Sie sind auf Seite 1von 7

ON PV-PQ BUS TYPE SWITCHING LOGIC IN POWER FLOW

COMPUTATION
Jinquan Zhao,
Hohai University
Nanjing, P.R. China
jqzhao2@tom.com

Hsiao-Dong Chiang, Hua Li,


Cornell University
Ithaca, NY, U.S.A

Abstract A comprehensive investigation of bus PV-PQ


type switching logic in power flow computation considering generator reactive power limits is presented in this
paper. Beside the numerical blow-up divergence, another
kind of divergence called bus type identification divergence can be met in the power flow computation. The bus
type identification divergence denotes a phenomenon of
voltage instability, which is corresponding to the limit
induced bifurcation voltage collapse. Using several supplementary logics, even though they are seemingly reasonable, will result in finding an unstable solution. The numerical results of a practical system and IEEE 118 bus test
system are shown to describe this phenomenon. Then a
theoretical analysis is given and the Stability Constraints
of a load flow solution are proposed for the first time. At
last, the numerical results are shown that the wrong results could be obtained by using unreasonable supplementary PV-PQ switching logics in continuation power flow
computation for voltage stability assessment.

Keywords: Power Flow Computation, PV-PQ type


switching, Switching Logic, Voltage Stability, Unstable
Solution
1 INTRODUCTION
Power flow computation is an elementary task in
power system analysis. Many static or dynamic advanced applications such as transient stability analysis,
voltage stability analysis, optimal power flow and contingency security analysis are based on a basic power
flow solution. The accuracy and rapidity of power flow
computation used to be considered as an important issue
for the past forty years [1-8]. Another important issue
about power flow computation, handling of generator
reactive power limits, is paid much less attention for a
long term.
In static analysis of power system the reactive power
outputs of generators and switchable shunts are modeled to vary instantaneously within their physical limits
to maintain some buses voltage magnitudes. Those
buses called voltage regulated buses. In power flow
computation, if the voltage magnitude of a bus can be
regulated by the automatic voltage regulator of generator or the continuously-switchable shunt, its type may
be switched between PV and PQ. Generally speaking,
the type of a bus is PV, which means the real power
injection and voltage magnitude of a bus are fixed while
its reactive power injection and voltage phase angle are
free. When its reactive power injection reached its upper or lower limit, the type of this bus becomes PQ,

16th PSCC, Glasgow, Scotland, July 14-18, 2008

Ping Ju
Hohai University
Nanjing, P. R.China

which means that the real and reactive power injections


are fixed while the voltage phase angle and magnitude
are free.
Most of the voltage regulated buses are local control
buses since the corresponding reactive power injection
equipments located at them. Only less of them are remote control buses since the corresponding reactive
power support equipments located at the neighboring
buses instead of themselves. For some special cases,
one bus may belong to both of them. That is to say, its
responsible reactive power support equipments are
located at both themselves and some neighboring buses.
The process of power flow computation is mathematically an iterative solution process of a set of linear
algebraic equations. The type identification of voltage
regulated bus is done between two iterations. If some
buses types are switched, then the dimension of the
Jacobian matrix is changed. Without loss of generality,
in this paper we use the Newton-Raphson method and
suppose that all voltage regulated buses are the local
control buses.
In this paper we are going to show that beside the
numerical blow-up divergence there is another kind of
divergence in power flow computation, in which the
power mismatches maintain around a small specific
value but the bus type of the voltage regulated buses
keep switching. This kind of divergence phenomena are
ignored usually. Are they ture unsolvable cases or an
unreasonable bus-type switching logic is used? Can the
switching logic be modified to solve these divergence
problems? What relationship between this kind of divergence and the Limit Induced bifurcation in voltage
stability assessment? Are there any mathematical ways
to handle the inequality constraints in power flow computation instead of the heuristic type-switching logic?
We try to answer these questions in this paper.
An investigation about the impacts of the bus type
switching logic on power flow solution was presented
in this paper. The paper is organized as follows: In
Section II, the bus PV-PQ type switching logic in load
flow computation is given in detail. In section III three
supplementary logics used usually for handling the type
identification difficulties are given. The numerical test
results of a practical system and IEEE 118 bus system
are shown that a voltage unstable solution could be
obtained by using the supplementary logics in Section
IV. A theoretical analysis is presented in Section V. In
Section VI, the impacts of unreasonable switching lo-

Page 1

gics on computation of continuation power flow are


described.
2

DESCRIPTION OF PV-PQ SWITCHING


LOGIC
Mathematically two tasks need to be done in power
flow computation. One is to decrease the mismatches to
a very small value through an iterative process. Another
is the type identification of the voltage regulated buses.
After k th power flow iteration, for a voltage regulated bus i , the computed value of its reactive power
injection can be solved as:
(1)
Q i = f Q (V k , k )
i

where V and are the bus voltage magnitude and


k

angle vectors respectively after k th power flow iteration.


There are three possibilities for each voltage regulated bus:
1. Bus i is a PQ bus in the previous iteration and its
reactive power was fixed at its lower limit:
If its voltage magnitude V i V i , s e t , then it is still a
PQ bus at current iteration and set Qi = Qi min . If
V i < V i , set , then to compare Qi with the upper and

lower limits. If Q i Q i m ax , then it is still a PQ bus but


set Qi = Qi max . If Q i Q i m in , then it is still a PQ bus
and set Qi = Qi min . If Qi min < Qi < Qi max , then it is
switched to PV bus, set V i , n e w = V i , s e t .
2.

Bus i is a PQ bus in the previous iteration and


its reactive power was fixed at its upper limit:
If its voltage magnitude V i V i , s e t , then bus i still

a PQ bus and set Qi = Qi max . If V i > V i , set , then the


comparison between Q i and its upper/lower limits
need to be done. If Qi Qi max , then it is still a PQ bus
and set Qi = Qi max . If Q i Q i m in , then it is still a PQ
bus but let Qi = Qi min in current iteration. If

Qi min < Qi < Qi max , then it is switched to PV bus and set


Vi ,new = Vi , set .
3.

Bus i is a PV bus in the previous iteration


Compare Qi with its upper and lower limits. If

Qi Qi max , then it is switched to PQ and set


Qi = Qi max . If Q i Q i m in , then it is switched to PQ
and set Qi = Qi min . If Qi min < Qi < Qi max , then it is still a
PV bus.
Note: When a voltage regulated bus is switched to
PV from PQ, there is a mismatch with its voltage magnitude V = Vi , set Vi ,old . If it is too big, the numerical
oscillation would happen, which produce the frequent
oscillation of bus type and slow convergence. There-

16th PSCC, Glasgow, Scotland, July 14-18, 2008

fore, an acceleration factor ( 0 < 1) could be introduced to damp the numerical oscillation.
Vi ,new = Vi ,old + (Vi , set Vi ,old )
(2)
3 SUPPLEMENTARY SWITCHING LOGICS
There are two kinds of divergences with power flow
computation. One is the numerical divergence, in
which the power mismatches increase rapidly with
iterations. Another is the identification divergence, in
which the power mismatches maintain around a specific
value but the bus type identification of the voltage regulated buses meet difficulty. The former usually has less
iteration times than the latter one.
Those buses, which types are switched frequently in
the computation process, are called the hard identification buses in this paper. To handle this identification
difficulty, one seemingly reasonable idea is to stop the
oscillation. The following supplementary logic is usually used.
Supplementary logic 1: If the number of type
switching from PV to PQ for some buses is bigger than
N (say N = 3 ) in the solution process, then these
buses are fixed at PQ type in latter iterative computation
and the switching criterion above is never checked
again.
Generally, one solution can be obtained by using the
above supplementary logic. Furthermore, we hope that
the number of those buses, which are handled by this
supplementary logic, is as small as possible. Therefore,
the following supplementary logic could be an improved one:
Supplementary logic 2: When the supplementary
logic 1 is used, only one of the hard identification buses
is fixed in each iteration while let the others alone. In
general the most limit violated bus is selected to be
fixed.
Another advantage of the supplementary logic 2 is
that the numerical zigzag phenomena in the power flow
computation can be avoided. Its disadvantage is that the
total iteration times would increase.
Since both the bus type identification and the numerical solution are executed alternately in the power
flow computation, it is hard to know if the numerical
convergence process has a bad impact on the correct
identification of bus type. The following supplementary
logic is used based on this idea in which these two tasks
are separated.
Supplementary logic 3: First keep the type of all
voltage regulated buses, run purely numerical iteration
until a converged solution is got, then check if the type
of some buses need to be switched. If there are this kind
of buses, then switch them according to the original
logic described in section II and continue the numerical
solution process. If there is no such kind of bus, then
stop the computation.

Page 2

Table 1: Power Flow Solutions with Different PV-PQ


Switching Logics.

the tie line 5001 outage is an unstable contingency and


there is no an equilibrium point of post-contingency
system. The nose point is a Limit Induced Bifurcation
Point (LIBP) [10]. The key reactive power limit which
results in the immediate instability is the one corresponding to generator bus 111. Bus 111 is coincidently
the only bus fixed as PQ bus by using supplementary
logic 2.
Nose curvebus190

Voltagep.u.

4 NUMERICAL RESULTS
4.1 A practical power system
The numerical tests on power flow convergence of a
practical 640 bus power system are shown here. Using
the original bus type switching logic (OL) and the supplementary logics (SL) above we run the power flow
computation for base case, the case of 500KV tie line
#5031 outage and the case of #5001 outage respectively. The numerical results with respect to convergence are shown in table 1.
Logic
Base case
5001 outage 5031 outage
OL Convergence Divergence Convergence
SL1 Convergence Convergence Convergence
SL2 Convergence Convergence Convergence
SL3 Convergence Divergence Convergence

0.97
0.96
0.95
0.94
0.93
0.92
0.91
0.9
0

0.2

0.4

0.6

0.8

branch parameter

For base case system and the case of tie line 5031
outage, the converged solutions were obtained by using
the original logic and three supplementary logics. Actually the hard identification buses didnt arise in the
solution process. For the case of tie line 5001 outage,
the power flow diverges by using the original logic,
which is a type identification divergence. One convergent solution was obtained by using supplementary
logic 1, but this solution is an abnormal solution, since
five of all voltage regulated buses are fixed as PQ type
while their reactive powers are fixed at their upper limits but their voltage magnitudes are higher than their set
point values. One abnormal solution was also obtained
by using the supplementary logic 2, since the voltage
regulated bus 111 is fixed as PQ type while its reactive
power is fixed at its upper limit but its voltage magnitude is higher than its set point. By using supplementary
logic 3, the power flow computation is still diverged.
Tie line
5001 outage
SL1
SL2

Bus ID(fixed at PQ
by SL1 or SL2)

Abnormal bus

27, 43, 91, 93, 111 27, 43, 91, 93, 111
111
111

Table 2: Comparison of Two Infeasible Solutions.

One question is whether there is a steady stable solution with the case of tie line 5001 outage. In other
words, whether this post contingency load flow is solvable or not? The reference [9] presented a branch parameterized continuation power flow tool to investigate
the nonlinear effect of branch outages. It is used to
verify the above analysis and the nose curve obtained is
shown in Fig. 1.
The branch parameter value max corresponding
to the nose point of the V Curve is smaller than 1
( max 0.921) as shown in Fig. 1, which indicates that

16th PSCC, Glasgow, Scotland, July 14-18, 2008

Figure 1: V curve of contingency 5001 tie line outage.

Logic
OL
SL1
SL2
SL3

Base case
Convergence
Convergence
Convergence
Convergence

Case 1
Divergence
Divergence
Convergence
Divergence

Case 2
Divergence
Convergence
Convergence
Divergence

Table 3: Power Flow Solutions with Different PV-PQ


Switching Logics.

4.2 IEEE118 bus system


The following numerical tests on IEEE118 bus system were done. In base case, the load at bus 59 is
277MW/113MVar. We define Case 1 as increasing the
load in bus 59 to 1055.66MW/502.33MVar. We define
Case 2 as increasing the load in bus 59 to
990.66MW/470.33MVar and line 59-60 outage. The
slack bus 69 takes the real power differences induced by
load increases for both cases. Using the original logic
and three supplementary logics, the power flow to base
case, case 1 and case 2 are solved respectively. The
convergences are shown in table 3.
For base case, the solutions obtained by using the
original logic and three supplementary logics. In addition, since the hard-identification buses are not found in
the solution process, the solutions are same. For Case 1,
when the original logic is used, power flow is diverged
and it is an identification divergence. When the SL1 or
SL3 are used, solution is still diverged. When the SL2 is
used, power flow is converged. For Case 2, the solution
is diverged by using the original logic, which is a numerical divergence but there are about 30 bus type
switching iterations before the numerical blowup. When
we used SL1 and SL2, we got power flow solutions.

Page 3

However, these two solutions are infeasible solutions.


When we used SL3, the power flow is still diverged. In
table 4 we give the comparison of three infeasible solutions in table 3. According to table 4, not all the buses
fixed at PQ by supplementary logic are abnormal buses.
Bus ID(fixed at PQ by Abnormal bus
SL1 or SL2)
34, 36, 46, 54, 66
34, 46
SL2 case 1
34, 66
66
SL1 case 2
66
66
SL2 case 2
Table 4: Comparison of Three Infeasible Solutions.

We did the following numerical test for checking


whether there is a power flow solution corresponding to
Case 1. The maximal transfer power from the slack bus
69 to bus 59 is computed. The load at bus 59 is increased at a proportion 2 to 1 of real power w.r.t. reactive power. The maximal load at bus 59 obtained is
1051.66MW/500.33Mvar (smaller than Case 1). The
corresponding nose point is a LIBP and the key limit is
the generator reactive power limit at bus 66. In table 4,
bus 66 is right one of the buses which are fixed at PQ
type by SL2. The PV curve obtained by using
CPFLOW [11] is given in Fig. 2, which is shown that
there is no a feasible power flow solution for Case 1.
PV Curve(bus 59)

5 THEORETICAL ANALYSIS
First of all, we give the mathematical expression
which is equivalent with the heuristic PV-PQ switching
logic in power flow computation. They are a set of
complementarity constraints between the terminal voltage magnitude and the reactive power of each generator.
(Qi Qi min )(Vi Vi,set )(Qi max Qi ) = 0 (i = 1,2,", ng ) (3)
where ng is the number of generator buses. There are
two variables at each bus in load flow problem. They
are bus voltage magnitude and angle. If the reactive
powers of all generators are considered as variables
either, then there are totally 2n + ng variables. Therefore

ng equations in (3) plus 2n real and reactive power


equations for all buses ( n is the number of buses),
totally 2n + ng equations. They can be written in vector

1.2
Voltagep.t.

For some cases without feasible solutions, infeasible


solutions can be obtained by using SL1 or SL2 while no
solution can be got by using the original logic. The
reason is that these supplementary logics are unreasonable. The original logic excludes automatically the possibility of infeasible solutions. The SL3 has no help for
identification of bus type of voltage regulated buses.
Actually its resulting numerical zigzagging is very severe. The bus type identification divergence in power
flow solution generally corresponds to the LIBP phenomena.

form as:

0.8

f ( , V , Qg ) = 0

0.6

(4)

0.4

A set of inequality constraints about generator reac-

0.2

tive powers need to be satisfied simultaneously. They


are:

0
0

200

400

600

800

LoadMW

Figure 2: P-V curve with load increase at bus 59 (base case)

The numerical test for checking whether there is a


power flow solution corresponding to Case 2 is done as
follows. The line 59-60 is cut off first, then the maximal
load point of bus 59 is computed. The results show that
the maximal load at bus 59 is 985.51MW and
467.25MVar which is smaller than the one in Case 2.
The obtained maximal load point is a saddle-node bifurcation point. This result shows that there is no a power
flow solution with Case 2.
4.3 Analysis and Conclusions
The following conclusions can be made according
to the numerical results of the above practical system
and IEEE 118 bus system. For base case or light load
cases of power system, good solutions can be obtained
by using original logic and three supplementary logics.

16th PSCC, Glasgow, Scotland, July 14-18, 2008

Qi min Qi Qi max

(i = 1, 2,", ng )

(5)

The complementarity constraints (3) and inequality


constraints (5) are not enough for obtaining a feasible
load flow solution. Some important constraints were
missing here which can guarantee that the power flow
solution is a stable solution.

Q
Qmax
Stable
branch

Unstable
branch

Vset

Qmin

Unstable
branch

Stable branch

Figure 3: QV modeling curve of generator

Page 4

A classical criterion for voltage stability is that[12],


at a given operating condition for every bus in the system, the bus voltage magnitude increase as the reactive
power injection at the same bus is increased. A system
is voltage unstable if, for at least one bus in the system,
the bus voltage magnitude decreases as the reactive
power injection at the same bus is increased. In other
words, a system is voltage stable if V Q sensitivity is
positive for all buses while it is voltage unstable if
V Q sensitivity is negative for at least one bus (see the
reference [12]).
The QV modeling curve of generator is shown in
Fig. 3. The arrowhead denotes the load increase direction. It is undifferentiable at the turning points where
the Newton method does not work. Therefore the heuristic PV-PQ bus type switching logic has to be used
there for keeping the efficiency of Newton method. The
real lines denote the stable branches while the broken
lines denote the unstable cases in fig. 3. If the supplementary logics are used in load flow solution, then the
solution points of some generators may be located on
either the unstable branches or the stable branches. If
the original PV-PQ switching logic is used only, then
the solution points of all generators can be located on
the stable branches only. A hyperbolic or sigmoid function is proposed to smooth the QV curve of generator
by Kataoka in [13]. However, it is unsuitable in many
applications such as voltage stability assessment since it
masks the true face of this physical problem.
The stability constraints, which a stable power
flow solution has to subject to, can be written as:
(Vi Vi , set )(Qi max Qi ) 0
(6)

(Vi Vi , set )(Qi Qi min ) 0

(7)

Constraints (6) denote that if the voltage magnitude


of a bus is smaller than its setting point, the corresponding reactive power must be fixed at is upper limit. Constraints (7) denote if the voltage magnitude of a bus is
higher than its setting point, the corresponding reactive
power must be fixed at its lower limit. When a bus is
switched to PQ type from PV type, there is a small
perturbation Q with its reactive power injection. If a
positive perturbation produces a positive response of its
voltage magnitude or a negative perturbation produces a
negative response, then it will satisfy with the original
switching logic and the counter-switching does not
happen. Or else the inequality constraints (6), (7) are
not satisfied, it needs to be switched from PQ to PV.
Therefore according to the classical criterion for static
voltage stability, the original bus type switching logic
without any supplements is an only reasonable logic. It
is equivalent to the voltage stability constraints (6) and
(7). It avoids the possibility that the power flow computation is converged to a voltage unstable solution.
The power flow problem with generator reactive
power limits could be considered as a special optimal
power flow problem. However, the way of control in
this problem is a decentralized local control instead of

16th PSCC, Glasgow, Scotland, July 14-18, 2008

the global control. The objective function is a set of


decentralized and independent objectives which are to
minimize the adjustments of voltage magnitudes of all
generators as the following formula:
1
(8)
M in
(V i V i , set ) 2 i = 1, 2, ..., n g
2
where V i is the generator terminal voltage. Therefore,
the complete problem formulation is composed of the
objective function (8), power flow equation, and the
inequality constraints (5), (6) and (7) together. It is
necessary to note that the complementary conditions (3)
are not required any more. This problem seemingly can
be solved by using a multiple objective nonlinear programming tool without help of the heuristic logics.
However, the inequality constraints (6) and (7) are not
easy to be handled. This formulation is not appropriate
for a direct application of optimization methods. Whenever Qi is at a limit and Vi = Vi , set for a bus, the
Hessian matrix will become singular. A set of complementarity constraints on generator excitation limits,
which are equivalent to stability constraints (6) and (7),
are given by Vournas in loadability limit problem formulation [14]. The decentralized objective functions (8)
are replaced by a scalar objective function. However,
the generator underexcitation limits are not considered
in [14]. A penalty function based approach is used to
solve this optimization problem and to overcome the
complementarity constraints induced solution difficulty.
A more complete formulation of OPF with the complementarity constraints is given by Rosehart in [15,16].
The nonlinear interior point is used to solve the OPF
and a criterion is given for eliminating the singularity of
Hessian matrix or the ill-conditioning of reduced Hessian matrix.
In this paper we focused on the load flow problem instead of general OPF problem. A better way of solving
this special OPF problem is that to solve the basic load
flow equations iteratively while the constraint checking
and equations switching are executed by using a heuristic logic between two iterations. Therefore the sparsity
characteristics are used well to decrease the computational demands. The only requirement is that the switching logic is reasonable and correct, especially for constraints (6) and (7).
According to the sensitivity theory of voltage stability, the bus with key reactive power constraint, which
results in a limit induced bifurcation, is the most sensitive controllable bus for preventing voltage collapse and
increasing the load margins. The area where this bus
located is the most sensitive area. Therefore, if we can
locate quickly the hard identification voltage regulated
buses while the bus type identification divergence happens in load flow solution process, it would give the
valuable information for control.

Page 5

IMPACT ON THE COMPUTATION OF


CONTINUATION POWER FLOW
The Continuation Power Flow has been used
widely in voltage stability analysis [17]. Its mathematical formulation:

f ( x, ) = 0

Q g min Q g ( x , ) Q g max

(9)

where x R n is the n dimensional state variable


vector, R is a parameter, f R n is the power
flow equations, Q g R n g is ng -dimensional generator reactive power vector. f ( x, ) is a piece wise
differentiable function because of generator reactive
power constraints. Therefore it has a piece wise continuous Jacobian matrix. Same as in the power flow
computation, handling the reactive power inequality
constraints in the continuation power flow computation
is done by the PV-PQ bus type switching logic. Therefore, if an unreasonable bus type switching logic is
used, the trajectory tracing may start from an unstable
equilibrium point and follow a wrong trajectory to get
totally wrong results. Sometimes the abnormal branch
switching phenomena would be happened.
P-V Curve(bus 190)
1.3

Voltage(p.u.)

1.25
1.2
1.15
1.1
1.05
1
0.95
0

Lamda

10

15

Figure 4: PV curve obtained by SL 2 under tie line 5001


outage

Voltage

PV Curve(bus 50)
1.06
1.04
1.02
1
0.98
0.96
0.94
0

10

20

30

40

LoadMW

Figure 5: PV curve obtained by SL 2 of IEEE 118 bus


system

16th PSCC, Glasgow, Scotland, July 14-18, 2008

A PV curve obtained by using supplementary logic


2 is shown in Fig. 4. The abscissa is the parameter , 1
p.u. equals to 100MW. For IEEE 118 test system, we
got Fig. 5 by using SL2 in whole CPF computation. The
starting point of this PV curve is an infeasible solution
of Case 1. The load at bus 50 is increased while the
generation at slack bus is increased. Different with the
normal PV curve, the curves in fig. 4 and fig. 5 are
starting from their lower parts to their upper parts. The
bus voltage magnitudes increase instead of decrease
with load increasing. The maximal transfer powers
obtained are much bigger than the actual values. These
results are definitely wrong.
7 CONCLUSIONS
The impacts of bus type switching logic on power
flow computation are investigated in this paper. When
the limitation of the convergence region of the Newton
method and the bad impacts of the unsuitable initial
state are ignored, the power flow divergence indicates
that the corresponding operational state of power system is unstable in static sense. The type identification
divergence of the voltage regulated buses is a kind of
divergence of power flow computation in which the
power mismatches maintain at a small value while the
types of some voltage regulated buses keep switching.
The numerical results of a practical system and IEEE
118 standard test system are shown to indicate that
using some supplementary logics will result in converging to an unstable solution. The bus type identification
divergence in power flow computation is an important
index of limit induced bifurcation voltage instability.
When the unreasonable PV-PQ switching logics are
used in CPF computation, the wrong PV curves and
load margins can be obtained, which are dangerous for
on-line voltage stability assessment. According to a
classical static voltage stability criterion, the stability
constraints for load flow solution are given. Even
though load flow problem considering generator reactive power limits can be written a special form of optimization problem, we suggest that it is solved by the
traditional way with help of heuristic switching logic.
REFERENCES
[1] W. F. Tinney, C. E. Hart, Power Flow Solution by
Newtons Method, IEEE Trans. PAS, Vol.86, no.
11, 1967,pp.1449-1460.
[2] Stott B, Alsac O, Fast Decoupled Power Flow,
IEEE Trans. PAS vol 93, no. 3, 1974, pp. 859-869.
[3] S. Abe, N. Hamada, A Isono, et al, Load flow convergence in the vicinity of a voltage stability limit,
IEEE Trans On PAS, Vol.97, no.6, 1978, pp.19831993.
[4] S. Iwamoto, Y. Tamura, A Load flow calculation
method for ill conditioned power systems, IEEE
Trans On PAS, vol.100, no.4, 1981, pp. 1736-1743.
[5] Y. Tamura, H. Mori, S. Iwamoto, Relationship
between voltage instability and multiple load flow

Page 6

solutions in electric power systems, IEEE Trans On


PAS, 1983, PAS-102(5): 1115-1125.
[6] M. Dehnel, H W Dommel, A method for identifying weak nodes in non-convergent load flows,
IEEE Trans. On Power systems, Vol.4, No. 2, Nov.
1989, pp.801-807.
[7] P. R. Bijwe, S M Kelapure, Nondivergent Fast
Power Flow methods, IEEE trans. On power systems, Vol.18, no. 2, May, 2003, pp.633-638.
[8] J. E. Tate, T. J. Overbye, A comparison of the
optimal multiplier in polar and rectangular coordinates, IEEE trans. On power systems, vol. 20, no.
4, Nov. 2005, pp. 1667-1674.
[9] A. J. Flueck, J. R. Dondeti, A New Continuation
Power Flow Tool for Investigating the Nonlinear Effects of Transmission Branch Parameter Variations,
IEEE Trans. On Power Systems, Vol. 15, No. 1,
Feb. 2000, pp. 223~227.
[10]I. Dobson, L. M. Lu. "Voltage Collapse Precipitated
by the Immediate Change in Stability When Generator Reactive Power Limits are Encountered", IEEE
Trans. On Circuits and Systems-I: Fundamental
Theory and Applications. Vol. 39, No. 9, September, 1992, pp. 762~766.
[11]H. D. Chiang, A.J. Flueck, K.S. Shah and N. Balu,
''CPFLOW: a practical tool for tracing power system
steady-state stationary behavior due to load and gen-

16th PSCC, Glasgow, Scotland, July 14-18, 2008

eration variations,'' IEEE Trans. on Power Systems,


Vol. 10, No. 2, May 1995, pp. 623-634.
[12]P. Kundur, Power System Stability and Control,
McGraw-Hill Companies, Inc. ISBN 0-07-035958.
[13]Y. Kataoka, A smooth power flow model of electric power system with generator reactive power
limits taken into consideration. IEEE International
Symposium on Circuits and Systems, ISCAS 2005.
23-26 May 2005 pp.5286-5289.
[14]C. D. Cournas, M. Karystianos, and N. G. Maratos,
Bifurcation points and loadability limits as solutions of constrained optimization problems, Proc.
IEEE PES summer Meeting, vol.3 July.16-20,2000,
pp.1883-1888.
[15]W. Rosehart, C. Roman, A. Schellenberg, Optimal
power flow with complementarity constraints,
IEEE trans. On power systems, vol. 20, no. 2, May,
2005, pp.813~822.
[16]C. Roman, W. Rosehart, Complementarity model
for generator buses in OPF-based Maximum loading
problems. IEEE trans. On power systems, vol. 20,
no. 1, Feb. 2005, pp.514~516.
[17]J. Zhao, H-D Chiang, H. Li, A new contingency
parameterization CPF model and sensitivity method
for voltage stability control, IEEE PES General
meeting 2005, July, San Francisco.

Page 7