Method for load-flow solution of radial distribution

networks
S.Ghosh and D.Das
Abstract: A simple and efficient method for solving radial distribution networks is presented. The
proposed method involves only the evaluation of a simple algebraic expression of receiving-end
voltages. Computationally, the proposed method is very efficient. The effectiveness of the proposed
method is demonstrated through three examples.

List of symbols
NB
LN1
j
PL(9
QL(z)
V(z)
R(ii)
Xh]

=
=
=
=
=

Zh)
Zu)

IL(9
LPh)
LQ(ii)

IScii)

IRW
yo(z)

=
=

ZC(9

=
=

=
=
=
=

DVMAX =
1

total number of nodes

total number of branches
branch number,j = 1,2, 3, ..., LN1
real power load at ith node
reactive power load at ith node
voltage of ith node
resistance o f a h branch
reactance o f a h branch
impedance of branchjj
current that flows through branchj
load current of node i
real-power loss of branchjj
reactive-power loss of branchj
sending-end node of branchj
receiving-end node of branchj
charging admittance at node i
charging current at node i
maximum voltage dfference

Introduction

Little attention has been given to load-flow analysis of distribution systems, unlike load-flow analysis of transmission
systems. However, some work has been carried out on
load-flow analysis of distribution networks, but the choice
of a solution method for a practical system is often dsicult. Generally, distribution networks are radial and the R/
Xratio is very high. For t h s reason, conventional NewtonRaphson (NR) [l] and fast decoupled load-flow [2] methods do not converge. Many researchers have suggested
modfied versions of the conventional load-flow methods
for solving power networks with hgh R/X ratio [3-51.
Kersting and Mendive [q and Kersting [I have developed a load-flow techque for solving radial distribution
networks using ladder-network theory. They have
OIEE, 1999
ZEE Proceedhgs online no. 19990464
DO1 l O . l 0 4 9 / i p g t d : l ~
Paper fmt received 5th November 1998 and in revised form 26th March 1999
The authors are with the ElectricalEngineering Department, Indian Institute of
T E ~ o I o ~Kharagpur
Y,
- 721302, WB, India
IEE Proc.-Gener. Transm. Distrib.. Vol. 146, No. 6, November 1999

developed the ladder technique from basic ladder-network

theory into a working algorithm, applicable to the solution
of radial load-flow problems. Stevens et al. [8] have shown
that the ladder technique is found to be fastest but did not
converge in five out of 12 cases studied. Shirmohammadi er
al. [9] have proposed a method for solving radial distribution networks based on the direct application of Kirchhoff's voltage and current laws. They have developed a
branch-numbering scheme to enhance the numerical performance of the solution method. They have also extended
their method for solution of weakly meshed networks.
Baran and Wu [lo] have obtained the load-flow solution in
a distribution system by the iterative solution of three fundamental equations representing real power, reactive power
and voltage magnitude. They have computed the system
Jacobian matrix using a chain rule. In their method, the
mismatches and the Jacobian matrix involve only the evaluation of simple algebraic expressions and no trigonometric
functions. They have also proposed decoupled and fast
decoupled distribution load-flow algorithms. Chiang [l 11
has also proposed three different algorithms for solving
radial distribution networks based on the method proposed
by Baran and Wu [lo]. He has proposed decoupled, fast
decoupled and very fast decoupled distribution load-flow
algorithms. In fact decoupled and fast decoupled dlstribution load-flow algorithms proposed by Chiang [I 13 are similar to that of Baran and Wu [lo]. However, the very fast
decoupled distribution load flow proposed by Chiang [111
is very attractive because it does not require any Jacobian
matrix construction and factorisation. Renato [12] has proposed one method for obtaining a load-flow solution of
radial distribution networks. He has calculated the electrical equivalent for each node summing all the loads of the
network fed through the node including losses and then,
starting from the source node, the receiving-end voltages of
all the nodes are calculated. Goswami and Basu [13] have
presented a direct method for solving radial and meshed
distribution networks. However, the main h t a t i o n of
their method is that no node in the network is the junction
of more than three branches, i.e. one incoming and two
outgoing branches. Jasmon and Lee [14, 151 have proposed
a new load-flow method for obtaining the solution of radial
distribution networks. They have used the three fundamental equations representing real power, reactive power and
voltage magnitude derived in [lo]. They have solved the
radial distribution network using these three equations by
reducing the whole network into a single h e equivalent.
641

Das et al. [16] have proposed a load-flow technique for

solving radial distribution networks by calculating the total
real and reactive power fed through any node. They have
proposed a unique node, branch and lateral numbering
scheme which helps to evaluate exact real- and reactivepower loads fed through any node and receiving-end voltages.
In this paper, the main aim of the authors has been to
develop a new load-flow technique for solving radial distribution networks. The proposed method involves only the
evaluation of a simple algebraic expression of receiving-end
voltages. The proposed method is very efficient. It is also
observed that the proposed method has good and fast convergence characteristics. Loads in the present formulation
have been presented as constant power. However, the proposed method can easily include composite load modelling,
if the composition of the loads is known. Several radial distribution 'feeders have been solved successively by using the
proposed method. The speed requirement of the proposed
method has also been compared with other existing
methods.
2

Assumption

It is assumed that the three-phase radial distribution networks are balanced and can be represented by their equivalent single-line diagrams.
1

similarly for branch 2,

V ( 3 )= V ( 2 )- 1 ( 2 ) 2 ( 2 )

(2)

As the substation voltage V(1) is known, so if Z(l) is

known, i.e. current of branch 1, it is easy to calculate V(2)
from eqn. 1.
Once V(2) is known, it is easy to calculate V(3) from
eqn. 2, if the current through branch 2 is known. Similarly,
voltages of nodes 4, 5, ..., NB can easily be calculated if all
the branch currents are known. Therefore, a generalised
equation of receiving-end voltage, sending-end voltage,
branch current and branch impedance is

V(m2)= V(rn1) - I ( j j ) Z ( j j )
where j j is the branch number.
m2 = I R ( j j )

(3)

(4)

ml = IS(jj)

(5)
Eqn. 3 can be evaluated forjj = 1, 2, ..., LN1 (LN1 = NB
- 1 = number of branches). Current through branch 1 is
equal to the sum of the load currents of all the nodes
beyond branch 1 plus the sum of the charging currents of
all the nodes beyond branch 1, i.e.
LN1

LN1

IL(2) +

I(1)=

IC(2)

(6)

i=2

i=2

The current through branch 2 is equal to the sum of the

load currents of all the nodes beyond branch 2 plus the
sum of the charging currents of all the nodes beyond
branch 2, i.e.

12

+ IL(6)

+ I C ( 3 ) + IC(4) + I C ( 5 ) + I C ( 6 )
+

Fig. 1 Single-line diagram of radial distribution network

I(2)= IL(3) IL(4) IL(5)

+ IL(10) + IL(11)

11

Solution methodology

Fig. 1 shows single-line diagram of a distribution feeder.

The branch number sending-end and receiving-end node of
this feeder are given in Table 1. Consider branch 1. The
receiving-end node voltage can be written as

V ( 2 )= V ( l ) - 1(1)2(1)

IC(10) IC(11)
(7)
Therefore, if it is possible to identify the nodes beyond all
the branches, it is possible to compute all the branch currents. Identification of the nodes beyond all the branches is
realised through an algorithm as explained in Section 4.
The load current of node i is

(1)

Table 1: Branch number (jj,

sending-end (ml /S(jj))node, receiving-end
node (m2 = (/R(jll) and nodes beyond branches 1,2,3. ..., 11 of Fig. 1
Branch
number ( j j

642

Sending end

rnl = /S(hl

2
3
4
5
6
7
8
9
10
11

2
3
4
5
2
7

a
4
10
8

Receiving end
m 2 = /R(jj

2
3
4
5
6
7
8
9
10
11
12

Nodes beyond branch j j

2,3,7,8,5,10,9,12,6,11
3, 4,5,IO, 6,1 1
4,5, IO, 6,1 1
5,6
6
7,8,9,12
8,9,12
9
IO, 1 1
11
12

Total number of
nodes Mjl)
beyond branch j

11

6
5
2
1
4
3
1
2
1

IEE Proc.-Gener. Transm. Disfrib., Vol. 146, No. 6,November 1999

First define the variables:

j j = 1, 2, 3, ..., LN1 cij indicates branch of Fig. 1, see also
Table 1);
ip is the node count (identifies the number of nodes beyond
a particular branch);
ZK(ip) is the node identifier (helping to identify nodes
beyond all the branches);
Mjj] is the total number of nodes beyond branchjj; and
IEGJ, ip + 1) is the receiving-end node.
ZE(J, ip + 1) will now be explained. Consider the fxst
branch in Fig. 1, i.e. j j = 1; the receiving-end node of
branch 1 is 2, i.e. ZRb] = ZR(1) = 2. Therefore ZEb, ip + 1)
= ZE(1,ip + 1) will help to identify all the nodes beyond
branch 1. This will help to find the exact current flowing
through branch 1. S i a r l y , consider branch 2, i.e. jj = 2;
the receiving-end node of branch 2 is 3, i.e. IR(jj] = ZR(2) =
3. Therefore, E ( J , ip + 1) = ZE(2, ip + 1) will identify all

The charging current at node i is

i = 2 , 3 , . . . ,N B

I C ( i )= yo(i)V(i)
(9)
Load currents and charging currents are computed iteratively. Initially, a flat voltage of all the nodes is assumed
and load currents and charging currents of all the loads are
computed using eqns. 8 and 9. A detailed load-flow-calculation procedure is described in Section 5.
The real and reactive power loss of branchjj are given
by:
Wjj)=

IW)12Wj)

L Q ( j j ) = Il(jj)12X(jj)
4

(10)

(11)

Identificationof nodes beyond all the branches

Before the detailed algorithm is given, the details of the

methodology of identifying the nodes beyond all branches
will be discussed. This wdl help in finding the exact current
flowing through all the branches.

(7)
from A

read sending-end
and receiving-end
nodes and total
number of nodes
and branches

in 5 ip
to F
no

ni
iil
nc=l

k=jj+l

no

I II

from B

.(
IE(jj, ip+l)=lR(jj)

from E

iP =i P+1
IK(ip)=i

from C

Li
nc=O

IE(jj, ip+l)=iR(jj)
N(jj) =ip+ 1

to E

L, d

r -

i=i+l

in=i

f
IE(jj, ip+l)=IR(jj)
N(jj)=ip+1

IE(LN,I)=IR(LNI)
N(LNl) = 1

stop

Fig.2
IEE Proc.

643

the nodes beyond branch 2. This will help to compute the

exact current flowing through branch 2. For each node
identitication beyond a particular branch, '@' will be incremented by 1. Note here that before identification of nodes
beyond a particular branch, '@' has to be reset to zero.
For jj = 1 (first branch of Fig. 1 , Table I), ZRG) = ZR(1 )
= 2; check whether IR(1)= ZS(i) or not for i = 2, 3, 4, ...,
LN1. It is seen that ZR(1)= ZS(2) = 2, IR(1)= IS(6) = 2;
the corresponding receiving-end nodes are ZR(2) = 3 and
ZR(6) = 7.
Therefore, IE(1,1) = 2, IE(1, 2) = 3 and ZE(1, 3) = 7 .
Note that there should not be any repetition of any node
while identifying nodes beyond a particular branch, and
this logic has been incorporated in the proposed algorithm
and further explained in the flowchart given in Fig. 2.
From the above discussion, it is seen that node 2 is connected to nodes 3 and 7. Smilarly, the proposed logic will
identify the nodes which are connected to nodes 3 and 7.
First it w d check whether node 3 appears in the left-hand
column of Table 1 . It is seen that node 3 is connected to
node 4. Therefore, ZE(1, 4) = 4. Then it will check whether
node 7 appears in the left-hand column of Table 1 . It is

seen that the node 7 is connected to node 8. Therefore,

ZE(1, 5 ) = 8.
From the above dlscussion, it is again seen that node 3 is
connected to node 4 and node 7 is connected to node 8.
Similarly, the proposed logic will check whether nodes 4
and 8 are connected to any other nodes. Thls process will
continue unless all nodes are identified beyond branch 1.
The nodes beyond branch 1 are also given in Table 1.
The total current flowing through branch 1 is equal to
the sum of the load currents of all nodes beyond branch 1
plus the sum of the charging currents of all the nodes
beyond branch 1 .
For jj = 2 (second branch in Fig. 1; Table l), ZRG] =
ZR(2)= 3, check whether ZR(2) = ZS(z] or not for i = 3,
4, ..., LN1. It is seen that IR(2) = IS(3) = 3. The corresponding receiving-end node is IR(3) = 4. Therefore, IE(2,
1) = 3 and ZE(2, 2) = 4.
From the above discussion, it is seen that node 3 is connected to node 4. The proposed logic will identify the nodes
which are connected to node 4. It will check whether node
4 appears in the left-hand column of Table 1. It is seen that
node 5 and node 10 are connected to node 4. Therefore

-+
from A

start

-1
read substation voltage V(l),
line data and load data.

jj 5 LN1

assume a flat voltage starI,i.e. V(i)

=V(l)=I~O'for i=2,3 ,........, NB
set W(i)=V(i) for i=2,3 ,......., NB
set lSS(ij)=lS(jj) and IRR(jj)=
IR(jj) for jj=1,2,3 ,........, LNl
set iteration count k=O

calculate IL(i) and IC(i)

using eqns. 8 and 9
for i=2,3, .... NB

calculate branch currents

by using eqn. 12

no

,-,

L7J
converged

W(m2)=V(m2)
for m2=2,3,.... NB

calculate line
losses, line
flows etc. and
print required
dyta

to B

set m l =ISS(jj) and rnS=IRR(jj)

compute receiving-end voltage V(m2)
by using eqn. 3
calculate absolute change in voltage at
node m2, DV(m2)=ABS(/V(m2)/-/W(m2)/)

I
Fig.3
644

jj=jj+l

-1

to A
Fibwehartfor load-jhw calculation of r&
dimibution network
IEE Proc-Gener. Transm. Distrib., Vol. 146, No. 6,November 1999

ZE(2, 3) = 5 and IE(2, 4) = 10. The proposed logic will

check whether nodes 5 and 10 are connected to any other
nodes. This process will continue unless all nodes are identified beyond branch 2. The nodes beyond branch 2 are
given in Table 1.
Similarly it is necessary to consider the receiving-end
node of branch 3, branch 4, ..., branch LN1 in Fig. 1 and,
in a similar way to that discussed above, the nodes have to
be identified beyond these branches. The nodes beyond all
the branches are also given in Table 1.
Note that, if the receiving-end node of any branch in
Fig. 1 is an end node of a particular lateral, the total current of this branch is equal to the load current of this node
plus the charging current of this node itself. For example,
consider node 6 in Fig. 1 (branch 5, Table 1); this is an end
node. Therefore, the branch current 1(5)is equal to the load
current of node 6 plus the charging current of node 6 (i.e.
4 5 ) = ZL(6) + ZC(6)). Sirmlarly, 9, 11 and 12 are end nodes
of Fig. 1. The proposed computer logic will identify all the
end nodes automatically.
The concept of identifying the nodes beyond all the
branches, whch helps in computing the exact current flowing through all the branches, has been realised using an
algorithm (Fig. 2) and applied in the load-flow technique
as shown in the flowchart in Fig. 3.
5

Load-flow calculation

Once all nodes beyond each branch are identified, it is very

easy to calculate the current flowing through each branch
as described in Section 3. For this purpose, the load current
and charging current of each node are calculated by using
eqns. 8 and 9. Once the nodes are idenflied beyond each
branch, the expression of branch current is given as
N(33)

I ( j j )=

I L { I E ( j j ,i)}
2=1

Nb3)

I C { I E ( j j ,i)}
2=1

(12)
Initially, a constant voltage of all the nodes is assumed and
load currents and charging currents are computed using
eqns. 8 and 9. After load currents and charging currents
have been calculated, branch currents are computed using
eqn. 12. The voltage of each node is then calculated by
using eqn. 3 with eqn. 4. Real and reactive power loss of
each branch is calculated by using eqns. 10 and 11, respectively. Once the new values of the voltages of all the nodes
are computed, convergence of the solution is checked. If it
does not converge, then the load and charging currents are
computed using the most recent values of the voltages and
the whole process is repeated.
The convergence criterion of the proposed method is that
if, in successive iterations the maximum difference in voltage magnitude (DVMAX) is less than O.OOOlp.u., the solution has then converged. The convergence analysis of the
proposed method is given in Appendjx 1 (Section 9.1).
The proposed distribution load-flow algorithm for solving radial distribution networks is given in the form of a
flowchart in Fig. 3.
6

Example

To demonstrate the effectiveness of the proposed method,

three examples are selected. The first example is a 29-node
11kV rural distribution feeder of India. Data for this 29node system are given in Appendix 2 (Section 9.2). Table 2
gives the load-flow results of a 29-node radial distribution
network. The real- and reactive-power loss of this system
are 303.78kW and 124.74kVAr, respectively.
IEE Proc -Gener Transm Distrcb , Vol 146, No. 6, November 1999

Table 2: Load-flow solution of example 1: 29-node radial distribution network

~

Node
number

Voltage
magnitudes (P.u.)

Node
number

Voltage
magnitudes (P.u.)

1
2
3
4

1 .ooooo
0.95884
0.91877
0.89598
0.83736
0.85068
0.82855
0.82189
0.81233
0.80821

0.88945

12

0.94989
0.94655

15
16
17
18
19
20
21
22
23
24
25
26

13
14

0.93980
0.93684

27
28

5
6
7
8
9
10
11

0.88560
0.88296
0.87556
0.86914
0.86378
0.86240
0.86120
0.81710
0.81079
0.80725
0.80467
0.80884
0.80774

Table 3: Load-flow solution of example 2:69-node radial distribution network

Node
number

Voltage
magnitudes (P.u.)

Node
number

Voltage
magnitudes (P.u.)

1
2
3
4
5
6
7
8
9

1 .ooooo
0.99997
0.99993
0.99984
0.99902
0.99009
0.98079
0.97858
0.97745

36
37
38
39
40
41
42
43
44

10

0.97245
0.97135
0.96819
0.96526
0.96237
0.95950
0.95897
0.95809
0.95808
0.95761
0.95731
0.95683
0.95683
0.95675
0.95660
0.95643
0.95636
0.95634
0.99993
0.99985
0.99973
0.99971
0.99961
0.99935
0.99901
0.99895

45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69

0.99992
0.99975
0.99959
0.99954
0.99954
0.99884
0.99855
0.99851
0.99850
0.99841
0.99840
0.99979
0.99854
0.99470
0.99415
0.97854
0.97853
0.97466
0.97142
0.96694
0.96257
0.94010
0.92904
0.92476
0.91974
0.91234
0.91205
0.91166
0.90976
0.90919
0.97129
0.97129
0.96786
0.96786

11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35

645

Table 4 Load-flow results of example 3: 33-node radial distribution network

Status of the network

Real-power
loss (kW)

Reactive-power
loss (kVAr)

Minimum
voltage (P.u.)

Without charging admittance

202.67

135.14

Vi8 = 0.91309

With charging admittance

196.54

131.03

V78 = 0.91513

Table 5: Comparison of speed of proposed method with other existing methods

Proposed method and other four

existing methods

Example 1: 29-node
distribution network

Example 2: 69-node
distribution network

Example 3: 33-node
distribution network

CPU time

Iteration
number

CPU time

(s)

Iteration
number

CPU time

(S)

(S)

Iteration
number

Proposed method

0.07

0.16

0.09

Chiang [ I l l very fast decoupled

distribution load flow

0.09

0.24

0.11

Load flow using single-line

equivalent [14, 151

0.1 1

0.29

0.13

Renato [I21 load flow using

forward-sweeping method

0.12

0.33

0.14

Kersting [71 load flow using ladder 0.14

technique

0.37

0.16

The second example is a 69-node radial distribution network. Data for this system are available in [lo]. Load-flow
results of this system are given in Table 3. Real- and
reactive-power losses of this system are 224.96kW and
114.15kVAr, respectively.
The third example is a 33-node radial distribution network. Data for this system are available in [17, 151. T h ~ s
33-node radial network is solved with and without considering charging admittance. Table 4 gives real- and reactivepower losses and the minimum voltage with and without
charging admittance.
The proposed method is also compared with four other
existing methods. Table 5 shows the CPU time and
number of iterations of all three examples. All these three
examples were simulated on a Meteor 400VT with a
66MHz clock. From Table 5, it is seen that the proposed
method is better than the other four existing methods.
It is worth mentioning here that the authors have tried to
solve the above three examples using the Newton-Raphson
(NR) and Gauss-Seidel (GS) methods. However, for all
these three examples, the NR and GS methods did not
converge.
7

Conclusions

A simple and efficient load-flow technique has been proposed for solving radial distribution networks. It has been
found from the cases with which the method was tested
that the method has good and fast convergence characteristics compared with some other existing methods. The proposed method has been implemented on a Meteor 400VT
with a 66MHz clock. Several radial distribution networks
have been solved successfully using the proposed method.
8

References

1 TINNY, W.F., and HART, C.E.: Power flow solution by Newtons

method, IEEE Tram., 1967, PAS86, pp. 1449-1456

2 SCOlT, B., and ALSAC, 0.: Fast decoupled load flow, ZEEE
Trans., 1974, PAS-93, pp. 859-869
3 RAJICIC, D., and TAMURA, Y.: A modification to fast decoupled
power flow for networks with hgh R/X ratios, IEEE Tram., 1988,
PWRS-3, pp. 743-746
4 IWAMOTO, S., and TAMURA, Y.: A load flow calculation method
for illconditioned power systems, IEEE Trans., 1981, PAS-100, pp.
17361713
646

TRIPATHY, S.C., DURGAPRASAD, G., MALIK, O.P., and

HOPE, G.S.: Load flow solutions for ill-conditioned power system by
a Newton like method, ZEEE Tram., 1982, PAS-101, pp. 36843657
KERSTING, W.H., and MENDIVE, D.L.: An application of ladder
network theory to the solution of three phase radial load flow problem. IEEE PES winter meeting, New York, January 1976
KERSTING, W.H.: A method to the design and operation of a distribution system, ZEEE Trans., 1984, PAS-103, pp. 1945-1952
STEVENS, R.A., RIZY, D.T., and PURUCKER, S.L.: Performance
of conventional power flow routines for real-time distribution automation application. Proceedings of 18th southeastern symposium on
System theory, 1986, (IEEE Computer Society), pp. 196200
SHIRMOHAMMADI. D.. HONG. H.W.. SEMLYEN. A.. and
LUO, G.X.: A compensa6on-based power flow method for weakly
meshed distribution and transmission networks, IEEE Trans., 1988,
PWRS-3, pp. 753-743
10 BARAN, M.E., and WU, F.F.: Optimal sling of capacitors placed on
a radial distribution system, IEEE Trans., 1989, P-2,
pp. 735743
11 CHIANG, H.D.: A decoupled load flow method for distribution
power network algorithms, analysis and convergence study, Electr.
Power Energy Syst., 1991, 13, (3), pp. 13C138
12 RENATO, C.G.: New method for the analysis of distribution networks, IEEE Trans., 1990, PWRD-5, (l), pp. 9-13
13 GOSWAMI, S.K., and BASU, S.K.: Direct solution of distribution

systems, IEE Proc. C. , 1991, 188, (I), pp. 78-88

14 JASMON, G.B., and LEE, L.H.C.C.: Distributionnetwork reduction
for voltage stability analysis and load flow calculations, Electr. Power
Energy Syst., 1991, 13, ( I ) , pp. 9-13
15 JASMON. G.B.. and LEE. L.H.C.C.: Stabilitv of load flow techniques fordistribution system voltage stability ahysis, IEE Proc. C,
1991, 138, (6), pp. 479484
16 DAS, D., NAGI, H.S., and KOTHARI, D.P.: Novel method for
solving radial distribution networks, IEE Proc. C, 1994, 141, (4), pp.
391-3911
,
I

I
,
-

17 BARAN, M.E., and WU, F.F.: Network reconfiguration in distribution systems for loss reduction and load balancing, ZEEE Tram.
Power Deliv., 1989, PWRD-4, pp. 1401-1407

Appendices

9.1 Convergence analysis

It has been explained in Section 3 that the proposed loadflow algorithm starts with an initial set of node voltages
and load currents and charging currents are then computed
using eqns. 8 and 9.
Now, for the convergence analysis of the proposed
method, consider the electrical equivalent of Fig. 1. Fig. 4
shows the electrical equivalent of Fig. 1.
In Fig. 4,
j j = branch number
m l = ISfij) = sending-end node
m2 = IRfij) = receiving-end node
IEE Proc-Gener. Transm. Distrib., Vol. 146, No. 6, November 1999

P(m2) = total real power load fed through node m2

Q(m2) = total reactive power load fed through node m2

From Fig. 4, the following equations can be written:

) { W j )+ j X ( j j ) }

(13)

P(m2)- jQ(m2)= V * ( m 2 ) l ( j j )
In fact, eqn. 13 is similar to eqn. 3:

(14)

V(m2)= V(m1) - W

V(m1)

sending end

Mi)

V(r2)

receiving end

I 1P(mZ)+jQ(mZ)

RW+iXUi)

mi =IS(ij)

Fig. 4

fore, if e(m2) for m2 = 2, 3, ..., NB converges then, in

eqn. 3, V(m2),for m2 = 2, 3, ..., NB, will also converge.
Also note that, when one is solving the receiving-end voltage, the sending-end voltage is known.

rnP=IR(jj)

Electrical equivalent of Fig. I

P(m2) = sum of the real-power loads of all the nodes

beyond node m2 plus the real-power load of the node m2
itself plus real-power losses of all the branches beyond node
m2.
Q(m2) = sum of the reactive-power loads of all the nodes
beyond node m2 plus the reactive-power load of the node
m2 itself plus reactive-power losses of all the branches
beyond node m2.
From eqns. 13 and 14 one obtains

To study the convergence analysis of the proposed

method, assume that

From eqns. 23 and 22 one obtains

V(m2)= V(m1)

(15)
Now the voltages at nodes m2 and ml are expressed as

+
V(m1)= e(m1)+ j f ( m 1 )

V(m2)= e(m2) j f ( m 2 )

(16)

(17)

Substituting eqns. 16 and 17 in eqn. 15 and separating real

and imamary parts gives

e(m2)= g{e(m2)}
(24)
Now consider the graph (Fig. 5) of g(e(m2)). In Fig. 5,
e0(m2)is the initial approximation of e(m2), 5(m2) is the
final solution of e(m2), and ~,(rn2) and .en+,(m2)are the
errors in the nth and (n+l)th iterations. From Fig. 5,

en(m2)= t(m2) ~ , ( m 2 )

en+l (m2)= ((ma) ~ ~ + l ( m 2 ) (26)

The solution of e(m2) at various iterative stages can be
written using eqn. 24:

e2(m2) f(m2) = {e(ml)e(m2) f ( m l ) f ( m 2 ) )

el (m2)= g(eo(m2))
e2 (ma)= d e l (4)

- { W j ) P ( m 2+
) X(jj)Q(m2)1
(18)

First let us consider eqn. 18. In fact in a distribution system the voltage angle is extremely small. Hence, the imaginary parts of the voltages are extremely small. Therefore,
the termsf2(m2) andjfmlMm2) in eqn. 18 can be dropped.
Therefore, eqn. 18 can be written as

(25)

en+l(m2)= g{en(m2))
From eqns. 27 and 25 one obtains

(27)

en+l(m2)= g(r(m2)+ En(m2))

From eqns. 28 and 26 one obtains

(28)

[(ma) Enfl(m2)= g(E(m2) En(m2)) (29)

Expanding the right-hand term of eqn. 29 by Taylor series
and considering the first three terms, one obtains
t(m2)+ En+l(m2)= t(m2)+ &n(m2)g{t(m2)}

+ *g{t(m2)}
2
Assuming that

en+1(m2)= E n ( m 2 ) g { t ( m 2 ) } + ~ g r { F ( m 2 ) }
2

(30)
From eqns. 20 and 21 one obtains

Now the first and second derivatives of eqn. 23 are:

4 j j )
g(e(m2))= ez(m2)
Now consider eqn. 19. In eqn. 19, the terms on the righthand side are independent of jfm2), and jfm2) can be
obtained directly after computing e(m2) iteratively. ThereIEE Proc.-Gener. Transm. Distrib., Vol. 146, No. 6, November 1999

g(e(m2))=

-2Au(jj)
e3 (m2)
641

negligible and can be dropped from eqn. 36.

when e(m2) = gm2)

(33)
and
(34)

From eqns. 30, 33 and 34, one obtains

From eqns. 35 and 33, one obtains

During the iterative process, it has been observed that, for

all the examples, (~,(m2)/5(m2)}for m2 = 2, 3, ..., NB is

.. E n S l ( m 2 ) = &n(m2)g{l(m2))

(37)

From eqn. 37 it is seen that the error at the (n + 1)th iteration is proportional to the error at the nth iteration. Hence
convergence of the proposed method is linear. When r, x, P
and Q are expressed in P.u., the term Aubj) (eqn. 21) is
2)}
extremely small and hence g(Krn2)) = { A u ( ~ j ] / ~ ( mis
extremely small, i.e. g(gm2)) << 1.0. Therefore, the convergence of the proposed method is always guaranteed. It
has already been mentioned that solutions of the node voltages are obtained iteratively by using eqn. 3.
In fact several problems were solved using the proposed
method and an interesting point is that, for all the examples
tested, the proposed method has taken three iterations to
converge. Therefore, it may be concluded that the number
of iterations is independent of the size of the problem.

9.2 Data for29-node system

Table 6: Line data and load data of example 1: 29-node radial distribution network
Branch
number

Sending
end

Receiving
end

R (8)

X(Q)

receiving-end
node (kW)

QL of
receiving-end
node (kVAr)

PL of

1.8216

0.7580

140.00

90.00

2.2270

0.9475

80.00

50.00

1.3662

0.5685

80.00

60.00

0.9 180

0.3790

100.00

60.00

3.6432

1.5160

80.00

50.00

2.7324

1.1370

90.00

40.00

1.4573

0.6064

90.00

40.00

2.7324

1.1370

80.00

50.00

10

3.6432

1.5160

90.00

50.00

10

10

11

2.7520

0.7780

80.00

50.00

11

11

12

1.3760

0.3890

80.00

40.00

12

12

13

4.1280

1.1670

90.00

50.00

13

13

14

4.1280

0.8558

70.00

40.00

14

14

15

3.0272

0.7780

70.00

40.00

15

15

16

2.7520

1.1670

70.00

40.00

16

16

17

4.1280

0.7780

60.00

30.00

17

17

18

2.7520

0.7780

60.00

30.00

18

19

3.4400

0.9725

70.00

40.00

19

19

20

1.3760

0.3890

50.00

30.00

20

20

21

2.7520

0.7780

50.00

30.00

21

21

22

4.9536

1.4004

40.00

20.00

22

23

3.5776

1.0114

50.00

30.00

23

23

24

3.0272

0.8558

50.00

20.00

24

24

25

5.5040

1.5560

60.00

30.00

25

26

2.7520

0.7780

40.00

20.00

26

26

27

1.3760

0.3890

40.00

20.00

27

27

28

1.3760

0.3890

40.00

20.00

Base voltage = 1l.OkV, base MVA = 100.00

648

IEE Proc.-Gener. Transm. Distrib., Vol. 146, No. 6, November 1999