You are on page 1of 9

ELEOTRIC

POI,tJER $

ELSEVIER Electric Power Systems Research 32 (1995) 35-43

A fast method for determining the voltage stability limit of


a power system
M.H. Haque
Department o[" Electrical and Computer Systems Engineering, Monash University', Clayton, Vic. 3168, Australia
Accepted 19 August 1994

Abstract

The voltage stability problem of a power system is associated with a rapid voltage drop due to heavy system load, and it occurs
because of inadequate reactive power support at some critical buses. One of the serious consequences of the voltage stability
problem is a system blackout, and this problem has received much attention in recent years. This paper proposes a fast method
for finding the maximum load, especially the reactive power demand, at a particular load bus before reaching the voltage stability
limit. The method uses the base-case system information to find special two-bus equivalents of the system for analyzing the voltage
stability problem. The method was tested on the IEEE 14-, 30-, and 118-bus systems and the results obtained were compared with
those found by some other methods.

Keywords: Voltage stability; Network reduction; Power system security

1. Introduction a certain degree of voltage security. When the voltage


of a system starts to decrease, the current, and hence
Transmission lines in a power system are loaded the reactive power loss in transmission lines and trans-
more heavily than ever before to avoid the capital cost formers, is increased. On the other hand, a decrease in
of building new lines. For a short line, the loading voltage reduces the reactive power supply by the line
capability may be restricted by the thermal limit. charging and shunt capacitors. Thus the voltage reduc-
However, for a long line, the loading capability may tion has a cumulative effect unless ample reactive power
be dictated by the voltage stability rather than the sources or some appropriate controls are available to
thermal or transient stability limit. When a power sys- regulate the voltage and maintain the reactive power
tem approaches the voltage stability limit, the voltage balance.
of some buses reduces rapidly for small increments in Even though the symptom of imminent voltage col-
load and the controls or operators may not be able to lapse is rapid voltage reduction, the voltage magnitude
prevent the voltage decay. In some cases, the response itself is not a good indicator of the proximity of voltage
of controls or operators may aggravate the situation collapse [3]. Several approaches for analyzing the
and the ultimate result is voltage collapse. Voltage voltage instability problem have been reported in the
collapse has become an increasing threat to power literature. Venikov et al. [4] estimated the voltage stabil-
system security and reliability. Many incidents of sys- ity limit from the convergence of the N e w t o n - R a p h s o n
tem blackouts because of voltage stability problems (NR) load flow calculation. However, in the vicinity of
have been reported worldwide (see Refs. [1,2] for a the voltage instability point, the divergence of the NR
sample of the literature). Nowadays, a proper analysis method may be caused either by numerical problems or
of the voltage stability problem has become one of the by the fact that the instability condition has already been
major concerns in power system operation and plan- reached. The method is very time consuming but it gives
ning studies. reasonably good results. The minimum singular value of
The main reason for voltage instability in a power the Jacobian matrix of the N R load flow equations has
system is inadequate reactive power support at some also been used by some researchers to determine the
critical buses. Voltage instability is a reactive power voltage stability index [5,6].
problem. Unlike active power, it is very difficult to The nonlinear load flow equations have many solu-
estimate the reactive power margin required to achieve tions. One of the solutions is called the stable or

0378-7796/95/$09.50 ~ 1995 Elsevier Science S.A. All rights reserved


SSDI 0378-7796(94) 00893 -9
36 M.H. Haque /Electric Power Systems Research 32 (1995) 35-43

high-voltage solution. The rest of the solutions are R jx V~LS~


called the unstable or low-voltage solutions. Some re-
2
searchers have used multiple load flow solutions to
determine the voltage stability limit [7,8]. A formal
mathematical way of analyzing the voltage stability
S=P +jQ
problem using multiple load flow solutions is based on
bifurcation phenomena [9]. Refs. [10,11] used energy Fig. I. A simple two-bus system.
methods to assess the voltage stability. Energy methods
require both the high- and low-voltage solutions of the
transmission line having an impedance of Z = R + j X
load flow equations. The voltage stability problem can
to a load center at bus 2. Bus 1 is considered as a swing
also be analyzed by the Q - V and P - V curves [3]. The
bus where both the voltage magnitude V, and angle 6,
relationship between the energy based voltage security
are kept constant. For a given value of V,, the relation-
and these curves is demonstrated in Ref. [12].
ship between the load voltage magnitude V2 and the
Chebbo et al. [13] used the concept of the two-bus
load power S = P + j Q can readily be written as
theory to estimate the maximum loading capability of a
p z + Q2
particular load bus in a power system. In this method
V12 = II22 + 2 ( R P + X Q ) + (R 2 + X 2) - - (1)
the power system is first replaced by the Th6venin V2
theory to get the two-bus equivalent model. To include
By assuming x = V22, the above equation can be writ-
the effects of nonlinearity of loads and generators in the
ten in a quadratic form as follows:
equivalent circuit, some repetitive computation in the
original system is required to get the ultimate results. In alx 2 + b , x + Cl = 0 (2)
each step, the method requires the solution of the load
where
flow equations and computation of the Z matrix for the
entire system. To find the loading capability of a differ- al=l
ent load bus, the entire procedure has to be repeated.
bl = 2 ( R P + X Q ) - V, 2
This seems to be very time consuming and may not be
very useful in practice. When a power system is repre- c, = (R 2 + X2)(P 2 + Q2)
sented by such an equivalent model, the method of
The positive voltage magnitudes of bus 2 can be ob-
Refs. [14,15] can also be used to estimate the maximum
tained from the solution of Eq. (2) and are given by
loading capability.
This paper determines the maximum loading capa- V~ = ( -b, ~-dl/2) 1/2
bility of a particular load bus in a power system 2al (3a)
through the Th6venin equivalent circuit. The equivalent V~ ( - b l - d l / 2 ~ 1/2
circuits of all load buses are obtained in a single shot. = \ 2a~ / (3b)
Special care has been taken in modeling the generators
to reflect actual operation, even for a change in operat- where the discriminant d is given by
ing conditions. Unlike the other methods [12,13], the
d = b, 2 - 4 a ' c l
proposed approach can provide very good results with
less computation using the base-case system informa- = V. 4 + 4 [ 2 P Q R X - V,Z(RP + X Q ) - RZQ 2 - X 2 P 21
tion. The method has been tested on three different
(4)
IEEE standard test systems for a number of cases.
Here, V2" is called the high-voltage or stable solution
while V~ is called the low-voltage or unstable solution.
2. Problem formulation For zero load (P = Q = 0), V~ and V~ become Vl and
0, respectively. As the load (at normal power factor) is
The objective of this section is to demonstrate the increased from zero, V2n decreases while V~ increases.
concept of the voltage stability problem in a very simple This process continues until a point is reached where
two-bus system. The same concept is then applied to a both V~ and V~ become the same. This occurs when
general power system to determine the maximum load- the value of d in Eq. (4) becomes zero. The load power
ing capability of a particular load bus within the for which V~ = V~ is called the critical power and the
voltage stability limit. This requires a special two-bus corresponding voltage is called the critical voltage. It is
equivalent model of the general power system. A proce- said that the system has reached the voltage stability
dure for finding the equivalent model of a power system limit and it is not capable of transferring any additional
is described in the following sections. power. For higher load power, the real solution of Eq.
Consider a simple two-bus system as shown in Fig. (2) (and hence the magnitude of I/'2) will cease to occur
1. The generator at bus 1 transfers power through a because of the negative value of d.
M.H. Haque/ Electric Power Systems Research 32 (1995)35 43 37

1.2 Vcr= I V l 2 - 2Sm(R Cos O + X sin O)lJ/2


0.90eo,~) 2 (7)

Eqs. (6) and (7) represent the maximum load apparent


power and the corresponding voltage, respectively, for a
~e,-O.8 ~ 0 (1~9) unity /~/ given load PF angle 0. The maximum active power
loading Pm (with Q = 0) and the corresponding voltage

+ °++-2!, can be obtained from Eqs. (6) and (7), respectively, by


setting 0 = 0:
/
I I iI j ~
VIz(Z -- R)
0.4 /
i
l/
t/
/]
/ J

Pm- 2X 2 (8a)
/ / / /

i/,,/ //

,~.I ]
= (8b)
,¢¢'+
O*O I I I I i i i i i [ ~ i i i i i i i i i i i i i i i i 4 i i i i i i i i i i i I
Similarly, the maximum reactive power loading Qm at
0.0 2.0 4.0 6.0 8.0 lagging PF (with P = 0) and the corresponding voltage
Load apparent power, pu
can be written as (by setting 0 = 90 °)
Fig. 2. Variation of load voltage against the load apparent power for
various power factors. VI2(Z - X)
Qm - 2R 2 (9a)
Typical variations of load voltage against the load
Vcr = ( V'2 --~2QmX) '/z (9b)
apparent power for various power factors (PFs) are
shown in Fig. 2. The figure is plotted for V1 = 1.0 p.u.,
R = 0.01 p.u. and X = 0.1 p.u. The high-voltage or sta- It can also be noticed in Fig. 2 that the critical voltage
ble solution is represented by full curves while the for unity and lagging PF loads is significantly less than
low-voltage or unstable solution is represented by bro- the nominal value of 1.0 p.u. and it may not be accept-
ken curves. These two curves or voltages meet at the able in practice. In power system operation, it may be
critical point (marked by "< in the figure). It can be necessary to know the maximum load at which the load
observed in the figure that both the maximum load voltage should not drop below a specified value V2sp
apparent power and critical voltage increase as the load ( > Vor). The corresponding load Ssp can be obtained
PF changes from lagging to leading. The objective is to from Eq. (1) by setting V2 = Vzsp. This involves the
find the maximum load apparent power and the corre- solution of the following quadratic equation:
sponding voltage. Obviously, the condition of the max- a3Ssp 2 -}- b 3 S s p -F C3 : 0 10)
imum load apparent power (Sin) can be obtained by
setting the value of d in Eq. (4) to zero. This gives the where
following quadratic equation: a3= R 2 _]_ X2
a2 Sm 2 _L b2 Sm + C2 = 0 (5)
b 3 = 2 V2w2(R cos 0 + X sin O)
where
C3 = V2sp4-- V 12 V2sp2
a 2 = 4[RX sin(20) - R 2 sin20 - X 2 cos20]
Typical variations of load apparent power S+p against
b2 = - 4 V ~ ( R cos 0 -- X sin 0) the load PF angle for various specified load voltages are
shown in Fig. 3. It can be seen in the figure that the
C2= V14
load apparent power decreases as the lagging PF angle
In deriving Eq. (5), it is considered that P = S cos 0 is increased. For a given PF, the load apparent power
and Q = S sin 0, where 0 is the PF angle. The value of increases as the specified load voltage is decreased.
Sm can be obtained from the solution of Eq. (5):
Vt 2 Z - (R cos 0 + X sin 0)
Sin- 2 (R sin 0 + X c o s 0) 2 (6) 3. Voltage stability limit of a general power system

Here Z = (R2+ X 2) ~/2. Note that the other solution of In general, the determination of the voltage stability
Eq. (5) is not feasible because of its negative value. limit of a general power system is a very difficult task.
Once the value of Sm is known, the corresponding A quicker way to find the voltage stability limit of a
critical voltage Vcr can be obtained from either Eq. (3a) power system is through the two-bus or Th6venin
or (3b) by setting d = 0 and evaluating the coefficient bl equivalent circuit of the system. Note that the operating
at the maximum load apparent power Sin. This gives point of the generators at the verge of voltage stability
38 M.H. Haque /Electric' Power Systems Research 32 (1995) 35 43

5.0
jx,~
m

~4.0

Q)

0
~.3.0
JXgl
V=0.75 pu
m+l
~2.0 V=0.80 pu

1.0
V=0,85

V=0.90

V=0.95
pu

pu

pu
~ Es~

(a)
0.0 ''''''''ll'''l'l'llllll'l'l'lll'~ll,,,,ll,,,lllll ]
0 10 20 30 40 50
Lagging pf angle, degree
Fig. 3. Variation of load apparent power against the load PF angle
for various specified values of load voltage. jXs. - 0
ggm
m n---~
may differ significantly from the base-case operating
point. Thus, the Th6venin equivalent circuit obtained at k
the base case with a conventional generator model may
not represent a good equivalent circuit to determine the jX¢I = 0 Vc,
voltage stability limit unless some special care is taken m+l
in modeling the generators. Chebbo et al. [13] overcome
this problem first by finding an equivalent circuit at the
I
base case and predicting the maximum loading capabil-
ity of a particular load bus in the equivalent circuit. The
load of that bus is then increased to the predicted value
and a second equivalent circuit is obtained for the new (b)
operating point. The process is repeated until the load Fig. 4. Generator model in a power system.
flow solution of the original network for the predicted
load fails to converge. The method ultimately produces
an equivalent circuit that has been obtained for a load to maintain constant terminal voltages Vg as long as the
condition which is very close to the voltage stability reactive power generation of the generators is within
limit. The repetitive computations in this method seem the limits. Because of the variable nature of Eg, the
to be very time consuming. In this paper special atten- Th6venin equivalent circuit obtained for a given operat-
tion has been given in modeling the generators to avoid ing point, with the above generator model, may not be
these repetitive computations and the Th6venin equiva- used for a different operating point unless they are
lent circuits obtained at the base-case operating point very close. However, the voltage instability point is, in
may be faithfully used to determine the voltage stability general, far away from the base-case operating point.
limit. Some researchers have modeled the generators by
constant impedances with an appropriate sign. Those
3.1. Generator model impedances are again obtained from the base-case load
flow solution. A better generator model for the analysis
Consider a general power system as shown in Fig. of the voltage stability problem is proposed in the
4(a). Buses 1 to m are the generator buses where the following.
voltage magnitudes are kept constant. Buses m -4- 1 to n As mentioned earlier, the generator internal voltages
are the load buses. The aim is to find the maximum Eg are adjusted with the system load to maintain con-
loading capability of a particular load bus k within the stant terminal voltages Vg. The difference between Eg
voltage stability limit. In general, the generators in a and Vg is the voltage drop across Xg. With this in mind,
power system are modeled by internal voltage sources the generators can be considered as Eg = Vg = constant
Eg with series reactances Xg, as shown in the figure. with Xg~-0, as shown in Fig. 4(b). When reactive
When the system load or operating point changes, the power generation is changed due to the change in
generator internal voltages Eg are adjusted accordingly reactive power demand in the system, the angle of Vg
M.H. Haque /Electric Power Systems Research 32 (1995) 35 43 39

(or Eg) will not change significantly because the angle


jXg~ =--0
mainly depends on active power. Thus the generators in m N
Fig. 4(b) can be considered as ideal voltage sources
where both the voltage magnitudes and angles are more :b

or less constant and internal impedances are negligible.


This model preserves the terminal characteristics of the
jXsj = 0
generators even for a change in operating conditions. m+l
Such a model can be faithfully used to determine the
voltage stability limit due to the change in reactive
power demand as long as the reactive power generation
of the generators is within the limits. However, if it
exceeds the limits for some generator buses, those buses
can be considered as load buses by setting the value of Z~,
Q to the limiting values.
Fig. 5. Th~venin i m p e d a n c e of bus load k.
When the active power limit or the apparent power
limit (for nonzero PFs) is determined with this genera-
tor model, some erroneous results may occur. In this
case the variation of active power generation will procedure has to be repeated. This ~s a very time
change the angle of Vg. Thus, a generator can be consuming process.
considered as a source with constant terminal-voltage A very fast approach to determine the Th6venin
magnitude but variable angle. Simulation results indi- equivalent circuits of all load buses in a single shot is
cated that the error involved in this case is not signifi- proposed in the following sections. The proposed ap-
cant because the change in angle is not as high as in the proach uses the results of a single load flow solution
case of angle instability. It may again be mentioned and the system Z matrix. Both the load flow solution
here that the voltage stability problem is a reactive and the Z matrix are obtained by considering all the
power problem and the main objective of this paper is loads in the system. The voltage and impedance of the
to determine the reactive power limit. Th6venin equivalent circuit are then obtained by
slightly modifying the load flow solution and the diago-
nal elements of the Z matrix in order to nullify the
3.2. Thkvenin equivalent circuit behind a load bus effects of load impedance at the candidate bus.

The no-load or Th6venin equivalent voltage Vth of 3.2.1. Determination of the Thkvenin impedance
load bus k can be obtained from the load flow solution. Let Zkk be the kth diagonal element of the Z matrix
The load flow solution is to be determined by consider- when all loads are considered. Our aim is to find the
ing all loads in the system except at bus k. The Th6v- Th6venin impedance Zth of bus k when its load is
enin equivalent impedance Zth of bus k is the kth ignored. These two impedances (Zkk and Zth) are
diagonal element of the Z matrix. In calculating the Z shown in Fig. 5. It can be observed in the figure that
matrix, loads of all buses (except at bus k) are to be L
replaced by constant impedances and the generators are Zk Z,~ (12)
= z#.llZ,,, z#. + z,,,
to be replaced by negligible reactances. The relationship
-

between the voltage, impedance and load of bus i is Here Z~ is the load impedance of bus k Thus, the
Th6venin impedance Z,h can readily be written as
Iv'12 (ll)
P, - JQi Zth = 2~ (13)
k
Note that, at the verge of voltage stability, the con-
stant-impedance load model may represent less load 3.2.2. Determination of the Thkvenin voltage
than it should because of voltage reduction at some Let Vx. be the voltage at bus k obtained from the
buses. Thus, the Th6venin equivalent circuit with the load flow solution when all loads in the system are
above load model may provide pessimistic results at the considered. The objective is to find the Th+venin
verge of voltage stability. However, the error is not voltage V~h of bus k when its load is ignored. Fig. 6(a)
significant because the load impedances are, in general, and (b) shows the Thavenin equivalent circuits of Fig. 5
much higher than the line impedances or generator at points aa' and bb', respectively. By comparing Fig.
impedances (Xg ~ 0). Thus, the load impedances have 6(a) and (b), the value of Vth can readily be written as
very little effect on the diagonal element of the Z matrix
or the Th6venin impedance. In order to find the Th6v- Z~h\
V,h= l+~y|Vk (14)
enin equivalent circuit behind a different bus, the entire Z, /
40 M.H. Haque /Electrie Power Systems Research 32 (1995) 35-43

Z~ Z~ Table 1
a a
I I /------] M a x i m u m reactive power loading of the IEEE 14-bus system

Bus no. V~r (p.u.) Qm (MVAr) obtained Error (%)

Proposed NR
method method

a'
4 0.5156 622 615-620 1.1
(a) (b) 5 0.5144 611 600 605 1.8
Fig. 6. Th~venin equivalent circuits of load bus k. 7 0.5316 348 335 340 3.9
9 0.5421 262 255 260 2.7
10 0.5375 193 185 190 4.3
Fig. 6(b) represents the Th6venin equivalent circuit of
11 0.5416 201 195-200 3.1
bus k and the maximum loading capability of this bus 12 0.5531 180 175-180 2.9
can be determined by varying the load impedance Z~. 13 0.5476 295 290 295 1.7
When the Th6venin equivalent circuit of a general 14 0.5410 127 125 130 1.6
power network behind a particular load bus is ob-
tained, all the equations derived in Section 2 can be
used to analyze the voltage stability problem associated
system was determined through the Th6venin equiva-
with that bus.
lent circuit as described in Section 3. Values of Qm and
Vcr obtained from Eqs. (9a) and (9b), respectively, are
4. Simulation results given in Table 1. The maximum reactive power load-
ings, for these buses, obtained by the NR method are
The proposed method of determining the maximum also given in Table 1. The limit on the reactive power
reactive power loading of a load bus in a general power generation of the voltage control buses was not consid-
system has been vigorously tested on the following ered in both the proposed and the NR method. It can
three power systems: the IEEE 14-bus system, the IEEE be observed in Table 1 that the maximum reactive
30-bus system, and the IEEE 118-bus system. power loading obtained by the proposed method is
The results obtained by the proposed method were slightly higher, as expected, then the corresponding
compared with those found by the NR method. In the value found by the NR method. The maximum error
NR method, the maximum reactive power loading of a (less than 4.5%) was observed at bus 10. For this bus,
load bus is determined by gradually increasing the the maximum reactive power loading obtained by the
reactive power demand at the candidate bus in the NR method was between 185 and 190 MVAr. That is,
original unreduced system until the method fails to the method converged for 185 MVAr and it failed to
converge in solving the load flow problem. The actual converge for 190 MVAr. The corresponding value
maximum reactive power demand should not be less found by the proposed method was 193 MVAr.
than the value obtained by the NR method because of It was observed from the load flow solution of the
the numerical problems of the method in the vicinity of original system that the reactive power generation of all
the voltage instability point. As mentioned earlier, the voltage control buses exceeded the limits. This hap-
maximum loading capability of a bus estimated by the pened at Qm for all load buses. In order to consider the
proposed method is slightly higher than the correspond- limits on the reactive power generation, all voltage
ing actual value because of the constant-impedance control buses (except the swing bus) were considered as
load model. Thus, the results obtained by the proposed
and NR methods and the actual value can be ranked as
Table 2
follows: M a x i m u m reactive power loading of the IEEE 14-bus system when
Result obtained Actual Result estimated by the reactive power generation limits are considered
< <
by NR method value proposed method Bus no. Vcr (p.u.) Qm (MVAr) obtained Error (%)
The actual error involved in the proposed method is
Proposed NR
less than that obtained by comparing the results of the
method method
proposed and NR methods.
9 0.5861 135 128 130 5.5
4.1. The I E E E 14-bus system 10 0.5694 110 97 99 13.4
11 0.5676 97 86--88 12.8
12 0.5745 80 73---75 9.6
The single-line diagram and data of the IEEE 14-bus
13 0.5767 99 92 94 7.6
system are obtained from Ref. [16]. The maximum 14 0.5723 84 77-79 9.1
reactive power loading Qm of all load buses in the
M.H. Haque ~Electric Power Systems Research 32 (1995) 35 43 41

Table 3 Table 4
Maximum apparent power loading (at PF =0.8 lagging) of the Maximum reactive power loading of the IEEE 30-bus system
14-bus system
Bus no. V~ (p.u.) Qm (MVAr) obtained Error (%)
Bus no. V~, (p.u.) Sm (MVA) obtained Error (%)
Proposed NR
Proposed NR method method
method method
7 0.5139 452 445 450 1.6
9 0.5797 299 275 280 8.7 14 0.5390 127 120 125 5;.8
II 0.5494 206 200 205 3.0 21 0.5323 189 180 185 5.0
14 0.5480 131 125 130 4.8 28 0.5089 446 435 440 2.5
30 0.5251 38.7 3738 4.6

load buses after setting the value of Q for those buses


to the upper limits. The proposed and N R methods 3. The active power generation of bus 2 was considered
were then applied to the modified system to determine to be 80, 70, and 60 MW in determining the value of Sm
the value of Qm for some remote buses (9-14). The at buses 9, 11, and 14, respectively. The above MW
results obtained are summarized in Table 2. It can values were considered in order to increase the genera-
again be observed in Table 2 that the results obtained tion of all generators by almost the same factor. In this
by the proposed method are slightly higher, as usual, case, the angle of the generator bus voltage changes
than those found by the N R method. The maximum because of the change in active power generation. Thus,
error found in this case was less than 13.5% (or the Th6venin equivalent circuit obtained in Section 3
13 MVAr), which is slightly higher than the previous may give some erroneous results, as mentioned earlier.
case. Because of the absence of a voltage control bus, However, the maximum error observed in this case was
the voltage of all buses (except the swing) at Qm was not significant and in fact was less than 10%.
found to be much less than the nominal value of The actual voltage of bus 4 for various reactive
1.0 p.u. However, voltages at the base case (around power demands at that bus (up to Qm) was determined
1.0 p.u.) were used to replace the loads by constant by the N R method in the original system. The corre-
impedances in determining the Th6venin equivalent cir- sponding voltage was also estimated from Eq. (3a) after
cuit. Because of a significant reduction of voltage in the replacing the rest of the system by the Th6venin equiva-
entire system, those impedances represent lower loads lent circuit. Fig. 7 shows the variation of the actual and
at the critical points. This results in a higher estimation estimated voltages of bus 4 against the reactive power
of reactive power demand obtained by the proposed demand. It can be observed in the figure that there is
method, as mentioned earlier. hardly any difference between these two voltages. This
The maximum apparent power loading, Sin, at 0.8 validates the excellent accuracy of the generator model
lagging PF and the corresponding voltage Vcr were also used in Section 3.1 in determining the Th~venin equiva-
determined from Eqs. (6) and (7), respectively. The lent circuit.
results for some buses in this system are given in Table
4.2. The I E E E 30-bus system
1.2
The single-line diagram and data of the IEEE 30-bus
system are given in Ref. [16]. The maximum reactive
power loading, Qm, and the corresponding voltage Vc,.,
for some load buses, obtained by the proposed method
,~0.8 are given in Table 4. The maximum reactive power
found by the N R method is also given in Table 4. It is
clear from Table 4 that the values of Qm obtained by
the proposed method are very close to the correspond-
-~o.4 ing values found by the N R method, but slightly higher,
as expected. The maximum error observed in this sys-
- - Actuol
tem was less than 6%.
. . . . Estimoted

4.3. The I E E E 11S-bus system


0.0
0.0 ' ~ ' ' ' r ~ ' 2 I ' ' ' ' ' ' ' ' ' l ' ' '0l."'4' ' ' ' l ' ' ' ' ' ' ' ' ' l . O6.0 8.()
Load reactive power, pu The single-line diagram and data of the IEEE ll8-
Fig. 7. Voltage at bus 4 in the IEEE 14-bus system for various bus system are obtained from Ref. [17]. The proposed
reactive power loadings. and N R methods of determining the maximum reactive
42 M.H. Haque /Electric Power Systems Research 32 (1995) 35-43

Table 5 very close to the corresponding actual or true value,


Maximum reactive power loading of the IEEE ll8-bus system but slightly higher, as expected. The error was be-
tween 2.8% and 6.7%. On the other hand, the error
Bus no. Vcr (p.u.) Qm (MVA) obtained Error (%)
obtained by Overbye's method [12] was significantly
Proposed NR high (between - 3 0 % and 72%). It is worth mention-
method method ing here that the error in Overbye's method decreases
as the system load approaches the stability limit.
2 0.4948 495 485-490 2.1 However, this load limit is not known in advance
22 0.4911 184 175-180 5.1
30 0.4893 1128 1080 1085 4.4
and the objective is to find it from the base-case infor-
44 0.5045 180 170-175 5.9 mation.
56 0.4913 708 650 655 8.9
62 0.5046 1249 1245 1250 0.3
88 0.5014 477 470 475 1.5 5. Conclusions

A simple and fast method for analyzing the voltage


power loading were applied to some load buses in this stability problem of a general power system through a
system. A summary of the results is given in Table 5. The two-bus equivalent has been described. The generator
error observed in this system was between 0.3% and model used in this paper is very insensitive to the
8.9%. change in operating conditions. Thus the two-bus
The maximum loading capability of bus 44 in this equivalents obtained at the base-case operating point
system is very interesting. In a recent article by Overbye through the Th6venin theorem can be faithfully ap-
et al. [12], the active, reactive, and apparent power (at plied to determine the voltage stability limit. Unlike
PF = 0.894 lagging) margins of bus 44 were found to be the other methods, the Th6venin equivalent circuits of
420 MW, 114 MVAr, and 165 MVA, respectively. These all load buses are efficiently obtained in a single shot.
values were obtained by a different method for a base This requires the results of the base-case load flow
load of about 4000 MW. Note that the method requires solution and computation of the Z matrix when all
both the high- and low-voltage solutions of the load flow loads in the system are considered. A minor modifica-
equations. The corresponding actual or true values tion to the bus voltages and the diagonal elements of
reported in the paper were 237 MW, 167 MVAr, and the Z matrix is required to exclude the effects of the
183 MVA, respectively. In this paper, the values of load at the candidate bus. Determination of the vari-
Vth and Zth behind bus 44 were found to be 0.99734 p.u. ous quantities at the verge of voltage stability involves
and (0.04256+j0.13495) p.u., respectively. For this the solution of simple quadratic equations. The maxi-
Th6venin equivalent circuit, the maximum active, reac- mum demand at a load bus to ensure a minimum
tive, and apparent power loading of the bus were found specified voltage can also be determined from the solu-
to be 270 MW, 180 MVAr, and 207 MVA, respectively. tion of another quadratic equation. Because of the
Note that the original load at bus 44 is (16 +j8) MVA. constant-impedance model of the system load, the
Thus, the margins obtained by the proposed method Th6venin equivalent circuit used in this paper may
were 254 MW, 172 MVAr and 189 MVA. These results provide slightly overestimated results at the verge of
are summarized in Table 6. voltage stability.
It can be observed in Table 6 that the maximum The proposed method of determining the voltage
loading capability obtained by the proposed method is stability limit through the two-bus equivalent has been
tested on the IEEE 14-, 30-, and ll8-bus systems for a
number of cases. Simulation results reveal that the
Table 6
Maximum loading capability of bus 44 of the ll8-bus system ob-
voltage stability limit obtained by the proposed method
tained by the proposed and Overbye's methods due to the change in reactive power demand is very
close to the corresponding actual value. The method
Active Reactive Apparent may provide slightly erroneous results when the active
power power power
power demand of the system is also changed. However,
limit (MW) limit (MVAr) limit (MVA)
the error found in this case was not significant. It is
Actual value" 253 175 201 worth mentioning here that the errors found in the
Overbye's method 436 122 182.9 simulation results are higher than the actual values
(Percentage error) (72%) ( - 30%) ( - 9%) because of numerical problems with the NR method at
Proposed method 270 180 207
the verge of voltage stability. Unlike the other methods,
(Percentage error) (6.7%) (2.8%) (3.0%)
the proposed method can provide much better and
a The actual values are obtained from Ref. [12] after considering reliable results using the base-case system information
the original load of (16 +j8) MVA. with significantly less computation.
M.H. Haque /Electric" Power Systems Research 32 (1995) 35-43 43

References collapse in power systems, IEEE Circuits Devices, 8 (3) (1992)


40-45.
[1] IEEE System Dynamic Performance Subcommittee, Voltage [10] C.L. DeMarco and T.J. Overbye, An energy based security
Stability o f Power Systems: Concepts, Analytical Tools and In- measure for assessing vulnerability to voltage collapse, IEEE
dustrial Experiences, IEEE Publ. No. 90TH0358-2-PWR, IEEE, Trans. Power Syst., 5 (1990) 419 427.
New York, 1990. [11] T.J. Overbye and C.L. DeMarco, Improved techniques for
[2] A. Tiranuchit, L.M. Ewerbring and R.A. Duryea, Towards a power system voltage stability assessment using energy methods,
computationally feasible on-line voltage instability index, IEEE IEEE Trans. Power Syst., 6(1991) 1446 1452.
Trans. Power Syst., 3 (1988) 669-675. [12] T.J. Overbye, I. Dobson and C.L. DeMarco, Q v curve inter-
[3] H.K. Clark, New challenge: voltage stability, IEEE Power Eng. pretation of energy measures for voltage security, IEEE Trans.
Rev.. lO (4) (1990) 33 37. Power Syst., 9 (1994) 331 340.
[4] V.A. Venikov, V.A. Stroev, V.I. Idelchick and V.I. Tarasov, [13] A.M. Chebbo, M.R. Irving and M.J.H. Sterling, Voltage col-
Estimation of electric power system steady state stability in load lapse proximity indicator: behavior and implications, lEE Proc.
flow calculation. IEEE Trans. Power Appar. Syst., PAS-94 C, 139 (1992) 241 252.
(1975) 1034 1041. [14] M.H. Haque, Maximum power transfer capability within the
[5] P.A. Lof, T. Smed, G. Andersson and D.J. Hill, Fast calculation voltage stability limit of series and shunt compensation schemes
of a vollage stability index. IEEE Trans. Power Syst., 7 (1992) for AC transmission systems, Eh'ctr. Power Svst. Res., 24 (1992)
54 64. 227 235.
[6] A. Tiranuchit and R.J. Thomas, A posturing strategy against [15] C.S. Indulkar, B. Viswanathan and S.S. Venkata, Maximum
voltage instabilities in electric power systems, IEEE Trans. power transfer limited by voltage instability in series and shunt
Power Syst., 3 (1988) 8 7 93. compensated schemes for AC transmission systems. IEEE Trans.
[7] Y. Tamura, H. Mori and S. lwamoto, Relationship between Power Deliveo', 4 (1989) 1246 1252.
voltage instability and multiple load flow solutions in electric [16] L.L. Freris and A.M. Sasson, Investigation of the load-
power systems, IEEE Trans. Power Appar. Syst., PAS-102 flow problem, Proc. Inst. Electr. Eng., 115 (1968) 1459-1470.
(1983) I115 1123. [17] T. Athay, V.R. Sherkat, R. Podmore, S. Virmani and C. Puech,
[8] A. Yokoyama and T. Kumano, Static voltage stability index Transient energy stability analysis, Conf ,~vstem Engineering for
using multiple load-flow solutions, Electr. Eng. Jpn., 111 (3) Power: Emergency Operating State Control Secthm IV, Davos,
(1991) 69 79. Switzerland, 1980, US Department of Energy, Publ No. CONF-
[9] 1. Dobson, H. Glavitsch, C.C. Liu, Y. Tamura and K. Vu. Voltage 790904-PI.

You might also like