IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 33, NO.

6, NOVEMBER 2018 6655

Loss Allocation in Distribution Networks

Based on Aumann–Shapley
Hortensia Amaris , Member, IEEE, Yuri P. Molina , Monica Alonso , Member, IEEE, and Jaime E. Luyo

Abstract—This paper outlines a procedure for loss allocation which correspond to energy that is supplied but is not measured
in both radial and meshed distribution networks with distributed in client installations (because of errors in the meters or acts of
generation that could be regulated in various ways. The method electricity fraud).
is analytically developed based on the theory of electrical circuits
combined with game theory based on Aumann–Shapley, which Several publications in the scientific community propose
guarantees both the electrical principles and the fair axioms of measures for reducing the total values of losses [3], such as
game theory. The proposed method obtains unitary participa- controlling the tap position of substation transformers or using
tion coefficients for each network user based on the currents power inverters for the regulation of reactive power.
demanded/injected by each user and the network topology. The In general, the method for reducing losses is by applying
proposed allocation method based on Aumann–Shapley has been
compared with other traditional allocation methods, is adaptable to global control solutions at the Distribution System Operator
distribution networks, and shows great potential and ease of imple- (DSO) centers [4]. However, DSOs do not have any knowledge
mentation. Moreover, it can be applied to any kind of distribution about the users (consumers, generators, prosumers) who are
network (radial or meshed) with distributed energy resources. causing the increase in network losses. Knowing the locations
Index Terms—Distributed energy resources, distributed gener- of the customers who are responsible for the network losses will
ation, game theory, meshed networks. allow the DSO to apply specific local actions in those customer
installations in order to reduce the network losses. These actions
I. INTRODUCTION
can be based on Demand Response strategies [5].
NERGY losses in electrical networks can be determined
E based on the difference between the energy generated at
power stations and the energy measured at consumers’ installa-
The research proposal sets out in this article is focused on allo-
cating distribution network losses among network users, which
is not a trivial process, as indicated in [6] and [7]. This prob-
tions. In recent years, the difference between these values has lem has been studied extensively in the scientific literature and
increased, which entails an increase in electrical network losses. has been primarily applied to transmission networks, where the
According to data from the 2014 World Bank Database [1], it greatest interest lies in sharing the costs of the network among
can be observed that during the past decade, the average value of network users for economic purposes and electricity markets
world energy losses has been slowly decreasing (8.85% in 2000, [8]. Currently, loss allocation in distribution networks are usu-
8.47% in 2007 and 8.26% in 2014). These losses translate into ally considered through electricity tariffs. In many countries,
a loss of energy efficiency in electrical networks and result in tariff designs are based on volumetric charges where each cus-
an increase in electricity prices for end users, who are the ones tomer pays for the kilowatt-hours consumed without including
who must cover the network losses [2]. As such, it is imperative the costs that each user inflicts on the grid. In general, there is
to apply measures that will help reduce network losses, since no single accepted loss allocation method that can be applied
this improves the efficiency of electrical systems and the useful in different countries. Moreover, in the majority of countries,
life of electrical infrastructures and at the same time translates network losses are hidden within the fixed and variable elec-
into economic benefits for users. tricity fee. Very few countries, such as Hungary, include a loss
Losses can be classified as technical losses due to dissipation distribution component in their electricity bill [9].
of energy in lines and transformers and non-technical losses, Differences in the distribution tariff schemes can be es-
tablished even within one country. This is the case in Swe-
Manuscript received October 17, 2017; revised March 19, 2018 and May 22,
2018; accepted June 2, 2018. Date of publication June 6, 2018; date of cur- den, where each DSO is free to use the tariff design they
rent version October 18, 2018. Paper no. TPWRS-01577-2017. (Corresponding prefer as long as that methodology is nondiscriminatory and
author: Hortensia Amaris.) objective [10].
H. Amaris and M. Alonso are with the Department of Electrical Engineer-
ing, University Carlos III of Madrid, Madrid 28911, Spain (e-mail:, hortensia. It is not easy to compare the loss allocation methodologies
[email protected]; [email protected]). that are applied in different countries due to the great diversity
Y. P. Molina is with the Department of Electrical Engineering, Fed- of tariff schemes and the lack of transparency about the method-
eral University of Paraı́ba, João Pessoa 58051–900 Brazil (e-mail:, molina.
[email protected]). ology for calculating tariffs that each country uses [9]. However,
J. E. Luyo is with the Universidad Nacional de Ingenierı́a, Lima 15333, Perú some general conclusions can be highlighted:
(e-mail:,[email protected]). r The majority of the distribution tariff schemes are based
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. on the pro rata method, which is the simplest method but
Digital Object Identifier 10.1109/TPWRS.2018.2844740 does not consider the network loss caused by each user.

0885-8950 © 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KANPUR. Downloaded on June 28,2022 at 05:43:08 UTC from IEEE Xplore. Restrictions apply.
 6656 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 33, NO. 6, NOVEMBER 2018

r Distributed generation (DG) is not required to pay distri- apply the reconciliation and thus contains a factor of
bution charges, as is the case in Germany, the Netherlands, arbitrariness.
Norway, Belgium, Poland and Portugal. The exception oc- r It does not consider the contribution of reactive power in
curs in Great Britain, where embedded generators can be PV generation nodes.
allocated negative charges if they are providing a benefit The same authors proposed the method of Direct Loss Co-
to the network [10]. efficients (DLC) [15], the goal of which is to obtain a direct
r Losses due to reactive power flow through lines are not relationship between line losses, L, and the power injection,
considered in the network losses. avoiding the reconciliation process of the MLC method. This
The main difficulty for selecting the distribution allocation method proposes using a Taylor series expansion around the
method is that network losses are non-linear functions of the in- operation point to obtain the relation of losses, L, with nodal
jection power and power demand, which complicates the iden- voltages: module (U ) and phase (θ). Its problems include the
tification of the exact contribution of each network user to the computational complexity being high, as it is necessary to calcu-
total network losses. late the Hessian matrix, and the result of the allocation of losses
As introduced in [11], “fair” distribution is difficult to containing inherent imprecisions due to the truncation error of
achieve because almost all allocation methods entail a degree of the Taylor expansion.
arbitrariness. The method based on impedance matrices was originally pro-
The goal of a method for allocating losses is to share the posed for transmission networks by [16]. It is based on calcu-
total cost (losses) of the network among network users (both lating the losses as the sum of nodal network losses, know-
generators and consumers), in a fair way that reflects the real ing the admittance matrix of the circuit and the solution of
contribution of each user to the total network losses. Tradition- the load flow at a specific operation point. This method al-
ally, methods of loss allocation have been applied to transmis- locates 50% of the losses to generators and 50% to loads.
sion networks, covering a wide range of methodologies with However, this formulation cannot be applied in distribution
different levels of complexity. The simplest method is the pro networks.
rata method, where the cost is allocated to a client in propor- Several studies related to loss allocation in distribution net-
tion to their power connected to the network [12]. This method works have been published in recent years. Some offer high com-
presents the problem of unfair distribution because it does not putational complexity, which hinders their application to real-
take into account the locations of the network clients, and as a time situations [17]. Others are based on logarithmic schemes
consequence, it does not consider their contributions to the total [18], which provide different results for the same problem when
losses. they are compared to other allocation methods [19]. The fact that
For this reason, other methods based on circuits that consider different loss allocation methods provide different results for the
the locations of users in the transmission networks have been same network means that some methods are not providing an
proposed [13], [14]. However, the application of these methods accurate solution.
to the distribution networks is not immediate, as the role of the The Branch Current Decomposition (BCDLA) method was
swing/slang node is different in both networks. proposed by [19], and it is developed specifically for radial dis-
Another proposal is the method of Marginal Loss Coefficients tribution networks. This method only requires the information
(MLC), which is based on distributing total line losses among of the network topology and the solution of the load flow. How-
the units connected to the network according to coefficients that ever, this methodology cannot be applied to weakly meshed
consider the contribution of each unit (generator or consumer) to distribution networks.
the total network losses [15]. These coefficients (MLCs) express This article sets outs a method for loss allocation in distri-
the variation in the total active losses, L, due to the marginal bution networks based on game theory. More specifically, it is
variation in active power Pi and reactive power Qi in each based on Aumann-Shapley theory and circuit laws, offering an
network node i. However, the application of MLC produces analytical solution that is easy to implement and results in a fair
a nodal distribution loss that is approximately double that of share among participants, such as: consumers, prosumers, Dis-
the total network losses. As such, it is necessary to apply a tributed Energy Resources (DER). These two aspects allow the
reconciliation factor so that the total allocation loss to all users loss allocation proposal to be applied in distribution networks
equals the total network loss. (radial or meshed) where the real-time operation is needed.
The original proposal by [15] was intended for transmission The main contributions of the proposed method are as follows:
networks, in which 50% of the total losses were allocated to r The proposal of a new formula for allocating losses among
the swing node. The main problems of applying this method to participants (consumers, prosumers, DGs and DER) based
distribution networks are the following: on Aumann-Shapley in the distribution networks. The re-
r In distribution networks, the lines are short, with a high sult is an analytical solution that is easy to implement in
R/X ratio, so the Newton-Raphson load flow solution distribution networks (radial and meshed), offering low
methods might not converge. complexity.
r The computational complexity is high, as it is necessary to r It considers the real and imaginary components of the cur-
operate with the Jacobian matrix of the load flow. rent injections, so it identifies and quantifies the individual
r The resulting allocation does not correspond to a di- contribution of each user in terms of active and reactive
rect result of the formulation because it is necessary to power losses. This offers the advantage of differentiating

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KANPUR. Downloaded on June 28,2022 at 05:43:08 UTC from IEEE Xplore. Restrictions apply.
 AMARIS et al.: LOSS ALLOCATION IN DISTRIBUTION NETWORKS BASED ON AUMANN–SHAPLEY 6657

the participation of the DER that is operating in voltage The Shapley value is defined by:
control mode from another that works in power control
 (|S| − 1)! (n − |S|)!
mode. Φi (v) = [v (S) − v (S − {i})]
r Losses allocated to DER and loads are independent, even n!
S,i∈S
when they are in the same bus. (1)
r It considers circuit laws and at the same time has desirable where:
characteristics in terms of economic coherence, because r S: Coalition, defined as the group of agents that unite to
it is based on circuit laws in combination with Aumann- obtain greater benefits than they would have if they went
Shapley. alone
r The methodology offers consistency with different load r |S|: the number of elements of coalition S
levels, simultaneously allowing great accuracy and fair r v(S): value of the coalition
billing to the different types of users connected to the r i: player i-th
distribution network, such as consumers, prosumers and r n: number of players
DER. Thus, the Shapley value of a player is defined as the av-
erage value of the allocations in all possible orders of player
II. METHODS BASED ON GAME THEORY incorporation in the coalitions [23].
The problem of allocating losses among network users can
be considered a cooperative game, which consists of sharing B. Methods of Loss Allocation by Aumann–Shapley
total network losses among a group of agents that either reduce Robert J. Aumann and Lloyd Shapley proposed the Aumann-
or cause them. The goal of the organization is to find an ef- Shapley method to solve the problem of cost allocation by ad-
ficient and fair allocation procedure, that is, with none of the dressing smaller units of the same participant [24]. The idea of
participants favored at the expense of others and without any- allocation by the Aumann-Shapley method consists of dividing
one coming out of it harmed. Cooperative game theory has been the contribution of each player into infinitesimal parts. The num-
applied to loss allocation in transmission networks [20]. To do ber of permutations increases, so applying the previous Shapley
this, generators, consumers and, in general, users of the electri- method, [25], would complicate it.
cal network are modeled as rational agents or players interested However, working with incremental, infinitesimal units
in forming groups and coalitions to obtain the maximum bene- shows that it is possible to obtain a closed, direct solution
fit in the final result. The solution provided by game theory is without having to perform any permutation of players in the
fair, efficient and stable [21]. Within this discipline, the most coalitions.
used methods are based on the Shapley value and the Aumann- Now, the incremental cost of an infinitesimal agent can be
Shapley method, described below. approximated to its marginal cost. Furthermore, the number of
permutations allows for the assumption that all of them have the
A. Methods of Loss Allocation Based on the Shapley Value same uniform probability of appearing, respecting the propor-
The Shapley value method finds an expected marginal con- tion of the size of the agents.
tribution to each player in the game with respect to a uniform The Aumann-Shapley solution [26] is defined in the following
distribution on the set of all permutations in the entry order of equation:
players to analyze all possible combinations in the game. The  1
∂SL O S S (λI)
cost, the profit, the benefits or the participation of each agent πk = dλ (2)
is calculated when it is the first, second, third and so on. The 0 ∂Ik
average value of the incremental costs calculated in each permu- where SL O S S is the complex electrical losses, Ik is the injection
tation determines the cost that corresponds to each agent. Thus, current at node k, πk represents the unitary Aumann-Shapley
the influence of the player entry order on the cost allocation is participation of element k, and λ is the integration variable,
eliminated [22]. which ranges from 0-1. The losses associated with this agent
The Shapley value can be interpreted as being the average are defined by:
value of the incremental costs of including the agent. It consid-
ers all sub-coalitions that do not contain this particular agent, Lk = Ik πk (3)
including the empty sub-coalition. Assuming that the probabil-
ities of the occurrences of sub-coalitions of several sizes are the where Lk is the allocation of losses to agent k.
same, the allocation is formally defined through the analytical It can be noted that Aumann-Shapley calculates the average
expression (1). value of each player’s incremental cost when it grows uniformly
The Shapley value is based on the concepts of fairness and from zero to its current value. The parameters vary constantly in
efficiency [21]: the interval [0,1], so the same proportion of division is applied
r Fairness: Each player must receive an allocation according in all the agents.
to their global contribution to the game. The Aumann-Shapley solution is fair and economically effi-
r Efficiency: The sum of the distribution of losses among the cient since it uses information about marginal costs and fulfills
participants must coincide with total network losses. the following axioms of game theory:

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KANPUR. Downloaded on June 28,2022 at 05:43:08 UTC from IEEE Xplore. Restrictions apply.
 6658 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 33, NO. 6, NOVEMBER 2018

r Symmetry: two players that have the same contribution in where the voltages can be referred to the swing node,
the total network losses receive the same unitary partici- obtaining:
pation.
r Efficiency: The sum of all losses allocated to each individ- 
N
Ii = Y(i,j ) (Uj − U1 ) (9)
ual player coincides with the total network losses.
r Additivity: The sum of the losses allocated to a player that j =2

decides to play two games separately is equal to the loss Solving the voltage vector Ui , i ∈ {2, . . . , N }, the following
allocated to this player when both games are played at the expression is obtained:
same time. 
N
r Monotonicity: If network losses increase/diminish, losses Ui = U1 + Z(i,j ) Ij (10)
allocated to players increase/diminish. j =2
Several researchers have used Aumann-Shapley for cost dis-
where matrix Z represents the impedance matrix of the cir-
tribution in transmission networks [22], [27], sharing 50% of
cuit without including node 1. Keep in mind that Z does not
the total network losses among the slack bus and generators and
correspond to Z-bus [16], so it can be applied to any distribution
the other 50% among consumers. In distribution networks, this
network, avoiding the problem of the inversion of the singular
formulation is not acceptable because the slack node in these
admittance matrix.
networks is the connection point to the grid and consequently
has a loss allocation that is assumed to be zero.
A. Loss Allocation Due to the Real Component of the Current
The method proposed in this article solves the problem of
loss allocation in distribution networks via Aumann-Shapley, This section explains the allocation of losses (active, reactive)
identifying the individual participation of the power components among participants who are connected to distribution network
(active and reactive) of the loads and DER in each node and nodes k ∈ {2, . . . , N }. The distribution will be done consider-
each branch of the network, considering each component as an ing the nodal injection currents, expressed in Cartesian compo-
independent agent. Moreover, the properties of the Aumann- nents, that correspond to the consumption/generation by agents
Shapley method guarantee equitable allocations. connected to node k.
The unitary participation (2) of the active component of the
III. LOSS ALLOCATION AMONG PARTICIPANTS nodal current in bus k is:
 1
Complex power losses, SL O S S (active, reactive) of a network ∂SL O S S (λI)
re
πloss,k = dλ (11)
with N nodes can be obtained from the following equation: 0 ∂Ikr e


N On the basis of (7), it can be expressed as:
SL O S S = Ui Ii∗ (4)  1 N 
∂ 
∗
i=1 πloss,k =
re
re (Ui − U1 ) (λIi ) dλ (12)
0 ∂Ik
where Ui is the voltage in the i-th node and the injection i=2

current to the node is Ii . Taking node 1 as the slack/swing node Keeping in mind (10) and decomposing the nodal current
and the voltage reference, (4) can be expressed as shown in the at node k in its real and imaginary components, the partial
following equation: derivatives of the voltage with respect to the current can be
obtained as:

N
⎛ ⎞
SL O S S = U1 I1∗ + Ui Ii∗ (5)  N
∂Uk ∂ ⎝
i=2 = U1 + Z(i,j ) Ij ⎠ = Z(k ,i) (13)
∂Ikr e ∂Ikr e j =2
considering that the nodal current I1 is related to the nodal
currents {2, . . . , N }, as follows: Substituting in (12):

N  1 
N
I1 = − Ii (6) re
πloss,k = 2 Z(j,k ) λIjr e dλ (14)
i=2 0 j =2

and substituting in (5), results in the following expression: results in:


N 
N

SL O S S = (Ui − U1 )Ii∗ (7)

re
πloss,k = Z(j,k ) Ijr e (15)
i=2 j =2

The nodal injection current can be calculated from the nodal Finally, the total losses (3) due to the active component of the
voltages and the network admittance matrix: nodal current k can be calculated by the following expression:


N 
N

Ii = Y(i,j ) Uj (8) Lrke = re

πloss,k Ikr e = Ikr e Z(j,k ) Ijr e (16)
j =2
j =1

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KANPUR. Downloaded on June 28,2022 at 05:43:08 UTC from IEEE Xplore. Restrictions apply.
 AMARIS et al.: LOSS ALLOCATION IN DISTRIBUTION NETWORKS BASED ON AUMANN–SHAPLEY 6659

B. Loss Allocation Due to the Imaginary Component of the

Current
Likewise, the unitary participation of the reactive component
of the current in bus k is:
 1
∂SL O S S (λI)
im
πloss,k = dλ (17)
0 ∂Ikim Fig. 1. Node k with load and generation.

Keeping in mind (7) and substituting the partial derivatives

of the voltage with respect to the current: re
Applying the same nodal unitary participation (πloss,k ,
 1 
N im
πloss,k )to the generation and the load, the following holds:
im
πloss,k = 2 Z(j,k ) λIjim dλ (18) r Losses associated with the DG: Losses owing to the real
0 j =2
component of the current injected by the generator are
integrating: LrGek = πloss,k
re
IGr ek , and those that correspond to the imag-
inary component are Lim G k = πloss,k IG k , hence, the result-
im im

N
ing total loss for the DG source will be:
im
πloss,k = Z(j,k ) Ijim (19)
j =2 LG k = LrGek + Lim
Gk (23)
Therefore, the expression for total losses due to the imaginary r Losses associated with the load: Likewise, losses ow-
component of the nodal current k is obtained: ing to the real load current are LrLek = πloss,k re
ILr ek , and

N the contribution owing to the imaginary component is
L k = πloss,k IL k , so the total loss allocated to the load is:
Lim im im
Lim
k = im
πloss,k Ikim = Ikim Z(j,k ) Ijim (20)
j =2
LL k = LrLek + Lim
Lk (24)
C. Complex Loss Allocation Note that at node k, it holds that the total loss associated with
Once the contributions of losses due to the active and reactive node k (21) is the sum of the losses allocated to the generation
components of the nodal currents are known, the total nodal loss and the load:
based on the sum of both individual contributions Lrke and Lim k
Lk = LG k + LL k (25)
can be obtained:

N
IV. BRANCH LOSS ALLOCATION AMONG PARTICIPANTS
Lk = Lrke + Lim
k = Z(j,k ) Ikr e Ijr e + Ikim Ijim (21)
j =2 In this case, branch losses are allocated among participants,
(26) where Ul is the voltage drop of the l branch and zl is the
It is important to highlight that there is a connection between branch series impedance:
the active and reactive power in line losses. That is, the flow
of active (reactive) power through the lines will produce not Sloss,l = Ul Il∗ = zl Il Il∗ (26)
only active (reactive) losses but also reactive (active) losses. As
such, the flow of active power influences reactive losses like the The branch current Il is defined as a linear combination of nodal
flow of reactive power influences active losses. This connection injection currents, that is:
cannot be ignored in the process of allocating losses [28]. 
N
Il = αl,2 I2 + . . . + αl,N IN = α(l,j ) Ij (27)
D. Loss Allocation at Nodes With Load and Generation j =2

At those network nodes where both loads and generators are Consequently, complex power losses associated with branch
connected, it will be necessary to distribute the loss allocated to l (26) can be expressed according to the nodal currents:
the node among the different agents that are connected to it. ⎛ ⎞⎛ ⎞∗
Considering that two agents in node k are connected (load 
N 
N
Sloss,l = zl ⎝ α(l,j ) Ij ⎠ ⎝ α(l,j ) Ij ⎠ (28)
and generation), it can be said that SL k is the demand for the
j =2 j =2
load connected at node k and that SG k is the power of the
DG unit located at the same node (see Fig. 1). The resulting
A. Branch Loss Allocation Due to the Real Component of
complex power injection at node k will be Sk = Pk + jQk and
Nodal Currents
corresponds to the sum of the power demanded by the load and
the power injected by the DG source. The nodal injection current The active unitary participation associated with branch l due
Ik will be the sum of the injection current due to the generation to the nodal current in bus k is expressed as:
and the load as:  1
∂SL O S S (λI)
πloss,l,k =
re
dλ (29)
Ik = IG k + IL k (22) 0 ∂Ikr e

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KANPUR. Downloaded on June 28,2022 at 05:43:08 UTC from IEEE Xplore. Restrictions apply.
 6660 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 33, NO. 6, NOVEMBER 2018

Deriving (28) with respect to Ikr e allows the unitary partici-

Algorithm 1: Distribution Loss Allocation Aumann-
pation to be expressed as:
⎡ ⎤ Shapley
1  N  N Input: Network Topology Data:
re
πloss,l,k = zl ⎣α(l,k ) ∗
α(l,j ∗ ∗
) Ij + α(l,k ) α(l,j ) Ij ⎦ N number of Nodes
2 j =2 j =2 Np number of Users (Consumers, DG, D-STATCOM,
(30) etc.)
Nb number of Branches
B. Branch Loss Allocation Due to the Imaginary Component Z Network impedance matrix (without the slack bus)
of Nodal Currents Input: Users Currents Matrix: Ij,u (N x Np )
 Nodal
In the same way, for the imaginary component of the current: 1: Ijr e ← u Ij,u
re
total nodal real current
 im
 1 2: Ij ← u Ij,u
im
total nodal imag current
∂SL O S S (λI)
im
πloss,l,k = dλ (31) 3: if is Node Loss Allocation required then
0 ∂Ikim 4: call Node_Loss_Allocation ()
Deriving (28) with respect to Ikim , the unitary participation is 5: if is Branch Loss Allocation required then
expressed as: 6: call Branch_Loss_Allocation ()
⎡ ⎤
1  N  N
im
πloss,l,k = j zl ⎣α(l,k ) ∗
α(l,j ∗ ∗
) Ij − α(l,k ) α(l,j ) Ij ⎦
2 j =2 j =2
Algorithm 2: Node_Loss_Allocation ()
(32) 1: for node = 2, N do
2: Compute (15) real nodal unitary
C. Complex Branch Loss Allocation participation
3: Compute (19) imag nodal unitary
Once the active and reactive unitary participation coefficients participation
for branch l are known, the branch loss allocation can be ob- 4: for node = 2, N do
tained by (3). Equations (30) and (32) depend on the coefficients 5: for user = 1, Np do
α, which are directly calculated by using the nodal voltages and 6: Compute (21) complex loss allocation
branch impedance, according to (33): 7: return
1 
N
1
Il = (Uk − Um ) = Z(k ,j ) − Z(m ,j ) Ij (33)
zl zl j =2
Algorithm 3: Branch_Loss_Allocation ()
Equating the previous equation to (27), it can be deduced that 1: for branch = 1, Nb do
coefficients α depend only on the network topology as: 2: for node = 2, N do
1 3: Compute (34) branch coefficients
α(l,i) = Z(k ,i) − Z(m ,i) (34) 4: Compute (30) real branch unitary
zl
participation
Consequently, the loss allocation for l branch due to the i-th
5: Compute (32) imag branch unitary
nodal current is obtained by the following expression:
participation
Lloss,l,i = πloss,l,i
re
Iir e + πloss,l,i
im
Iiim (35) 6: for branch = 1, Nb do
7: for user = 1, Np do
which only depends on the impedance branch and nodal
8: Compute (35) branch loss allocation to
currents.
user
9: return
V. FEATURES OF THE LOSS ALLOCATION FORMULATION
A. Algorithm and Computational Complexity
The proposed loss allocation formulation offers two possibil- Shapley value-based allocation methods requires 2n computa-
ities, to compute the loss allocation among network users (con- tions, where n is the number of players. Methods based on
sumer, DG, DER) and/or to compute the branch loss allocation derivatives [15] rely on computations of Jacobian and Hessian
among users. Both possibilities are shown in the pseudo-code matrices which increase the order complexity when compared
detailed in Algorithm 1. to the proposed method.
Notably, the aforementioned method provides an analytical If the currents (real and imaginary components) are not avail-
solution, requiring only the nodal currents from each agent, and able, it is necessary to first compute them, by applying load flow
the data of the network. From a complexity point of view, the or state estimation algorithms. The impedance matrix (with-
loss allocation algorithm requires 3(N − 1) multiplications and out the slack bus) is computed only once for a fixed topology.
2(N − 1) additions, N being the number of nodes of the net- Whenever the topology changes, the impedance matrix has to
work, which leads to an order of complexity O(1). In contrast, be updated.

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KANPUR. Downloaded on June 28,2022 at 05:43:08 UTC from IEEE Xplore. Restrictions apply.
 AMARIS et al.: LOSS ALLOCATION IN DISTRIBUTION NETWORKS BASED ON AUMANN–SHAPLEY 6661

B. Assumptions
The main assumptions considered are the following:
r The active and reactive currents of users independently
participate in the allocation of losses.
r Distribution network losses are allocated among network
participants (customers, prosumers, DER units).
r Loss allocations at nodes with load and generation will be Fig. 2. Four-node network with generation and three loads.
considered different agents.

C. Application advantage of guaranteeing isonomy with respect to the size of

agents.
It must be highlighted that the computational complexity of The proposed method is designed specifically for distribution
the proposed formula depends only on the number of nodes. It networks with DER, and it assumes that 100% of the losses are
is not affected by the number of existing loops in the network or the responsibility of the network users, such as: DER, customers
by the number of network users. Consequently, it can be easily and prosumers. Therefore, every individual user is treated as a
applied to large-scale networks (radial or/and meshed). complex current/voltage source, and the analytical solution ob-
The proposed loss allocation method can be applied to tained starts from (2). It considers the real and imaginary compo-
distribution networks with many DERs, such as: convert- nents of the current injections and then identifies and quantifies
ers, distributed energy storage devices and reactive power the individual contribution into active and reactive power losses.
compensators. Once the DER nodal currents are available Moreover, this complex loss allocation formulation considers
(by measurement or power flow solutions) the proposed loss the real current component contribution to reactive losses, as
allocation method will assign the corresponding losses to the well as the imaginary current component contribution to active
DER owner using only the currents (real, imaginary) and the losses.
network topology. However, the method presented in [29] has been designed for
The formulation is designed to grid connected distribution transmission networks and involves two steps; in the first step,
networks. When microgrids are operating in grid-connected the generators are modeled as current sources and the loads
mode, the proposed loss allocation method can be directly ap- as constant admittances and in the second step, the generators
plied. However, the operation mode of microgrids in islanded are modeled as constant admittances and the loads as current
mode requires a different formulation in order to maintain the sources. In this way, it is possible to allocate losses by 50% for
distinction among the slack node and the other nodes when loads and 50% for generators, which is one of the assumptions
disconnecting from the main grid. of [29].

D. Consistency for Different Load Levels VI. CASE STUDIES

The formulation offers consistency for different load levels, In this section, the proposed method will be applied to two
which can be proved mathematically as follows: examples of distribution networks, and the validity of the method
Considering a fixed network topology and knowing the nodal will be verified by comparing it with other existing methods.
currents (real, imag), complex losses can be obtained by (21).
If all users increase/reduce their current in the same factor ρ,
such as, I  A. Four-Node Distribution Network With Distributed
k = ρIk and Ik = ρIk , then losses allocated to
re re im im
Generation
each user increase/reduce their value proportionally:
Fig. 2 shows a four-node distribution network with a tie-line

N   switch (between node 2 and node 4), which can be operated
k = Z(j,k ) I
re  im 
im = ρ2 L
k Ij + Ik Ij
L re (36)
k in radial form (switch open) or meshed form (switch closed).
j =2
Node 1 is the swing node with a voltage Usw = 1.01 [p.u]. At
which means that the share of losses among all network users nodes 2, 3 and 4, three equal loads (P Q) are connected with
for different loads levels is determined proportionally by the power demand SL = 0.5 + j0.3 [p.u]. At node 4, a DG source
factor ρ2 , whatever their original current value. is connected and can operate in two modes of operation. The
four branch impedances are equal to a value zl = 0.02 + j0.01
[p.u].
E. Comparison With Other Loss Allocation Methods Based on This four-node network will be used as a network type to
Game Theory explain the procedure for executing the proposed method by
In terms of equity, the proposed method guarantees an equi- means of the following cases:
table solution by complying with the axioms of symmetry, ef- r case a) radial network (switch open): DG behaves like node
ficiency, additivity and monotonicity, which are inherent when P Q, it only injects P = 1.0 [p.u], (Q = 0 [p.u])
the Aumann-Shapley method is applied. Compared with the r case b) radial network (switch open): DG behaves like
Shapley methods [25], the Aumann-Shapley method offers the node P V , P = 1.0 [p.u], U4 = 1.01 [p.u]

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KANPUR. Downloaded on June 28,2022 at 05:43:08 UTC from IEEE Xplore. Restrictions apply.
 6662 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 33, NO. 6, NOVEMBER 2018

Fig. 3. Branch losses case a: radial network, (solid) nodal active current Fig. 5. Branch losses case c: meshed network, (solid) nodal active current
contribution, (hatched) nodal reactive current contribution. contribution, (hatched) nodal reactive current contribution.

TABLE I
NODAL LOSS ALLOCATION [kW]

closed, and consequently, the network is meshed with a loop.

In Fig. 5, it can be seen that closing the switch increases the
real losses of branches 2 − 3 and 3 − 4 due to the load demand
Fig. 4. Branch losses case b: radial network, (solid) nodal active current
contribution, (hatched) nodal reactive current contribution.
connected at node 3. In case a), the load connected at node 3
increases losses only in the first branch 1 − 2. Meanwhile, in
this case, the load connected at node 3 is also increasing losses
r case c) meshed network (switch closed): DG behaves like in branches 2 − 3 and 3 − 4. However, the load at node 3 is
helping to reduce losses in branch 2 − 4. In contrast, the power
node P Q, it only injects P = 1.0 [p.u], (Q = 0 [p.u])
at node 4 (load and DG) is responsible for increasing the losses
1) Case a: In this case, the tie-line switch is open, and the
at branch 2 − 4.
network is operated in radial form. The DG connected at node 4
4) Comparative Analysis: At this point, it is interesting to
injects only active power, but loads at nodes 2, 3 and 4 demand
compare the results of the proposed loss allocation method based
both active and reactive power. It can be noted in Fig. 3 that
on Aumann-Shapley with those obtained when applying other
total active power demand at node 4 (load and DG) increases
methods, such as BCDLA, MLC and DLC. These results are
losses through branch 3 − 4, but its contribution to branch 1 − 2
shown in Table I.
losses is negative, which means that the DG injection at node 4 r When dealing with radial networks and DGs operating in
is helping to reduce the losses in this branch 1 − 2. The contri-
power control mode (case a), the four allocation methods
bution of DG to branch 2 − 3 losses is irrelevant, as it neither
produce quite similar results.
increases nor decreases losses in this branch. Because DG is in- r Derivative methods such as MLC and DLC offer equivalent
jecting only active power, reactive power demand from loads 2,
results, provided that generation is treated as node P Q, but
3 and 4 must be totally provided by the grid. Consequently, this
they do not behave well with P V nodes in voltage support
reactive power demand by load 4 increases losses at branches
mode (case b).
1 − 2, 2 − 3 and 3 − 4. The same occurs with the reactive power r In meshed networks (case c), BCDLA cannot be applied.
demand from loads 2 and 3. r The proposed loss allocation methodology is the one that
2) Case b: In this case, the network is operated in radial
can be applied to meshed networks (case c), in contrast to
form. It can be observed (Fig. 4) that in case b), there is a
BCDLA, and to voltage support buses (case b) in contrast
reduction in losses because the generator at node 4 operates as a
to MLC and DLC.
node P V , injecting reactive power to maintain the nodal voltage
at the set point (U4 = 1.01 [pu]). In this case, reactive power
B. 33-Node Distribution Network With Distributed Energy
generation by DG connected at node 4 slightly increases the
losses at branch 3 − 4, but it has a negative influence in branch Resources
1 − 2 losses. In this case, we use the IEEE 33-node distribution network
3) Case c: This third case is similar to case a), with the proposed in [30], where the basic network has been modified
exception that the tie-line switch between nodes 2 and 4 is to include DER units: DG and Distribution Static Synchronous

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KANPUR. Downloaded on June 28,2022 at 05:43:08 UTC from IEEE Xplore. Restrictions apply.
 AMARIS et al.: LOSS ALLOCATION IN DISTRIBUTION NETWORKS BASED ON AUMANN–SHAPLEY 6663

TABLE III
IEEE-33 TIE-LINES

r Case B: Radial network with DGs.

r Case C: Meshed network with DGs and D-STATCOM.
r Case D: Meshed network with DGs, D-STATCOM and
Fig. 6. IEEE-33 node distribution network with DER. different load levels.
1) Case A: In this case, there is no DER unit connected to the
network, and all the tie-lines are open. In case A network losses
TABLE II are 202.677 kW and all the losses allocated to nodes are positive
IEEE-33 RADIAL DISTRIBUTION NETWORK LOSS ALLOCATION
(Table II), which indicates that users contribute to increasing
them. The three methods (the proposed method, BCDLA and
[31]) give similar results when the network is operated radially
and there is no DER.
2) Case B: In case B, the network is radial, but there are three
DGs units located at nodes 6, 25 and 31. In case B network losses
diminish substantially (43.430 kW), so loss allocations to each
user are, in general, reduced due to the participation of DGs. It
is seen at node 6 (Table II), that both the load demand and the
generation contribute to reduce losses and their loss allocation
is negative. Similar results are obtained by the BCDLA. How-
ever, it must be noted that the method proposed by [31] gives
higher losses to the DG compared to the losses allocated by the
proposed method and by the BCDLA.
At node 25, negative losses (indicating gain) are allocated to
the DG, but the allocation of losses to the load of that node is
still positive.
The location of the third DG source at node 31, which injects
only active power, has a positive influence in the network losses,
which indicates that part of the power it injects produces losses
in the network. It is important to note that the method proposed
by [31] allocates higher negative losses to DG connected at
node 31, showing an opposite behavior compared to BCDLA
and with the proposed method in this paper.
3) Case C: This case corresponds to the 33-node meshed
distribution network, which can include up to five loops (Ta-
ble III). It must be highlighted that because the distribution
network is meshed, the BCDLA and [31] allocation methods
cannot be applied, making the proposed method the only one
that can provide the loss allocation solution.
The load and production data coincide with the data used in
case B with the exception of the DG located at node 31, which
is replaced by a D-STATCOM of 500 kvar in order to prove how
the proposed method behaves with reactive power compensators
and meshed networks. DG data located at nodes 6 and 21 have
Compensator (D-STATCOM), as indicated in Fig. 6. The net- the same values of case B.
work is a 12.66 kV system, and it includes five looping branches Table IV shows the total network losses and the allocation
(tie-lines) that allow the network to be operated in radial form results for the different meshed situations when there is only a
(open switches) or meshed form (closed switches). The data for single loop up to five loops (totally meshed). In general, as the
demand and generation are shown in Table II and coincide with number of loops increases, total network losses are reduced, and
the data indicated in [31]. consequently, losses allocated to each participant are reduced.
The following cases of operation are considered: The more relevant results appear at nodes 6, 21, and 31 where
r Case A: Radial network without DER. there are DER units. It can be noted that the D-STATCOM

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KANPUR. Downloaded on June 28,2022 at 05:43:08 UTC from IEEE Xplore. Restrictions apply.
 6664 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 33, NO. 6, NOVEMBER 2018

TABLE IV
IEEE-33 MESHED DISTRIBUTION NETWORK LOSS ALLOCATION

Fig. 7. Loss allocation for different load levels (IEEE-33 meshed distribution
network with DER).

TABLE V
NETWORK LOSSES FOR DIFFERENT LOAD LEVELS

are assigned the same proportion of losses for each load and
generation percentage. This last case shows the consistency of
the proposed formulation for different load levels.
5) Discussion of Results: From previous results, the follow-
ing conclusions can be inferred:
r The solution provided by the proposed Aumann-Shapley
methodology offers the characteristics of fairness, effi-
ciency, and stability, which are required for the correct
allocation of distribution losses among participants. More-
over, for radial distribution networks, the proposal solution
behaves similarly to BCDLA [19], which confirms that the
proposal provides accurate loss allocation solutions.
r It is demonstrated that the proposed formulation can be
located at bus 31 reduces network losses for all situations from
applied straightforwardly to meshed distribution networks,
one loop to five loops. In contrast, DG located at bus 6 increases
which is an important advantage compared to the BCDLA
losses, and consequently, its loss allocation is always positive.
methodology [19]. The BCDLA methodology has been
According to the customers, it can be noted that customers
specifically formulated for radial distribution networks and
near the D-STATCOM and DG located at bus 25 are allocated
cannot be applied to meshed distribution networks.
negative losses according to the increased number of loops, r Furthermore, the proposed loss allocation methodology
which means that their demand is reducing network losses.
satisfies the axioms of fairness and consistency for differ-
4) Case D: In this situation, the topology that has been con-
ent loads levels. It has been shown that every single agent
sidered corresponds to the totally meshed network (five closed
(customer, DER) is responsible for the network losses pro-
tie-lines) with the D-STATCOM located at node 31 and two
portionally to the load level.
DGs connected at nodes 6 and 21. The load and generation are r It has been shown that the proposed formulation can be ap-
varied from 10% to 100% of the data used in case B. A similar
plied to distribution networks (radial and meshed) with DG
load demand variation is performed in [32]. Fig. 7 shows the
and reactive power compensators such as D-STATCOM
loss allocation at the nodes located in the main feeder (from
without requiring any modification or assumption in the
node 9 to node 18) for different load and generation percent-
formulation.
ages (referred in case B) and total network losses for each load
variation is shown in Table V.
VII. CONCLUSION
It is important to emphasize that for different load levels, the
variation in losses assigned to every single user is the same. This paper demonstrates the potential of the Aumann-Shapley
This means that all participants (loads, DG and D-STATCOM) game theory methodology to allocate losses among network

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KANPUR. Downloaded on June 28,2022 at 05:43:08 UTC from IEEE Xplore. Restrictions apply.
 AMARIS et al.: LOSS ALLOCATION IN DISTRIBUTION NETWORKS BASED ON AUMANN–SHAPLEY 6665

users (consumers, generators, prosumers, etc.) for distribution [9] European Commission, “Policies for DSOs, Distribution tariffs and data
networks. handling,” 2016.
[10] European Commission, “Study on tariff design for distribution system,”
The allocation of losses proposed in this study combines cir- 2015.
cuit theory with game theory through the analytical application [11] A. J. Conejo, J. M. Arroyo, N. Alguacil, and A. L. Guijarro, “Transmission
of the Aumann-Shapley method in power distribution networks. loss allocation: A comparison of different practical algorithms,” IEEE
Trans. Power Syst., vol. 17, no. 3, pp. 571–576, Aug. 2002.
The results obtained corroborate the influence of active and re- [12] M. Ilic, F. Galiana, and L. Fink, Power Systems Restructuring: Engineering
active components in the allocation of complex losses among and Economics (Kluwer International Series in Engineering). New York,
different agents. For the distribution of losses among network NY, USA: Springer, 1998.
[13] J. Daniel, R. Salgado, and M. Irving, “Transmission loss allocation through
users, it must be considered that the results of the allocation of a modified Y-bus,” IEE Proc.—Gener., Transmiss. Distrib., vol. 152, no. 2,
negative active losses do not entail a cross-subsidy among par- pp. 208–214, 2005.
ticipants because the Aumann-Shapley results reflect the partic- [14] W. L. Fang and H. W. Ngan, “Succinct method for allocation of network
losses,” IEE Proc.—Gener., Transmiss. Distrib., vol. 149, no. 2, pp. 171–
ipation of agents that reduce total network losses with benefits 174, Mar. 2002.
for all network users. [15] J. Mutale, G. Strbac, S. Curcic, and N. Jenkins, “Allocation of losses
The proposed method of allocating losses can be directly in distribution systems with embedded generation,” IEE Proc.—Gener.,
Transmiss. Distrib., vol. 147, no. 1, pp. 7–14, 2000.
applied for both radial and meshed networks with or without DG, [16] A. J. Conejo, F. D. Galiana, and I. Kockar, “Z-bus loss allocation,” IEEE
offering a full range of applications to distribution networks. Trans. Power Syst., vol. 16, no. 1, pp. 105–110, Feb. 2001.
This is one of the advantages that the methodology offers relative [17] J. Savier and D. Das, “An exact method for loss allocation
in radial distribution systems,” Int. J. Elect. Power Energy Syst.,
to other methods of allocation, such as the BCDLA, which can vol. 36, no. 1, pp. 100–106, 2012. [Online]. Available: http://www.
be applied only to radial networks. Likewise, it is able to handle sciencedirect.com/science/article/pii/S0142061511002791
sources of DER operating in power control mode or voltage [18] K. M. Jagtap and D. K. Khatod, “Novel approach for loss allocation
of distribution networks with DGs,” Electr. Power Syst. Res., vol. 143,
control mode, where methods based on the Jacobian matrix pp. 303–311, 2017.
(MLC, DLC) cannot be directly applied. [19] E. Carpaneto, G. Chicco, and J. S. Akilimali, “Branch current decom-
The simplicity of the final formulation of the proposed method position method for loss allocation in radial distribution systems with
distributed generation,” IEEE Trans. Power Syst., vol. 21, no. 3, pp. 1170–
is noteworthy, as it permits its implementation in distribution 1179, Aug. 2006.
networks where it is necessary to allocate the losses in real- [20] J. Contreras and F. F. Wu, “Coalition formation in transmission expansion
time. planning,” IEEE Trans. Power Syst., vol. 14, no. 3, pp. 1144–1152, Aug.
1999.
The methodology can be used to obtain the coefficients of [21] P. Young, “Cost allocation,” in Handbook of Game Theory with Economic
losses of different clients for tariff purposes in Smart Grids. Applications, vol. 2, R. J. Aumann and S. Hart, Eds. Amsterdam, The
Additionally, it can be applied to Demand Response situations Netherlands: North Holland, 1994.
[22] M. Junqueira, L. C. da Costa, L. A. Barroso, G. C. Oliveira, L. M. Thome,
offering the possibility to select users that show a greater po- and M. V. Pereira, “An Aumann-Shapley approach to allocate transmission
tential for reducing losses. With this knowledge, local actions service cost among network users in electricity markets,” IEEE Trans.
can be focused on key clients to increase network efficiency, Power Syst., vol. 22, no. 4, pp. 1532–1546, Nov. 2007.
[23] L. Shapley, “A value for n-person,” in Theory of Games. Annals of Mathe-
making it not necessary to implement central control actions in matical Studies, vol. II, H. Kuhn and A. Tucke, Eds. Princeton, NJ, USA:
the DSO centers. Princeton Univ. Press, 1953, pp. 307–317.
[24] R. Aumann and L. Shapley, Values of Non-Atomic Games. Princeton, NJ,
USA: Princeton Univ. Press, 1974.
[25] S. Sharma and A. Abhyankar, “Loss allocation for weakly meshed distri-
REFERENCES bution system using analytical formulation of Shapley value,” IEEE Trans.
Power Syst., vol. 32, no. 2, pp. 1369–1377, Mar. 2017.
[1] World Bank Database, “Electric power transmission and distri- [26] Y. P. Molina, O. R. Saavedra, and H. Amaris, “Transmission network
bution losses,” 2017. [Online]. Available: http://data.worldbank.org/ cost allocation based on circuit theory and the Aumann-Shapley method,”
indicator/EG.ELC.LOSS.ZS IEEE Trans. Power Syst., vol. 28, no. 4, pp. 4568–4577, Nov. 2013.
[2] Energy-Regulators, “Treatment of Losses by Network Operators. Euro- [27] D. A. Lima, J. Contreras, and A. Padilha-Feltrin, “A cooperative
pean Regulators Group Position Paper for public consultation,” 2008. game theory analysis for transmission loss allocation,” Electr. Power
[3] J. Savier and D. Das, “Impact of network reconfiguration on loss allocation Syst. Res., vol. 78, no. 2, pp. 264–275, 2008. [Online]. Available:
of radial distribution systems,” IEEE Trans. Power Del., vol. 22, no. 4, http://www.sciencedirect.com/science/article/pii/S0378779607000314
pp. 2473–2480, Oct. 2007. [28] S. M. Abdelkader, “Transmission loss allocation through complex power
[4] L. Mokgonyana, J. Zhang, L. Zhang, and X. Xia, “Coordinated two-stage flow tracing,” IEEE Trans. Power Syst., vol. 22, no. 4, pp. 2240–2248,
volt/var management in distribution networks,” Electr. Power Syst. Res., Nov. 2007.
vol. 141, pp. 157–164, Dec. 2016. [29] Y. P. Molina, R. B. Prada, and O. R. Saavedra, “Complex losses allocation
[5] K. Han, J. Lee, and J. Choi, “Evaluation of demand-side management over to generators and loads based on circuit theory and Aumann-Shapley
pricing competition of multiple suppliers having heterogeneous energy method,” IEEE Trans. Power Syst., vol. 25, no. 4, pp. 1928–1936, Nov.
sources,” Energies, vol. 10, no. 9, 2017, Art. no. 1342. 2010.
[6] A. G. Exposito, J. R. Santos, T. G. Garcia, and E. R. Velasco, “Fair [30] M. E. Baran and F. F. Wu, “Network reconfiguration in distribution systems
allocation of transmission power losses,” IEEE Trans. Power Syst., vol. 15, for loss reduction and load balancing,” IEEE Trans. Power Del., vol. 4,
no. 1, pp. 184–188, Feb. 2000. no. 2, pp. 1401–1407, Apr. 1989.
[7] L. Olmos and I. Perez-Arriaga, “A comprehensive approach for compu- [31] K. M. Jagtap and D. K. Khatod, “Loss allocation in distribution net-
tation and implementation of efficient electricity transmission network work with distributed generations,” IET Gener., Transmiss. Distrib., vol. 9,
charges,” Energy Policy, vol. 37, pp. 5285–5295, 2009. no. 13, pp. 1628–1641, 2015.
[8] A. Elmitwally, A. Eladl, and S. M. Abdelkader, “Efficient algorithm for [32] F. Flaih, X. Lin, M. Abd, S. Dawoud, Z. Li, and O. Adio, “A new
transmission system energy loss allocation considering multilateral con- method for distribution network reconfiguration analysis under differ-
tracts and load variation,” IET Gener. Transmiss. Distrib., vol. 9, no. 16, ent load demands,” Energies, vol. 10, no. 4, Apr. 2017, Art. no. 455. doi:
pp. 2653–2663, 2015. 10.3390/en10040455.

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KANPUR. Downloaded on June 28,2022 at 05:43:08 UTC from IEEE Xplore. Restrictions apply.
 6666 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 33, NO. 6, NOVEMBER 2018

Hortensia Amaris (M’97) received the M.Sc. and Ph.D. degrees in electrical Monica Alonso (M’11) received the M.Sc. degree in electrical engineering and
engineering from the Polytechnic University of Madrid, Madrid, Spain, in 1992 the Ph.D. degree in electrical engineering, electronics, and automation from the
and 1995, respectively. In 1996, she joined the Department of Electrical En- Universidad Carlos III de Madrid, Madrid, Spain, in 2003 and 2010, respectively.
gineering, University Carlos III of Madrid, Madrid, where she is currently a She is currently a Lecturer with the Department of Electrical Engineering,
Professor. She has been involved in several international research projects re- Universidad Carlos III de Madrid. Her research interests include distributed
lated to renewable energy sources, optimization, and control of smart grids. Her energy resources, smart grids, power quality, and power protection.
research interests include control from renewable energy sources, optimization
of power systems, and smart grids.

Yuri P. Molina was born in 1978. He received the Diploma in electrical en-
gineering from the National University of Engineering, Lima, Peru, in 2003,
the M.Sc. degree in power systems from the Federal University of Maranhão, Jaime E. Luyo received the M.Sc. degree in electrical engineering from the
São Luı́s, Brazil, in 2005, and the Ph.D. degree in electrical engineering from Rensselaer Polytechnic Institute, Troy, NY, USA, and the Ph.D. degree in eco-
the Catholic University of Rio de Janeiro, Rio de Janeiro, Brazil, in 2009. He nomics (applied to energy and public services regulation) from the Universidad
joined the Department of Electrical Engineering, Federal University of Paraı́ba, Nacional Mayor de San Marcos, Lima, Peru. He is currently the Director of the
João Pessoa, Brazil, in 2012. His research interests include control, operations, Ph.D. Energy Program with the Universidad Nacional de Ingenierı́a, Lima. His
planning, and economics of electric energy systems, as well as the development research interests include energy planning and policy, electric power systems,
of new electricity markets. sustainable energy development, energy economics, and energy markets.

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KANPUR. Downloaded on June 28,2022 at 05:43:08 UTC from IEEE Xplore. Restrictions apply.