . It is thereby possible to extend power-angle analysis to circuits with signicant
resistance; however, the analysis based on is only approximate in the case of long lines or when
there is constant-admittance load. When load or generator power is constant, it is found that the
critical angle
C
at which voltage collapse occurs is intimately related to the critical voltage ratio

C
by the relation
C
cos
C
= 1/2, independent of other circuit parameters or conditions.
Sometimes an alternative stability criterion is spoken of, based on the sensitivity of voltage to
reactive power. By this criterion, the system is voltage stable whenever the injection of VArs raises
the voltage and absorption of VArs lowers the voltage, and voltage unstable when the reverse is
the case. In almost all practical cases, this voltage stability criterion is equivalent to operation in
the positive branch of voltage: if the system is stable by one criterion it will be stable by the
other. For this rule to be violated requires that either P or Q be very large, similar in magnitude
to V
2
S
/Z where Z is the circuit impedance. Other considerations generally preclude operating the
circuit at such high power for a given circuit impedance.
Appendix: Derivation of the Two-Bus Load Flow Solution
Plain Series Impedance
In this section an equation is derived for the voltage magnitude V in Figure 3, given the source
voltage magnitude V
S
, the power P and Q absorbed at the load, and the series impedance R+jX.
In the following, a tilde (e.g.

V ) denotes a complex phasor quantity, while the same symbol
without a tilde (e.g. V ) denotes the magnitude of the phasor quantity. Complex conjugation is
indicated with an asterisk, thus

Z = R +jX,

Z

= R jX.
Our starting point is Ohms Law, whereby the complex voltage dierence across the series
impedance equals the impedance times the current

I:

V
S


V =

Z

I. (93)
Now multiply this through by the conjugate voltage

V

to obtain

V
S

V

V
2
=

Z

I

V

. (94)
The complex power

S is dened as

V

I

, and its conjugate is



S

=

V

I. So the above can be


rewritten as

V
S

V

= V
2
+

Z

S

. (95)
23
We can now multiply this whole equation by its own conjugate, to obtain
V
2
S
V
2
=
_

V
S

V

__

V
S

V

=
_
V
2
+

Z

S

__
V
2
+

Z


S
_
= V
4
+
_

Z

S

+

Z


S
_
V
2
+Z
2
S
2
= V
4
+ 2Re
_


S
_
V
2
+Z
2
S
2
. (96)
Finally, taking Z = R +jX and S = P +jQ gives


S = (R jX)(P +jQ) = (RP +XQ) +j(RQXP) (97)
Z
2
S
2
= (R
2
+X
2
)(P
2
+Q
2
) (98)
and substitution in (96) gives
V
4
+
_
2(RP +XQ) V
2
S
_
V
2
+ (R
2
+X
2
)(P
2
+Q
2
) = 0. (99)
This immediately leads to the load ow equation (7) given in the text, upon dividing through by
V
4
S
and setting r = R/V
2
S
, x = X/V
2
S
.
Series Impedance with Shunt Loads
If the circuit includes a shunt admittance

Y = G + jB in addition to the constant-power element
P +jQ, then the complex power in the above derivation must be modied to
S = P +jQ+

Y

V
2
= (P +GV
2
) +j(QBV
2
). (100)
With this change, the derivation goes through as above. We now have


S = (R jX)
_
P +GV
2
+j(QBV
2
)
_
= RP +XQ+ (RGXB)V
2
+j
_
RQXP (RB +XG)V
2
_
(101)
Z
2
S
2
= (R
2
+X
2
)(P
2
+Q
2
+ 2(GP BQ)V
2
+ (G
2
+B
2
)V
4
) (102)
and substituting into (96) gives, after rearranging
_
1 + 2(RGXB) + (R
2
+X
2
)(G
2
+B
2
)
_
V
4
+
_
2(RP +XQ) + 2(R
2
+X
2
)(GP BQ) V
2
S
_
V
2
+ (R
2
+X
2
)(P
2
+Q
2
) = 0. (103)
Now we normalise by V
2
S
, setting
r =
R
V
2
S
, x =
X
V
2
S
, P
sh
= GV
2
S
, Q
sh
= BV
2
S
, (104)
and divide through by V
4
S
, resulting in
_
1 + 2(rP
sh
+xQ
sh
) + (r
2
+x
2
)(P
2
sh
+Q
2
sh
)
_

4
+
_
2(rP +xQ) + 2(r
2
+x
2
)(P
sh
P +Q
sh
Q) 1
_

2
+ (r
2
+x
2
)(P
2
+Q
2
) = 0. (105)
Finally, collecting the coecients of P and Q in the
2
term and dening r
sh
, x
sh
and a
sh
as in
Section 11 results in equation (67) as given in the text.
Long Lines
The derivation for a long line follows a similar process: only the algebra is slightly more involved.
Start with the voltage equation for a long transmission line [1, 3]:

V
S
=

V cosh l +

Z
C

I sinh l. (106)
For brevity, we will write this as

V
S
=

W

V +

Z

I,

W = cosh l,

Z =

Z
C
sinh l. (107)
24
As before, multiply through by

V

V
S

V

=

WV
2
+

Z

I

V

=

WV
2
+

Z

S

(108)
and multiply this equation by its conjugate:
V
2
S
V
2
=
_

WV
2
+

Z

S

__

V
2
+

Z


S
_
= W
2
V
4
+ 2Re
_

W

Z


S
_
V
2
+Z
2
S
2
. (109)
Before we substitute the actual expressions for

W and

Z, we will work through to the actual solution
to show how the general form is preserved. First we normalise Z, setting
z =
Z
V
2
S
. (110)
Then dividing (109) through by V
4
S
and rearranging, we get
W
2

4
+
_
2Re
_

W z


S
_
1
_

2
+z
2
S
2
= 0. (111)
For the solution of this quadratic equation, resolve the quantity

W z


S into its real and imaginary
components as

W z


S = j. (112)
Then the solution is

2
=
1
2W
2
_
1 2 +
_
(1 2)
2
4W
2
z
2
S
2
_
=
1
2W
2
_
1 2 +
_
1 4 + 4
2
4(
2
+
2
)
_
=
1
2W
2
_
1 2 +
_
1 4( +
2
)
_
. (113)
We see that the solution has the form (60) if we dene a

= W
2
.
It remains only to substitute the actual line parameters. For a

, we have
a

= W
2
= |cosh l|
= sinh
2
l + cos
2
l =
cosh2l + cos 2l
2
. (114)
For the remaining parameters, rst calculate

W z

=

Z

C
V
2
S
sinh

l cosh l
=
1
2V
2
S
(R
C
jX
C
)(sinh2l j sin2l)
=
R
C
sinh2l X
C
sin2l
2V
2
S
j
R
C
sin2l +X
C
sinh2l
2V
2
S
(115)
= r

jx

and nally multiply by



S:
j =

W z


S = (r

jx

)(P +jQ) = r

P +x

Qj(x

P r

Q). (116)
We see that this matches the denitions of a

, r

, x

,
l
and
l
in the text.
Long Lines with Shunt Loads
To account for shunts, we repeat the long-line derivation, but replacing

S with

S +

Y

V
2
. Thus
(108) becomes

V
S

V

=

WV
2
+

Z

S

+

Z

Y V
2
(117)
25
and (109) in its turn becomes
V
2
S
V
2
=
__

W +

Z

Y
_
V
2
+

Z

S

___

+

Z

_
V
2
+

Z


S
_
=
_
W
2
+ 2Re
_

W

Z

_
+Z
2
Y
2
_
V
4
+ 2Re
__

W +

Z

Y
_

Z


S
_
V
2
+Z
2
S
2
. (118)
As before, we normalise

Z using (110), and we also normalise

Y along the lines of (104), setting
y =

Y V
2
S
= GV
2
S
+jBV
2
S
= P
sh
jQ
sh
. (119)
After rearranging and dividing by V
4
S
, equation (118) now reads
_
W
2
+ 2Re
_

W z

_
+z
2
y
2
_

4
+
_
2Re
__

W + z y
_
z


S
_
1
_

2
+z
2
S
2
= 0. (120)
Now, if one denes
_

W + z y
_
z


S = j (121)
a

sh
= W
2
+ 2Re
_

W z

_
+z
2
y
2
(122)
then it follows by multiplying (121) by its conjugate that

2
+
2
=
_
W
2
+ 2Re
_

W z

_
+z
2
y
2
_
z
2
S
2
= a

sh
z
2
S
2
(123)
and hence the solution of this (real) quadratic equation again agrees with the formula in the text:

2
=
1
2a

sh
_
1 2 +
_
1 4 + 4
2
4a

sh
z
2
S
2
_
=
1
2a

sh
_
1 2 +
_
1 4 + 4
2
4(
2
+
2
)
_
=
1
2a

sh
_
1 2 +
_
1 4( +
2
)
_
. (124)
Finally one substitutes the line and load parameters, noting that

W and z are the same as in
the long-line derivation above. Hence
W
2
= a

W z

= r

jx

z
2
=
W
2
z
2
W
2
=
(r

)
2
+ (x

)
2
a

W z

= (r

jx

)(P
sh
+jQ
sh
) = r

P
sh
+x

Q
sh
j(x

P
sh
r

Q
sh
)

W z


S = r

P +x

Qj(x

P r

Q)
z y z


S = z
2
y

S =
(r

)
2
+ (x

)
2
a

(P
sh
P +Q
sh
Qj(Q
sh
P P
sh
Q))
and so
a

sh
= a

+ 2 (r

P
sh
+x

Q
sh
) +
1
a

_
(r

)
2
+ (x

)
2
_ _
P
2
sh
+Q
2
sh
_
(125)
= r

P +x

Q+
1
a

_
(r

)
2
+ (x

)
2
_
(P
sh
P +Q
sh
Q) (126)
= x

P r

Q+
1
a

_
(r

)
2
+ (x

)
2
_
(Q
sh
P P
sh
Q) (127)
which again are seen to match the formulae in the text.
References
[1] Turan Gonen. Electric Power Transmission System Engineering: Analysis and Design. Wiley,
1988.
[2] P. Kundur. Power System Stability and Control. McGraw-Hill, 1994.
[3] W.D. Stevenson. Elements of Power System Analysis. McGrawHill, fourth edition, 1982.
26

Das könnte Ihnen auch gefallen