2006 IEEE PES Transmission and Distribution Conference and Exposition Latin America, Venezuela

A New Approach to Load Flow Solutions for

Radial Distribution System
T. Thakur, and Jaswanti Dhiman

solutions for radial distribution system, in which the choice of the

switches to be opened is based on the calculation of voltage at the
buses, real and reactive power flowing through lines, real power
losses and voltage deviation, using distribution load flow (DLF)
program. In the process of load flow calculation, two developed
matrices bus injection to bus current (BIBC) and branch current
to bus voltage (BCBV), and a simple matrix multiplication are
used to obtain load flow solutions. The solution converges very
early on; therefore execution time is very small. Particular
attention is paid to the relationship between electrical parameters
of the distribution system and the mathematical parameters that
influence the convergence properties of the algorithm. The
results reveal the speed and the effectiveness of the proposed

Index Terms---Distribution network reconfiguration, Loss

minimization, Power flow control, Radial distribution network
(RDN).

I. INTRODUCTION

OSS minimization in power distribution system is one of

the biggest challenges before power engineers.
Transmission and distribution losses are as high as 20 to 30
percent of total power generation. Therefore, the challenge is
more pronounced in case of distribution systems. Basic reason
behind these huge power losses is resistive loss, as well as
distribution systems are operated at much lower voltages as
compared to transmission systems. So, in totality, we can say
that, optimal operation of a distribution power networks has
become an engineering challenge.
In real time applications that are Supervisory Control and
Data Acquision System (SCADA), Distribution Automation
(DA) for the management of radial distribution system (RDS)
such as network optimization, VAR planning, Switching, state
estimation and so forth require a robust and efficient load flow
method. The traditional load flow methods such as GaussSeidel and Newton Raphson fail to meet the above
requirement. Therefore, there is need to develop a load flow
solutions to meet the properties of RDS with less
computational time.
Several load flow algorithms specially designed for
distribution systems have been proposed and published so far

___________________________________
T.Thakur and Jaswanti are with the Department of Electrical Engineering,
Punjab Engineering College (Deemed University), Chandigarh, India -160012
(e-mail: tilak20042005@yahoo.co.in; jaswanti98@yahoo.co.in)

1-4244-0288-3/06/$20.00 2006 IEEE

time to time. In [1]-[5] optimal load flow methods for losses

reduction in distribution system are proposed. In [6] a simple,
fast and approximate heuristic based algorithm for
determining the minimum loss configuration of RDN.
Numerous other [7]-[17] reported works on load flow are for
radial system with a focus on reduction of Kilowatt losses,
and/or load balancing in feeder reconfiguration, energy
restoration and combinatorial optimization problem by using
mathematical programming and traditional Artificial
Intelligent approaches on a similar problem with improvement
in loss reduction in less computational time. The commonly
used methods [18] using Gauss Implicit Z-matrix does not
exploits the radial and weakly meshed network structure of
distribution systems which requires the solution of a set of
equations of the size proportional to the number of buses.
However, [19] provides compensation based technique to
solve distribution load flow problems, where branch power
flow rather than branch current were used in improved version
in new data format. By using the special topological
characteristics of distribution network, the bus injection to
branch current matrix and the branch current to bus voltage
matrix and a simple multiplication are used to obtain the load
flow problem solutions technique [20]. This has been applied
for two different topologies under distribution reconfiguration
[21] for loss minimization.
Therefore, the paper presents a new approach load flow
program to solve the load flow problem in radial distribution
networks, in which the choice of the switches to be opened is
based on the calculation of voltage at the buses, real and
reactive power flowing through lines, real power losses and
voltage deviation, using distribution load flow (DLF)
program. An IEEE 33-bus radial distribution test system is
taken as a study system for performing the test of DLF
program. The proposed load flow method has been found to
give better result than those obtained by some other recent
methods reported in [21]. The results reveal the speed and the
effectiveness of the proposed method for solving the
problems.
II. MATHEMATICAL PROBLEM FORMULATION
A. Computation of Voltages at the Buses
In order to obtain load flow solutions, first objective is to
obtain voltages at the buses. If Vk is the voltages of the buses
after kth iteration, then voltages at the buses after (k+1)th
iteration is given by
(1)
V k 1 V k  'V k
Here 'V
iterations.

is change in bus voltages after two successive

B. Real Power Flow

Pij Real[Vi {(Vi  V j ) yij }*]

(2)

Here Pij is the real power flowing through the line

connecting ith and jth buses, Vi and Vj are the voltages of ith
and jth bus respectively and yij is the admittance of the line
between ith and jth buses.
C Reactive Power Flow

Im ag [Vi {(Vi  V j ) y ij }*]

Qij

(3)

Here Qij is the reactive power flowing through the line

connecting ith and jth buses.
D. Real Power Loss

Loss = Real Vss

>(V

ss

j 1

@ PD j

 V j ) y ss , j * 

jss

(4)

Where Vss and Vj in Eq. (4) refers to the voltages at main

substation and bus j, respectively, yss,j refers to the line
admittance between the main substation bus and bus j, PDj
refers to the real power load at bus j and N the number of
buses in the radial distribution system (RDS).
E. Voltage Deviation Index (VDI)
In order to quantify the extent of violation of limits
imposed on voltages at buses in a RDS, the following Voltage
Deviation Index (VDI) has been defined.
NVB

(V

VDI

Li

 VLiLIM ) 2

(5)

i 1

Subject to V jMIN d V j d V jMAX

j 1 to N
Where NVB is the number of buses that violates the
prescribed voltage limits and VLiLIM is the upper limit of the Ith
load bus voltage if there is upper limit violation or lower limit
if there is a lower limit violation.
During reconfiguration, if the state of the system has
voltage limit violations; the given solution must try and
minimize the index VDI.
ALGORITHM DEVELOPMENT

III.

A sample distribution system drawn below is taken here to

illustrate the methodology [20].
Bus
4

Bus
1
S/S

I4

B3

Bus
2

B4

Bus
55

I5

Bus
3

B1

The algorithm is used to calculate the load flow is a novel

one in. The only input data of this algorithm is the
conventional bus-branch oriented data used by most utilities.
Here, the objective is to develop a formulation, which takes
advantages of the topological characteristics of distribution
systems, and solve the distribution load flow directly.
The distribution Networks, the equivalent-currentinjection-based model is more practical as shown in Fig. 1.
For bus i, the complex load Si is expressed by
Si = (Pi + jQi) i =1N.

(6)

Corresponding equivalent current injection at the kth

iteration of solution is

Iik = Iik(Vik) + jIii(Vik) = Pi  jQi
(7)

Vi k

Where Vi and Ii are the bus voltage and equivalent current

injection of bus I at the kth iteration, respectively. Iir and Iii are
the real and imaginary parts of the equivalent current injection
of bus i at the kth iteration respectively.
The power injection can be connected to the equivalent
current injections by using (7) and relationship between the
bus current injections and branch current can be obtained by
applying Kirchoffs current law (KCL) to the distribution
network. The branch currents can then be formulated as
functions of equivalent current injections. For example the
branch currents B1, B2 and B5 can be expressed by equivalent
current injections as
B1 = I2 + I3 + I4 + I 5 + I6
B3 = I4 + I5
B5 = I6

(8)

Therefore the relationship between the bus current

injections and branch currents can be expressed as
[B] = [BIBC] [I]

(9)

The constant BIBC matrix is an upper triangular matrix

and contains values of 0 and +1 only.
The relationship between branch currents and bus voltages
can be obtained as follows
V2 = V1 B1 Z12
V3 = V2 B2 Z23
V4 = V3 B3 Z34

(10.1)
(10.2)
(10.3)

Where Vi is the voltage of bus i, and Zij is the line

impedance between bus i and bus j. Substituting (10.1) and
(10.2) into (10.3), the equation (10.3) can be written as

B2

V4 = V1 B1 Z12 B2Z23 B3 Z34

I2

I3
B5

Bus
6

I6
Fig. 1. Equivalent current injection based model of distribution network

(11)

From (11), it can be seen that the bus voltage can be

expressed as a function of branch currents, line parameters and
the substation voltage. Similar procedures can be performed
on another buses; therefore the relationship between branch
currents and bus voltages can be expressed as
Delta V = [BCBV] [B]

(12)

The algorithm can easily be expanded to a multiphase line

section or bus. For example, if the line section between bus i
and bus j is a three phase line section, the corresponding
branch current will be a 3x 1 vector and the +1 in the BIBC
matrix will be a 3 x 3 identity matrix. Similarly, if the line
section between bus i and bus j is a three phase line section,
the Zij in the BCBV matrix is a 3 x 3 impedance matrix.
The BIBC and BCBV matrices are developed based on the
topological structure of distribution systems. The BIBC matrix
represents the relationship between bus current injections and
branch currents. The corresponding variations at branch
currents, generated by the variations at bus current injection
can be calculated directly by the BIBC matrix. The BCBV
matrix represents the relationship between branch current and
bus voltages. The corresponding variations at bus voltage,
generated by the variations at branch currents can be
calculated directly by the BCBV matrix. Combining equation
(9) and (12), the relationship between bus current injections and
bus voltages can be expressed as

Start

Form BIBC and BCBV matrices according to the algorithm.

Calculate DLF matrix

Calculate change in voltages

Set iteration k=3 or 4

Calculate new values of bus voltages

['V] = [BCBV] [BIBC] [I]

['V] = [DLF] [I]

(13)

DLF is a multiplication matrix of BCBV and BIBC

matrices and the solution for distribution load flow can be
obtained by solving (8) iteratively as
Iik = Iik (Vik) + j Iii (Vik) = Pi  jQi

Vi k

['V K+1] = [DLF] [Ik]

[Vk+1] =[V0] + ['Vk+1]

(14.1)

Iteration = k
No
Yes

Final bus voltages

(14.2)
(14.3)

It means that the time consuming L-U decomposition and

forward backward substitution of the Jacobian matrix or the Y
admittance matrix, required in the traditional NewtonRaphson and Gauss implicit Z matrix algorithms, are not
necessary in the new development.
According to the research, the arithmetic operation, number
for LU factorization is approximately proportional to N3. For a
large value of N, the LU factorization will occupy a large
portion of the computational time. Therefore if the LU
factorization can be avoided, the load flow method can save
tremendous computational resource.
IV. FLOW CHART
Fig. 2 show a flowchart of the proposed method, based on
the algorithm as discussed in the previous section for two
different configurations the base case and the optimal case for
a 33 bus radial Distribution System Reconfiguration.

End

Fig. 2. Proposed algorithm for distribution system reconfiguration

V. TEST RESULTS
The proposed algorithm for distribution system
reconfiguration has been tested on 33-bus RDS given in
Fig. 3. Substation voltage is 12.66 KV and base MVA has
been taken as 10 MVA; the load data, line details and the tie
lines available for switching in [21].
System has five tie lines. The two configurations are
termed as Base Configuration and Optimal Configuration
respectively. Using DLF program voltages at the buses, real
and reactive powers flowing through lines, real power loss and
voltage deviation index (VDI) were calculated for the two
configurations. Electrical Transient Analyzer Program (ETAP)
simulation was performed and results obtained from the
simulation proved the authenticity of the program developed.
Results are shown here in three parts as base case, optimal
case and comparisons between various results.

Solutions to the voltages at the buses obtained show that at

each bus, voltage in case DLF program is better than those in
case of ETAP simulation results. Worst bus voltage in case of
ETAP is 0.908 and that in case of DLF method it is
0.914.Also the best voltage is higher in case of DLF solutions.
Once the voltages become higher, the losses are bound to be
reduced. For the same load, power drawn in case of ETAP
solutions is higher as compared to that obtained by DLF
method. This only signifies the fact that losses in latter case
have been reduced.
B. Solution Graph for Base Case
Following are the graphs, which show voltages at the buses
after each iteration. Graphs after each iteration are almost
looking alike. This signifies the fact that solutions get
converged very early on.
Fig. 3. A 33-bus radial distribution system

Reduction in the real power loss, improved voltage

deviation and increased bus voltages are the merits shown by
the method used. This can be under stood by having a look on
the following tables:
Table I
Case

Loss(DLF/

VDI(DLF

Worst

Lines switched

given

in

/given in

Voltage

out

[21])

in

[21])

(DLF/given
in [21]in p.u.

KW
Base

Optimal

201.42/

0.0174/

0.9143/

33 - 34 -35 -

211

0.0248

0.9038

36 - 37

158.24/

0.0039/

0.9388/

06 -14 - 09 - 32

178

0.0041

0.9378

- 37

Tie Lines, Substation Voltage = 12.66 kV

Table II
Case
Base

%Loss
Reduction
4.5

%
VDI
Improvement
30

% Increment in
Worst Voltage
1.16

Optimal

11.1

4.8

0.106

Fig. 5. Comparative voltage variations along the lines after subsequent

iterations

The valley point or lower most point shown in the Fig. (5)
shows the lowest voltage bus and the lowest bus voltage is on
bus number 18 of the system investigated. This bus is situated
at extreme end of the network. There is no significant change
in shapes of the graphs for different iterations which was
plotted for all the iterations simultaneously shows that there is
very little deviation in bus voltages in first iteration and
second iteration. It only signifies the fact that load flow
solutions get converged very early on, even in second
iteration. Because of the above fact, execution is very fast.
Best voltage is at bus number 2 and worst at bus number 18.
C. Comparison of Base and Optimal Voltages

A. Comparison Between the Bus Voltages

Bus Voltages for Base & Optimal Cases

DLF
1.02
1. 0 2

1
Bus Voltages in p.u

ETAP
Softw are

0 . 9 8

0 . 9 6

0 . 9 4

0.98

Optimal
Case
Base
Case

0.96
0.94
0.92
0.9
0.88

0 . 9 2

0.86
1

0 . 9
0

10

15

2 0

B uses

Fig. 4. Voltage comparison

2 5

3 0

3 5

11 13 15 17 19 21 23 25 27 29 31
Buses

Fig. 6. Base and optimal voltage comparison

Voltage Comparision
Bus Voltages in p.u

1.02
1
0.98

Optimal Bus
Voltages(DLF)

0.96
0.94

Optimal Bus
Voltages[6]

0.92
0.9
0.88
0.86
1

10 13 16 19 22 25 28 31
Buses

Fig. 7. Voltage comparison for optimal cases

Fig. (6) compares the results obtained for the two cases
considered .It is concluded that voltages at the buses in case of
optimal case is much better than that in the base case for
majority of buses. Few buses have lower voltages (in case of
optimal case) than that in base case. Fig. (7) compare the
results of optimal bus voltage and [6]. Again the results of
optimal DLF case are better than [6] optimal case.
D. Comparison of Base and Optimal Real and Reactive Power
Real Power Comparision
4000
Re3500
al
3000
Po
wer2500
(k
W) 2000

Real Power
(Base Case)
Real Power
(Optimal Case)

1500
1000
500
0
1

9 11 13 15 17 19 21 23 25 27 29 31
Line No.

Fig. 8. Real power comparison

Fig. (8) and (9) are showing the comparisons between real
and reactive power respectively flowing through the lines in
the two cases. In optimal case lesser real power is required
because the loss has been decreased. This was the objective to
be achieved through reconfiguration. This discussion is
equally applicable to reactive power.

Results obtained indicate that the approach to load flow

solutions is much superior to the previous approaches such as
used in [6], [21] and ETAP software. For example even in
base configuration the worst voltage is better than the worst
voltage obtained through ETAP simulation. Also, voltages at
the majority of buses are greater than those obtained by the
other methods such as in [6] and ETAP simulation. Also, real
and reactive powers drawn are lower for the same demand.
This aspect leads the system to have lower losses and better
voltage deviations index (VDI) as shown by the results.
The voltage was improved by 4.5% and 11.1% in base case
and optimal case respectively. Voltage deviation index was
improved by 30% and 4.8% in base case and optimal case
respectively. Similarly worst voltage was improved by 1.16%
and 0.106% in base case and optimal case respectively.
VI. DISCUSSION AND CONCLUSION
In this a new approach, load flow program to solve the load
flow problems in radial distribution system has two matrices,
which are developed from the topological characteristics of
distribution systems. The BIBC matrix represents the
relationship between bus current injections and branch
currents, and BCBV matrix represents the relationship
between branch current and bus voltages. These two matrices
are combined to form a direct approach for solving load flow
problems. The execution time is extremely smaller as
compared to known methods of other load flow methods for
radial distribution systems, such as fast decoupled and Gauss
Implicit Z-matrix method. Here, we do not require to compute
Z-matrix or jacobian.
This method is extremely fast, so this can be used in online
application. This merit of the proposed method can be
exploited in distribution automation, this will require
interfacing of an improved form of this program and
Supervisory Control and Data Acquisition (SCADA) systems.
Utilization of this approach of load flow and developed load
flow program in radial distribution system for optimal
operation is that it is very effective and solutions get
conversed even in second iteration; therefore, the execution
time of the proposed method is quite small. This is the
advantage for distribution system where the load varies
indiscriminately. Limitation of the program is that it can be
used only for the radial distribution system and, not for
meshed distribution systems and transmission systems.

Reactive Power Comparision

Reactive Power (kVAR)

2500
2000
Reactive Power
(Base Case)

1500

Reactive Power
(Optimal Case)

1000
500
0
1

10 13 16 19 22 25 28 31
Line No.

Fig. 9. Reactive power comparison

VII. REFERENCES
[1] C.S.Cheng and D.Shirmohammadi, A Three Phase Power Flow Method
to Real-Time Distribution System Analysis, IEEE Trans. Power Syst.,
vol.10, pp 671-679, Nov.1995.
[2] R.D.Zimmermann and H.P.Chiang, Fast Decoupled Power Flow for
Radial Distribution System, IEEE Trans. Power Syst., vol.10,
Nov.1995.
[3] Antonio Gomez Exposito Esther Romero Ramos, Reliable Load Flow
Technique for Radial Distribution Networks, IEEE Trans. Power Syst.,
vol. 14, no.3, pp. 1063-1070, 1999.
[4] Mesut E.Baran & Felix F. Wu, Network Reconfiguration in Distribution
Systems for Loss Reduction & Load Balancing, IEEE Trans Power
Del., vol.4, no.2, pp. 1401-1407, April 1989.
[5] Koichi Nara, Atsushi Shiose, Minoru Kitagawa & Toshihisa Ishihara,
Implementation of Genetic Algorithms for Distribution Systems Loss

6
Minimum Reconfiguration, IEEE Trans. Power Syst., vol. 7, no. 3, pp.
1044-1051,Aug.1992.
[6] S.K.Goswami & S.K.Basu, A New Algorithm for the Reconfiguration
of Distribution Feeders for Loss Minimization, IEEE Trans. Power
Del., vol.7, no.3, pp. 1484-1491, July1992.
[7] Hoyong Kim, Yunseok Ko & Kung-Hee Jung, Artificial Neural-Network
Based Feeder Reconfiguration for Loss Reduction in Distribution
Systems, IEEE Trans. Power Del., vol.8, no.3, pp. 1356-1366,
July1993.
[8] Ji-Yuan Fan, L.Zhang & John D. McDonald, Distribution Network
Reconfiguration: Single Loop Optimization, IEEE Trans. Power Syst.,
vol. 11, no. 3, pp. 1643-1647,Aug.1996.
[9] R.J.Sarfi, M.M.A.Salama, and A.Y. Chikhani, Distribution system
reconfiguration for system loss reduction: an algorithm based on
network partitioning theory, IEEE Trans. Power Syst. vol. 11, no. 1, pp.
504-510, Feb.1996.
[10] Jin-C Wang, Hsiao-Dong Chiang & G.R.Darling, An Efficient
Algorithm for Real Time Network Reconfiguration in Large Scale
Unbalanced Distribution Systems, IEEE Trans. on Power Syst. vol. 11,
no. 1, pp. 511-517, Feb.1996.
[11] G.J.Peponis, M.P.Papadopoulas and N.D.Hatziragyrioue, Optimal
operation of distribution networks, IEEE Trans. Power Syst., vol. 11,
no. 1, pp. 59-67,February 1996.
[12] Tsai-Siang Chen & Jen-T. Cherng, Optimal Phase Arrangement of
Distribution Transformers Connected to a Primary Feeder for System
Unbalance Improvement & Loss Reduction Using a Genetic
Algorithm, IEEE Trans. Power Syst., vol. 15, no. 3, Aug. 2000.
[13] K.N.Miu, H.D.Chiang, and R.J.Mcnulty, Multitier service restoration
through network reconfiguration and capacitor control for large scale
radial distribution networks, IEEE Trans. Power Syst., vol.15, pp
1001-1007, Aug.2000.
[14] Raman A Gallego, A.J.Manticelli & Ruben Romero, Optimal Capacitor
Placement in Radial Distribution Networks, IEEE Trans Power Syst.,
vol.16, no.4, pp 630-637, Nov. 2001.
[15] Antonio Augugliaro, L.Dusonchet, M.G.Ippolito & E.R Sanseverino,
Minimum Loss Reconfiguration of MV Distribution Networks Through
Local Control of Tie Switches, IEEE Trans. Power Del., vol. 8, no.3,
pp 762-712, July2003.
[16] A.Moussa, M.El-Gammal, E.N.Abdallah, and A.I.Attia, A Genetic
Based Algorithm For Loss Reduction in Distribution Systems, IEEE
Trans. Power Del., vol. 4, no., 2, pp. 447-453, May 2000.
[17] Yasuhiro Hayashi & Junya Matsuki, Loss Minimum Configuration of
Distribution System Considering N-1 Security of Dispersed
Generators, IEEE Trans. Power Syst., vol.19, no14, pp. 636-642, Nov
2004.
[18] D.Shirmohammadi, H.W.Hong, A.Semlyen & G.Y.Luo, A
Compensation Based Power Flow Method for Weakly Meshed
Distribution & Transmission Networks, IEEE Trans. on Power Syst.,
vol.3, pp. 753-762, May 1988.
[19] T.H.Chen, M.S.Chen,K..J.Hwang, P.Kotas & E.A.Chebli, Distribution
System Power System Analysis-A Rigid Approach, IEEE Trans. on
Power Del., vol. 6, no.4, pp. 1146-1152, July1991.
[20] Jen-Hao Teng, A Direct Approach for Distribution System Load Flow
Solutions, IEEE Trans. Power Del., vol. .8, no..3, pp. 882-887,
July2003.
[21] B.Venkatesh, Rakesh Ranjan & H.B.Gooi, Optimal Reconfiguration of
Radial Distribution Systems to Maximize Loadability, IEEE Trans. on
Power Syst., vol.19, no.1, pp. 260-266, Feb. 2004.

VIII.

BIOGRAPHIES

Dr. Tilak Thakur is born in 1963. He

graduated from B.I.T. Sindri, in Electrical
engineering in 1987. He completed his
Post graduation in Power System from the
same institute and achieved his Ph.D in
Electronic Instrumentation from Indian
School of Mines, Dhanbad in the area of
SCADA in 1999. He served as a lecturer
in B.I.T., Sindri and NERIST Arunachal
Pardesh. Presently, he is Assistant
Professor in the Department of Electrical Engineering, Punjab
Engineering College (PEC), Chandigarh, India.. He has a teaching

experience of more than 15 years. He is involved in active Research

in Power System Automation and Control

Mrs. Jaswanti graduated in Electrical

Engineering
from Punjab Engineering
College, Chandigarh, India in 1993. She
got her master of Engineering from same
institute in 1997. She is currently doing
her Ph.D in power system. She is also a
associate member of IEI and ISTE. Her
main research interests are power
distribution system operation, analysis and
control.