A Three-Phase Power Flow Solution Method for

Unbalanced Distribution Networks

Salem Elsaiah, Student Member, IEEE, Mohammed Benidris, Student Member, IEEE, and
Joydeep Mitra, Senior Member, IEEE
Electrical and Computer Engineering Department- Michigan State University
East Lansing, Michigan 48824, USA
email: {elsaiahs,benidris,[email protected]}
Abstract This paper introduces a methodology for solving
power flow problem of both balanced and unbalanced radial
distribution networks. Two matrices
the line Primitive
Impedance Matrix (PIM) and the Branch-Current Matrix
(BCM) are developed to obtain the power flow solution for the
given network. The special topological characteristics of the
distribution network have been taken into consideration while
developing these matrices. Due to its distinct features, the
optimal node ordering which is necessary in the forwardbackward substitution method, LU-decomposition, or Y-bus
formulation and factorization are no longer needed in the
proposed approach. The proposed method has been tested on a
balanced 33-bus network and an 8-bus unbalanced network.
The results of the proposed method were found similar when
compared with some other methods presented in the literature,
while requiring less computational time.

I.

INTRODUCTION

Power flow solution is a crucial tool for the steady-state

analysis of any electric power system. The main objective of
the load flow solution is to determine the voltage phase and
magnitude at each bus as well as the real and reactive powers
flowing in each line for specific loading conditions. In the
literature three conventional methods have been widely used
to solve the power flow problem. These methods are the
Gauss-Seidel (GS) iterative method, the Newton-Raphson
(NR) method, and the fast-decoupled (FDLF) method [1,2].
It has been observed, however, that these conventional power
flow methods, which were essentially developed for solving
power flow problem at the transmission level, can encounter
convergence problems when applied directly to distribution
systems unless some treatments are introduced into them. The
problem of convergence can be attributed in part to the
special features of the electric power distribution network.
Some of these prominent features include, but are not limited
to, the following:

Radial structure or nearly radial structure (weakly

meshed).
High
ratio that may cause the NR method and
the FDLF methods to diverge.

Untransposed or rarely transposed lines that make

the ignorance of the mutual coupling between
phases unacceptable for certain applications.
Unbalanced loads along with single-phase and
double-phase laterals.
Unbalanced distributed loads.

These features combined with the large number of nodes and

branches of the distribution network make such a system fall
into the category of ill-conditioned power networks. These
ill-conditioned systems should be analyzed on three-phase
basis rather than single-phase basis.
A significant amount of research has been done to
develop robust and reliable distribution power flow
algorithms over the past two decades [3-5]. Such algorithms
are required by several real-time engineering applications that
include the operational and operating stages as reported in the
IEEE tutorial course on Power Distribution Planning [6]. For
example, distribution feeder reconfiguration, optimum
location of dispersed sources, and reactive power control and
optimization need robust, flexible, and efficient power flow
programs. Therefore, an acceptable power flow program
should be quick enough to deal with real-time applications in
which repeated power flows are required, the memory space
should also be suitable, and it should attain some sort of
versatility and simplicity as well.
Zimmerman and Chiang presented a fast decoupled load
flow method in [7]. In this approach, a set of nonlinear power
mismatch equations are formulated and then solved by
Newtons method. The advantage of this method was that it
ordered the laterals instead of buses; hence the problem size
has been reduced to the number of systems laterals. Use of
laterals as variables instead of nodes makes this algorithm
more efficient for a given system topology, yet it may add
some difficulties if the network topology is changed
regularly, which is the case in distribution networks as a
result of the switching operation. In [8], Zhang and Cheng
used Newtons method to solve the distribution power flow
problem. Even though the Jacobian matrix has to be

computed only once, shunt capacitors were not included in

this method. Baran and Wu [9] developed a method for
solving distribution system power flow by solving three
equations representing the voltage magnitude, real power, and
reactive power. In this approach, only simple algebraic
equations are utilized in the developing of the Jacobian
matrix and the power mismatches. The formulation of the
Jacobian matrix in every iteration, however, turns to be
computationally cumbersome in terms of execution time and
storage requirement.
Kersting and Mandive [10] and Kersting [11] suggested a
methodology to solve the distribution system power flow
problem based on Ladder-Network theory in the iterative
routine. This approach has the advantage that it is derivativefree and uses basic circuit theory laws. However, ladder
network method computes the bus voltages twice for a single
iteration. Goswami and Basu presented a direct approach to
obtain a solution for the distribution system power flow for
both radial and meshed networks in [12]. The method was
applied also for balanced and unbalanced practical networks.
The main advantage of this approach is that convergence is
guaranteed for any realistic distribution system. The
disadvantage, however, is that no node in the system can
serve as a junction for more than three network branches,
which limits the use of this approach.
A compensation based method is proposed in [13] to solve
the balanced distribution system power flow problem based
on Kirchhoffs laws. In this approach, a simple numbering
scheme is implemented to enhance the computational efforts.
Then, forward sweep/backward sweep is applied to obtain the
power flow solution. The same method was extended to
include the weakly meshed networks. In doing so, the grid
was first broken to a number of points- breakpoints- to
have a simple radial network. The radial network was then
solved by the direct application of Kirchhoffs current and
voltage laws. The effectiveness of this method, however,
diminishes as the number of breakpoints goes up. As a result,
the application of this method to the weakly meshed networks
was practically restricted. In [14], Cheng and
Sharimohammadi have directly extended the work done in
[13] to include the case of unbalanced distribution networks.
The method emphasized on the modeling of various
distribution system components and was successfully applied
for large systems.
Teng has proposed numerous methods for solving
distribution load flow [15-17]. One method [15] is based on
the optimal ordering scheme and triangular factorization of
the bus-admittance matrix, (Ybus). Inherently, this approach is
a combination of the modified GS-method and the Z-implicit
method. Basically, the approach used the Y-bus factorization,
hence large computational time is required by this method.
The method presented in [16] is based on the equivalent
current injection technique. One advantage of this method is
the constant Jacobian matrix that needs to be converted only
once. The reactances of the distribution network feeders have
not been taken into consideration by assuming that
.
However, distribution networks are characterized by a wide
range of resistance and reactance which indicates that the

method may fail in the case of

. A direct approach to
obtain the distribution power flow solution was presented in
[17] by same author. Two matrices, the bus injection to
branch current BIBC and the branch current to bus voltage
BCBV and direct matrix multiplication are used to obtain the
distribution power flow solution, DLF. The solution involves
direct matrix multiplication, thus large memory space is
needed especially if it applied for large-scale distribution
systems. Further, these matrices contain many zero elements,
so the memory space is not economically utilized, and large
CPU time is required. Prakash and Sydulu [18] have
introduced certain modifications to previous method so that
the DLF matrix is obtained directly without the formulation
of either the BIBC or the BCBV matrices. However,
balanced conditions are assumed, which is not the case in
realistic systems. The network topology based method has
also been modified to include the weakly meshed networks in
[19].
In this paper a methodology for solving power flow
problem for radial distribution networks is presented. Two
matrices the line Primitive Impedance Matrix (PIM) and the
Branch-Current Matrix (BCM) are developed to obtain the
power flow solution. In this method, optimal node ordering
which is necessary in the forward-backward substitution
method, LU-decomposition, or Y-bus formulation and
factorization are no longer needed. The proposed method has
been tested on several distribution networks.
This paper is organized as follows: in section II various
components models are given. Section III described the
solution methodology. Two case studies are explained in
Section IV. Test results and discussion are given in Section
V. Concluding remarks and future work are finally presented
in Section VI.
II.

COMPONENTS MODELING

The models of various distribution network components have

a great influence on the accuracy of the results of the power
flow program. This section discusses briefly the network
components that were used in the presented power flow
solution.
A. Modeling of Unbalanced Three-Phase Line Section
A four-wire distribution network is assumed in this paper as
this type of systems is widely used worldwide. Fig.1 shows a
three-phase line section connected between two buses
and
.
Zaa

VA

Va
Zab

Zac

VB

Zbb

VC

Vb
Zbc

Zan
Zbn

Zcc

Vc
Zcn

VN

Znn

Fig. 1 Three-phase line section model

Vn

The parameters of the line can be found by the method

developed by Carson and Lewis [6]. A 4 4 matrix that takes
the effect of the self-and-mutual coupling between phases can
be expressed as,

i.
-connected loads
Fig. 2 shows a three-phase Y-connected unbalanced load
model.
ILc

c
Vcn
b
ILb

(1)

Vbn
SLc

SLb
n

However, it is convenient to represent (1) as a 3 3 matrix

instead of the 4
4 matrix by using Krons method. The
effect of the ground conductor is still included in the resultant
matrix, that is
(2)

Van

SLa

a
I La

Fig.2 Unbalanced Y-connected load

From this figure, the load current injections at the
bus for
three-phase Y-connected or single-phase connected line-toneutral can be expressed as,

Now by applying the KVL to the circuit model of Fig.1, the

relationships between the bus voltages and branch currents
can be simply written as,
=

(3)

It should be noted, however, that single phase and two phase

line sections are most common in distribution networks.
Hence, in this research for any phase that fails to present, the
corresponding row and column in (3) will have zeros entries.
For example, for two phase line section with and phase,
equation (3) will be reduced to,
(4)

(6)

ii.
-connected loads
Referring to Fig. 3 shown below, the current injections at the
bus for three-phase loads connected in or single-phase
connected line-to-line can be expressed as,
c
Vbc

SLb

b
ILbc

Also, for single phase line section, equation (3) will be

reduced to,

ILab
Vca

Vab
S La

(5)

SLc
ILca

A. Modeling of Loads
Loads in distribution network can be spot loads or distributed
loads or both.

Fig.3 Unbalanced -connected load

1. Spot Loads
Equivalent current injection technique was used in this
research to represent the distribution network loads. This is
attributed in the first place to the nature of the loads in the
distribution networks which are inherently unbalanced loads.
It is assumed that all three-phase loads are connected or ,
and all single-phase and two-phase loads have connections
between line and neutral and line-to-line, respectively.
Further, a constant power model at each bus was assumed
during the realization of this work.

(7)

2. Distributed Loads
Distributed loads are assumed to be uniformly distributed
along the line and lumped at point F, the midpoint of Fig. 4.

However, due to the inclusion of this factious bus, the

computational speed will probably be increased. Therefore, a
load transfer technique based on KVL and KCL has been
used to reduce the power flow equations. In Fig. 4 shown
below,
represents the line impedance while
represents
the load.
ZL/2

Bus m

AC

Bus F

ZL/2

ImF

VF

(9)

Bus n

Where,
Vn

AC

Vm

From Fig. 6, the current injections due to line charging at bus

can be written as,

(10)

InF

SL

III.

Fig. 4 Representation of distributed load

After the load transfer is applied, the distributed loads can be
represented as shown below in Fig. 5,
Bus m
Vm

Bus n
Vn
ILn

ZL

ILm
SL/2

ALGORITHM DEVELOPMENT

The proposed approach is based on the development of

two matrices the line Primitive Impedance Matrix (PIM) and
the Branch-Current Matrix (BCM) to obtain the power flow
solution. This section describes the development procedure of
the proposed method.
A. Equation Development
Consider the radial distribution system with
and -branches shown below in Fig. 7

SL/2

A
Z34

Fig. 5 Load transfer of distributed loads

-buses
5

Z 45

B3

B4

B. Modeling of Capacitors
Capacitors are assumed of Y-connected with ground
conductor. The current injections are given as [20],

ABC

IL4

ABC

Z12

IL5

Z23

B1

BC

B2

SS Bus
IL2
B5

Z 36

IL3
6

(8)
IL6

Fig.7 6-bus distribution system

C. Modeling of Shunt Capacitances
Charging capacitors have a noticable influence on systems
voltage profile. In section A, a model for unbalanced threephase line section is introduced. This model can be improved
by including the line shunt (charging) capacitors. A model
takes into account the effect of the self and mutual
capacitance is shown in Fig. 6.
Bus m
Va

Bus n
Va

yab/2

yab/2

Vb

Vb
yca/2

ybc/2

yca/2

ybc/2

Vc
ycc/2

Vc
ybb/2

yaa/2

yaa/2

ybb/2

Fig. 6 Shunt capacitance line model

ycc/2

The complex load at bus can be expressed in terms of the

real and reactive power as,
(11)
Then, the corresponding current injected at the same node
can be further described as,
(12)
The relationships between the bus current injections,
branch currents, and voltage at the various system buses can
be obtained by applying Kirchhoffs current law KCL and
Kirchhoffs voltage law KVL, respectively. The branch
currents can be expressed by equivalent current injections as,

3. Calculate node currents from the following equation,

(13)
Where is the node number, is load complex power and
is the node shunt admittance
4. Calculate the branch currents vector as follows:
The branch currents can generally be described in vector form
as,

where

is the nodes injected currents vector,

(14)
Similarly, the branches primitive impedance vector between
buses and can be given as,
(15)
Now, the relationship between branch currents and bus
voltages can be determined by applying the KVL, as will be
given later. The step-by-step development of the BrachCurrent-Matrix BCM is extensively explained in section B.

5. Calculate the branches voltage drops,

Where
is the branch number,
is current matrix of
branch and
is primitive impedance matrix of branch
6. Starting from the substation nodes towards the subsequent
nodes, calculate the new node voltages as follows: for nodes
and which are connected through branch
,
7. After calculating all node voltages, calculate the maximum
error between the new voltages and the voltages of the
previous step.

B. Algorithm Development
BCM Formulation:
1. Number the system buses from
to
where is the
number of system buses and let 1 denotes the substation bus.
2. Label the branch currents such that the branch between
buses and is
.
3. Form the branch currents matrix,
,
as follows:
i. Start from bus
phase
, fill the element
in the branch current matrix with 1.
ii. Search for bus phase C in the line section data in the
receiving end buses column and determine the number
of the bus that is connected to bus phase
in the
sending end buses column, e.g., .
iii. If
, fill the element
in the branch
current matrix with
and let
and go to ii,
otherwise go to iv.
iv. Reduce by one and check if
, repeat substeps i, ii and iii, otherwise stop and go to sub-step v
v.
Repeat for phases and such that for phase use
and
; for
phase
use
and
.
Power Flow Solution:
1. Form the branch currents matrix,
2. For flat start, all nodes voltages are assumed to be
except for the voltages of the substation nodes
which are specified.

If the maximum error is less than or equal to the specified

tolerance, go to step 9, otherwise continue to step 8.
8. Use new voltage values and repeat steps 3 to 7.
9. Print out the power flow results.
Bus
2

A B C
A 1 0 0

B1 B 0 1 0

C 0 0 1

A 0 0 0

B2 B 0 0 0
C 0 0 0

A 0 0 0

BCM B3 B 0 0 0

C 0 0 0

A 0 0 0

B4 B 0 0 0
C 0 0 0

A 0 0 0

B5 B 0 0 0
C 0 0 0

IV.

Bus
3

A B C
1 0 0

0 1 0
0 0 1

1 0 0

0 1 0
0 0 1

0 0 0

0 0 0
0 0 0

0 0 0

0 0 0
0 0 0

0 0 0

0 0 0
0 0

Bus
4

A B C
1 0 0

0 0 0
0 0 0

1 0 0

0 0 0
0 0 0

1 0 0

0 0 0
0 0 0

0 0 0

0 0 0
0 0 0

0 0 0

0 0 0
0 0 0

Bus
5

A B C
1 0 0

0 0 0
0 0 0

1 0 0

0 0 0
0 0 0

1 0 0

0 0 0
0 0 0

1 0 0

0 0 0
0 0 0

0 0 0

0 0 0
0 0 0

Bus
6

A B C
0 0 0

0 1 0
0 0 1

0 0 0

0 1 0
0 0 1

0 0 0

0 0 0
0 0 0

0 0 0

0 0 0
0 0 0

0 0 0

0 1 0
0 0 1

CASE STUDIES

For any new method, it is very important to check its validity

with other methods that exist in the literature. To verify the
convergence characteristics and the accuracy of the proposed
method, it is implemented on balanced and unbalanced
distribution systems. The first test system is taken from Ref.
[21] and is given below in Fig. 8 for explanation purposes.

The system nominal voltage is 12.66KV with two feeder

substations, 33 bus, and 5 tie lines. The total substation loads
for the base configuration are 5084.26 KW and 2547.32
KVARs. The branch data, loading, and voltage profile for the
base configuration can be found in the same reference.
25
24

27

28

29

30

31

32

33

10

11

12

13

14

23

1
SS Bus

26

2
3

15

16

17

18

19
20
21
22

Fig. 8 A 33 bus distribution system

The second test system was acquired from Ref.[19] and is
shown below in Fig. 9. In this system, the single-phase and
double-phase laterals have been lumped to form unbalanced
loads for testing purposes. This system is characterized by its
high
ratio (app.1.8), which makes it an ill-conditioned
distribution network and should be analyzed on three-phase
basis rather than single-phase ones. The base values for this
system are chosen to be 14.4 KV and 100 KVA.
3
3a
SS Bus

5b

1a
2a
4
1b
4b

2b
1c
2c

4c
7

6c
7c

8
8c

Fig. 9 An 8 bus distribution system

V.

TEST RESULTS

The proposed three-phase power flow solution is carried out

using Matlab 7.9 environment with a tolerance of 0.0001.
Two methods have been used in the tests for comparison
purposes. These methods are,
Method 1: The proposed method.
Method 2: The method reported by Jen-Hao Teng [19].
The above two methods were compared on the basis of their
performance parameters.
A. Power Flow Solution of the 33-bus System
The power flow solution of the 33-bus system is given in
Table I. It is observed that the proposed method took three

iterations to converge for an accuracy of 0.0001. Method II

took same number of iterations, however the proposed
method has utilized less memory space as only dot product
was performed to obtain the power flow solution. The
maximum deviation is almost zero for five decimal places.
B. Power Flow Solution of the 8-System
The power flow results of the 8-bus system of Fig.9 are given
in Table II. The method converged in three iterations.
Method II took three iterations to converge too. The
maximum deviations in both voltage magnitude and angle
from Method 2 are 0.0001 and 0.0020, respectively. Though
Method II has the same number of iterations, the memory
space and, hence, the computational time taken by this
method is higher than the proposed method since Method II
involves the formulation of a matrix whose dimensions of
to obtain the distribution power flow
solution. The product of the PIM and BCM in the presented
method, however, results in a matrix with dimensions of
to obtain the distribution power flow solution. This
means that the saving in memory space memory will be
. For example, for an 11-bus system, the
matrix that is involved in the power flow solution of Method
2 will have dimensions of
. However, for our
approach i.e. Method 1, the matrix will have dimensions of
, therefore, the saving in memory space is
.
C. Robustness Test of the Proposed Method
Divergence of the solution of the power flow problem usually
occurs when certain ill-conditioned are presented in the
system. As it is mentioned earlier in Section I, distribution
networks with high
ratio are inherently ill-conditioned
power networks. Ill-conditioned often takes place when the
system involves short-lines, which is the case in this paper, or
even very long lines. In order to prove that the proposed
method can be used for such harsh situations, a robustness
test was done on the distribution network depicted in Fig. 8.
Firstly, selected primitive impedances were taken from each
section of the system sections at random basis. Secondly,
among the selected impedances, one is divided by a factor of
10 to represent a short-line while the other is multiplied by 10
to have a very long line representation. For example, sections
2-4, and 4-5 are multiplied by 10 whereas sections 4-6 and 78 are divided by 10. Test result shows that the method
converged slower than the base case by an amount of 21.5%,
which is an acceptable figure in such severe conditions and
the method has proven to be suitable for ill-conditioned
distribution networks.
VI.

CONCLUSION

An approach for solving radial distribution system power

flow problem has been proposed in this paper. The method
utilizes only basic circuit theory fundamentals. Unlike other
traditional methods, the presented work has not utilized any
Y-bus formulation or factorization. Also, the LUdecomposition has not been used in the presented study. The

approach has proven to be economically-effective as it

utilizes less memory space to store the necessary data. The
method is tested on balanced and unbalanced systems and the
test results were found to be closed to other methods in the
same field. It is suggested that the proposed method could be
used for distribution network planning and expansion in
which repeated power flows are required. The method could
also be modified to accommodate the changes that usually
happen in the existing network by modifying the PIM and the
BCM matrices. Moreover, the proposed approach can easily
be modified to incorporate constant current, constant
impedance or exponential load models. The application of
the proposed method to include large systems is in progress
and results will be reported in due course.

TABLE I I LOAD FLOW RESULTS OF THE 8-BUS SYSTEM

Bus
Number
&
Phase
1-A
1-B
1-C
2-A
2-B
2-C
3-A
4-B
4-C
5-B
6-C
7-C
8-C

TABLE I LOAD FLOW RESULTS OF THE 33-BUS SYSTEM

Node no.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
KW-losses
KVAr-losses
P-supplied
Q-supplied

Method 1
Node Vol.(p.u)
1.00000
0.99701
0.98288
0.97537
0.96795
0.94947
0.94594
0.93229
0.92596
0.92010
0.91923
0.91771
0.91154
0.90925
0.90782
0.90644
0.90439
0.90378
0.99649
0.99291
0.99220
0.99157
0.97930
0.97263
0.96930
0.94754
0.94497
0.93353
0.92531
0.92175
0.91759
0.91668
0.91639
210.987
143.128
3925.99
2443.13

Method 2
Node Vol.(p.u)
1.00000
0.99701
0.98288
0.97537
0.96795
0.94947
0.94594
0.93229
0.92596
0.92010
0.91923
0.91771
0.91154
0.90925
0.90782
0.90644
0.90439
0.90378
0.99649
0.99291
0.99220
0.99157
0.97930
0.97263
0.96930
0.94754
0.94497
0.93353
0.92531
0.92175
0.91759
0.91668
0.91639
210.987
143.128
3925.99
2443.13

Method 1

Method 2

(p.u.)

Angle
(Deg.)

(p.u.)

Angle
(Deg.)

1.0000
1.0000
1.0000
0.9839
0.9711
0.9697
0.9832
0.9652
0.9668
0.9640
0.9649
0.9683
0.9671

0.0000
-120.0
120.0
0.1830
-119.76
119.97
0.1790
-119.73
119.93
-119.74
119.92
119.96
119.96

1.0000
1.0000
1.0000
0.9839
0.9712
0.9697
0.9832
0.9652
0.9669
0.9640
0.9650
0.9683
0.9671

0.0000
-120.0
120.0
0.1830
119.76
119.97
0.1780
-119.73
119.93
-119.74
119.92
119.96
119.95

REFERENCES
[1]
[2]
[3]

[4]

[5]

[6]
[7]

[8]

[9]

[10]

[11]

[12]
[13]

[14]

[15]

B. Stott, Review of load-flow calculation methods, Proc. IEEE, vol.

62, no. 7, pp. 916929, Jul. 1974.
B. Stott, Decoupled Newton load flow, IEEE Trans. Power App.
Syst., vol. PAS-91, no. 5, pp. 19551959, Sep. 1972.
M. F. AlHajri and M.E. El-Hawary,Exploiting the radial distribution
structure in developing a fast and flexible radial power flow for
unbalanced three-phase networks, IEEE Trans.Power Del., vol.25,
no.1, pp 378-389, Jan. 2009.
T.-H. Chen, M.-S. Chen, K.-J. Hwang, P. Kotas, and E. A. Chebli,
Distribution system power flow analysisA rigid approach, IEEE
Trans. Power Del., vol. 6, no. 4, pp. 16711678, July 1991.
T.-H Chen and N.-C. Yang,Three-phase power-flow by direct ZBR
method for unbalanced radial distribution systems,IET Gener. Transm.
Distrib., vol.3, Iss.10, pp. 903-910, March 2009.
IEEE Tutorial Course on Power Distribution Planning.
R. D. Zimmerman and H.D. Chiang,Fast decoupled load flow for
unbalanced radial distribution systems,IEEE-PES Winter meeting,
Paper no.95, New York, 1995.
F. Zhang and C. S. Cheng, A modified Newton method for radial
distribution system power flow analysis, IEEE Trans. Power Syst.,
vol.12, no. 1, pp. 389397, Feb. 1997.
M.E. Baran and F.F. Wu,Optimal capacitor placement on radial
distribution systems, IEEE Tranactions On Power Delivery vol.4,
no.1, 1989, pp 725-734.
W. H. Kersting and D. L. Mendive, An application of ladder network
theory to the solution of three-phase radial load-flow problems,
presented at the IEEE Winter Power Meeting Conf., New York, 1976.
W. H. Kersting, A method to teach the design and operation of a
distribution system, IEEE Trans. Power App. Syst., vol. PAS-103, no.
7,pp. 19451952, Jul. 1984.
S. K. Goswami and S. K. Basu,Direct solution of distribution
systems, IEE Proc. vol. 188, no.1, pp 78-88, 1991.
D. Shirmohammadi, H. W. Hong, A. Semlyen, and G. X. Luo, A
compensation- based power flow method for weakly meshed
distribution and transmission networks, IEEE Trans. Power Syst., vol.
3, no. 2,pp. 753762, May 1988.
D. Shimohammadi, and C. Cheng, A Three-Phase power flow method
for real-time distribution systems analysis, IEEE Tran. Power Syts.,
vol.10,no.2, May 1995.
J. H. Teng, A Modified Gauss-Seidel Algorithm ofThree-phase power
flow Analysis in Distribution Networks,Electrical Power and Energy
Systems, vol. 24, pp. 97-102, 2002.

[16] W. M. Lin and J. H. Teng, "Three-Phase distribution networks fast

decoupled power flow solutions," Electric Power and Energy Systems,
vol. 22, no. 5, pp.375-380, Jun. 2000.
[17] J. H. Teng, A network-topology based three-phase load flow for
distribution systems, Proceedings of National Science Council ROC
(A), Vol. 24, no. 4, pp. 259- 264,2000.
[18] K. Prakash and M. Sydulu, An effective topological and primitive
impedance based distribution load flow method for radial distribution
systems, TENCON 2008 - IEEE Region 10 Conference pp 10441049, 2008.
[19] J. H. Teng, "A direct approach for distribution system load flow
solution," IEEE Trans. Power Del., vol. 18, no. 3, pp. 882-887, July
2003.
[20] J. Vieira Jr, W. Freitas, and A. Morelato, Phase-decoupled method for
three-phase power-flow analysis of unbalanced distribution systems,
IEE Proc.-Gener., Transm., Distrib., vol. 151, no. 5, pp. 568-574, Sep.
2004.
[21] M. E. Baran and F. F. Wu, "Network reconfiguration in distribution
systems for loss reduction and load balancing," IEEE Trans. Power
Del., vol. 4, no. 2, pp. 1401-1407,1989.

VII. BIOGRAPHIES
Salem Elsaiah (S09) received Bsc. (Engrg) degree in Electrical and
Electronic Engineering in 1999 from the University of Garyounis, Benghazi,
Libya. After his graduation, he joined the department of Electrical and
Electronic Engineering of the same university as an assistant lecturer. He
earned Msc. Degree in Electrical Engineering specialized in the Electrical
Machines and Power in 2006 from Garyounis University too. Currently, he is
pursuing a Ph.D program at Michigan State University, USA. His area of
interest includes distribution system design and planning, and power system
reliability. (email: [email protected])
Mohammed Benidris (S10) received Bsc. (Engrg) degree in Electrical and
Electronic Engineering in 1998 from the University of Garyounis, Benghazi,
Libya. After his graduation, he joined the department of Electrical and
Electronic Engineering of the same university as an assistant lecturer. He
earned Msc. Degree in Electrical Engineering specialized in Power System in
2005 from Garyounis University too. Currently, he is pursuing a Ph.D
program at Michigan State University, USA. His area of interest includes
power system reliability, and power system planning. (email:
[email protected])
Joydeep Mitra (S 94, M 97, SM 02) is an Associate Professor of
Electrical Engineering at Michigan State University, East Lansing. Prior to
this, he was Associate Professor at New Mexico State University, Las
Cruces, Assistant Professor at North Dakota State University, Fargo, and
Senior Consulting Engineer at LCG Consulting, Los Altos, CA. He received
a Ph.D. in Electrical Engineering from Texas A&M University, College
Station, and a B.Tech. (Hons.), also in Electrical Engineering, from Indian
Institute of Technology, Kharagpur. His research interests include power
system reliability, distributed energy resources, and power system planning.
(Email: [email protected])