You are on page 1of 22

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/232724788

Distribution Systems Forward/Backward Sweep-based Power Flow


Algorithms: A Review and Comparison Study

Article in Electric Power Components and Systems · January 2009


DOI: 10.1080/15325000802322046

CITATIONS READS

175 5,298

2 authors, including:

Mehmet Hakan Hocaoglu


Istanbul Ticaret University
128 PUBLICATIONS 1,952 CITATIONS

SEE PROFILE

All content following this page was uploaded by Mehmet Hakan Hocaoglu on 28 May 2014.

The user has requested enhancement of the downloaded file.


This article was downloaded by: [Hocaoglu, M. Hakan]
On: 11 December 2008
Access details: Access Details: [subscription number 906617037]
Publisher Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Electric Power Components and Systems


Publication details, including instructions for authors and subscription information:
http://www.informaworld.com/smpp/title~content=t713399721

Distribution Systems Forward/Backward Sweep-based Power Flow Algorithms:


A Review and Comparison Study
U. Eminoglu a; M. H. Hocaoglu b
a
Department of Electrical and Electronics Engineering, Nigde University, Nigde, Turkey b Department of
Electronics Engineering, Gebze Institute of Technology, Kocaeli, Turkey

Online Publication Date: 01 January 2009

To cite this Article Eminoglu, U. and Hocaoglu, M. H.(2009)'Distribution Systems Forward/Backward Sweep-based Power Flow
Algorithms: A Review and Comparison Study',Electric Power Components and Systems,37:1,91 — 110
To link to this Article: DOI: 10.1080/15325000802322046
URL: http://dx.doi.org/10.1080/15325000802322046

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf

This article may be used for research, teaching and private study purposes. Any substantial or
systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or
distribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses
should be independently verified with primary sources. The publisher shall not be liable for any loss,
actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly
or indirectly in connection with or arising out of the use of this material.
Electric Power Components and Systems, 37:91–110, 2009
Copyright © Taylor & Francis Group, LLC
ISSN: 1532-5008 print/1532-5016 online
DOI: 10.1080/15325000802322046

Distribution Systems Forward/Backward


Sweep-based Power Flow Algorithms:
A Review and Comparison Study

U. EMINOGLU1 and M. H. HOCAOGLU2


1
Department of Electrical and Electronics Engineering, Nigde University,
Nigde, Turkey
2
Department of Electronics Engineering, Gebze Institute of Technology,
Kocaeli, Turkey
Downloaded By: [Hocaoglu, M. Hakan] At: 10:28 11 December 2008

Abstract Because of different characteristic features of distribution networks, such


as radial structure and high R/X ratio, distribution system load flow analysis differs
significantly from transmission systems. Accordingly, a number of methods that are
specially designed for the solution of the power flow problem in radial distribution
systems are found in the literature. Most of the developed methods are based on the
forward/backward sweep processes for the solution of the ladder networks. In this arti-
cle, various distribution system load flow algorithms, based on the forward/backward
sweeps, are reviewed, and their convergence ability is quantitatively evaluated for
different loading conditions, R/X ratios, and sub-station voltage levels. Moreover, the
effects of static load modeling on the convergence characteristics of algorithms are
also investigated.

Keywords radial distribution system, forward/backward sweep algorithms, load flow


analysis

1. Introduction
In the last few decades, efficient and reliable load flow solution techniques, such as Gauss-
Seidel (G-S), Newton-Raphson (N-R), and fast decoupled load flow [1–3], have been
developed and widely used for power-system operation, control, and planning. However,
it has repeatedly been shown that these methods may become inefficient in the analysis
of distribution systems due to the special features of such networks—i.e., radial structure,
high R/X ratio, un-transposed lines, and unbalanced loads along with single-phase and
two-phase laterals [4, 5]. In addition to these, distribution network matrices are generally
ill-conditioned and this may cause numerical problems for the conventional power flow
algorithms [4–6]. These characteristic features make the distribution systems power
flow computation different and somewhat difficult to analyze when compared to the
transmission systems’ load flow analysis.
Recently, many researchers have paid attention to obtain the load flow solution of
distribution networks and, accordingly, various methods are available in the literature to

Received 6 November 2007; accepted 19 May 2008.


Address correspondence to Dr. M. H. Hocaoglu, Department of Electronics Engineering,
Gebze Institute of Technology, Kocaeli, 41400, Turkey. E-mail: hocaoglu@gyte.edu.tr

91
92 U. Eminoglu and M. H. Hocaoglu

carry out the analysis of balanced and/or unbalanced radial distribution systems [7–40].
Methods developed for the solution of ill-conditioned radial distribution systems may
be divided into two categories. The first type of methods [7–11] is utilized by proper
modification of existing methods such as N-R and G-S. On the other hand, the second
group of methods [12–40] is based on forward and/or backward sweep processes using
Kirchhoff’s laws [12–24] or making use of the well-known bi-quadratic equation [25–40].
This article qualitatively reviews and quantitatively analyzes balanced and unbalanced
distribution system forward/backward sweep-based power flow algorithms. In particu-
lar, the convergence ability of these algorithms are evaluated under different loading
conditions and R/X ratios, and the effect of static load models on the convergence
characteristics of the algorithms are analyzed.

2. Voltage Formulation for Distribution Networks


Let us consider a distribution line model as given in Figure 1. The real and reactive
power at the receiving end can be written as
Downloaded By: [Hocaoglu, M. Hakan] At: 10:28 11 December 2008

Vs Vr Vr2
Pr D cos.z ıs C ır / cos.Z /; (1a)
Z Z
Vs Vr Vr2
Qr D sin.z ıs C ır / sin.Z /; (1b)
Z Z
where Vs and Vr stand for the phase voltage magnitudes at bus s and bus r , respectively,
and Z is the magnitude of the line impedance. z , ıs , and ır stand for the phase angle
of the line impedance and bus voltages, respectively. Equations (1a) and (1b) can be
rewritten as
Pr Z Vr
cos.z ıs C ır / D C cos.z /; (2a)
Vs Vr Vs
Qr Z Vr
sin.z ıs C ır / D C sin.z /I (2b)
Vs Vr Vs
using the trigonometric identity,

cos2 .z ıs C ır / C sin2 .z ıs C ır / D 1; (3)

and substituting Eqs. (2a) and (2b) into Eq. (3), we get the general form of the well-
known bi-quadratic equation given in Eq. (4a). The maximum real root of the equation
gives the line receiving-end voltage magnitude. The line receiving-end voltage can also

Figure 1. A two-bus distribution network.


Distribution Systems Power Flow Algorithms 93

be written in terms of line sending-end powers and sending-end voltage magnitude as


given in Eq. (4b):

Vr4 C 2Vr2 .Pr R C Qr X/ Vs2 Vr2 C .Pr2 C Qr2 /Z 2 D 0; (4a)


s
.Ps2 C Qs2 /Z 2
Vr D Vs2 2.Ps R C Qs X/ C : (4b)
Vs2

For the same line, given in Figure 1, Kirchhoff’s Voltage Law (KVL) can be written in
complex form as

V s D V r C I s Z; (5a)

Vr D Vs I s Z: (5b)
Downloaded By: [Hocaoglu, M. Hakan] At: 10:28 11 December 2008

These equations, Eqs. (4a), (4b), (5a), and (5b), are frequently used for the calcu-
lation of branch sending-end or receiving-end node voltages for radial networks in
forward/backward sweep-based algorithms.

3. Brief Descriptions of Forward/Backward


Sweep-based Algorithms
Forward/backward sweep-based power flow algorithms generally take advantage of the
radial network topology and consist of forward and/or backward sweep processes. In
these type of algorithms, developed in [12–40], the forward sweep is mainly the node
voltage calculation from the sending end to the far end of the feeder and laterals, and
the backward sweep is primarily the branch current and/or power summation from the
far end to the sending end of the feeder and laterals. In addition to the branch current
and/or power summation, the node voltages are also computed in backward sweeps in
some algorithms [19, 20, 24, 29, 33, 36]. These sweep-based power flow algorithms can
be classified in two groups—Kirchhoff’s formulation and the quadratic equation-based
algorithms.

3.1. Kirchhoff’s Formulation-based Sweep Algorithms


Most of the distribution system power flow algorithms [12–24] employ KVL and Kirch-
hoff’s Current Law (KCL) to calculate the node voltages in the forward and backward
processes. In [12], the authors presented a compensation-based power flow method
for radial distribution networks and/or for weakly meshed networks using a multi-port
compensation technique and basic formulations of Kirchhoff’s laws. The radial part is
solved by a straightforward two-step procedure in which the branch currents are first
computed (backward sweep) and then the bus voltages are updated (forward sweep) by
using Eq. (5a) for each branch. Maximum real and reactive power mismatch of loads
are used for the convergence criteria in the algorithm. This algorithm is also applied to
unbalanced radial and/or weakly meshed distribution networks in [13].
The same approach, with slight modifications, is used in [14–17] for balanced and
unbalanced distribution networks. In [14], the radial part of the networks is solved
using backward current and forward voltage calculation as given in [12], additionally;
94 U. Eminoglu and M. H. Hocaoglu

a three-phase transformer model, developed in [41], is incorporated into the algorithm.


The main difference of the method developed in [15] from that given in [13] is that
each branch current is computed with its receiving-end node voltage and transferred
power calculated in the backward process instead of simply applying KCL to the load
current. The same approach is also used in [16]. The radial part of the system is solved
using backward current and forward voltage calculation as given in [12], and maximum
node voltage mismatch is used for the convergence criteria. In addition, authors of [16]
demonstrated a computerized building algorithm for identifying the nodes beyond all
branches that helps to find the exact current flow through all branches. The algorithm
given in [15] is further developed for the balanced radial networks by using sub-iterations
on laterals in [17]. In that study [17], sub-iterations are used to calculate the lateral total
current, which is added to the feeder at the main iteration. The maximum node voltage
mismatch is used for the convergence criteria in the sub and main iterations, and the
effect of load modeling on the power flow results is also analyzed.
In [18], a modified version of the forward/backward sweep method that works for
balanced and unbalanced radial networks was proposed. In the backward process, each
Downloaded By: [Hocaoglu, M. Hakan] At: 10:28 11 December 2008

branch current is calculated by using KCL. Then, using these currents, the node voltages
are computed by using Eq. (5b) in the forward process. The voltage magnitudes at each
bus are compared with the previous iteration. If the error is within the tolerance limits,
the procedure is stopped. Otherwise, the forward and backward processes are continued
until the convergence for all node voltages is achieved. The ladder network theory [19]
is similar to the forward/backward sweep method [18]. In the ladder network theory,
the currents in each branch are computed in backward process and, in addition to the
branch currents, the node voltages are also computed by using Eq. (5a) in the algorithm.
Thus, the magnitude of the source bus voltage is also calculated. The computed voltage
value of the source bus is compared with its specified value. If the error is within the
limits, the procedure is stopped.
An algorithm, based on the forward and backward voltage updating process, was
developed by Liu et al. in [20] for balanced radial and/or weakly meshed networks. In the
algorithm, the authors have proposed the ratio-flow method, which is based on the forward
and backward ladder equations by using a voltage ratio for the convergence control. This
method is applied with standard N-R method for complex distribution systems, which
have multiple sources or relatively strong connected loops with extended long radial
feeders including laterals. The radial part is solved by using forward voltage calculation
and backward voltage and current calculation as in the ladder network theory [19]. The
main difference of this method from the ladder network theory is that node voltages
are adjusted by using a ratio obtained from the ratio of the source voltage and its
computed value after the backward process. In addition, sub-iterations are applied for
the laterals to calculate their bus voltages and total current. Consequently, they are added
to the feeder as a spot load with their computed total current during the main iteration
on the feeder. For the convergence criteria, maximum node voltage mismatch is used in
the algorithm.
An unbalanced distribution system power flow algorithm is given in [21] by Teng. In
this algorithm, a topographical approach has been used to tackle the load flow problem.
Two matrices are developed—the bus injection to branch current (BIBC) matrix and the
branch current to bus voltage (BCBV) matrix. Consequently, the distribution load flow
(DLF) matrix, which is obtained by multiplication of these two matrices, is computed;
the voltage drop on each branch is then computed using the DLF matrix and load currents
as given in Eq. (6a). After the calculation of the voltage drop on the branches, the node
Distribution Systems Power Flow Algorithms 95

voltages are computed by using the source bus voltage, and the computed voltage drops
as given in Eq. (6b):

VkC1 D DLF  Ik ; (6a)

VkC1 D Vs VkC1 : (6b)

For the convergence criteria, maximum load current mismatch is used in the algorithm.
This algorithm is further improved in [22] for radial and/or weakly meshed networks by
modifying the BIBC and BCBV matrices. A similar method is given in [23] for balanced
radial networks. In the algorithm, the voltage drops of each branch are computed using
their current, ib, and series impedances, zb , as given in Eqs. (7a) and (7b). Branch currents
are calculated by using load currents and the branch-to-node matrix, C. Then, each node
voltage is calculated by using computed voltage drop and the C matrix using Eq. (7c).
For the convergence criteria, the maximum difference of the calculated node voltages
and their previous values is used in the algorithm:
Downloaded By: [Hocaoglu, M. Hakan] At: 10:28 11 December 2008

ib D C  Iload; (7a)

vb D zb  ib ; (7b)
X
vi D vo Cj i vj : (7c)
j

A KVL-based load flow algorithm for the solution of the single-phase distribution
networks was also developed in [24]. In the algorithm, the load flow problem is solved
by considering the loads as impedances at each iteration in the backward process. Each
node voltage is calculated by applying Eq. (5a) to the passive network in the backward
process. Then, the ratio between the voltage imposed at the source node and its new
computed value is used to calculate the actual value of the bus voltages, load, and branch
currents. In addition, the authors demonstrated a PV-bus modeling method to keep its
bus voltage at the desired value by calculating required reactive power of the PV nodes.

3.2. Quadratic Equation-based Sweep Algorithms


The quadratic equation relates the voltage magnitude at the receiving end to the branch
power and the voltage at the sending end. Several power flow methods, given in [25–40],
have been developed by using the quadratic equation in Eqs. (4a) and (4b). In the study
of [25], Cespedes presented an algorithm solving the quadratic equation, Eq. (4a), for each
branch. The method has forward voltage calculation and backward power summation. In
the backward process, load power and branch power losses are used for branch power
calculation, and the voltage calculation is ended when the system’s total active and
reactive power loss mismatches are lower than a specified error value. The quadratic
equation was also used for voltage calculation in the forward process by Haque in [26]
to develop a load flow algorithm for balanced radial networks. In the algorithm, branch
sending-end power is used, and receiving-end node voltages of each branch is calculated
by using Eq. (4b) in the forward process. Node phase angles are also computed in the
forward process. For the convergence criteria, maximum node voltage mismatch is used.
Moreover, load flow results for various voltage-dependent load models were also analyzed
in the study. The algorithm was further enhanced and applied to the solution of both radial
and/or weakly meshed networks in [27] and with more than one source bus in [28].
96 U. Eminoglu and M. H. Hocaoglu

A method that uses the quadratic equation was given by Asfari et al. in [29]. The
node voltages are computed in the backward process by solving Eq. (4a) for each branch
starting from the end of the feeder to the source bus, and this process is also applied to
the laterals in sub-iterations in the algorithm. The difference of the source bus voltage
magnitude and its computed value (after the backward process) is added to the last
node voltage magnitude of the feeder. This process is continued when the difference is
lower than the specified error value in the main iteration. The same voltage adjustment
is also used in sub-iterations for laterals, but in this case, the lateral’s first node voltage
magnitude is used instead of source bus voltage magnitude. The quadratic equation was
also used by Ranjan and Das [30] for balanced distribution system power flow analysis. In
the algorithm, the node voltages are computed by solving Eq. (4a) for each branch in the
forward process, and the transferred active and reactive power, including power losses,
from each branch is computed in the backward process for the whole network by using
a computerized node and branch numbering technique. For the convergence criteria,
the maximum difference of the transferred active and reactive power from the source
bus with their previous value is used. The same approach was also used by Das et al.
Downloaded By: [Hocaoglu, M. Hakan] At: 10:28 11 December 2008

[31, 32]. In these algorithms, the node voltages are calculated by solving the quadratic
equation, given in Eq. (4a), for each branch in the forward process, and branch active
and reactive power, including losses, is calculated in the backward process automatically
using a computerized algorithm developed by using a special node and branch numbering
technique. For the convergence criteria, the maximum branch active and reactive power
losses mismatch and the maximum difference of total active and reactive power with
their previous values were used in [31] and [32], respectively.
A quadratic equation-based load flow algorithm was also given in [33] for balanced
radial networks. In the algorithm, node voltages are calculated by solving quadratic
equation, Eq. (4a), for each branch in the forward sweep. Then the backward calculation
is applied to obtain new voltages using Kirchhoff’s formula, given in Eq. (5a). The
node voltages are adjusted using a voltage ratio, which is obtained as in the ratio-flow
method [20]. In addition, the authors incorporated the voltage-dependent loads into the
quadratic equation and evaluated the effect of load models on the convergence ability of
the method. The maximum node voltage mismatch is used for the convergence criteria
in the algorithm.
In the literature, there are some sweep based algorithms [34–40] that have been
developed by using the branch flow equation, which includes the quadratic equation
or reformulation of the quadratic equation in different form due to the requirements of
special solution techniques that may cause computational burden. In the method proposed
by Baran and Wu [34], for each branch of the network, three non-linear equations (called
the branch flow equation and given in Eqs. (8a)–(8c)) are written in terms of branch
power flows and bus voltages, and then Newton’s method is applied to solve this set of
equations. The algorithm was developed by Mekhamer et al. [35] by solving only the bi-
quadratic equation given in Eq. (8c) for the node voltage calculation, and the transferred
power from the source bus is updated by adding the computed transferred power of the
last node of the main feeder and laterals. The transferred power from each branch is also
calculated by using Eqs. (8a) and (8b) in the forward process. In addition, the authors
demonstrated a method for the calculation of the lateral total active and reactive powers.
For the convergence criteria, the maximum transferred active and reactive powers from
the last node of the feeder and laterals are used in the algorithm. Baran and Wu [36] also
presented another power flow method for balanced radial distribution networks. Branch
flow equations, which are given in Eqs. (8a)–(8c) and Eqs. (9a)–(9c), are used in the
Distribution Systems Power Flow Algorithms 97

forward and backward processes, respectively. First, the bus voltages and transferred
active and reactive power from each branch is calculated by solving the backward branch
equations, given in Eqs. (9a)–(9c), in the backward process. Then, the forward process
is applied by solving Eqs. (8a)–(8c) for each branch to compute each node new voltage
starting from the source bus to the far end. The difference of the fixed source bus
voltage and its computed value, after the backward sweep, is used for the convergence
criteria in the algorithm.

Forward branch flow equations:

Ps2 C Qs2
PsC1 D Ps R PLr ; (8a)
Vs2

Ps2 C Qs2
QsC1 D Qs X QLr ; (8b)
Vs2
Downloaded By: [Hocaoglu, M. Hakan] At: 10:28 11 December 2008

.Ps2 C Qs2 /jZj2


Vr2 D Vs2 2.RPs C XQs / C : (8c)
Vs2

Backward branch flow equations:

Pr2 C Qr2
Ps D PsC1 C R C PLr ; (9a)
Vr2

Pr2 C Qr2
Qs D QsC1 C X C QLr ; (9b)
Vr2

.Pr2 C Qr2 /jZj2


Vs2 D Vr2 C 2.RPr C XQr / C : (9c)
Vr2

Luo and Semlyen [37] developed an algorithm for radial and/or weakly meshed
networks. Longitudinal and transversal voltage drop equations, given in Eqs. (10a)–
(10d), are used for the calculation of the node voltage magnitudes and phase angles
in the forward process, and branch power, including losses, is updated in the backward
process. For the convergence criteria, the maximum node voltage mismatch is used in
the algorithm. In addition, the algorithm is improved for the modeling of distributed
generations (DGs) as a PV-bus in the load flow analysis:

RPs C XQs
V 0 D ; (10a)
Vs
XPs RQs
V 00 D ; (10b)
Vs
p
Vr D .Vs V 0 /2 C V 002 ; (10c)

V 00
 
ır D ıs tan 1 : (10d)
Vs V 0

Jasmon and Lee [38] proposed a new load flow algorithm for balanced radial distribution
systems. The authors developed two fundamental equations, given in Eqs. (11a) and (11b),
98 U. Eminoglu and M. H. Hocaoglu

representing real and reactive powers without any knowledge on bus voltages, and then
solved by reducing the whole network to the single-line equivalent. The calculation
process ends when the maximum active and reactive transferred power mismatch in each
branch is lower than a specified error value. After the convergence for each branch, node
voltages are calculated by using line active or reactive power losses:

2X 2 Pr 2RXQr C R
Ps D
2.R2 C X 2 /
p
.2X 2 Pr 2RXQr C R/2 4.R2 C X 2 /.X 2 Pr C R2 Qr 2RXPr Qr C RPr /
;
2.R2 C X 2 /
(11a)

2R2 Qr 2RXPr C X
Qs D
2.R2 C X 2 /
Downloaded By: [Hocaoglu, M. Hakan] At: 10:28 11 December 2008

p
.2X 2 Pr 2RXQr C R/2 4.R2 C X 2 /.X 2 Pr C R2 Qr 2RXPr Qr C RPr /
:
2.R2 C X 2 /
(11b)

In the study of [39], the authors developed a sweep algorithm using Eqs. (12a)–(12d)
for the node voltage calculation in the forward process and each branch power, including
power losses, updated in the backward process. Moreover, DGs are modeled as a complex
current source, and its current is adjusted at each iteration in the algorithm. The authors
also demonstrated a computerized building algorithm for identifying the nodes beyond all
branches, which helps to find the exact current and power flowing through all branches:

Re.V r / D a Re.V s / bIm.V s /; (12a)

Im.V r / D b Re.V s / C aIm.V s /; (12b)

RPs C XQs
aD1 ; (12c)
Vs2
RQs XPs
bD : (12d)
Vs2

In the study of [40], the authors developed an algorithm for balanced radial distri-
bution networks. The algorithm consists of forward voltage calculation and backward
power and current summation processes. In the forward process, the node voltages of
each branch are computed using Eqs. (13a)–(13c) in the network after updating each
branch current and power, including power losses in the backward process. Maximum
node voltage mismatch is used for the convergence criteria in the algorithm:
 
1 Qr
r D tan ; (13a)
Pr

K D Vs2 2Pr .R C X tan r /; (13b)


s p
K C K 2 4.R2 C X 2 /Pr2 sec2 r
Vr D : (13c)
2
Distribution Systems Power Flow Algorithms 99

4. DG Modeling
DG is an effective means of increasing energy efficiency and reducing energy costs.
Accordingly, the number of DGs integrated to the distribution systems has been greatly
increased in recent years. Hence, DGs must be included to the power flow analysis
with their representative features, i.e., the voltage controlled bus (PV-bus). Inclusion of
DGs to the radial system results in networks that may not be strictly radial, and their
feeders may carry power whose direction may change as a function of the loading and
distributed generation levels [42]. In distribution system power flow analysis, DGs are
generally classified as constant PQ- or PV-nodes. The constant PQ models are identical
with constant power (CP) load models, except that the current is injected into the bus. In
these systems, the problem of how to handle the PV buses is the key to the power flow
solution of the distribution systems.
The conventional methods and their modifications perform well for the power flow
solution with PV-modeled buses. However, the problem of conventional power flow al-
gorithms still persists for ill-conditioned systems. Some sweep-based algorithms have
Downloaded By: [Hocaoglu, M. Hakan] At: 10:28 11 December 2008

also been made to incorporate Q-sources into the radial load flow scheme based on the
current injection model. Even in these methods, only a few of them properly handled
controlled Q-sources (PV-bus) [12, 24, 37, 39] for the balanced distribution networks. In
the study of [12], amount of injected reactive power was calculated by using the secant
method as follows:
.k 1/ .k 2/
.k/ QP v Qpv .k/ .k 1/ .k 1/
Qpv D .k 1/ .k 2/
.jVpv j jVpv j/ C Qpv ; (14)
jVpv j jVpv j

where k shows the iteration number. From Eq. (14), it is seen that the secant method
heavily depends upon initial conditions, and it may cause the convergence difficulties on
the load flow algorithms when appropriate voltage and reactive power is not chosen at the
initial state. Due to the convergence difficulties on the secant method, authors improved
their algorithm by using the linearized relationship between injected power and a sensi-
tivity matrix for the breakpoints power injections as well as PV-node power injections
in [37]. The incremental power change is applied with the assumption of all bus voltages
being close to 1.0 p.u., and its value is calculated as

Z.S / D V ; (15)

where Z shows the sensitivity matrix and is computed by obtaining the Thevenin equiv-
alent seen from the breakpoints. After the calculation of S , it is added to the total
power, which is obtained at the last iteration. The same approach was generalized for
any controlled Q-sources (i.e., synchronous generators, static VAr compensators etc.)
in [43]. A new approach was developed in [24]. In that study, DGs are modeled as
shunt capacitors and to bring their bus voltages to the specified value; their reactance
are computed at each iteration using Eq. (16), which is obtained by using the Thevenin
equivalent circuit seen from the DG bus. Then the computed shunt reactance is added in
parallel to the existing one, which is computed at the previous iteration:
v !
u 2
u
2 2
Eth
Xth t Xth C Zth 1
Vsp2
Xc D 2
: (16)
Eth
1
Vsp2
100 U. Eminoglu and M. H. Hocaoglu

Most of the unbalanced sweep-based power flow algorithms modeled the DGs as a
CP model, such as is given in [44]. In this model, DGs are modeled as injected current
sources, and currents are calculated by using their terminal voltages and total active and
reactive powers that are constant. A few of three-phase sweep algorithms, given in [13]
and [14], modeled DGs as PV-nodes, and their bus voltages are controlled at the setting
value by improving the model given in [37]. In the method developed in [13], the required
reactive power generation is calculated using the PV-node voltage and its injected current,
which is computed by using Eq. (17) as follows:

ŒZŒI q  D ŒV ; (17)

where Z is a constant real matrix and can be selected as equal to the real part of the
sensitivity matrix given in Eq. (15). After the calculation of the required reactive power
of the DG, it is modeled as a PQ-bus with its current injected into the bus at the next
iteration. Similar methods were used for modeling DGs as PV-nodes in unbalanced [14]
and balanced [39] systems load flow analysis. In these studies, DGs are modeled as
Downloaded By: [Hocaoglu, M. Hakan] At: 10:28 11 December 2008

a current source and their values are adjusted at each iteration after the node voltages
calculation. In [14], the DG’s reactive current is adjusted using calculated current, which
is computed by using a PV-node sensitive matrix and voltage mismatch as given in [13]
at each iteration. The computed current is added to the PV-bus total current, which is
computed at the last iteration. Then, this adjusted current is used as a load current, except
that the current is injected into the bus in node voltage calculations at the next iteration.
The same approach was used in [39] for balanced distribution systems. In this method,
PV-node active and reactive current components are adjusted by using its incremental
current changes at each iteration. After the calculation of incremental current changes
(I ) as given in [13], the node voltages are re-calculated by adding voltage correction,
which is obtained by applying the incremental current changes at the PV-nodes to the
network as current sources, without any other sources, and setting the voltage at the slack
node to zero.
The general structure and main features of all forward/backward sweep-based algo-
rithms reviewed in this article are tabulated in Table 1. In the table, Column 2 shows the
algorithms that can also be applied to the unbalanced distribution networks. In Column 3,
how the node voltages of the system are calculated is briefly explained. The convergence
criterion of each algorithm is given in Column 5, and the algorithms in which the DG
is modeled as PV-bus are shown in Column 6. Column 7 classifies the three-phase
algorithms in which three-phase transformers are modeled.

5. Application of Sweep-based Algorithms to Radial


Distribution Networks
Three distribution networks (12-bus, 33-bus, and 69-bus radial networks) are used to
test the convergence ability of the sweep-based algorithms. They have all been coded in
Matlab. The traditional power flow algorithms (classical N-R and G-S) were coded by
Saadat in Matlab [45], and a standard power flow program (PFLOW [46]), designed by
using Newton’s method and sparse matrix techniques, are also applied for the power flow
analyses. A single-line diagram of the test systems and their data can be obtained from
the references which are given in Table 2, in which the iteration number of algorithms
for solving the load flow of the test systems are provided. In load flow analyses, initial
voltage magnitude at all buses is considered to be the same as the source bus (1.0 p.u.)
Distribution Systems Power Flow Algorithms 101

Table 1
General features of forward/backward sweep-based algorithms
Voltage calculation Backward sweep

Ref. Thr. Forward Backward Branch Branch Convergence DGs Tr.


no. phs. sweep sweep current power criterion (PV) M.

[12] — Eq. (5b) — 3 — Maximum active and reactive 3 —


power mismatch in each node
[13] 3 Eq. (5b) — 3 — Maximum active and reactive 3 —
power mismatch in each node
[14] 3 Eq. (5b) — 3 — Maximum active and reactive 3 3
power mismatch in each node
[15] 3 Eq. (5b) — — 3 Maximum node voltage — —
mismatch
[16] — Eq. (5b) — 3 — Maximum node voltage — —
mismatch
[17] — Eq. (5b) — 3 — Maximum node voltage — —
mismatch
Downloaded By: [Hocaoglu, M. Hakan] At: 10:28 11 December 2008

[18] 3 Eq. (5b) — 3 — Maximum node voltage — —


mismatch
[19] 3 Eq. (5b) Eq. (5a) 3 — Source node voltage mismatch — —
[20] — Eq. (5b) Eq. (5a) 3 — Maximum node voltage — —
mismatch
[21] 3 Eqs. (6a) and (6b) — 3 — Maximum load current — —
mismatch
[22] 3 Eqs. (6a) and (6b) — 3 — Maximum load current — —
mismatch
[23] — Eqs. (7a)–(7c) — 3 — Maximum node voltage — —
mismatch
[24] — — Eq. (5a) 3 — Source node voltage mismatch 3 —
[25] — Eq. (4a) — — 3 Total power losses mismatch — —
[26] — Eq. (4b) — — 3 Maximum node voltage — —
mismatch
[27] — Eq. (4b) — — 3 Maximum node voltage — —
mismatch
[28] — Eq. (4b) — — 3 Maximum node voltage — —
mismatch
[29] — — Eq. (4a) — 3 Source node voltage mismatch — —
[30] — Eq. (4a) — — 3 Total transferred active and — —
reactive power mismatch
[31] — Eq. (4a) — — 3 Maximum branch active and — —
reactive power losses
mismatch
[32] — Eq. (4a) — — 3 Total transferred active and — —
reactive power mismatch
[33] — Eq. (4a) Eq. (5a) 3 3 Maximum node voltage — —
mismatch
[34] — Eqs. (8a)–(8c) — — 3 Maximum transferred active and — —
reactive power of last nodes
[35] — Eq. (8c) — — — Maximum transferred active and — —
reactive power of last nodes
[36] — Eqs. (8a)–(8c) Eqs. (9a)–(9c) — 3 Source node voltage mismatch — —
[37] — Eqs. (10a)–(10d) — — 3 Maximum node voltage 3 —
mismatch
[38] — Eqs. (11a) and (11b) — — 3 Maximum transferred active and — —
reactive power mismatch of
each branch
[39] — Eqs. (12a)–(12d) — — 3 Maximum node voltage 3 —
mismatch
[40] — Eqs. (13a)–(13c) — — 3 Maximum node voltage — —
mismatch

Thr. phs.: Can also be applied to the unbalanced distribution network.


Tr. M.: Modeling of the three-phase transformer in the algorithm.
102 U. Eminoglu and M. H. Hocaoglu

Table 2
Iteration number of load flow algorithms for different test systems

Number of iterations

Radial test systems

No. Algorithm type Ref. no. 12-bus [31] 33-bus [36] 69-bus [30]

1 Kirchhoff’s [12] 5 8 8
formulation-based
sweep algorithms
2 [13] 5 8 8
3 [14] 5 8 8
Downloaded By: [Hocaoglu, M. Hakan] At: 10:28 11 December 2008

4 [15] 5 6 6
5 [16] 5 6 6
6 [17] 5 6 5
7 [18] 5 6 6
8 [19] 5 6 6
9 [20] 4 6 5
10 [21] 6 8 8
11 [22] 6 8 8
12 [23] 5 6 6
13 Quadratic [25] 4 5 5
equation-based sweep
algorithms
14 [26] 3 4 4
15 [27] 3 4 4
16 [28] 3 4 4
17 [29] 5 6 5
18 [30] 4 5 5
19 [31] 3 4 4
20 [32] 4 5 5
21 [33] 3 3 3
22 [34] 4 4 5
23 [35] 8 11 10
24 [36] 4 4 4
25 [37] 3 4 4
26 [39] 3 4 5
27 [40] 3 4 4
28 Conventional algorithms G-S 530 N/A N/A
29 N-R 4 5 N/A
30 PFLOW 4 5 N/A
N/A: Not available.
Distribution Systems Power Flow Algorithms 103

for all test system, and a tolerance of 10 6 on convergence criteria is used for the ending
of load flow solutions.
From the required iteration number given in Table 2, it can be said that the quadratic
equation-based algorithms, except the algorithm developed in [35], have generally good
convergence ability when compared with Kirchhoff’s formulation-based algorithms. Al-
though a comprehensive comparison on the computation time is not undertaken, the
initial results suggest that the computational efficiency of the quadratic equation-based
algorithms are somehow slower than Kirchhoff’s formulation-based algorithms because
of the solution of the quadratic equation. Moreover, even though the branch power flow-
based methods developed in [34] and [36] have good performance, the formulation and
evaluation of Jacobians during each iteration require large computer memory storage
and cause a decrease of their computational efficiency as evaluated in [35]. In addition,
classical G-S, N-R, and standard power flow program (PFLOW) face the convergence
problem when they are applied to the large radial distribution networks, as can be seen
from the results of 69-bus test system in Table 2.
The convergence characteristic of the methods is also analyzed for different R/X
Downloaded By: [Hocaoglu, M. Hakan] At: 10:28 11 December 2008

ratios, loading conditions, tolerance values, and different load models on a 69-bus radial
distribution network. Different loading conditions are considered by multiplying each
node’s power by a load factor () as S D   Sb , and results are depicted in Figure 2(a)
for the variation of  from 1 to 2.5. R/X ratio of branch impedances are varied by
multiplying each branch resistance by a coefficient (k) as Z D k  R C jX, and results
are depicted in Figure 2(b) for the variation of k from 1 to 3. All sweep algorithms
that are given in Table 2 are considered in the parametrical analyses and given with
its reference number in the figures. From Figure 2, it is clearly seen that for the heavy
loading conditions and high R/X ratios, the iteration number of the quadratic equation-
based methods are generally small when compared with the methods that are based on
KVL and KCL. Moreover, they are less sensitive to the loading levels and R/X ratios of
the systems. On the contrary, the convergence speed of most of the KCL- and KVL-based
algorithms are heavily under the influence of the system parameters.
Convergence tolerances for all methods are varied from 10 5 to 10 10 , and results de-
picted in Figure 3(a). Likewise, it is seen that methods in which the quadratic equation is
employed have better convergence characteristics than KVL-based methods for different
tolerance values. Finally, the performance of the sweep-based methods are evaluated for
CP, constant current (CI), and constant impedance (CZ) load models, and the number of
iterations are given in Figure 3(b). From this figure, it is seen that an increase on voltage
dependency of the loads causes convergence difficulties on the quadratic equation-based
algorithms; this is due to the variation of the load and branch power with the node voltage
magnitudes at each iteration. On the contrary, the convergence speed of the KVL-based
algorithm is less affected with the increase of the load voltage dependency. In addition,
because the KVL-based methods are based on current summation, and when the loads
are modeled as a CI, the required iteration is in the minimum level for these type of
algorithms, as can be seen from Figure 3(b).
Accordingly, although the branch power flow-based methods given in [34] and [36]
have good performance and are less affected with the variation of loading conditions, R/X
ratios of the lines, convergence tolerances, and voltage dependency of load, they require
a large computer memory storage and computational time due to the formulation and
evaluation of Jacobians. From Figures 2 and 3, it is also seen that the required iteration
number of the method, developed in [35] by using quadratic equation, is quite high,
and its convergence properties are excessively influenced with the system parameters.
Downloaded By: [Hocaoglu, M. Hakan] At: 10:28 11 December 2008

(a)
104

(b)

Figure 2. Variation of the iteration number of the algorithms for: (a) different loading conditions ( D 1; 1:25; : : : ; 2:5) and (b) different R/X ratios (k D
1; 1:25; : : : ; 3).
Downloaded By: [Hocaoglu, M. Hakan] At: 10:28 11 December 2008

(a)
105

(b)

Figure 3. Variation of the iteration number of the algorithms for: (a) different tolerance values (eps D 10 5 ; 10 6 ; : : : ; 10 10 ) and (b) different static load
models.
106 U. Eminoglu and M. H. Hocaoglu

This because of the source bus transferred power updating process, which also causes
the convergence problem in the algorithm when loads are modeled as CI and CZ load in
the network. From Figures 2 and 3, it is seen that the KVL-based algorithms developed
in [17] and [20] also have good convergence ability, and their performances are less
affected with the increase of loading levels, R/X ratios, and the tolerance values when
compared with the other KVL-based algorithms.
The forward/backward sweep algorithms are developed by considered different con-
vergence criteria, such as node voltages, load currents, line losses, or total power mis-
match, and the convergence criteria may affect the convergence ability as a result of it-
eration number of the algorithms. For this reason, the convergence characteristics of
the methods are analyzed for the same convergence criteria (node voltage and load
current mismatch) on a 69-bus distribution test system. First, the convergence ability
of the algorithms is examined for the criteria of node voltage mismatch and variation of
the maximum node voltage mismatch with the iteration number; these are given in the
logarithmic scale in Figure 4(a) for different sweep algorithms. It is noted that most
of the KVL-based algorithms have nearly the same characteristics with the algorithm
Downloaded By: [Hocaoglu, M. Hakan] At: 10:28 11 December 2008

given in [12]; therefore, they are not considered in the comparison. Similarly, quadratic
equation-based algorithms that show nearly the same convergence characteristic with the
algorithms given in [25], [37] and [39] are not given in the figure. From Figure 4(a),
it is seen that the quadratic equation-based algorithms, except the algorithm given in
[35], converged more quickly than the KVL-based algorithm. As a result, they have
faster convergence ability than the KVL-based algorithms when maximum node voltage
mismatch is considered for the convergence criteria. Similarly, the same observation could
be made from Figure 4(b) when the maximum load current mismatch is considered for
the convergence criteria. It is also seen from these figures that the KVL-based algorithm
developed in [20] has fast convergence ability when compared with the other KVL-based
algorithms. Similarly, the algorithm developed in [33] has the best convergence charac-
teristics, while other algorithms developed in [37, 40] have relatively faster convergence
characteristics than the other quadratic equation-based algorithms.

6. Conclusions
In this article, various balanced and unbalanced distribution system load flow algorithms,
based on the forward and/or backward sweeps, are reviewed, and their convergence
ability is compared for different loading conditions, R/X ratios, convergence tolerances,
and different static load models for radial distribution networks. From the analyses, it
is concluded that iteration number of the methods based on the quadratic equation is
generally quite small when compared with the KVL-based sweep algorithms. Moreover,
they are less sensitive to the system parameters (i.e., different loading conditions and
R/X ratios of the networks). On the other hand, the convergence speed of the KCL-
and KVL-based algorithms are heavily under the influence of system parameters.
In addition, it is observed that an increase on voltage dependency of the loads causes
the convergence difficulties on the quadratic equation-based algorithm. This is due to the
variation of loads as a result of branch powers with the node voltage magnitudes at each
iteration. On the contrary, convergence speed of the KVL-based algorithm is less affected
with the increase of the load voltage dependency. It is also observed that the KVL-based
algorithm developed in [20] and also in [17] have fast convergence ability as opposed to
the other KVL-based algorithms. Similarly, the algorithm developed in [33], has the best
convergence characteristics, while other algorithms developed in [26, 37, 40] have fast
Distribution Systems Power Flow Algorithms 107
Downloaded By: [Hocaoglu, M. Hakan] At: 10:28 11 December 2008

(a)

(b)

Figure 4. Variation of maximum node voltage mismatch: (a) maximum node voltage mismatch
and (b) maximum load current mismatch; iteration number for KVL-based algorithms (- - -) and
quadratic equation-based algorithms (——).
108 U. Eminoglu and M. H. Hocaoglu

convergence ability when compared with the other quadratic equation-based algorithms.
Although a comprehensive comparison on the computation time is not undertaken, the
initial results suggest that the computational efficiency of the quadratic equation-based
algorithms are somehow slower than Kirchhoff’s formulation-based algorithms. On the
other hand, the quadratic equation-based algorithms, specially developed in [37, 40], are
more suitable in terms of iteration number as well as computation time when the system
parameters, such as loading level and branch R/X ratio, are increased.

References
1. Stevenson, W. D., Elements of Power System Analysis, India: McGraw-Hill, 1982.
2. Tinney, W. G., and Hart, C. E., “Power flow solutions by Newton’s method,” IEEE Trans.
Power Apparatus Syst., Vol. PAS-86, pp. 1449–1457, 1967.
3. Stott, B., and Alsac, O., “Fast decoupled load flow,” IEEE Trans. Power Apparatus Syst.,
Vol. PAS-93, No. 3, pp. 859–869, 1974.
Downloaded By: [Hocaoglu, M. Hakan] At: 10:28 11 December 2008

4. Tripathy, S. C., Prasad, G. D., Malik, O. P., and Hope, G. S., “Load flow solutions for ill-
conditioned power systems by a Newton-like method,” IEEE Trans. Power Apparatus Syst.,
Vol. PAS-101, No. 10, pp. 3648–3657, 1982.
5. Iwamoto, S., and Tamura, Y., “A load flow calculation method for ill-conditioned power
systems,” IEEE Trans. Power Apparatus Syst., Vol. PAS-100, No. 4, pp. 1736–1743, 1981.
6. Stagg, G. W., and El-Abiad, A. H., Computer Methods in Power System Analysis, New York:
McGraw Hill, 1968.
7. Zhang, F., and Cheng, C. S., “A modified Newton method for radial distribution system power
flow analysis,” IEEE Trans. Power Syst., Vol. 12, No. 1, pp. 389–397, 1997.
8. Da Costa, V. M., Martins, N., and Pereira, J. L., “Development in the Newton Raphson
power flow formulation based on current injections,” IEEE Trans. Power Syst., Vol. 14, No. 4,
pp. 1320–1326, 1999.
9. Garcia, P. A. N., Pereira, J. L. R., Carneiro, S., Da Kosta, V. M., and Martins, N., “Three-phase
power flow calculations using the current injection method,” IEEE Trans. Power Syst., Vol. 15,
No. 2, pp. 508–514, 2000.
10. Lin, W.-M., and Teng, J.-H., “Three-phase distribution network fast-decoupled power flow
solutions,” Elect. Power Energy Syst., Vol. 22, No. 5, pp. 375–380, 2000.
11. Teng, J. H., “A modified Gauss-Seidel algorithm of three-phase power flow analysis in distri-
bution networks,” Elect. Power Energy Syst., Vol. 24, No. 2, pp. 97–102, 2002.
12. Shirmohammadi, D., Hong, H. W., Semlyen, A., and Luo, G. X., “A compensation-based
power flow method for weakly meshed distribution and transmission networks,” IEEE Trans.
Power Syst., Vol. 3, No. 2, pp. 753–762, 1988.
13. Cheng, C. S., and Shirmohammadi, D., “A three-phase power flow method for real time
distribution system analysis,” IEEE Trans. Power Syst., Vol. 10, No. 2, pp. 671–679, 1995.
14. Zhu, Y., and Tomsovic, K., “Adaptive power flow method for distribution systems with
dispersed generation,” IEEE Trans. Power Delivery, Vol. 17, No. 3, pp. 822–827, 2002.
15. Ranjan, R., Venkatesh, B., Chaturvedi, A., and Das, D., “Power flow solution of three-phase
unbalanced radial distribution network,” Elect. Power Compon. Syst., Vol. 32, No. 4, pp. 421–
433, 2004.
16. Ghosh, S., and Das, D., “Method for load-flow solution of radial distribution networks,” IEE
Proc. Generat. Transm. Distrib., Vol. 146, No. 6, pp. 641–648, 1996.
17. Mok, S., Elangovan, S., Longjian, C., and Salama, M., “A new approach for power flow analysis
of balanced radial distribution systems,” Elect. Machines Power Syst., Vol. 28, pp. 325–340,
2000.
18. Thukaram, D., Banda, H. M. W., and Jerome, J., “A robust three-phase power flow algorithm
for radial distribution systems,” Elect. Power Syst. Res., Vol. 50, No. 3, pp. 227–236, 1999.
Distribution Systems Power Flow Algorithms 109

19. Kersting, W. H., Distribution System Modeling and Analysis, Boca Raton, FL: CRC Press,
2002.
20. Liu, J., Salama, M. M. A., and Mansour, R. R., “An efficient power flow algorithm for
distribution systems with polynomial load,” Int. J. Elect. Eng. Educat., Vol. 39, No. 4, pp. 371–
386, 2002.
21. Teng, J. H., “A network-topology based three-phase load flow for distribution systems,” Proc.
Natl. Sci. Council ROC (A), Vol. 24, No. 4, pp. 259–264, 2000.
22. Teng, J. H., “A direct approach for distribution system load flow solutions,” IEEE Trans. Power
Delivery, Vol. 18, No. 3, pp. 882–887, 2003.
23. Aravindhabuba, P., Ganapathy, S., and Nayar, K. R., “A novel technique for the analysis of
radial distribution systems,” Elect. Power Energy Syst., Vol. 23, No. 3, pp. 167–171, 2001.
24. Augugliaro, A., Dusonchet, L., Favuzza, S., Ippolito, M. G., and Sanseverino, E. R., “A new
backward/forward method for solving radial distribution network with PV nodes,” Elect. Power
Syst. Res., Vol. 78, No. 3, pp. 330–336, March 2008.
25. Cespedes, G. R., “New method for the analysis of distribution networks,” IEEE Trans. Power
Delivery, Vol. 5, No. 1, pp. 391–396, 1990.
26. Haque, M. H., “Load flow solution of distribution systems with voltage dependent load
Downloaded By: [Hocaoglu, M. Hakan] At: 10:28 11 December 2008

models,” Elect. Power Syst. Res., Vol. 36, No. 3, pp. 151–156, 1996.
27. Haque, M. H., “Efficient load flow method for distribution systems with radial or mesh
configuration,” IEE Proc. Generat. Transm. Distrib., Vol. 143, No. 1, pp. 33–38, 1996.
28. Haque, M. H., “A general load flow method for distribution systems,” Elect. Power Syst. Res.,
Vol. 54, No. 1, pp. 106–114, 2000.
29. Asfari, M., Singh, S. P., Raju, G. S., and Rao, G. K., “A fast power flow solution of radial
distribution networks,” Elect. Power Compon. Syst., Vol. 30, No. 3, pp. 1065–1074, 2002.
30. Ranjan, R., and Das, D., “Simple and efficient computer algorithm to solve radial distribution
networks,” Elect. Power Compon. Syst., Vol. 31, No. 1, pp. 95–107, 2003.
31. Das, D., Nagi, H. S., and Kothari, D. P., “Novel method for solving radial distribution
networks,” IEE Proc. Generat. Transm. Distrib., Vol. 141, No. 4, pp. 291–298, 1994.
32. Das, D., Kothari, D. P., and Kalam, A., “Simple and efficient method for load flow solution of
radial distribution networks,” Elect. Power Energy Syst., Vol. 17, No. 5, pp. 335–346, 1995.
33. Eminoglu, U., and Hocaoglu, M. H., “A new power flow method for radial distribution systems
including voltage dependent load models,” Elect. Power Syst. Res., Vol. 76, No. 1–3, pp. 106–
114, 2005.
34. Baran, M. E., and Wu, F. F., “Optimal sizing of capacitors placed on a radial distribution
system,” IEEE Trans. Power Delivery, Vol. 4, No. 1, pp. 735–742, 1989.
35. Mekhamer, S. F., Soliman, S. A., Moustafa, M. A., and El-Hawary, M. E., “Load flow solution
of radial distribution feeders: A new contribution,” Elect. Power Energy Syst., Vol. 24, No. 9,
pp. 701–707, 2002.
36. Baran, M. E., and Wu, F. F., “Network reconfiguration in distribution systems for loss reduction
and load balancing,” IEEE Trans. Power Delivery, Vol. 4, No. 2, pp. 1401–1407, 1989.
37. Luo, G. X., and Semlyen, A., “Efficient load flow for large weakly meshed networks,” IEEE
Trans. Power Syst., Vol. 5, No. 4, pp. 1309–1316, 1990.
38. Jasmon, G. B., and Lee, L. H. C., “Distribution network reduction for voltage stability analysis
and load flow calculations,” Elect. Power Energy Syst., Vol. 13, No. 1, pp. 9–13, 1991.
39. Rajicic, D., Ackovski, R., and Taleski, R., “Voltage correction power flow,” IEEE Trans. Power
Delivery, Vol. 9, No. 2, pp. 1056–1062, 1994.
40. Satyanarayana, S., Ramana, T., Sivanagaraju, S., and Rao, G. K., “An efficient load flow
solution for radial distribution network including voltage dependent load models,” Elect. Power
Compon. Syst., Vol. 35, No. 5, pp. 539–551, 2007.
41. Baran, M. E., and Staton, E. A., “Distribution transformer models for branch current based
feeder analysis,” IEEE Trans. Power Syst., Vol. 12, pp. 698–703, 1997.
42. Willis, H. L., and Scott, W. G., Distributed Power Generation: Planning and Evaluation, New
York: Marcel Dekker, Power Engineering Series, 2000.
110 U. Eminoglu and M. H. Hocaoglu

43. Nagendra Rao, P. S., and Deekshit, R. S., “Radial load flow for systems having distributed
generation and controlled Q sources,” Elect. Power Compon. Syst., Vol. 33, No. 6, pp. 641–655,
2005.
44. Chen, T. H., Chen, M. S., Inoue, T., Kotas, P., and Chebli, E. A., “Three-phase cogenerator
and transformer models for distribution system analysis,” IEEE Trans. Power Delivery, Vol. 6,
No. 4, pp. 1671–1681, 1991.
45. Saadat, H., Power System Analysis, New York: McGraw-Hill, 2002.
46. Canizares, C. A., Mithulananthan, N., Salama, M. M. A., and Reeve, J., “PFLOW: Continuation
and direct methods to locate fold bifurcations in AC/DC/FACTS power systems,” available at:
http://www.power.uwaterloo.ca.
Downloaded By: [Hocaoglu, M. Hakan] At: 10:28 11 December 2008

View publication stats

You might also like