You are on page 1of 6

384 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 19, NO.

1, FEBRUARY 2004

Loss Allocation in Distribution Networks With


Embedded Generation
Paulo Moisés Costa and Manuel A. Matos, Member, IEEE

Abstract—This paper addresses the allocation of electrical II. EFFICIENT LOSS ALLOCATION
losses in distribution networks with embedded generation, in a
liberalized environment. The nonlinear nature of the issue, the loss
changes due to voltage variation and, specially, the contribution A large number of different methods for loss allocation have
of embedded generation to loss variation are considered. The been published. Most of them are dedicated to transmission net-
proposed method is based on tracing the real and imaginary works, and can be split in three main families: 1) pro rata proce-
parts of the currents and has two steps. First, the losses in the dures, 2) marginal procedures, and 3) proportional procedures
distribution network, in the absence of embedded generation,
are allocated to the consumers (or their providers). Second, the
[3].
variations in the losses that result from the influence of embedded In the pro rata procedures, the losses allocated to a gener-
generation are allocated to the generators. These variations are ator or a consumer are related to its level of power, disregarding
a measure of the avoided or added costs related to losses. In the the relative location within the network [3], [4]. Therefore, re-
allocation process, made in a branch basis, both real and reactive
motely located generators or demands benefit, at the expense of
powers are considered. The methodology presented in this paper
can be used to evaluate embedded generation incentives or to all others [4]. On the other hand, procedures based on marginal
design tariffs for the use of the distribution network. costs result in over-recovery of losses, and a process of normal-
Index Terms—Deregulation, dispersed generation, distribution ization is thus needed [3], [4]. Proportional methods provide a
networks, loss allocation. computationally efficient procedure for loss allocation [4], but
they allocate all the losses solely to the consumers or to the
generators (unless the values of losses to be allocated to gen-
I. INTRODUCTION erators and consumers are fixed arbitrarily [4]). This is also a
frequent situation in the pro rata and marginal methods. Most
U NLIKE generation and sale of electricity, transmission
and distribution activities are generally considered as
a natural monopoly. The costs of these activities need to be
of the methods of these three families also ignore the reactive
power flow in the process of loss allocation.
allocated to the network users, namely through network use In conclusion, application of these methods in distribution
tariffs, with a focus on the true impact they have on these costs networks can be difficult, due to two reasons.
[1]. Among others, distribution power losses are one of the
1) If the reactive power flow is not considered in the al-
costs to be allocated. This allocation has two main difficulties:
location process, situations of cross-subsidization will
the nonlinearity between losses and the delivered power, which
emerge.
complicates the determination of the impact of each user
2) Distinction between the contribution to the overall losses
consumption in the network losses [1]; and the definition of
of the two categories of users (consumers and embedded
the way losses should be shared between users of the network
generators) is essential to achieve an efficient technical
(embedded generators and consumers) [2]. No matter the
and economical allocation.
method of loss allocation, a set of fundamental principles must
be respected, in order to guarantee a correct allocation. The last reason implies the allocation to each user of its
First, the method must be objective, easy to understand and impact in the losses of each network element. Therefore, loss
based on real data of the network [2]. It also has to be econom- allocation must be made in a branch basis. Several methods
ically efficient, avoiding crossed subsidization between users which use this principle have been proposed [1], [5], [6].
and sending out economical signals that may lead to the increase Some of these methods use the proportional sharing principle,
of the efficiency of the network [2]. assuming that the flows entering any node are distributed
Finally, the method has to recover the total value of losses, proportionally between the outflows. Generally, these methods
avoid discrimination between users, be consistent when applied allocate the losses only to the consumers and allow cross-sub-
to different situations and be applicable to open electricity mar- sidization of two kinds: one related to the fact that the power
kets [2]. losses of a specific branch are a quadratic function with non
negligible crossed terms; the other related to the reactive
Manuscript received March 6, 2003. power flow influence, once these methods do not consider this
P. M. Costa is with the Department of Electrical Engineering, School of
Technology of Viseu, Polytechnic Institute of Viseu, Viseu 3504-510, Portugal influence in the allocation process. In [1], several methods with
(e-mail: paulomoises@elect.estv.ipv.pt). a careful allocation of the crossed terms are proposed, but some
M. A. Matos is with the Faculty of Engineering of the University of Porto problems remain, namely the losses are allocated only to the
and INESC Porto, Campus da FEUP, Porto 4200-465, Portugal (e-mail:
mmatos@inescporto.pt). consumers and only the real power flow is considered in the
Digital Object Identifier 10.1109/TPWRS.2003.820698 allocation process.
0885-8950/04$20.00 © 2004 IEEE
COSTA AND MATOS: LOSS ALLOCATION IN DISTRIBUTION NETWORKS WITH EMBEDDED GENERATION 385

The method proposed in this paper tries to contribute to Some authors divide the tracing algorithm in two variants, the
overcome the mentioned difficulties. The benefits/costs of em- “downstream-looking algorithm” and the “upstream-looking al-
bedded generation in terms of real power losses are quantified. gorithm” [6], [10]. The first variant allows the determination of
Crossed terms are conveniently allocated and the influence the contribution of each consumer to the current components in
of the reactive power flow is also considered. In particular, each network branch. The second variant allows the evaluation
the method allocates the corresponding losses to a user that of the contribution of each generator or network connection (to
only influences the reactive power flow in a specific branch, a the transmission system) to each branch.
feature not possible in methods that only use the active power By using the two variants of the tracing algorithm, it is pos-
flow (or current magnitude) in the allocation process. sible to evaluate the contribution of each user to the current com-
In the present approach, the embedded generators may be re- ponents in each branch. Therefore, the losses can be allocated
warded for avoided losses or have to pay for increased losses. to the consumers and to the embedded generators, considering
This principle was also used in other approaches, as reported their impact on it. This leads to stronger economical signals, be-
in [7] and [8]. The method, described in [7], for loss allocation cause each user will realize the true costs that he causes in each
when embedded wind generation is present, uses marginal loss element of the network.
coefficients that relate the changes in the system losses to the in-
cremental change in nodal active and reactive power injections. III. PROPOSED METHOD
It was developed to quantify the benefits and costs of wind em- The algorithm of the proposed method operates in three
bedded generation in terms of real power losses in distribution phases: the first for loss allocation to consumers (or their
systems and it’s not suitable to do the complete allocation of the providers), the second for loss allocation to the embedded
losses between all users (generators and consumers). generators and the third for allocation of the remaining loss
In [8], two schemes for loss allocation in distribution net- variations (related to the change in the voltage profile).
works are proposed. The first is based on the concept of marginal
losses and needs a process of reconciliation (because it recovers A. Consumer’s Loss Allocation
approximately twice the amount of losses). The second relates In this phase, the losses on the network (without considering
losses directly to power injections and does not need the recon- the embedded generators) are evaluated and allocated to the con-
ciliation process. Both schemes do not seem to be able to do the sumers. First, an AC power flow is run without the embedded
allocation of the crossed terms of losses. Besides, both methods generators in the network (base case). This study allows the
allow the consumers to be rewarded in some situations, instead determination of the active and reactive power flows in each
of paying for losses. Another method that yields negative alloca- branch and of the voltages at each node. The real and imaginary
tions that rewards loads is proposed in [4]. This method, devel- components of the currents are then evaluated in each branch.
oped for transmission networks, uses the currents in the process Defining the current in branch i as , we will
of loss allocation. obtain, after applying the “downstream-looking algorithm”, a
These last three methods have a different philosophy than the separation of the contribution of each load j to the real and imag-
one proposed in this paper, which assumes that: inary parts of the current:
1) The consumers always have to pay losses (which results
(1)
from their use of the network when embedded generation
is not considered);
2) The variations in losses (positive or negative) that take So, the losses in branch i may be expressed as:
place after the consideration of embedded generators are
allocated to the producers.
(2)
This philosophy allows the transmission of adequate econom-
ical signals to the users of the network [2].
Once the currents are the dominant factor in the determina- where represents the resistance of branch i and Z is the
tion of losses [4], the contributions of each user to the real and number of consumers.
imaginary part of the currents of each element of network are The development of this expression shows that the consumer
used in the method proposed in this paper. These contributions j has impact in the following terms:
can be obtained by using tracing algorithms, like the ones de-
scribed in [6], [9] or [10]. These algorithms are based on the
proportional sharing principle previously mentioned.
Although they have been developed to trace powers, their ap- which are exclusively due to consumer j;
plication to trace currents is straightforward, since the propor-
tional sharing principle remains unchanged. The only require-
ment for the input data is that Kirchhoff’s Current Law must be
accomplished in all the nodes in the network [6]. As a result,
when the users of the network show different power factors, the which result of the simultaneous influence of consumer j and of
process of tracing currents must be made separately for the real the remaining consumers k in the components of the current in
and imaginary parts of the current. the element i.
386 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 19, NO. 1, FEBRUARY 2004

These crossed terms must be also allocated. The allocation The referred variations are then allocated to the embedded
can be made in a proportional, quadratic or geometric way, as generators responsible for them. However, we first need to cal-
mentioned in [1]. The quadratic scheme is adopted, once it is culate some auxiliary quantities, through the application of the
more consistent with the relationship between losses and cur- “upstream-looking algorithm” presented in [6]:
rents. Based on this assumption, the losses of the element i of contribution of generator k to the real or imaginary
the network allocated to consumer j are: current of consumer j;
contribution, in the inverse direction, of generator k
to the real or imaginary component of the current in
branch i;
contribution, in the direct direction, of generator k
to the real or imaginary component of the current in
branch i;
(3) The allocation is now calculated according to the type of vari-
ations [2]. If , then the variation of the real or imagi-
nary components of the current in branch i, due to consumer j,
The global value of losses to be supported by consumer j re- that is assigned to embedded generator k is given by:
sults from the sum of the losses allocated to it in each element i
of the network, as shown in the following equation:

(4) (7)

where N represents the number of network elements. If , then:


If the losses are allocated to providers, then each one will
support the losses that were allocated to their customers.
(8)
B. Loss Allocation to Embedded Generators
In this phase, the loss variations due to the presence of the where:
generators are evaluated and allocated to them. A new power is the set of embedded generators that contribute to the
flow is run, now with the presence of embedded generators, and real or imaginary current of consumer j without con-
the new values of the real and imaginary parts of the current tributing, in direct direction, for the same component
in each branch are calculated. The application of the “down- of the current in branch i;
stream-looking algorithm” shows how each consumer j now is the set of embedded generators that contribute, in in-
contributes to the components of the current in each element verse direction, to the real or imaginary components of
i of the network, and . the current in branch i, and, simultaneously, contribute
At this point, an important definition for the operation of the to the same component of current of the consumer j.
method must be made. The way current components flow in is the set of embedded generators that contribute, in
each element of the network in the base case is, by definition, the direct direction, to the real or imaginary current in
the direct direction, and therefore is considered positive. In some branch i and, simultaneously, contribute to the same
branches (e.g. connection branches of embedded generators to component of the current in consumer j.
the network), where power flows only exist when embedded Two particular cases of allocation of the variations in the
generation is considered, a particular situation takes place. In components of the currents may occur. The first case happens
these cases, the direct direction is defined by the components of when a meshed distribution network exists, as shown in Fig. 1.
the current after the insertion of the embedded generators. The presence of the embedded generator leads to a decrease of
Therefore the values and can be positive or neg- active power flow (and real current) in branch 1-2 and conse-
ative, according to: quently to the reduction of losses in this branch. The avoided
losses must be allocated to the generator. It is easy to see that
if is in the direct direction . Thus, .
(5) This variation must be allocated using one of the (7). How-
if is in the inverse direction
ever, the application of the “upstream-looking algorithm” im-
Note that, in these and in the following equations, “x” is used putes to the generator the responsibility of a part of the real cur-
to represent either “p” or “q”, in order to avoid the duplication of rent in branch 1-2, in the direct direction. So, the application of
formulas. Once these values are calculated, the following step (7) is not possible and the variation is not allocated to the gen-
is the determination of the variations in the components of the erator.
currents in each element of the network due to each consumer. To solve this problem, the power produced by the embedded
The variations are given by: generator, that reaches node 1, should be considered as be-
longing to the network connection. Obviously, the contributions
(6) of the generator to the power flow (and real current) of branch
COSTA AND MATOS: LOSS ALLOCATION IN DISTRIBUTION NETWORKS WITH EMBEDDED GENERATION 387

Network connection erators are given by (2). When the embedded generators are con-
1
sidered, the value of losses in the same branch change to:

6 MW

2 MW 3

2 (10)

4 where H is the number of embedded generators.


5 Using the same principles described in Section III-A., (11) is
C1 1 MW
C2
used to allocate the loss variation in branch i to the embedded
8 MW
generator k.
Network connection

1
1 MW
G
2 MW

2
1 MW
6 MW
4
(11)
5 C1 1 MW
C2 8 MW
This process must be applied to all generators and elements
Fig. 1. Particular case of loss allocation. of the network, in order to allocate all loss variations. The
global value of losses that are allocated to embedded generator
k equals:
2-5 and to the load should be reduced according to that.
Therefore, the contribution of the embedded generator to the
real current in branch 1-2 will be null, and the avoided losses (12)
will be allocated to it. This procedure does not imply increment
of costs for any entity.
C. Voltage-Related Loss Variation
The second case happens when, in low load situations, the
embedded generators do not produce reactive power. In these The procedure described above allocates the major part of the
situations, it can be possible for the reactive power to flow from loss variations due to the presence of the embedded generators,
the network to the embedded generators, and so the imaginary but does not account for variations in losses due exclusively to
component of the current will take the same direction. the change in the voltage profile. For example, in Fig. 2, in the
This situation implies the existence of variations in imaginary absence of embedded generation, only the network connection
currents of some network elements due to embedded genera- supplies load C. When the embedded generation is operating,
tors. Those variations must be allocated to the generators. The the received power flow in branch 2-3 stays unaffected, but, due
“downstream-looking algorithm” allows to determine and to al- to the improvement of the voltage profile of the network, the cur-
locate those variations to each generator, considering fictitious rent in this branch decreases, as well as its losses. This variation
loads to represent the demand of the generators of an imaginary in the losses is not allocated to the generators by the algorithm
current. steps described so far.
The global value of variations in the components of current The total of these voltage related loss variations can be cal-
in each network element, allocated to each generator, can be culated using:
calculated by: (13)

where and are the values of total losses in the network,


(9) respectively with and without the embedded generators, and
is the total value of losses allocated to the embedded generators
by the previous steps of the algorithm.
The next step consists on the allocation of the variations in In order to allocate these variations to the embedded gen-
the losses of each element of the network to the generators. The erators, several alternatives could be considered [2]. The
losses in branch i before the consideration of the embedded gen- most straightforward way is to distribute them proportionally,
388 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 19, NO. 1, FEBRUARY 2004

Network connection 1

G2 2
G1 3 4
28
3 MW G2 5
2 15
9 13
2 MW 6
7
5 MW
14
3 11
10 8
C
G1
10 MW
27
12 16 17
Fig. 2. Example of voltage related loss variation.

21
according to the apparent power of each generator, which leads 22
to: 18 19
26
23
(14)
24 25 20
where represents the apparent power of generator m.
Fig. 3. Test distribution network.
IV. APPLICATION
The methodology presented in the paper was applied to a test TABLE I
distribution network presented in Fig. 3, whose data can be seen LOSS ALLOCATION TO GENERATORS
at [2] or requested to the authors. The method was applied con- Loss variation Voltage
sidering two load scenarios: peak load and low load. In both Total loss
allocated to related loss
variation
situations, the total losses that exist in the network without em- variation
Branch (kW) G1 (kW) G2 (kW)
bedded generators are allocated to the consumers. These alloca- (kW)
1-2 -270.840 -143.928 -126.912
tions show that: 1-3 -54.779 -54.343 -0.436
1) Each consumer only has allocated losses at branches to 2-4 -194.268 -183.353 -10.915
which power flow it contributes. 15-2 -14.315 -14.315
3-9 -44.194 -43.968 -0.226
2) In some cases, only the reactive power and respective
4-5 -169.224 -161.324 -7.900
losses of a specific branch are allocated to consumers. 5-6 -36.521 -36.479 -0.042
3) Consumers placed on the right side of Fig. 3 had more 5-7 -2.810 -2.810
losses allocated than the ones placed on the left side. This 6-8 -38.715 -38.715
is due to the fact that these consumers use branches with 7-11 -1.534 -1.534
8-17 -6.726 -6.726
greater currents. 8-27 27.605 27.605
9-10 -22.605 -22.605
At peak hours, the insertion of embedded generators reduced 10-12 -1.990 -1.990
the total losses of network in 56.4%. However, losses increased 10-13 7.353 7.353
11-16 -0.235 -0.235
in branches 8-27, 10-13, 13-14, 14-15 and 28-15. 12-21 -0.421 -0.421
This increase reflects the positive variations that occur in the 12-22 -0.189 -0.189
components of the current in these branches. The second column 13-14 11.903 12.018 -0.115
of Table I shows the variations in the losses of each branch. The 14-15 29.371 30.127 -0.756
28-15 33.702 33.702
variations of losses that occur in branches 8-27 and 28-15 are
17-18 -0.682 -0.682
totally allocated to generators 1 and 2, respectively. This was 17-19 -6.615 -6.615
expected because these branches are the paths of each generator 19-20 -1.029 -1.029
to the network. 21-23 -0.535 -0.535
The third and fourth columns of Table I show the loss vari- 21-25 -0.043 -0.043
22-26 -0.043 -0.043
ation of each branch that is allocated to each generator. In the 23-24 -0.048 -0.048
fifth column the voltage related loss variations are presented (a Total -758.427 -536.194 -178.943 -43.29
non negligible 5.7% of the total loss variation). These remaining
variations are shared between the generators, according to their
apparent power ( and ). losses were credited to generator 1 (73.1%), by two main rea-
The final allocation reveals that both generators must be paid sons: first, the better location of generator 1 which allows it to
for avoided losses, as described in Table II. Most of the avoided avoid more losses; second, generator 2 increases the losses in
COSTA AND MATOS: LOSS ALLOCATION IN DISTRIBUTION NETWORKS WITH EMBEDDED GENERATION 389

TABLE II of avoided and increased losses allow those users to become re-
TOTAL LOSSES ALLOCATED TO GENERATORS sponsible for their impact in the losses of the network. Finally,
Generator Losses allocated (kW) it is essential to mention that the method is based on AC power
G1 -554.718 flow studies. This is important in distribution networks where
G2 -203.661 the assumptions of DC power flow studies may not be adequate,
mainly due to the importance of reactive power flow and to the
influence of the voltage profiles in the network.
some branches (besides its connection branch). In fact, the in-
fluence of generator 2 increases the current in branches 10-13,
REFERENCES
13-14 and 14-15 as Table I shows.
In the low load situation, the embedded generators were con- [1] A. Gómez, J. M. Riquelme, T. González, and E. Ruiz, “Fair allocation
of transmission power losses,” IEEE Trans. Power Syst., vol. 15, pp.
sidered to produce only active power and their production was 184–188, Feb. 2000.
greater than the sum of the power of the consumers and the [2] P. Moisés Costa, “Loss allocation in distribution networks with em-
losses. Therefore, in this situation, active power flows from the bedded generation,” M.Sc. dissertation (in Portuguese), FEUP, Jan.
2002.
distribution network to the transmission network. However, the [3] A. J. Conejo, J. M. Arroyo, N. Alguacil, and A. L. Guijarro, “Trans-
reactive power needed for the consumers and for the generators mission loss allocation: a comparison of different practical algorithms,”
comes from the transmission network. The introduction of em- IEEE Trans. Power Syst., vol. 17, pp. 571–576, Aug. 2002.
[4] A. J. Conejo, F. D. Galiana, and I.Ivana Kockar, “Z-bus loss allocation,”
bedded generators decreased the network losses in 21.65%, but IEEE Trans. Power Syst., vol. 16, pp. 105–110, Feb. 2001.
note that the embedded generators have different influences on [5] C. N. Macqueen and M. R. Irving, An Algorithm for the Allocation of
the network losses. Distribution System Demand and Energy Losses: Brunel Insti. Power
Syst.—Brunel Univ., 1995.
In the end, generator 1 has to be paid for avoided losses while [6] J. Bialek, “Tracing the flow of electricity,” Proc. Inst. Elect. Eng., Gen.
the generator 2 has to pay for increased losses. This is due to Transm. Dist., vol. 143, no. 4, 1996.
the geographical location of each generator. In fact, generator [7] G. Strbac, A. Jayantilal, R. N. Allan, and N. Jenkins, “Allocation of
losses in electrical distribution systems with wind generation,” in Proc.
1 can avoid a lot of losses, as for example those provoked by Europe. Union Wind Energy Conf., Goteborg, Sweden, 1996.
consumers 8, 17, 18, 19 and 20 in branches as 1-2, 2-4, 5-6 [8] J. Mutale, G. Strbac, S. Curcic, and N. Jenkins, “Allocation of losses in
and 6-8. Generator 2 avoids losses in some branches, like 1-3, distribution systems with embedded generation,” Proc. Inst. Elect. Eng.,
Gen. Transm. Dist., vol. 147, no. 1, Jan. 2000.
3-9 and 9-10, but, on the other hand, it increases the losses in [9] D. Kirschen, R. Allan, and G. Strbac, “Contributions of individual
some branches (15-14, 14-13 and 13-10). The result is that the generators to loads and flows,” in Proc. IEEE/Power Eng. Soc. Winter
increased losses are more significant than the avoided losses, so Meeting, 1996.
[10] F. Wu, Y. Ni, and P. Wei, “Power transfer allocation for open access
the generator 2 must pay for increased losses. using graph theory—fundamentals and applications in systems without
When embedded generators were considered, the power flow loopflow,” IEEE Trans. Power Syst., vol. 15, pp. 923–929, Aug. 2000.
analysis was performed defining each embedded generator as [11] J. C. Pidre, J. A. M. Velasco, J. A. P. Lopes, and F. P. M. Barbosa, “Mod-
eling of nonlinear nodal admittances in load-flow analysis,” in Proc.
a PQ bus. This corresponds to a situation where the Q/P coeffi- IFAC Symp. Power Plants Syst., 1992.
cient is imposed or known. Other approaches would be possible,
like the use of PX buses [11], with no change in the proposed
method.
Paulo Moisés Costa was born in 1975 in Viseu,
Portugal. He received the Bachelor degree in
V. CONCLUSION Electrical Engineering from the School of Tech-
nology of Viseu, Polytechnic Institute of Viseu,
The branch-based operation of the proposed method ensures Viseu, Portugal, in 1996. He received a graduation
that each user only has allocated losses at branches for which (“Licenciatura”) and the M.Sc. degrees in electrical
current it contributes. This fact, associated to the careful sharing engineering from the Faculty of Engineering of
the University of Porto, Porto, Portugal, in 1998
of crossed terms of the losses and the consideration of reactive and 2002, respectively. He is currently pursuing
powers, allows the minimization of the situations of cross-sub- the Ph.D. degree at the University of Porto, Porto,
sidization. Portugal.
The proposed method makes a differentiation among network
users, depending on the characteristics and state of load of used
network elements. So, a tariff with nodal characteristics may
Manuel A. Matos (M’94) was born in 1955 in Porto,
emerge. However, the methodology proposed can also be useful Portugal.
when the postage stamp principle is used for tariffs. In this situa- Currently, he is Full Professor at the Faculty of
tion, the distribution company that guarantees the public service Engineering of the University of Porto, Portugal, and
the Manager of the Power Systems Unit of Instituto
must pay the total losses of its customers. A structure of mar- de Engenharia de Sistemas e Computadores do
ginal tariffs in time can also be implemented, defining periods Porto (INESC), Porto, Portugal. He also collaborates
for the application of the methodology. with the Management School of the University of
Porto. His research interests include fuzzy modeling
Referring to the embedded generators, the process of loss al- of power systems, optimization, and decision-aid
location transmits appropriate economical signals. The concepts methods.

You might also like