Electrical Power and Energy Systems 75 (2016) 173–186

Contents lists available at ScienceDirect

Electrical Power and Energy Systems

journal homepage: www.elsevier.com/locate/ijepes

Loss allocation in radial distribution networks with various distributed

generation and load models
Kushal Manoharrao Jagtap ⇑, Dheeraj Kumar Khatod
Alternate Hydro Energy Center, Indian Institute of Technology Roorkee, Roorkee, Uttarakhand 247667, India

a r t i c l e i n f o a b s t r a c t

Article history: This paper proposes a new method for loss allocation in radial distribution networks (DNs) considering
Received 18 December 2014 different models of distributed generation (DG) and load in context of a deregulated environment. In the
Received in revised form 9 July 2015 proposed method, a direct relation between real/reactive power flow in a branch and its losses has been
Accepted 25 July 2015
developed without taking any assumption and approximation. Suitable expressions/relations for network
power flow have been developed employing power summation algorithm. The developed expressions do
not contain any cross-terms. For allocating the losses among network participants, the proposed method
Keywords:
uses a circuit based branch oriented approach. Using only power flow results, this method employs a
Distributed generation
Load models
backward sweep network reduction technique to allocate the network losses to load/DG at various nodes.
Power flow This method does not require additional step of normalization to collect the exact amount of total
Radial distribution network network losses. In the present study, different types of DG, e.g. DG injecting only real power, DG injecting
Types of DG only reactive power, DG injecting real power and absorbing reactive power, and DG injecting both real
and reactive power are considered to allocate losses. In addition to this, various load models based on
impact of voltage variation on real/reactive power consumption are also considered. To test the proposed
method, modified 9-node and 33-node radial DNs have been considered. In order to show the effective-
ness of the proposed method, its numerical results have been compared with those by other methods
available in the literature.
Ó 2015 Elsevier Ltd. All rights reserved.

Introduction of DN. Further, power loss in a branch of DN is a quadratic function

Nowadays, electrical power system is experiencing major
changes and is adopting deregulated operation of electricity mar-
ket. The vertically integrated systems are being restructured and
unbundled into generation, transmission, and distribution seg-
ments which has introduced the competition among the network
participants (consumers and generators). Unlike the sale of electri-
cal energy by generation companies, activities of transmission and
distribution are generally considered as natural monopoly. There-
fore, electricity market does not have any control over the cost of
services provided by transmission and distribution networks
(DNs). Like transmission network, power losses in the DN have
large share of service charges. Thus, distribution power losses are
to be allocated among network participants, fairly and justifiably.
Distributed generation (DG), when introduced in DN, changes
the losses depending on its location and rating [1,2]. Hence, DG
should be rewarded/penalized according to its impact on losses
⇑ Corresponding author.
E-mail addresses: jagtapkushal@gmail.com (K.M. Jagtap), dheerfah@iitr.ac.in
(D.K. Khatod).
of power flowing through it due to loads and DGs [3]. Hence, in
bundled power flow [3], it is difficult to trace the exact share of
load and DG in the network. The interdependency among network
participants is expressed by the cross-terms which also have
significant impact on allocated losses to loads and DGs. Hence,
the allocation of total network losses cannot be carried out among
consumers and DGs in the straightforward way. The critical nature
of the loss allocation problem is made evident by the fact that early
formulated loss allocation mechanisms, even adopted at the
regulatory level, have been found to be inconsistent [4].
Various methods in the literature dealing with the problem of
loss allocation are mentioned below:
Based on the proportional principle, Pro rata (PR) method [5,6]
allocates the network losses to consumers/DGs based on their real
power consumption/injection. While allocating the losses, this
method does not consider the location of consumer/DG with
respect to (w.r.t.) root node and hence produces unfair result of
loss allocation. MW-mile method [7,8] overcomes the drawback
associated with PR method by considering the power rating as well
as location of a load/DG w.r.t. root node. PR and MW-mile methods
are simple and easy to implement. However, these methods do not

http://dx.doi.org/10.1016/j.ijepes.2015.07.042
0142-0615/Ó 2015 Elsevier Ltd. All rights reserved.
174 K.M. Jagtap, D.K. Khatod / Electrical Power and Energy Systems 75 (2016) 173–186
Nomenclature
Pt and Q t real and reactive power, respectively, available at
receiving end of branch t
PD;t and Q D;t real and reactive power, respectively, of load at
receiving end of branch t
P0;t and Q 0;t real and reactive power, respectively, of load at
receiving end of branch t under rated condition
PG;t and Q G;t real and reactive power, respectively, of DG at
receiving end of branch t
V s;t and V r;t phasor voltages of sending and receiving nodes,
respectively, of branch t
a exponent for different load model
Nt set of branches incident to node t
Kt set of branches ahead of branch t
Bt susceptance of branch t
PSt and QSt real and reactive power loss, respectively, in branch t
take the power flow into account for loss allocation. Thus, to over-
come these limitations of PR and MW-mile methods, marginal loss
coefficient (MLC) method [9,10] came into existence for loss alloca-
tion. MLC method allocates the losses to a load/DG using the MLCs
and power rating of load/DG. This method does not allocate the
losses to the root node and therefore, results in over-recovery of
total network losses, which is compensated by using suitable
normalization procedure. Direct loss coefficient (DLC) method [9]
allocates total losses based on the direct relationship between
the node power injection and network losses. Z-bus method [11]
considers network parameters for loss allocation. It can yield
negative allocation to those loads and DGs, which contribute to
reduce network losses due to their strategically well positioned
in the system. Both MLC and DLC methods are based on the results
of Newton–Raphson (NR) power flow, while Z-bus method
depends on formation of Z-bus matrix in order to allocate losses.
Since a distribution lines have higher R/X ratio in comparison with
transmission lines, many times NR method fails to converge for
load flow analysis of radial DNs. Also distribution lines have negli-
gible shunt admittance which offers difficulty in formulation of
Z-bus. Due to these facts, MLC and DLC methods cannot be applied
to radial DN [4].
In the absence of shunt admittance of lines, succinct method
[12] is able to calculate allocated losses. However, this method is
not able to provide equitable loss allocation in terms of reactive
power loads, when the ratio of reactance to resistance of a line is
greater than that of reactive to real power available at its receiving
node. Substitution method [9] calculates the allocated loss to a
consumer/DG by taking the difference of network losses before
and after connecting it to the network. In this method, the sum
of allocated losses to consumers/DGs is not equal to the total
network losses, and therefore additional step of normalization is
required.
Proportional sharing method [13,14] uses the results of power
flow and linear proportional sharing principle which states that
the power flow reaching a bus from the incoming lines is
distributed among the outgoing lines proportionally to their
corresponding power flows. However, this method does not
Table 1
Different values of exponent.
Load models Values of a
CP a=0
CC a=1
CI a=2
DPSD;u
t and DQSD;u
t allocated real and reactive power losses,
respectively, of branch t to load connected at receiving
end of branch u
DPSG;u
t and DQStG;u allocated real and reactive power losses,
respectively, of branch t to DG connected at receiving
end of branch u
0u
P0u
D;t and Q D;t updated value of PD,t and QD,t, respectively, when
connected at receiving end of branch u
0u
P0u
G;t and Q G;t updated value of PG,t and QG,t, respectively, when
connected at receiving end of branch u
DPSD;t and DQSD;t total allocated real and reactive power losses,
respectively, to load at receiving end of branch t
DPSG;t and DQSG;t total allocated real and reactive power losses,
respectively, to DG at receiving end of branch t
consider the interdependency of consumers and DGs, and allocates
entire network losses to consumers or DGs. The issues related to
loss allocation in radial DN with DG are addressed in [15]. It covers
the issues such as characteristic of loads and DGs, formulation of
the loss allocation problem for radial DNs with respect to transmis-
sion networks, and treatment of the root node in radial DNs. A
comparison of different practical algorithms is presented in [5]
for loss allocation in transmission networks.
In context of deregulated environment, Savier and Das [16] pre-
sented an exact method of loss allocation based on the relation
between node voltages and branch current in radial DN. They
implemented their method for traditional passive DN. Later, Savier
and Das [17] extended their method as in [16] for energy loss
allocation. Carpaneto et al. [18] presented a branch current decom-
position based loss allocation method by representing the power
loss in a branch as a function of branch current and load/DG
current at various nodes ahead of it in radial DN. Atanasovski
and Taleski [4] proposed a power summation method for loss
allocation (PSMLA) by establishing a direct relation between loss
in a branch and injected real and reactive power at various
nodes connected ahead of it. Further, they employed quadratic loss
allocation scheme in order to deal with cross-terms. Atanasovski
and Taleski [19] presented energy summation algorithm for
allocation of energy loss in DN with DG. It is a statistical approach
which uses daily load and generation curve. Using quadratic loss
allocation scheme for cross-terms, Costa and Matos [20] presented
a current based approach, which allocates entire variation of losses
to DGs by using upstream looking algorithm.
Brief literature review on various loss allocation techniques
presented above shows that these techniques deal with DG having
constant real and reactive power injection, and load having
constant power model for loss allocation in DNs. Practically, loads
normally encountered in low and medium voltage DNs are
dependent on the node voltage. Further, real and reactive power
injections by DG into network depend on technology employed
and resources available at the site. However, to the best of authors’
knowledge, the issue of loss allocation in radial DNs with various
DG types and voltage dependent load models has not been
addressed so far.
The present work proposes a new solution for the problem of
loss allocation considering the effects of various DG types and
voltage dependent load models. Based on power summation algo-
rithm, the proposed method adopts a branch oriented approach for
loss allocation in radial DNs. This method does not make any
assumptions and approximations, and hence it is an accurate,
simple, efficient, and useful methodology for loss allocation in
 K.M. Jagtap, D.K. Khatod / Electrical Power and Energy Systems 75 (2016) 173–186
(a)
(b)
(c)
Fig. 1. Various stages involved in backward sweep network reduction technique for n-node radial DN.
radial DNs. The developed method establishes a direct relation
between node voltages and injected power at a node to calculate
the losses in a branch and then employs a backward sweep
network reduction technique to allocate the network losses to
load/DG at various nodes. The network reduction technique starts
from terminal branches and ends at the root node. The proposed
method employs only outcomes of the power flow. This method
is tested on modified 9-node and 33-node radial DNs and the
obtained results are compared with those by power based PR
method and PSMLA for validation of the proposed method.
The rest of the paper consists of six sections. Section ‘Types of
DGs’ describes different types of DG. Section ‘Load modeling’
explains various voltage dependent load models. Section ‘Proposed
formulation for loss allocation’ deals with the mathematical
formulation of the proposed method of loss allocation.
175
Section ‘Results and discussion’ presents the results obtained
by proposed method on two radial DNs and comparison
of obtained results with those by other methods available
in the literature. Finally, Section ‘Conclusion’ draws the
conclusions.
Types of DGs
There are several bases to classify DG which is available in
[21]. In this paper, DGs are classified in the following four
groups depending on their real and reactive power delivering
capability:
1. Type-I DG: DG injecting only real power;
2. Type-II DG: DG injecting only reactive power;
176 K.M. Jagtap, D.K. Khatod / Electrical Power and Energy Systems 75 (2016) 173–186

CP load model CC load model CI load model Constant power (CP) load model: A CP load model is a static load
Manitude of voltage (in p.u.)

0.995
model in which the real and reactive power of a load does not vary
with the voltage magnitude.
0.985
Constant current (CC) load model: A CC load model is a static load
model in which the real and reactive power of a load varies linearly
0.975
with the voltage magnitude.
0.965 Constant impedance (CI) load model: A CI load model is a static
load model in which the real and reactive power of a load varies
0.955 with the square of voltage magnitude.
1 2 3 4 5 6 7 8 9
In general, for above mentioned static load models, the relation
Bus number between the real and reactive power of the load and the voltage
Fig. 2. Voltage profile of 9-node radial distribution network with various load
magnitude can be expressed by the following equations:
models and without DG.
PD;x þ j Q D;x ¼ ðP0;x þ j Q 0:x ÞjV r;x ja ð2Þ

3. Type-III DG: DG injecting real power and absorbing reactive The different value of exponent depends on the load models
power; and [23–26] and can be summarized in Table 1.
4. Type-IV DG: DG injecting both real and reactive power.

Type-I DG belongs to small power generation such as photo- Proposed formulation for loss allocation
voltaic, battery, and fuel cell, which are operated at unity power
factor (pf). This type of DG is connected to the network through The following considerations have been made to develop the
suitable power electronics based interface. proposed formulation:
Type-II DG belongs to synchronous condenser which is oper-
ated at zero pf to provide reactive power support to the system. 1. Starting from the root node which is always numbered as 1,
Type-III DG represents DG utilizing induction generators. This and different nodes are assigned a unique integer number
type of DG injects real power to the network but takes reactive sequentially;
power from network. 2. A branch is assigned a number equal to the one less than its
Type-IV DG is a generation employing synchronous generator receiving end node;
which is capable of delivering both real and reactive power to 3. Branch from substation to root node is marked as 0 and
the network. assumed as loss-less; and
In radial DNs, the apparent power supplied/absorbed by a DG 4. DGs are treated as a negative load, whereas consumers are
connected at receiving end of branch x can be expressed as follows: treated as a positive load.

8 The apparent power available at the receiving end of a branch x

> PG;x Type I DG Operating at unity pf
>
>
< jQ Type II DG Operating at zero pf
is the algebraic sum of all apparent power due to load and DG con-
G;x
SG;x ¼ ð1Þ nected at receiving end of branches ahead of it, power losses in dif-
>
> P G;x  j Q G;x Type III DG Operating at leading pf
>
: ferent branches supplied by it and reactive power injection due to
PG;x þ j Q G;x Type IV DG Operating at lagging pf half of shunt charging susceptances of different branches incident
to its receiving end as:
!
1 X
Load modeling Px þ j Q x ¼ ðPD;x  PG;x Þ þ j Q D;x  Q G;x  jV r;x j2 Bm
2 m2N x
(
The power consumption by majority of loads such as domestic, X
industrial, and commercial, encountered in radial DNs depends on þ ðPD;n  PG;n þ PSn Þ
voltage magnitude and frequency [22,23]. However for static n2K x
!)
analysis, the frequency deviation is insignificant, and thus only 1 X
the effects of the voltage variation on the real and reactive load þ j Q D;n  Q G;n þ QSn  jV r;n j2 Bn ð3Þ
2 m2N n
powers may be considered [24].
Depending upon the voltage dependency, following load Substituting real and reactive power of load models from Eq. (2)
models are suggested [24–26]: in Eq. (3), the following equation can be given:

Table 2
Performance of 9-node radial distribution network with various load models and DG types.

Without DG With Type-1 DG With Type-2 DG With Type-3 DG With Type-4 DG

CP CC CI CP CC CI CP CC CI CP CC CI CP CC CI
Real load (kW) 747.6 727.06 708.44 747.6 732.76 719.05 747.6 729.25 712.52 747.6 730.49 714.8 747.6 734.87 723.13
Reactive load (kVAr) 559.2 543.55 529.33 559.2 547.88 537.47 559.2 545.21 532.45 559.2 546.17 514.22 559.2 549.52 546.59
Real power loss (kW) 24.04 22.5 21.11 15.05 14.28 13.55 18.82 17.69 16.64 23.77 22.67 21.66 10.01 9.48 8.99
Reactive power loss 13.92 13.06 12.28 9.09 8.63 8.2 11.13 10.47 9.87 13.32 12.69 12.11 6.39 6.06 5.73
(kVAr)
Real power supplied 536.68 517.66 500.43 545.66 531.6 518.61 541.87 524.67 508.99 536.93 520.93 506.24 550.71 538.5 527.25
by root node (kW)
Reactive power 405.11 390.32 376.88 409.94 399.08 389.1 407.9 394.57 382.41 405.71 393.31 361.94 412.64 403.29 400.69
supplied by root
node (kVAr)
 K.M. Jagtap, D.K. Khatod / Electrical Power and Energy Systems 75 (2016) 173–186 177

10 To load without DG by proposed method To load with DG by proposed method

To DG by proposed method To load witout DG by PR method
8

Allocated losses (in kW)

To load with DG by PR method To DG by PR method
6 To load without DG by PSMLA To load with DG by PSMLA

0
1 2 3 4 5 6 7 8 9
-2
Bus number
-4

-6

(a)
10 To load without DG by proposed method To load with DG by proposed method
To DG by proposed method To load witout DG by PR method
8
Allocated losses (in kW)

To load with DG by PR method To DG by PR method

To load without DG by PSMLA To load with DG by PSMLA
6

0
1 2 3 4 5 6 7 8 9
-2
Bus number
-4

(b)
10 To load without DG by proposed method To load with DG by proposed method
To DG by proposed method To load witout DG by PR method
8
Allocated losses (in kW)

To load with DG by PR method To DG by PR method

6 To load without DG by PSMLA To load with DG by PSMLA

0
1 2 3 4 5 6 7 8 9
-2
Bus number
-4

-6

(c)
10 To load without DG by proposed method To load with DG by proposed method
8 To DG by proposed method To load witout DG by PR method
Allocated losses (in kW)

To load with DG by PR method To DG by PR method

6 To load without DG by PSMLA To load with DG by PSMLA
4
2
0
-2
1 2 3 4 5 6 7 8 9

-4
Bus number
-6
-8

(d)
Fig. 3. Loss allocation in 9-node radial distribution network for CP load model and with (a) Type-1 DG, (b) Type-2 DG, (c) Type-3 DG, and (d) Type-4 DG.
!
  1 X In later step, reactive power generated by half of shunt charging
Px þ jQ x ¼ P0;x jV r;x ja  PG;x þ j Q 0;x jV r;x ja  Q G;x  jV r;x j2 Bm susceptance is considered to be a part of reactive load at respective
2 m2N x
X n  ends of the branch. Now, the real and reactive power losses in a
þ P 0;n jV r;n ja  PG;n þ PSn branch x can be expressed by the following equation:
n2K x
!)
1 X V s;x  V r;x
þ j Q 0;n jV r;n ja  Q G;n þ QSn  jV r;n j2 Bn ð4Þ PSx þ j QSx ¼ ðPx þ j Q x Þ ð5Þ
2 m2N n V r;x
178 K.M. Jagtap, D.K. Khatod / Electrical Power and Energy Systems 75 (2016) 173–186

10 To load without DG by proposed method To load with DG by proposed method

To DG by proposed method To load witout DG by PR method
8
To load with DG by PR method To DG by PR method

Allocated losses (in kW)

6 To load without DG by PSMLA To load with DG by PSMLA

0
1 2 3 4 5 6 7 8 9
-2
Bus number
-4

-6
(a)
10 To load without DG by proposed method To load with DG by proposed method
To DG by proposed method To load witout DG by PR method
8 To load with DG by PR method To DG by PR method
Allocated losses (in kW)

To load without DG by PSMLA To load with DG by PSMLA

6

0
1 2 3 4 5 6 7 8 9
-2
Bus number
-4
(b)
10 To load without DG by proposed method To load with DG by proposed method
To DG by proposed method To load witout DG by PR method
8
To load with DG by PR method To DG by PR method
Allocated losses (in kW)

6 To load without DG by PSMLA To load with DG by PSMLA

0
1 2 3 4 5 6 7 8 9
-2
Bus number
-4

-6
(c)
10 To load without DG by proposed method To load with DG by proposed method
8 To DG by proposed method To load witout DG by PR method
To load with DG by PR method To DG by PR method
Allocated losses (in kW)

6 To load without DG by PSMLA To load with DG by PSMLA

4
2
0
-2 1 2 3 4 5 6 7 8 9

-4
Bus number
-6
-8
(d)
Fig. 4. Loss allocation in 9-node radial distribution network for CC load model and with (a) Type-1 DG, (b) Type-2 DG, (c) Type-3 DG, and (d) Type-4 DG.

   
Substituting V s;x ¼ cx þ j dx ; and V r;x ¼ ex þ j f x in Eq. (5) and PSx þ j QSx ¼ DPSD;x G;x
þ j DQSD;x G;x
x þ DPSx x þ DQSx
simplifying it results the following equation: X h   i
þ DPSD;n
x þ DPSx
G;n
þ j DQSD;n
x þ DQSx
G;n
ð7Þ
PSx þ j QSx ¼ ðcx Px  wx Q x Þ þ j ðwx Px þ cx Q x Þ ð6Þ
n2K x

where cx ¼ cx ee2x þd xf x
þf 2
cx f x
 1; and wx ¼ dxee2xþf 2 . Objective of any loss allocation problem states that summation
x x x x

The real and reactive power losses in the branch x are decom- of allocated losses to loads and DGs in the network must be equal
posed and assigned to loads and DGs connected at receiving end to total network losses. To collect exact amount of total network
of branches ahead of it as: losses without using normalization technique, proposed method
 K.M. Jagtap, D.K. Khatod / Electrical Power and Energy Systems 75 (2016) 173–186 179

10 To load without DG by proposed method To load with DG by proposed method

8 To DG by proposed method To load witout DG by PR method

Allocated losses (in kW)

To load with DG by PR method To DG by PR method
6 To load without DG by PSMLA To load with DG by PSMLA

0
1 2 3 4 5 6 7 8 9
-2
Bus number
-4

-6
(a)
10 To load without DG by proposed method To load with DG by proposed method
To DG by proposed method To load witout DG by PR method
8
Allocated losses (in kW)

To load with DG by PR method To DG by PR method

To load without DG by PSMLA To load with DG by PSMLA
6

0
1 2 3 4 5 6 7 8 9
-2
Bus number
-4
(b)
10 To load without DG by proposed method To load with DG by proposed method
To DG by proposed method To load witout DG by PR method
8
Allocated losses (in kW)

To load with DG by PR method To DG by PR method

To load without DG by PSMLA To load with DG by PSMLA
6

0
1 2 3 4 5 6 7 8 9
-2
Bus number
-4
(c)

10 To load without DG by proposed method To load with DG by proposed method

8 To DG by proposed method To load witout DG by PR method
Allocated losses (in kW)

To load with DG by PR method To DG by PR method

6 To load without DG by PSMLA To load with DG by PSMLA
4
2
0
-2 1 2 3 4 5 6 7 8 9

-4
Bus number
-6
-8
(d)
Fig. 5. Loss allocation in 9-node radial distribution network for CI load model and with (a) Type-1 DG, (b) Type-2 DG, (c) Type-3 DG, and (d) Type-4 DG.

employs backward sweep network reduction technique. This tech- n  1 (terminal branch). Using Eq. (3) for computation of power
nique begins with terminal branch of the network. In a n-node flow at the receiving end of branch n  1 and then substituting
radial DN as shown in Fig. 1(a), the load and DG connected at node the obtained expression in Eq. (6), the real and reactive power
n (end node) cause the power flow and thereby losses in the branch losses in branch n  1 can be given as follows:
180 K.M. Jagtap, D.K. Khatod / Electrical Power and Energy Systems 75 (2016) 173–186

1 CP load model
CC load model

Manitude of voltage (in p.u.)

0.99
CI load model
0.98
0.97
0.96
0.95
0.94
0.93
0.92
0.91
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
Bus number

Fig. 6. Voltage profile of 33-node radial distribution network with various load models and without DG.

Table 3
Summary of loss allocation (kW) for 9-node radial distribution network.

Proposed method Power based PR method PSMLA

Without DG With DG Without DG With DG Without DG With DG
Load Load DG Total Load Load DG Total Load Load DG Total
CP Type-1 24.04 20.02 4.97 15.05 24.04 19.07 4.02 15.05 24.04 18.8 3.75 15.05
Type-2 18.25 0.57 18.82 18.74 0.08 18.82 20.9 2.08 18.82
Type-3 24.43 0.66 23.77 27.44 3.67 23.77 22.12 1.65 23.77
Type-4 14.35 4.34 10.01 13.34 3.33 10.01 15.75 5.74 10.01
CC Type-1 22.5 19.04 4.76 14.28 22.5 18.19 3.91 14.28 22.5 17.99 3.71 14.28
Type-2 17.08 0.61 17.69 17.59 0.1 17.69 19.74 2.05 17.69
Type-3 23.4 0.73 22.67 26.21 3.54 22.67 21 1.67 22.67
Type-4 13.65 4.17 9.48 12.71 3.23 9.48 15.18 5.7 9.48
CI Type-1 21.11 18.15 4.6 13.55 21.11 17.34 3.79 13.55 21.11 17.23 3.68 13.55
Type-2 15.99 0.65 16.64 16.53 0.11 16.64 18.66 2.02 16.64
Type-3 22.78 1.12 21.66 25.09 3.43 21.66 19.97 1.69 21.66
Type-4 13.02 4.03 8.99 12.12 3.13 8.99 14.65 5.66 8.99

  
PSn1 þ jQSn1 ¼ cn1 ðP D;n1  PG;n1 Þ  wn1 Q D;n1  Q G;n1 After allocating the losses of terminal branch n  1 to load and
  
þ j wn1 ðPD;n1  PG;n1 Þ þ cn1 Q D;n1  Q G;n1 ð8Þ DG at its receiving node, these allocated losses are added to respec-
tive power ratings of load and DG so as to set their updated power
The losses in terminal branch n  1 should be allocated ratings. Then, load and DG with their updated ratings are con-
between the load and DG connected at its receiving end (i.e. node nected at receiving end of branch n  2 and branch n  1 is elimi-
n). Hence, using Eq. (7) for branch n  1 results in the following nated as shown in Fig. 1(b). The following equations are used to
relation: compute the updated ratings of load and DG:
   
PSn1 þ j QSn1 ¼ DPSD;n1 G;n1
þ j DQSD;n1 G;n1    
n1 þ DPSn1 n1 þ DQSn1 0n2
P0n2 D;n1
D;n1 þ jQ D;n1 ¼ P D;n1 þ DPSn1 þ j Q D;n1 þ DQSD;n1
n1 ð12Þ
ð9Þ
Now, rearranging different terms of RHS of Eq. (8) and equating    
0n2
it to RHS of Eq. (9), the allocation of losses of terminal branch n  1 P0n2 G;n1
G;n1 þ j Q G;n1 ¼ P G;n1 þ DPSn1 þ j Q G;n1 þ DQSG;n1
n1 ð13Þ
to the load and DG connected at its receiving end can be expressed
Now, branch n  2 becomes the terminal branch of reduced net-
by the following equations:
work (containing n  2 branches) as shown in Fig. 1(b) and its
 
n1 þ j DQSn1 ¼ cn1 P D;n1  wn1 Q D;n1
DPSD;n1 D;n1
power losses are allocated among load PD;n2 þ j Q D;n2 connected
  at its receiving node and new ratings of load and DG that are
þ j wn1 PD;n1 þ cn1 Q D;n1 ð10Þ
0n2 0n2
P0n2 0n2
D;n1 þ j Q D;n1 and P G;n1 þ j Q G;n1 , respectively, using a similar
  method as in case of branch n  1 of original network.
DPSG;n1
n1 þ j DQSG;n1
n1 ¼  cn1 PG;n1  wn1 Q G;n1
  After this, updated ratings of loads and DGs connected at receiv-
 j wn1 PG;n1 þ cn1 Q G;n1 ð11Þ ing end of branch n  2 are computed using equations similar to as

Table 4
Difference of allocated losses (kW) to nodes 4 and 7 of 9-node radial distribution network with CP load model.

Without DG Type of DG With DG

Proposed method Power based PR method PSMLA Proposed method Power based PR method PSMLA
0.03 0 0 Type-1 0.03 0 0
Type-2 0.02 0 0
Type-3 0.05 0 0
Type-4 0.02 0 0
 K.M. Jagtap, D.K. Khatod / Electrical Power and Energy Systems 75 (2016) 173–186 181

given in Eqs. (12) and (13). Now, loads and DGs at receiving end of
branch n  2 are connected to receiving end of branch n  3 with

75.62
55.91
3534.53
2191.39

1327.95
2530.16
their updated ratings and branch n  2 is eliminated. The process

CI
of allocation of losses; updating of loads and DGs power with their
allocated losses; and elimination of branches are continued till root

82.11
3167.79
2241.21

60.27
2156.94
1373.41
node occurred as shown in Fig 1(c). At the end of this procedure,

With Type-4 DG
there is no branch exists in the network, and all the loads and

CC
DGs with their updated ratings are connected at the root node as

2696.62
shown in Fig. 1(c).

1427.1
89.65
65.37
3715
2300
The power at the receiving end of branch 0 becomes equal to

CP
the updated ratings of all the loads and DGs connected at root
node. It is also equal to the sum of all loads, DGs and branch losses

3427.47

1148.47
2104.29

148.29
207.68

2291.05
in the network and can be expressed as follows:

CI
X n 00  
00 00
o
P0 þ j Q 0 ¼ P D;n  P00
G;n þ j Q D;n  Q G;n

3558.54
2193.17
224.99
160.04

1225.6
2404.8
With Type-3 DG
n2K 0
X  

CC
¼ ðPD;n  P G;n þ PSn Þ þ j Q D;n  Q G;n þ QSn ð14Þ

1317.11
n2K 0

247.54
175.36
2538.7
3715
2300
Now, the total real power losses allocated to load and DG

CP
connected at the receiving end of a branch can be determined by
taking the difference between corresponding updated ratings

3451.33
2124.25

2396.88
1230.72
125.68
connected at root node and their actual ratings. The allocated

86
CI
losses to load and DG connected at receiving end of branch x can
be calculated from the following equations:

3571.49

139.37

2503.39
1301.37
95.12
2204.02
With Type-2 DG
   

CC
00
DPSD;x þ j DQSD;x ¼ P00
D;x  P D;x þ j Q D;x  Q D;x ð15Þ

1386.26
2630.41
155.87
106.21
3715
2300
   

CP
00
DPSG;x þ j DQ G;x ¼ P 00
G;x  P G;x þ j Q G;x  Q G;x ð16Þ

3484.56

2445.75
1266.17
2150.57

76.87
110.05
The sum of allocated losses to all the loads and DGs in the
network should be equal to total network losses in order to get fair CI
and accurate loss allocation. Hence, the following relation must be

2540.51
120.82
2218.68
3590.06

84.06

1327.09
satisfied:
With Type-1 DG
CC

X
PS þ j QS ¼ ðPSn þ j QSn Þ 2652.66
1399.81
133.58

n2K 0
92.66
3715
2300

X n   o
Performance of 33-node radial distribution network with various load models and DG types.

CP

¼ DPSD;n þ DPSG;n þ j DQSD;n þ DQSG;n ð17Þ

n2K 0
3401.51
2083.53
155.69
103.38

1172.62
2317.06
CI

176.61
117.51
3543.26

2437.88
1255.98
2181.02

Results and discussion

CC

The effectiveness of the proposed method has been tested on

Without DG

two test networks, i.e., 9-node and 33-node radial DNs. Both of
2583.57
1357.33
135.14
202.67

these test networks have been considered for two conditions, i.e.,
3715
2300
CP

without DG and with DG. In order to calculate the losses in net-

work with DG, both test networks have been modified by adding
Reactive power supplied by root node (kVAr)

different types of DGs at suitable locations. The real power injec-

tion by Type-1, Type-3 and Type-4 DG is taken as 25% of total net-
Real power supplied by root node (kW)

work real load (obtained with CP load model) [27]. The reactive
power injection/absorption by Type-2, Type-3 and Type-4 DG is
taken as 25% of total network reactive load (with CP load model).
Reactive power loss (kVAr)

Type-3 and Type-4 DGs are operated at leading and lagging pf,
respectively. The size and location of DG are kept fixed in the net-
Real power loss (kW)
Reactive load (kVAr)

work for different load models. To analyze the impact of DG type

on allocated losses, one type of DG is connected to the network
Real load (kW)

at a time. To analyze the effect of various load models on the allo-

cated losses, both networks have also been tested with CP, CC, and
CI load models. The results obtained by the proposed method have
Table 5

been compared with those by other methods such as power based

PR method and PSMLA with different load models.
182 K.M. Jagtap, D.K. Khatod / Electrical Power and Energy Systems 75 (2016) 173–186

60 To load without DG by proposed method To load with DG by proposed method

To DG by proposed method To load witout DG by PR method

Allocated losses (in kW)

40 To load with DG by PR method To DG by PR method
To load without DG by PSMLA To load with DG by PSMLA
20

0
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
-20
Bus number
-40

-60
(a)
60 To load without DG by proposed method To load with DG by proposed method
To DG by proposed method To load witout DG by PR method
Allocated losses (in kW)

40 To load with DG by PR method To DG by PR method

To load without DG by PSMLA To load with DG by PSMLA

20

0
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
Bus number
-20

-40
(b)
80 To load without DG by proposed method To load with DG by proposed method
To DG by proposed method To load witout DG by PR method
Allocated losses (in kW)

60 To load with DG by PR method To DG by PR method

To load without DG by PSMLA To load with DG by PSMLA
40

20

0
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
-20
Bus number
-40

-60
(c)

60 To load without DG by proposed method To load with DG by proposed method

To DG by proposed method To load witout DG by PR method
Allocated losses (in kW)

40
To load with DG by PR method To DG by PR method
20 To load without DG by PSMLA To load with DG by PSMLA

0
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
-20
Bus number
-40
-60
-80
-100
(d)
Fig. 7. Loss allocation in 33-node radial distribution network for CP load model and with (a) Type-1 DG, (b) Type-2 DG, (c) Type-3 DG, and (d) Type-4 DG.

9-node test radial distribution network
A 11 kV, 9-node radial DN is used to observe the effects of var-
ious load models and DG types on the results of allocated losses.
The line and load data of this test network are taken from [16].
Fig. 2 shows the voltage profile of 9-node network with CP, CC
and CI load models without DG. It can be observed from this figure
that the lowest voltage is observed at node 9 for all the load mod-
els. Hence, DG is connected at node 9. For different load models,
Table 2 shows the real and reactive power of network load, net-
work losses and power supplied by the root node without and with
DG. It can be noticed from this table that the network load is
maximum with CP load model, while it is minimum with CI load
model. That is why among the three load models, CP load model
gives worst voltage profile, while CI load model provides best volt-
age profile as seen from Fig. 2. The voltage profile with CC load
model lies in between those with CP and CI load models. Because
of the lower load and better voltage profile, the network losses
with CI load model are also less as compared to those with CP load
model. Similar observation is made for supplied power from root
node for different load models from Table 2 irrespective of DG type
employed.
With different types of DG, Figs. 3–5 show the node-wise
allocated losses with CP, CC and CI load models, respectively.
 K.M. Jagtap, D.K. Khatod / Electrical Power and Energy Systems 75 (2016) 173–186 183

60 To load without DG by proposed method To load with DG by proposed method

To DG by proposed method To load witout DG by PR method
40

Allocated losses (in kW)

To load with DG by PR method To DG by PR method
To load without DG by PSMLA To load with DG by PSMLA
20

0
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
-20
Bus number
-40

-60
(a)
60 To load without DG by proposed method To load with DG by proposed method
To DG by proposed method To load witout DG by PR method
Allocated losses (in kW)

40 To load with DG by PR method To DG by PR method

To load without DG by PSMLA To load with DG by PSMLA

20

0
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
-20
Bus number
-40
(b)
80 To load without DG by proposed method To load with DG by proposed method
To DG by proposed method To load witout DG by PR method
60
Allocated losses (in kW)

To load with DG by PR method To DG by PR method

To load without DG by PSMLA To load with DG by PSMLA
40

20

0
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
-20
Bus number
-40

-60
(c)
60 To load without DG by proposed method To load with DG by proposed method
To DG by proposed method To load witout DG by PR method
40
Allocated losses (in kW)

To load with DG by PR method To DG by PR method

To load without DG by PSMLA To load with DG by PSMLA
20

0
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
-20
Bus number
-40

-60

-80
(d)
Fig. 8. Loss allocation in 33-node radial distribution network for CC load model and with (a) Type-1 DG, (b) Type-2 DG, (c) Type-3 DG, and (d) Type-4 DG.

In these figures, the loss allocation obtained by power based PR
and PSMLA methods is also plotted for the sake of comparison.
Table 3 summarizes the total allocated losses of consumer and
DG. Without DG integration in the network, total losses, supplied
by root node, are caused by various loads and hence positive losses
are allocated to all the loads by all the methods as seen from Table 3
and Figs. 3–5. This positive value of allocated loss indicates that
consumers should pay for losses caused by them. With DG integra-
tion in the network, total power demand of the network remains
unchanged, but network losses are reduced as in Table 2. For this
case, the proposed method allocates positive losses to all the
consumers and negative losses to Type-1, Type-3 and Type-4 DGs
as seen from Table 3 and Figs. 3–5. This negative value of allocated
loss to DG may be viewed as a reward to DG for its contribution
toward loss reduction in the network. From Table 3 and Figs. 3–5
(a), it is observed that the proposed method assigns more reward
to Type-1 DG due to only real power injection by it and provides
less cross subsidies to consumers as compared to power based
PR and PSMLA methods. As observed from Table 3 and Figs. 3–5
(b), proposed and power based PR methods do not assign any
reward to Type-2 DG as it injects only reactive power in the
network. However, in this case, PSMLA method assigns reward to
184 K.M. Jagtap, D.K. Khatod / Electrical Power and Energy Systems 75 (2016) 173–186

60 To load without DG by proposed method To load with DG by proposed method

To DG by proposed method To load witout DG by PR method
40 To load with DG by PR method To DG by PR method

Allocated losses (in kW)

To load without DG by PSMLA To load with DG by PSMLA
20

0
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
-20
Bus number
-40

-60
(a)

60 To load without DG by proposed method To load with DG by proposed method

To DG by proposed method To load witout DG by PR method
40 To load with DG by PR method To DG by PR method
Allocated losses (in kW)

To load without DG by PSMLA To load with DG by PSMLA

20

0
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
-20
Bus number

-40
(b)

60 To load without DG by proposed method To load with DG by proposed method

To DG by proposed method To load witout DG by PR method
40 To load with DG by PR method To DG by PR method
Allocated losses (in kW)

To load without DG by PSMLA To load with DG by PSMLA

20

0
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
-20
Bus number
-40

-60
(c)

60 To load without DG by proposed method To load with DG by proposed method

To DG by proposed method To load witout DG by PR method
40 To load with DG by PR method To DG by PR method
Allocated losses (in kW)

To load without DG by PSMLA To load with DG by PSMLA

20

0
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
-20
Bus number
-40

-60

-80
(d)
Fig. 9. Loss allocation in 33-node radial distribution network for CI load model and with (a) Type-1 DG, (b) Type-2 DG, (c) Type-3 DG, and (d) Type-4 DG.

Type-2 DG and provides less cross subsidies to consumers. With
Type-3 DG (injecting real power, but absorbing reactive power),
power based PR method assigns more reward to DG as compared
to proposed method, and PSMLA method does not provide any
reward to DG as seen from Table 3 and Figs. 3–5(c). In this case,
proposed method provides no cross subsidies but power based
PR method provides over cross subsidies to consumers which is
not acceptable in the electrical market. Now, from Table 3 and
Figs. 3–5(d), it can be observed that PSMLA method provides more
reward to Type-4 DG as compared to proposed and power based PR
methods. The reward assigned to DG by proposed method is
moderate and lies in between those provided by power based PR
and PSMLA methods. Also, with Type-4 DG, cross subsidies pro-
vided by proposed method are less and nearly equal to the PSMLA
method.
Now, the performance of different methods on the basis of dis-
crimination between the consumers of same ratings at different
electrical locations is also evaluated. The real power consumptions
at nodes 4 and 7 are same (there is a small mismatch on reactive
power consumption at these nodes), but electrical locations of
 K.M. Jagtap, D.K. Khatod / Electrical Power and Energy Systems 75 (2016) 173–186 185

Table 6
Summary of loss allocation (kW) for 33-node radial distribution network.

Proposed method Power based PR method PSMLA

Without DG With DG Without DG With DG Without DG With DG
Load Load DG Total Load Load DG Total Load Load DG Total
CP Type-1 202.67 179.15 45.57 133.58 202.67 175.23 41.65 133.58 202.67 190.49 56.91 133.58
Type-2 132.74 23.13 155.87 147.71 8.16 155.87 184.75 28.88 155.87
Type-3 258.8 11.26 247.54 300.36 52.82 247.54 222.85 24.69 247.54
Type-4 170.38 80.73 89.65 118.02 28.37 89.65 173.03 83.38 89.65
CC Type-1 176.61 161.35 40.53 120.82 176.61 160.12 39.3 120.82 176.61 173.8 52.98 120.82
Type-2 118.5 20.87 139.37 131.4 7.97 139.37 165.94 26.57 139.37
Type-3 231.69 6.7 224.99 274.63 49.64 224.99 197.43 27.56 224.99
Type-4 157.71 75.6 82.11 108.97 26.86 82.11 161.83 79.72 82.11
CI Type-1 155.69 146.54 36.49 110.05 155.69 147.23 37.18 110.05 155.69 159.78 49.73 110.05
Type-2 102.01 23.67 125.68 117.91 7.77 125.68 150.32 24.64 125.68
Type-3 211.65 3.97 207.68 254.81 47.13 207.68 177.82 29.86 207.68
Type-4 146.15 70.53 75.62 101.09 25.47 75.62 152.31 76.69 75.62

Table 7
Difference of allocated losses (kW) at selected nodes of 33-node radial distribution network with CP load model.

Difference of allocated losses of nodes 6 and 28 Difference of allocated losses of nodes 9 and 10
Without DG With DG Without DG With DG
Proposed Power PSMLA Type Proposed Power PSMLA Proposed Power PSMLA Type Proposed Power PSMLA
method based PR of DG method based PR method based PR of DG method based PR
method method method method
1.44 0 0.22 Type-1 0.7 0 0.21 0.37 0 0.27 Type-1 0.41 0 0.27
Type-2 0.51 0 0.22 Type-2 0.37 0 0.26
Type-3 1.31 0 0.23 Type-3 0.38 0 0.29
Type-4 0.46 0 0.2 Type-4 0.25 0 0.23

these nodes are different. The load connected at node 4 is electri-
cally closer to the root node as compared to that at node 7. Table 4
shows the differences of allocated losses to consumers at nodes 4
and 7 with CP load model. From this table, it is seen that the differ-
ence of allocated losses to consumers at nodes 4 and 7 without DG
by proposed method is 0.03 kW, while the same by power based PR
and PSMLA methods is 0 kW. Further, with different types of DG
also, the difference of allocated losses to consumers at nodes 4
and 7 by proposed method is more as compared to power based
PR and PSMLA methods. This indicates that the proposed method
is able to identify the electrical locations and contribution of con-
sumers in total network losses.
33-node test radial distribution network
The proposed method is also tested on a 12.66 kV, 33-node
radial DN. The line and load data of this test network are taken
from [28].
Fig. 6 shows the voltage profile of 33-node network with CP, CC
and CI load models without DG. From this figure, it can be observed
that among three load models, CI load model has better voltage
profile than CP and CC load models. The voltage profile with CC
load model lies in between those with CP and CI load models. DG
is connected to node 33 which has poor voltage and is far away
from the root node. For different load models, Table 5 shows the
real and reactive power of network load, network losses and power
supplied by root node without and with DG. From this table, it can
be noticed that the network load is maximum with CP load model,
while it is minimum with CI load model. Because of the lower load
and better voltage profile, the network losses with CI load model
are also less as compared to those with CP load model. Similar
observation is made for supplied power from root node with differ-
ent load models from Table 5 irrespective of DG type employed.
With different types of DG, Figs. 7–9 show the node-wise allo-
cated losses with CP, CC and CI load models, respectively. In these
figures, the loss allocation obtained by power based PR and PSMLA
methods is also plotted for the sake of comparison. Table 6 summa-
rizes the total allocated losses of consumer and DG. Without DG
integration in the network, positive losses are allocated to all the
loads by all the methods as seen from Table 6 and Figs. 7–9. With
DG integration in the network, the proposed method allocates pos-
itive losses to all the consumers and negative losses to Type-1,
Type-3 and Type-4 DGs as seen from Table 6 and Figs. 7–9. PSMLA
method assigns more reward to Type-1 DG and provides less cross
subsidies to consumers as compared to power based PR and pro-
posed methods. With Type-1 DG, the proposed method assigns
moderate reward and cross subsidies to DG and consumers,
respectively. As observed from Table 6 and Figs. 7–9(b), proposed
and power based PR methods do not assign any reward to Type-
2 DG as it injects only reactive power in the network. However,
in this case, PSMLA method assigns reward to Type-2 DG and
provides less cross subsidies to consumers. With Type-3 DG, the
network losses are increased as seen from Table 5, still power
based PR method assigns reward to DG and provides over cross
subsidies to consumers. PSMLA method does not provide any
reward to DG as seen from Table 6 and Figs. 7–9(c). In this case,
the performance of the proposed method is moderate. Now, from
Table 6 and Figs. 7–9(d), it can be observed that PSMLA method
provides more reward to Type-4 DG as compared to proposed
and power based PR methods. The reward assigned to DG by pro-
posed method lies in between those provided by power based PR
and PSMLA methods. Also, with Type-4 DG, cross subsidies pro-
vided by proposed method are nearly equal to the PSMLA method.
Now, the real and reactive power consumptions at nodes 6 and
28 are same, but electrical locations of these nodes are different.
The load connected at node 6 is electrically closer to the root node
186 K.M. Jagtap, D.K. Khatod / Electrical Power and Energy Systems 75 (2016) 173–186
as compared to that at node 28. Further, the power consumptions
at nodes 9 and 10 are same and these nodes are situated close to
each other. Table 7 shows the differences of allocated losses to con-
sumers at nodes 6 and 28; and at nodes 9 and 10 with CP load
model. From this table, it is seen that the differences of allocated
losses to consumer at nodes 6 and 28 without DG are 1.44 kW,
0 kW and 0.22 kW by proposed, power based PR and PSMLA meth-
ods, respectively. Similarly, with different types of DG, proposed
method creates more difference in allocated losses at nodes 6
and 28 as compared to power based PR and PSMLA methods. Fur-
ther, from Table 7, it is seen that the differences of allocated losses
to consumer at nodes 9 and 10 without DG are 0.37 kW, 0 kW and
0.27 kW by proposed, power based PR and PSMLA methods,
respectively. With different types of DG, proposed method is able
to differentiate between consumers of same rating and located
close to each other in a better way as compared to power based
PR and PSMLA methods.
From above mentioned results and discussion, it can be con-
cluded that the proposed method has potential to identify the
exact electrical location of consumers without DG. It is also able
to differentiate between the same rating of consumers located any-
where in the network and allocates the losses based on their power
consumption and contribution in the total network losses. It
always provides reward to DG which injects real power to the
network.
Conclusion
In the paper, a new method for loss allocation in radial DN with
DG has been presented. Based on power summation method, the
proposed method is a branch oriented approach which considers
the network power flow and establishes a direct relation between
real and reactive power at the receiving end of branch and its
losses. In order to eliminate the additional steps of normalization,
the proposed method employs backward sweep network reduction
technique. In order to get fair allocation of losses, this method does
not consider any assumptions and approximations; therefore, it is
a simple, efficient and easy method for loss allocation in radial
DNs. Further, this method does not require any additional informa-
tion other than the result of power flow. While allocating losses to
a participant, proposed method considers not only power injec-
tion/absorption by it but also its location, characteristic, and con-
tribution in the total network losses and thereby sends correct
economic signal. The proposed method has been tested on two test
networks and the obtained results are compared with those by
other existing methods to show its effectiveness.
References
[1] Jagtap KM, Khatod DK. Impact of different types of distributed generation on
radial distribution network. In: IEEE conf. ICROIT Faridabad India; 2014. p.
473–6.
[2] Khatod DK, Pant V, Sharma J. Evolutionary programming based optimal
placement of renewable distributed generators. IEEE Trans Power Syst
2013;28:683–95.
[3] Gómez A, Riquelme JM, González T, Ruiz E. Fair allocation of transmission
power losses. IEEE Trans Power Syst 2000;15:184–8.
[4] Atanasovski M, Taleski R. Power summation method for loss allocation in
radial distribution networks with DG. IEEE Trans Power Syst 2011;26:2491–9.
[5] Conejo AJ, Arroyo JM, Alguacil N, Guijarro AL. Transmission loss allocation: a
comparison of different practical algorithms. IEEE Trans Power Syst
2002;17:571–6.
[6] Gonzalez JJ, Basagoiti P. Spanish power exchange market and information
system. Design concepts, and operating experience, IEEE conf. PICA Santa Clara
USA; 1999. p. 245–52.
[7] Shirmohammadi D, Gribik PR. Evaluation of network capacity use for wheeling
transection. IEEE Trans Power Syst 1989;4:1405–13.
[8] Happ HH. Cost of wheeling methodologies. IEEE Trans Power Syst
1994;9:147–55.
[9] Mutale J, Strbac G, Curcic S, Jenkins N. Allocation of losses in distribution
systems with embedded generation. IEE Proc Gener Trans Distrib
2000;147:7–14.
[10] Galiana FD, Conejo AJ, Kockar I. Incremental transmission loss allocation under
pool dispatch. IEEE Trans Power Syst 2002;17:26–33.
[11] Conejo AJ, Galiana FD, Kochar I. Z-bus loss allocation. IEEE Trans Power Syst
2001;16:105–10.
[12] Fang WL, Ngan HW. Succinct method for allocation of network losses. IEE Proc
Gener Trans Distrib 2002;149:171–4.
[13] Bialek JW. Tracing the flow of electricity. IEE Proc Gener Trans Distrib
1996;143:313–20.
[14] Bialek JW, Kattuman PA. Proportional sharing assumption in tracing
methodology. IEE Proc Gener Trans Distrib 2004;151:523–6.
[15] Carpaneto E, Chicco G, Akilimali JS. Characterization of the loss allocation
techniques for radial systems with distributed generation. Elect Power Syst
Res 2008;78:1396–406.
[16] Savier JS, Das D. An exact method for loss allocation in radial distribution
systems. Int J Elect Power Energy Syst 2012;36:100–6.
[17] Savier JS, Das D. Energy loss allocation in radial distribution systems: a
comparison of practical algorithms. IEEE Trans Power Del 2007;22:2473–80.
[18] Carpaneto E, Chicco G, Akilimali JS. Branch current decomposition method for
loss allocation in radial distribution systems with distributed generation. IEEE
Trans Power Syst 2006;21:1170–9.
[19] Atanasovski M, Taleski R. Energy summation method for loss allocation in
radial distribution networks with DG. IEEE Trans Power Syst
2012;27:1433–40.
[20] Costa PM, Matos MA. Loss allocation in distribution networks with embedded
generation. IEEE Trans Power Syst 2004;19:384–9.
[21] Hung DQ, Mithulananthan N, Bansal RC. Analytical expression for DG
allocation in primary distribution network. IEEE Trans Energy Convers
2010;25:814–20.
[22] Liu J, Salama MMA, Mansour RR. An efficient power flow algorithm for
distribution systems with polynomial load. Int J Electr Eng Edu
2002;39:372–86.
[23] El-Hawary ME, Dias LG. Correspondence incorporation of load models in load
flow studies: form of model effects. IEEE Proc 1987;134:27–30.
[24] Haque MH. Load flow solution of distribution systems with voltage dependent
load models. Electr Power Syst Res 1996;36:151–6.
[25] Eminogle U, Hocaoglu MK. A new power flow method for radial distribution
system including voltage dependent load models. Electr Power Syst Res
2005;76:106–14.
[26] Ribeiro JR, Lang FJ. A new aggregation method for determining composite load
characteristics. IEEE Trans Power App Syst 1982;101:2869–75.
[27] Ackermann T, Andersson G, Soder L. Distributed generation: a definition. Electr
Power Syst Res 2001;57:195–204.
[28] Baran M, Wu FF. Network reconfiguration in distribution systems for loss
reduction and load balancing. IEEE Trans Power Del 1989;4:1401–7.