Electrical Power and Energy Systems 101 (2018) 491–499

Contents lists available at ScienceDirect

Electrical Power and Energy Systems

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

Contribution of transmission and voltage constraints to the formation of T

locational marginal prices
⁎
Tatiana Vaskovskayaa, , Priyanko Guha Thakurtab, Janusz Bialekb
a
V.A. Trapeznikov of Institute of Control Sciences of Russian Academy of Sciences and National Research University of Moscow Power Engineering Institute, Moscow,
Russia
b
Center for Energy Systems, Skolkovo Institute of Science & Technology, Moscow, Russia

A R T I C LE I N FO A B S T R A C T

Keywords: This paper develops analytical formulae for expressing Locational Marginal Prices (LMPs) as the sum of com-
Locational marginal pricing ponents due to each of transmission and/or voltage binding constraints using full non-linear Alternating Current
Optimal power flow (AC) power system model. Each LMP component at a node consists of a price-bonding factor (PBF) multiplied by
Lagrange multipliers the marginal price of a corresponding marginal node. Hence PBF represents a price linkage between a marginal
and a non-marginal node due to a certain system operating constraint. The aim of the methodology is not to
determine LMPs but to decompose them in order to identify particular constraints that are affecting the for-
mation, thereby identifying those lines that may need upgrade. The methodology has been tested on 5-node PJM
and 2746-node Polish systems.

Nomenclature Parameters

Sets B ∈ + number of transmission lines in operation

D binding transmission constraints
N all nodes in the system
mP ,(tP ) set of real power injection variables at marginal (non-mar-
ginal or price-taking) nodes for active power
mQ,(tQ ) set of reactive power injection variables at marginal (non-
marginal or price-taking) nodes for reactive power
mθ,(tθ ) set of voltage angle variables at marginal (non-marginal or
price-taking) nodes for active power
mV ,(tV ) set of voltage magnitude variables at marginal (non-marginal
or price-taking) nodes for reactive power
m unified set of active and reactive power injection variables mP
and mQ (mP ∪ mQ )
mX unified set of voltage angle and magnitude variables mθ and
mV (mθ ∪ mV )
t unified set of active and reactive power injection variables tP
and tQ (tP ∪ tQ )
tX unified set of voltage angle and magnitude variables tθ and tV
(tθ ∪ tV )
N ∈  + number of nodes in the system
CPg (CPd ) ∈ N generator (demand) real power bid prices at nodes
CQg (CQd ) ∈ N generator (demand) reactive power bid prices at nodes
F ∈ B maximum allowable real power flow through lines
Pg ( P g ) ∈ N vector of maximum (minimum) real power generator
outputs
Pd ( P d ,) ∈ N vector of maximum (minimum) real power demand
outputs
Qg ( Q g ) ∈ N vector of maximum (minimum) capacitive reactive
power generator outputs
Qd ( Q d ) ∈ N vector of maximum (minimum) inductive reactive power
demand and compensator outputs
V ( V ) ∈ N vector of maximum (minimum) allowed voltage magni-
tudes
θ ( θ ) ∈ N vector of maximum (minimum) allowed voltage angles
λP (λQ ) ∈ N vector of Lagrange multipliers corresponding to real
(reactive) power flow equations
μV ( μV ) ∈ N vector of Lagrange multipliers corresponding to max-
imum (minimum) voltage magnitude constraints
μθ ( μθ ) ∈ N vector of Lagrange multipliers corresponding to max-
imum (minimum) voltage angle constraints
νg ( ν g ) ∈ N vector of Lagrange multipliers corresponding to maximum

⁎
Corresponding author.
E-mail address: vaskovskyata@mpei.ru (T. Vaskovskaya).

https://doi.org/10.1016/j.ijepes.2018.04.004
Received 27 June 2017; Received in revised form 21 February 2018; Accepted 4 April 2018
0142-0615/ © 2018 Elsevier Ltd. All rights reserved.
T. Vaskovskaya et al.
(minimum) real generation constraints
νd ( ν d ) ∈ N vector of Lagrange multipliers corresponding to maximum
(minimum) real demand constraints
ρg ( ρ g ) ∈ N vector of Lagrange multipliers corresponding to max-
imum (minimum) reactive generation constraints
ρd ( ρd ) ∈ N vector of Lagrange multipliers corresponding to max-
imum (minimum) reactive demand and compensation con-
straints
σ ∈ B vector of Lagrange multipliers corresponding to real power
flow constraints
Optimization variables
Pg ,Pd ∈ N vector of generator (demand) real power injections
Qg ,Qd ∈ N vector of generator (demand) reactive power injections
V ∈ N vector of node voltage magnitudes
θ ∈ N vector of node voltage angles
Other variables
F ∈ B vector of real power flow through lines
P ∈ N vector of net real power injections
Q ∈ N vector of net reactive power injections
In the subscript of a matrix, an alphabet following a colon (:) implies
that all the columns of the matrix designated by the alphabet set are
considered, whereas a colon (:) following an alphabet implies that all
the rows of the matrix designated by the alphabet set are considered.
1. Introduction
In order to increase competition among different stakeholders
electricity sector has been decentralized in many parts of the world
leading to rapid changes in industry structure and policies. Nodal pri-
cing of electricity has gained an increased importance in order to utilize
generation and transmission resources efficiently in a competitive en-
vironment and manage system congestion, especially through the use of
locational marginal pricing (LMP) [1,2]. LMP is one of the most pow-
erful means of congestion management in many electricity markets
worldwide, such as in the United States, Australia, Russia, etc.
The LMP at a particular node can be defined as the marginal cost
that would be incurred to supply an additional increase of a load at that
particular node while respecting all the security limits of the system.
The values of LMPs differ among the nodes due to the system hitting its
security limits and/or transmission losses.
LMP decomposition and nodal pricing were first introduced in the
eighties [3–5]. Such a decomposition approach is termed classical, and
decomposes each LMP into energy, loss and congestion components.
LMP decomposition is based either on linearized1 or non-linearized2
optimization, and can either be reference node independent or depen-
dent. The aim of calculating components is twofold. Firstly, the com-
ponents are used to calculate LMPs [6]. Secondly, components are
calculated for the purpose of financial instruments such as Financial
Transmission Rights (FTRs) [7] and Loss Hedging Rights (LHRs) [8].
LMPs represent price signals and their components provide transpar-
ency of pricing. LMP components produce useful information for pro-
moting investment in transmission system [9] and congestion
1 4
Direct Current Optimal Power Flow (DCOPF).
2
Alternating Current Optimal Power Flow (ACOPF).
492
Electrical Power and Energy Systems 101 (2018) 491–499
management [10]. Particularly, they are used to introduce congested
zones [11], which are necessary, for example, for calculating certain
power market indices in market monitoring issues [12].
Expression for reference node dependent components of LMPs are
derived in [13–15]. LMP decomposition based on distributed slack is
proposed in [16]. However, the reference node dependent approach
(which also includes randomly selected distributed slack) has severe
limitations [17,18]. Such limitations are addressed with reference node
independent decomposition in [19–23] considering DCOPF and ACOPF
respectively. However, different results may be produced by DCOPF
and ACOPF, as shown in [24]. In [19,23], the congestion component is
obtained in a reverse manner based on KKT conditions and marginal
losses. Approaches introduced in [20–22] do not work with non-line-
arized version of economic dispatch. Additional components are also
included in the LMP decomposition such as ancillary services like
spinning reserve, voltage and security controls in [25] and additional
risk component in [26]. Quadratic target function is studied in [27] in
which a connection between cost function and LMPs is proposed for
LMP decomposition. Study on dual space of economic dispatch to cal-
culate LMPs is performed in [28]. An in-depth analysis of orthogonal
space of Lagrangian multipliers is performed in [29,30]. The energy
reference is chosen based on an optimization problem resulting in
participation factors, the target function being the welfare of FTR
payments [31,32].
The above-mentioned literature does not emphasize the contribu-
tion of each binding constraint with respect to the marginal nodes of the
system on the formation of LMP at a node, that is, the response of the
marginal nodes to the system binding constraints is not captured.
Hence, this paper attempts to measure such responses and bridge the
gap between the system binding constraints and power flow that gen-
erates LMPs in the system, that is to introduce a physical sense to LMPs
and their components. The proposed methodology is based on power
flow comprising of marginal nodes in the system.3
The specific contribution of this paper is the development of a de-
composition technique to calculate LMPs in terms of influences of each
binding transmission and/or voltage constraints for full AC OPF en-
vironment.4 The developed methodology works also for linearized op-
timization and expresses LMP at a given node as a linear combination of
weighted bids at marginal generators (i.e. generators that do not op-
erate at their limits). The weights are termed as price-bonding factors
(PBFs) in this paper and they define a price linkage between a marginal
node and a non-marginal node with respect to a certain system oper-
ating constraint.
The decomposed PBFs give a signal of sharing costs among the
market participants when e.g. a particular constrained line is re-
inforced. They can also help monitoring electrical energy markets and
identifying zones for which market power indices are introduced [33].
The paper is organized as follows. Section 2 gives a brief overview
of the mathematical background behind LMPs in a system. The pro-
posed methodology to decompose LMPs in this paper is shown in Sec-
tion 3. The methodology is applied on two test systems, the results of
which are shown in Section 4. Finally, Section 5 draws the conclusions
of the paper.
2. Mathematical background behind LMPs
This section reviews the mathematics behind LMP decomposition
and is required to understand the proposed methodology of this paper.
The basic security-constrained economic dispatch can be formulated as:
⊺ ⊺ ⊺ ⊺
max CPd Pd + CQd Qd−CPg Pg−CQg Qg (1)
3
It can be applied only after solving the economic dispatch problem.
In some countries, such as in Russian Federation, the wholesale power market utilizes
such a model both for day-ahead and intra-day markets.
T. Vaskovskaya et al. Electrical Power and Energy Systems 101 (2018) 491–499

s. t . ⊺
⊺ ∂f (V ,θ) ⎤⊺ ∂fQ (V ,θ) ⎤ ∂F (V ,θ) ⎤⊺
Pg−Pd−fP (V ,θ) = 0 : λP ⎛ ∂L ⎞ = ⎡ P λP + ⎡ λQ + ⎡ σ + μ V − μV
(2) ⎢ ⎢ ⎥
⎝ ∂V ⎠ ⎣ ∂V ⎥ ⎦ ⎣ ∂V ⎦ ⎣ ∂V ⎦
Qg−Qd−fQ (V ,θ) = 0 : λQ (3) (17)
where,
F (V ,θ) ⩽ F :σ (4)
∂f (V ,θ ) ∂fP1 (V ,θ) ∂fP1 (V ,θ)
⎡ P1 ⋯ ⎤
V ⩽V⩽V : μV ,μV ⎢ ∂θ1 ∂θ2 ∂θN
⎥
(5) ⎢ ∂fP 2 (V ,θ) ∂fP 2 (V ,θ) ⎥
∂fP 2 (V ,θ)
∂fP (V ,θ) ⋯
= ⎢ ∂θ1 ∂θ2 ∂θN ⎥
θ⩽θ⩽θ : μθ ,μθ (6) ∂θ ⎢ ⋮ ⋮ ⋱ ⋮ ⎥
⎢ ∂f (V ,θ) ∂fPN (V ,θ) ∂f (V ,θ)
⎥
⎢ PN ⋯ PN ⎥
P g ⩽ Pg ⩽ Pg : ν g ,νg (7) ⎣ ∂θ1 ∂θ2 ∂θN ⎦
∂F (V ,θ) ∂F1 (V ,θ) ∂F1 (V ,θ)
P d ⩽ Pd ⩽ Pd : ν d,νd (8) ⎡ 1∂θ ∂θ2
⋯
∂θN
⎤
1
⎢ ⎥
∂F (V ,θ) ⎢ ∂F2 (V ,θ) ∂F2 (V ,θ)
⋯ ∂F2 (V ,θ)
⎥
Q g ⩽ Qg ⩽ Qg : ρ g ,ρg (9) = ⎢ ∂θ1 ∂θ2 ∂θN ⎥
∂θ ⎢ ⋮ ⋮ ⋱ ⋮ ⎥
⎢ ∂FB (V ,θ) ∂FB (V ,θ) ∂FB (V ,θ) ⎥
Q d ⩽ Qd ⩽ Qd : ρd ,ρd (10) ⎢ ⋯
⎣ ∂θ1 ∂θ2 ∂θN ⎥ ⎦
The main objective of Eqs. (1)–(10) is to maximize economic social ∂fQ (V ,θ) ∂f (V ,θ) ∂fQ (V ,θ)
In a similar manner ∂θ , P∂V , ∂V and ∂F (V ,θ) can be represented.
welfare. The stated mathematical framework can be applied both for ∂V
Equating (12)–(17) to zeros yield the KKT conditions for optimality.
centralized (cost-based) and decentralized (market-driven) electricity
The classical Jacobian matrix of the power flow solution is re-
markets. Eq. (1) is the objective function that maximizes the social
presented as:
welfare in the system. Costs CPd,CQd,CPg , and CQg denote prices for
generating and consuming active and reactive power (both inductive ∂f (V ,θ) ∂fP (V ,θ)
⎡ P ∂θ ∂V
⎤
and capacitive). Eqs. (2) and (3) represent the real and reactive power J = ⎢ ∂f (V ,θ) ∂fQ (V ,θ)
⎥
balance constraints at each node of the system respectively. Eq. (4) ⎢ Q ⎥
⎣ ∂θ ∂V ⎦ (18)
represents the real power flow transmission constraint. The constraint is
considered such that Fij ⩽ Fij and Fji ⩽ Fji , in which i and j are the two Let,
end nodes of the transmission line. Eqs. (5) and (6) represent the upper
A = ⎡ ∂F ∂(Vθ ,θ) ∂F (V ,θ)
⎤
and lower bounds on system node voltage magnitudes and angles re- ⎣ ∂V ⎦ (19)
spectively. The angle limits are included in order to simply the re-
and
presentation of the PBF formulae. Eqs. (7)–(10) represent the upper and
lower bounds on generator (demand) active and reactive powers re- X = [θ⊺ V ⊺]⊺ (20)
spectively. It is to be noted that elements of the vectors Pg and Qg (Pd and Then, Eqs. (16) and (17) can be re-written in matrix form as:
Qd ) are zeros corresponding to the nodes where no generators (de-
⊺
mands) are present. The same applies for CPg ,CPd,CQg ,CQd, P g ,Pg , P d,Pd ⎛ ∂L ⎞ = ⎡ ∂L ∂L ⊺
⎤ = J ⊺λ + A⊺ σ + μ
Q g ,Qg , ν g ,νg , ν d,νd, ρ g ,ρg , ρd , and ρd . ⎝ ∂X ⎠ ⎣ ∂θ ∂V ⎦ (21)
The Lagrangian function for the above problem can be written as: where, λ =
⊺
[ λP λQ⊺ ]⊺
and μ = [ μθ⊺ μV⊺ ]⊺−[ μθ⊺
The zeros in the μ μV⊺ ]⊺.
⊺
L = CPg ⊺
Pg + CQg ⊺
Qg−CPd ⊺
Pd−CQd Qd vector correspond to the voltage angles (zero for slack node and the
nodes without constraints) in the system. Optimal solution to the pro-
− [Pg−Pd−fP (V ,θ)]⊺λP −[Qg−Qd−fQ (V ,θ)]⊺λQ
blem (1)–(10) is considered for further derivations. It is to be noted that
+ [V −V ]⊺μV + [ V −V ]⊺ μV + [θ−θ ]⊺μθ + [ θ−θ]⊺ μθ in further derivations the costs CPd,CQg and CQd are considered to be
zeros while P d = Pd, Q d = Qd . Eq. (21) forms the basis for the standard
+ [F (V ,θ)−F ]⊺σ
and proposed decomposition techniques.
+ [Pg−Pg ]⊺νg + [ P g−Pg ]⊺ ν g + [Qg−Qg ]⊺ρg + [ Q g−Qg ]⊺ ρ g
+ [Pd−Pd]⊺νd + [ P d−Pd]⊺ ν d + +[Qd−Qd]⊺ρd + [ Q d−Qd]⊺ ρd (11) 3. Proposed LMP decomposition

Taking partial derivatives of (11) with respect to Pg ,Qg ,θ,V yield: Solving the OPF problem (1)–(10) also identifies marginal gen-
⊺ erators besides calculating LMPs. LMP at a marginal node is always
⎛ ∂L ⎞ = C −λ + ν − ν equal to the bid at that node due to their ability to change their output
⎜ ⎟ Pg P g g
⎝ ∂Pg ⎠ (12) under a small system perturbation as they do not operate at their limits.
⊺
The LMPs at other nodes (called price taking nodes) are equal to a
⎛ ∂L ⎞ = C −λ + ρ − ρ linear combination of bids of marginal generators. The same theory
⎜ ⎟ Qg Q g g
⎝ ∂Qg ⎠ (13) applies for marginal demand.
The proposed methodology is based on the concept of marginal
⊺
nodes for active (mP ) and reactive (mQ ) power. Indeed, considering
⎛ ∂L ⎞ = −CPd + λP + νd− ν d
⎜ ⎟
active power, Lagrange multipliers νg , ν g for such generators or νd, ν d for
⎝ ∂Pd ⎠ (14)
such demands are equal to zeros. According to (12) and (14), LMP for
⊺ active power at such a node is the offer or bid price. This can be the case
⎛ ∂L ⎞ = −CQd + λQ + ρ − ρ
⎜ ⎟
d d for both generation and demand at one node simultaneously if they
⎝ ∂Qd ⎠ (15)
have same offer/bid prices. Same argument applies to reactive power.
⊺ ∂f (V ,θ) ⎤ ⊺
∂fQ (V ,θ) ⎤ ⊺ ⊺ The set of active (reactive) power variables is denoted as tP (tQ ) .
⎛ ∂L ⎞ = ⎡ P λ +⎡ λ +⎡
∂F (V ,θ) ⎤
σ + μθ − μ θ Notations mθ and tθ are used for voltage angle variables at the marginal
⎝ ∂θ ⎠ ⎢
⎣ ∂θ ⎦ ⎥ P ⎢ ∂θ ⎥ Q ⎣ ∂θ ⎦
⎣ ⎦ and non-marginal nodes for active power respectively, whereas nota-
(16) tions mV and tV are applied for voltage magnitude variables at the

493
T. Vaskovskaya et al. Electrical Power and Energy Systems 101 (2018) 491–499

marginal and non-marginal nodes for reactive power respectively. ⎛ pf ⎞

⊺ ⊺ ⊺ ⊺
Let Cm = [CmP CmQ ] = [Cm⊺ P 0⊺] be the vector of price-forming λj = ∑ k + ∑ k jid + ∑ k uji + ∑ k jil Ci
⎜ ji ⎟
bids at the marginal nodes. The Jacobian matrix J can be divided into i∈m ⎝ d∈D u∈V ( V ) l ∈ θ ( θ) ⎠ (30)
four parts based on the type of nodes m and t. Hence, Eq. (21) can be re- where Ci is the price of the i th bid (price forming offer). and k jil kjipf ,k jid,k uji
written as: represent the PBF of each bid initiated due to power flow (2) and (3)
⊺
μ and binding constraints (4)–(6). All the PBFs are computed analytically,
⎡ JmmX JmtX ⎤ ⎡Cm ⎤ + [ A: m A: t ]⊺σ + ⎡ mX ⎤ = 0
⎢ J J ⎥ λ X X
⎢ μ the algorithm of which is described in Appendix A. The sum of obtained
⎣ tmX tt X ⎢
⎦⎣ ⎦ t ⎥ ⎣ tX ⎥
⎦ (22)
components is equal to LMPs traditionally computed with OPF proce-
JmP mθ JmP mV ⎤ JmP tθ JmP t V ⎤ dure.
where JmmX = ⎡ , JmtX = ⎡ , JtmX =
⎢ JmQ mθ JmQ mV ⎥ ⎢ JmQ tθ JmQ t V ⎥ Marginal nodes address the power flow solution which is required
⎣ ⎦ ⎣ ⎦
for computation of the PBFs. In a power flow calculation the reduced
⎡ JtP mθ JtP mV ⎤, J = ⎡ JtP tθ JtP t V ⎤, λ = [ λ ⊺ λ ⊺ ]⊺, A
tP tQ : mX =[ A: mθ A: mV ], Jacobian matrix is obtained from its full counterpart by eliminating the
⎢ JtQ mθ JtQ mV ⎥ ttX ⎢ JtQ tθ JtQ t V ⎥ t
⎣ ⎦ ⎣ ⎦ corresponding rows and columns of voltage magnitudes and angles of
⊺ ⊺ ⊺ ⊺ ⊺ ⊺
A: tX = [ A: tθ A: t V ], μmX = [ μmθ μmV ] and μtX = [ μtθ μt V ] . slack and PV nodes. Applying the same for marginal nodes (as the
Considering the second row of Eq. (22) the LMPs at price-taking voltage magnitudes and angles are known) results in JttX , as obtained in
nodes can be expressed in terms of other nodes. Eq. (29). Increasing 1 MW of demand at a node requires increasing
generation at the marginal nodes. Thus the power flow PBF for node j
( )−1 (Jmt⊺
λt = − Jtt⊺X X Cm + A:⊺tX σ + μtX ) (23)
can be defined with respect to marginal node i as
It is to be noted that the sub matrix JttX represents the Jacobian
∂Pi
matrix for the Newton Raphson power flow calculation, when the rows kjipf = ,
∂Pj
and columns corresponding to marginal nodes for active and reactive
⏟
PF with MN (31)
power are eliminated as the outputs for them are not known. A careful
observation of Eq. (23) reveals that it can be decomposed into three where PF refers to power flow and MN refers to marginal nodes. In such
components: the power flow, a case the power flow PBFs form a participation vector for adjusted
marginal nodes.
λtpower flow = − Jtt⊺X( )−1Jmt⊺ X Cm (24) In this process the binding constraints get shifted. Power injections
at marginal nodes need to be changed in order to shift binding flows or
the transmission congestion
voltages within their allowable limits. For example, for d transmission
λttrans. cong. = − Jtt⊺X ( )−1A:⊺t X σ (25) constraint it can be expressed as

and the voltage congestion ⏞

OPF
∂Fd ∂Pi
k jid =
λtvolt. cong. = − Jtt⊺X
( ) −1μ
tX (26)
∂Pj ∂Fd
⏟
PF with MN (32)
By definition, Lagrange multipliers are equal to the derivatives of
Hence, a PBF is directly attributed to a particular constraint and shows
the objective function with respect to a constraint limit. Hence,
the appearance of corresponding LMP component. When congestion is
⊺
∂f ⎞ relieved ∂Pi/ ∂Fd becomes zero. In case of no congestion PBFs due to
σ=⎛
⎝ ∂F ⎠ (27a) them are also zeros. Proofs of (31) and (32) are given in Appendix B.
It is to be noted that the loss components are not dealt with in this
⊺
μtX = ⎛
∂f
⎜ −
∂f ⎞
⎟
paper. For both the unconstrained and constrained (OPF) power flows
⎝ ∂X t ∂ Xt ⎠ (27b) the changes in generation at marginal nodes already include losses to
cater the changes in demand at the corresponding node.
The piece-wise form of (1) and the ability to change the output only
for variables at marginal nodes aids in expressing σ and μtX in the
4. Illustrative examples
following forms5:
∂P ⊺ ∂X ⎞⊺ Simulations are performed on 5-node and 2746-node Polish systems
σ = −⎛ m ⎞ CPm = −⎛Jm : Cm
⎝ ∂F ⎠ ⎝ ∂F ⎠ (28a) in order to show the effectiveness of the proposed methodology.
⊺ ⊺
Matpower [34] is used to perform the simulations.
∂P ∂P ∂X ∂X ⎞
μtX = −⎛ m − m ⎞ CPm = −⎛Jm :
⎜ −Jm :
⎟ Cm ⎜ ⎟

⎝ ∂Xt ∂ X t ⎠ ⎝ ∂Xt ∂X t ⎠ (28b) 4.1. 5-node PJM system

∂X ∂X ∂X
where = ⎡ ∂θ ⎤. ∂X can be represented in a similar way. The
∂Xt
⎣ t ∂Vt ⎦ ∂ X t The proposed approach is demonstrated on the small system of [35]
∂P ∂P
derivatives ∂Ft , ∂Vt and ∂Pt at non-marginal nodes are zeros. Substituting which is often used for performing power economic studies. Results of
t ∂Vt
(28a) and (28b) in (23) yields the economic dispatch (1)–(10) are shown in Tables 1 and 2 for nodes
⊺
and lines respectively (see the initial case). LMPs differ from those in
∂X ⎞⊺ ⊺ ∂X ∂X ⎞ ⊺ ⎤
( )−1 ⎡⎢−Jmt⊺
λt = Jtt⊺X X + A:⊺tX ⎛
⎝ ∂F ⎠
Jm : + ⎛
⎜ − ⎟ Jm : Cm
⎥
[35] due to the difference in constraint (4) which in this paper refers to
⎣ ⎝ Xt ∂ X t ⎠
∂ ⎦ (29) real power rather than apparent power in [35]. Nodes 3 and 5 are
considered to be the marginal nodes for active power, the bidding
Calculation of the derivatives in (29) is performed by sensitivity ana-
prices of which are 30 and 10 $/MW respectively, and are colored pink.
lysis.
At the same time, nodes 4 and 5 balance the reactive power in the
Finally, for each price-taking node j Eq. (29) can be expressed as a
system. They are marginal for reactive power and colored orange.
sum of prices at marginal nodes multiplied by the PBF of power flow,
Hence, Pm = [ P3 P5 ]⊺, Qm = [Q4 Q5 ]⊺, Pt = [ P1 P2 P4 ]⊺,
transmission and voltage constraints. It can be represented as:
Qt = [Q1 Q2 Q3 ]⊺, θt = [θ1 θ2 θ4 ]⊺, Vt = [V1 V2 V3 ]⊺, λt =
[ λP1 λP 2 λP 4 λQ1 λQ2 λQ3 ]⊺, Cm = [Cg3 Cg5 0 0 ]⊺.
5
Whenever mp is used in conjunction with vector P, suffix p is omitted. The same is The solution of ACOPF results in one binding transmission con-
applied to other variables. straint between nodes 4 and 5 and a voltage constraint at node 3. The

494
T. Vaskovskaya et al. Electrical Power and Energy Systems 101 (2018) 491–499

Table 1 Table 3
Node data for ACOPF for the PJM 5-node system. PBFs and LMP components of price-taking nodes for the PJM 5-node system.
Ci ($/MW) Power flow Binding constraints Formed LMPa
kjipf k jiF5 − 4 k jiV 3 ∑ kji

Price-taking node 1
C3 = 30 0.1819 0.1641 0.0000 0.3459
C5 = 10 0.8181 −0.1666 0.0000 0.6516

LMP componentsb 13.64 3.26 0.00 16.89

Price-taking node 2
C3 = 30 0.7780 0.0465 0.0000 0.8246
C5 = 10 0.2258 −0.0473 −0.0000 0.1786

node at which the voltage magnitude is controlled is colored green. The LMP components 25.60 0.92 0.00 26.52
calculated PBFs for price-taking nodes are shown in Table 3.
Two marginal nodes define prices at price-taking nodes. Node 1 has Price-taking node 4
two infra marginal generators, node 2 has a load and node 4 has a load C3 = 30 0.4027 1.0948 −0.0000 1.4975
C5 = 10 0.6041 −1.1117 0.0001 −0.5075
and an extra marginal generator. Consider node 4 with the highest LMP
of 39.85 $/MWh. Its power flow component is obtained from regular LMP components 18.12 21.73 0.00 39.85
power flow with marginal nodes (31). The derivatives ∂P3/ ∂P4 and
a
∂P5/ ∂P4 are equal to 0.4027 and 0.6041 respectively. The nodal marginal The last column shows summarized PBFs for offers at marginal nodes and
power injection at node 4 leads to a shift of the flow through trans- the final LMPs.
b
mission line 4–5, the sensitivity ∂F4 − 5/ ∂P4 of which is 0.3526. To bring LMP components are obtained as weighted sum of marginal prices with
the flow back to the allowable limit re-dispatch of marginal nodes is corresponding PBFs.
required. Lowering transmitted power by 1 MW to respect the flow limit
requires the output of the cheaper marginal generator at node 5 to be
decreased by 3.1524 MW and increase the more expensive one at node Table 4
3 by 3.1046. Multiplying those numbers by 0.3526 according to (32) Node data for ACOPF for the modified PJM 5-node system.
the PBFs of transmission congestion 1.0948 and −1.1117 are obtained
for nodes 3 and 5 respectively. Thus, it can be concluded that the for-
mation of the LMP at node 4 stipulates the sensitivity of 0.4027 MW at
marginal node 3 due to power flow and 1.0948 MW more due to
transmission congestion. Marginal cost of this change is 30 $/MWh.
Also 1 MW rise in demand at node 4 requires changing of generated
power at node 5 by + 0.6041 MW due to power flow and by −1.1117 MW
due to transmission congestion. Bid price of such changes equals 10 $/
MWh, resulting in LMP of 39.85 $/MWh.
The proposed methodology also helps in identifying the impact of
individual transmission constraints on the LMPs in a congested area.
The effect of several congested lines is exemplified by lowering trans-
mission limit in line 1–2 to 250 MW (modified case). The results are obtained by first multiplying each PBF in corresponding columns by the
given in Tables 4 and 5. Since marginal generators have been changed, corresponding bids of marginal generators and then adding them. It
LMP representation in the system also changes. New PBFs for nodes are represents the transmission and voltage components of LMP at that
given in Table 6. A careful investigation of the table reveals that the particular node. In this example voltage component is insignificant.
influence on the LMP at price-taking node 2 is due to congestion at line The table confirms that the LMPs at price taking nodes appear as a
1–2, whereas the LMP at a price-taking node 4 is influenced by 5–4. It is linear combination of the bids at marginal nodes. The LMPs at the
evident that the contribution of line 5–4 is 46 % of the LMP in this case. marginal nodes are independent of each other and have no other
Hence, it needs an upgrade if it is a major contributor to increasing components.
LMPs on a regular basis due to daily generation load patterns. Hence,
the proposed methodology helps to identify the exact influence of a
particular system constraint on the determination of LMPs at price- 4.2. 2746-node Polish system
taking nodes.
It must be noted that the LMP components in Tables 3 and 6 are In order to show the effectiveness of the proposed methodology on a
bigger system 2746-node Polish system at winter peak is considered.

Table 2
Branch data for ACOPF for the PJM 5-node system.
Branch 1–2 1–4 1–5 2–3 3–4 4–5

Pij starting end (MWh) 253.11 187.68 −230.80 −48.43 −24.49 −238.45
Pji receiving end (MWh) −251.57 −186.63 231.15 48.63 25.08 240.00
Qij starting end (MVAr) −25.03 73.72 108.81 −138.27 153.40 −39.86
Qji receiving end (MVAr) 39.66 −63.95 −108.89 137.99 −148.23 54.56
σ ($/MW) – – – – – 61.61

495
T. Vaskovskaya et al. Electrical Power and Energy Systems 101 (2018) 491–499

Table 5
Branch data for ACOPF for the modified PJM 5-node system.
Branch 1–2 1–4 1–5 2–3 3–4 4–5

Pij starting end (MWh) 250.00 186.38 −236.21 −51.49 −23.58 −238.52
Pji receiving end (MWh) −248.51 −185.42 236.60 51.67 23.94 240.00
Qij starting end (MVAr) −16.92 50.61 123.80 −129.61 118.09 −13.79
Qji receiving end (MVAr) 31.00 −41.77 −123.60 129.15 −115.30 27.85
σ ($/MW) 8.93 – – – – 47.18

Table 6
PBFs and LMP components of price-taking nodes for the modified PJM 5-node
system.
Ci ($/MW) Power flow Binding constraints Formed LMP
kjipf k 2Fi1 − 2 k 2Fi5 − 4 ∑ k2i

Price-taking node 2
C1 = 15 0.2735 −1.2925 0.0013 −1.0177
C3 = 30 0.7289 0.4477 0.0001 1.1767
C5 = 10 0.0015 0.8415 −0.0013 0.8417

LMP components 25.98 2.46 0.01 28.45

Price-taking node 4
C1 = 15 0.3318 −0.0017 2.5538 2.8839
C3 = 30 0.3398 0.0006 0.1591 0.4996
C5 = 10 0.3350 0.0011 −2.7226 −2.3865 Fig. 1. Spatial LMP representation for the Polish system. Black circles represent
marginal generators, red lines represent congested transmission lines. (For in-
LMP components 18.52 0.00 15.86 34.38 terpretation of the references to color in this figure legend, the reader is referred
to the web version of this article.)

The corresponding marginal nodes for active power are 58, 382, 736,
neighborhood of the considered optimal solution. A close look at Fig. 3
1996, and 2096.6 The bids at the marginal nodes are 45, 4, 41, 36 and
reveals that the marginal node 382 has a negative influence on the LMP
62 $/MW respectively.
at node 417. A study of the effect on main components of the LMP at
The limits of the transmission lines between nodes 340–9, 389–20,
node 417 (Fig. 4) is conducted by changing the LMP at node C382 . Node
417–379, 1244–1359, 1559–1454 and 1964–1996 as well as the lower
382 remains marginal while changing C382 from 0 to 75 $/MWh. Be-
limits of voltage constraints at nodes 212, 301 and 505 are altered in
cause of close connection between node 417 and marginal node 382 the
order to create congestion in the system deliberately. The input data
power flow component of node 417 is approximately equal to C382 . As a
with the corresponding changes, ACOPF results and calculated PBFs can
result, the difference in LMPs before and after recalculation is
be found in [36].
∑ k 417 382 C382 . It is worth mentioning that in general the dependency is
LMPs, marginal generators, and congested lines are shown in Fig. 1.
nonlinear but according to tests this nonlinearity is not that significant.
In the figure, dotted lines show impact zones of the corresponding
In order to see the contribution of line 417–379 to the high value of
transmission constraints. High prices in the four zones in the middle are
LMP at node 417, the maximum permissible active power flow for line
caused by the generator at node 2096 which has the highest price
417–379 is increased from 67.5 MW to 82.5 MW. Upgrade of the line
among all the marginal ones. An increasing trend on prices in those
leads to decreasing transmission congestion component. Fig. 4 depicts
areas can be observed. The green areas with dots represents the nodes
the changes of λ417 P
and its main components: power flow and trans-
having significant impact of a voltage constraints.
mission line 417–389 components as an effect of increasing maximum
Fig. 2 shows the LMPs at all nodes in the system. It can be seen that
allowable real power flow through line 417–389. The power flow
a significant number of nodes have very high and low values of LMPs,
component of the LMP at node 417 is rather small and the value of LMP
out of which nodes 417 and 1559 have the highest 458.96 and lowest
is defined completely by the component of transmission line 417–382.
−17.16 values respectively. The conventional LMP decomposition is
When the limit increases generator 382 is able to increase its output
unable to detect the exact reason behind those values.
too. When the limit is greater than 73.4 MW the generator achieves its
Fig. 3 shows the reason behind such high and low values of LMPs at
capacity and becomes inframarginal. This explains the rise of power
the above-mentioned nodes. A close observation of the figure reveals
flow component. Continuing upgrading of the line leads to decreasing
that the bottleneck for the high value of LMP at node 417 is the
LMP at node 417 and its congestion component. The latter decreases to
transmission congestion on line 417–379. The LMP is determined by the
zero when the congestion is relieved.
marginal generators at nodes 382 and 2096. Similarly, the low value of
Applying the proposed methodology over an extensive period of
LMP at node 1559 is due to the congested line 1559–1454. The LMP is
time will be beneficial to make planning decisions of expanding
determined by the marginal generators at nodes 58 and 2096.
transmission network. For example, the transmission line 417–379
The LMP at node 212 is determined by the marginal generators 58,
needs to be upgraded if it continues contributing high LMP components
382 and 2096 due to voltage constraint at that node. The same gen-
for different loading patterns. Similarly voltage supporting device may
erators are responsible for the LMP at node 505 due to the voltage
need to be installed at node 212 if it continues to contribute to high
constraint at that node. Unlike transmission constraints the contribu-
values of LMP for an extensive period of time.
tion of voltage constraints to LMP formation is not very high.
A test is performed in order to show the linearity of Eq. (30) in the
5. Conclusion

6
The other generators are non-marginal and operating at their limits. This paper proposes a new methodology to decompose Locational

496
T. Vaskovskaya et al.
Fig. 2. LMPs at all nodes of the system.
Marginal Prices (LMPs) at price taking nodes into a linear combination
of components due to each transmission and/or voltage constraints in
the system. Each LMP component at a price-taking node consists of a
price-bonding factor (PBF) multiplied by the marginal price of a cor-
responding marginal node. Hence PBF represents a price linkage be-
tween a marginal and non-marginal node due to a certain system op-
erating constraint. This decomposition is done with respect to the
offers/bids at marginal nodes in the system and is independent of the
reference node. This is the advantage over classical reference node
dependent LMP decomposition. Moreover, the proposed decomposition
shows the changes in output of marginal generators and consequently
in marginal cost of the system due to the impact of transmission and/or
voltage constraints in the system.
Most of the commercial softwares use linear approach of the AC
497
Electrical Power and Energy Systems 101 (2018) 491–499
power flows in order to calculate LMPs in a system. Such approach has
severe limitations especially when the system is constrained due to
system voltages. This paper exploits the direct use of ACOPF in order to
calculate the LMPs which helps to reflect the voltage constraints in the
system in determining the LMPs. The developed methodology provides
transparency and can generate price signals that can support useful
information about the necessity of expanding the transmission infra-
structure to manage system congestion. The methodology can be ex-
tended in determining congestion zones due to each binding transmis-
sion constraints for either short or long terms. The decomposed PBFs
give a signal of sharing costs among the market participants when e.g. a
particular constrained line is reinforced. It can also help in monitoring
electrical energy markets. The methodology has been tested on a small
and a big system.
Fig. 3. PBFs for 4 price taking nodes. The x-axis
shows the congested lines or voltage constraints
whereas the y-axis represents the PBFs representing
the influence of each congested line in the two
subfigures on the left while the influence of voltage
constraints in the two subfigures on the right. The
color representation shows the contribution of each
corresponding marginal node. (For interpretation
of the references to color in this figure legend, the
reader is referred to the web version of this article.)
T. Vaskovskaya et al. Electrical Power and Energy Systems 101 (2018) 491–499

Fig. 4. LMP at node 417 and its main components – power flow and transmission line 417–389 components vs bid price of marginal generator 382 and transmission
limit in line 417–389.

Appendix A. Algorithm of the proposed methodology

This section describes the algorithm behind calculating the PBF that form the LMP at each non-marginal node of the system.

1. Run ACOPF (1)–(10) and obtain the Jacobian matrix J from the solution.
2. Filter JmtX and JttX sub-matrices out of J. These matrices represent an incremental change of X at non-marginal nodes on system injections at
marginal and non-marginal nodes respectively.
3. Compute kjipf , which is the dot product of the j th -row and i th -column of the matrix −(Jtt⊺)−1Jmt
⊺
∈ |t | × | m|. It represents the power flow PBF of i th
marginal node on the LMP of the j non-marginal node. Multiplying by Ci results in the component of the i th marginal node power flow price.
th
∂X ⊺
( )
4. Compute k jid . In order to calculate them two matrices need to be calculated separately: −(Jtt⊺)−1A:⊺t ∈ |t | × | D | and ∂F Jm⊺ : ∈ |D | × | m|. Dot product of
j th column of the former matrix and the i th row of the latter matrix yields a |t| × |m| matrix of PBFs due to d th transmission constraint. Hence, k jid is
the j th row and the i th column of that particular matrix.
∂X ⊺ ⊺
5. Compute k uji and k jil . In order to calculate them again two matrices need to be calculated separately: (Jtt⊺)−1 ∈ |t | × | t| and
⊺
( )J
∂Xt m: ∈ |t | × | m|

⎣ ( )
t
∂X
⎡or ∂ X Jm⊺ : ∈ |t | × | m|⎤7 For each tQ where active voltage constraints are observed, k uji or k jil are obtained by iterating each j th column of the former
⎦
matrix and the i th row of either of the latter matrices depending on the corresponding violations.
6. Finally, the LMPs at price taking nodes are obtained according to Eq. (30) by substituting the corresponding values of the PBFs computed
according to the steps stated above. The sum of LMP components should be equal to LMPs obtained in OPF solution.

Appendix B. Proofs of Eqs. (31) and (32)

Theorem 1. Every PBF kjipf at a price-taking node jcorresponding to a marginal bid Ci for a given OPF result represents per unit increment of output at a
marginal node i associated with increasing the demand at node j.
Proof. A small incremental change of power at price-taking nodes [ dPt⊺ dQt⊺]⊺ results in a change of voltages at the same nodes:

⎡ dθt ⎤ = J −1 ⎡ dPt ⎤.
ttX
⎢
⎣ dVt ⎥
⎦ ⎢
⎣ dQt ⎥
⎦ (B.1)
For solving PF problem equations for marginal nodes are not considered. Voltage angles are fixed and known from OPF solution. Voltage
magnitudes are fixed at nodes balancing reactive power. The output at marginal nodes then changes accordingly:

⎡ dPm ⎤ = Jmt ⎡ dθt ⎤ = Jmt J −1 ⎡ dPt ⎤.

X X ttX
⎢ dQm ⎦
⎣ ⎥ ⎢ dVt ⎦
⎣ ⎥ ⎢ dQt ⎦
⎣ ⎥ (B.2)
Representing the demand change as a negative identity matrix [ dPt⊺ dQt⊺]⊺ = −I yields

⎡ dPm ⎤ = −Jmt J −1 I = −Jmt J −1.

X ttX X ttX
⎢
⎣ dQm ⎥
⎦ (B.3)
Transposing (B.3) results in the factors in (24):
⊺
dP
( )−1Jmt⊺
λtpower flow = − Jtt⊺X X Cm = ⎡ m ⎤ Cm.
⎣ dQm ⎥
⎢ ⎦ (B.4)
Theorem 2. For a given OPF result, every PBF reflects the change in real power flow through a congested line d that resulted due to a change in power
k jid
demand at node j which in turn is met by change in generation at the marginal nodes.
Proof. According to (B.1) a small incremental change of power at price-taking nodes [ dPt⊺ dQt⊺]⊺ results in voltage changes at the same nodes. Voltage

7
Voltage at a particular node can violate either its upper or its lower limit.

498
T. Vaskovskaya et al. Electrical Power and Energy Systems 101 (2018) 491–499

angles (magnitudes) are fixed at the marginal nodes for active (reactive) power. The change of real power flow through the lines can be given as:
dθ dPt ⎤
dF = A: tX ⎡ t ⎤ = A: t Jtt−X1 ⎡ .
⎢
⎣ dVt ⎥
⎦ ⎣ dQt ⎥
⎢ ⎦ (B.5)
[ dPt⊺ dQt⊺]⊺ = −I yields:
dF = −A: tX Jtt−X1. (B.6)
Hence the following can be obtained for (25)
∂P ⊺
λttrans. cong. = − Jtt⊺X
( )−1A:⊺t X σ = dF ⊺ ⎛ m ⎞ CPm.
⎝ ∂F ⎠ (B.7)
Theorem 3. For a given OPF result, every PBF k jid
reflects the change in voltage magnitude (angle) of a node u (l) that resulted due to a change in power
demand at node j which in turn is met by change in generation at the marginal nodes.
Proof. This proof is similar to the one introduced above. According to (B.1) perturbation [ dPt⊺ dQt⊺]⊺ = −I exerts the change of voltages at price-
taking nodes:

⎡ dθt ⎤ = −J −1 I = −J −1.
ttX ttX
⎢ dVt ⎦
⎣ ⎥ (B.8)
Putting (B.8) into (26) yields
⊺ ⊺
dθ ∂P ∂P
λtvolt. cong. = − Jtt⊺X
( )−1μt = ⎡ t ⎤ ⎛ m − m ⎞ CPm.
⎜ ⎟
X ⎢ dVt ⎦
⎣ ⎥ ⎝ ∂Xt ∂ X t ⎠ (B.9)

Appendix C. Supplementary material

Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.ijepes.2018.04.004.

References
[1] Hogan WW. Contract networks for electric power transmission. J Reg Econ
1992;4:211–42.
[2] Hogan WW, Read EG, Ring BJ. Using mathematical programming for electricity spot
pricing. Int Trans Oper Res 1996;3:209–21.
[3] Schweppe FC, Caramanis MC, Tabors RD, Bohn RE. Spot pricing of electricity.
Norwell (MA): Kluwer; 1988.
[4] Caramanis MC, Bohn RE, Schweppe FC. Optimal spot pricing: practice and theory.
IEEE Trans Power App Syst 1982;PAS-101(9):3234–45.
[5] Bohn RE, Caramanis MC, Schweppe FC. Optimal pricing in electrical networks over
space and time. Source: RAND J Econ Rand J Econ 1984;15(3):360–76.
[6] Fu Y, Li Z. Different models and properties on lmp calculations. Proc. 2006 IEEE
power engineering society general meeting. IEEE; 2006.
[7] Quintana V, Bautista G. Financial transmission rights in electricity market. IEEE
mediterranean proc. electrotechnical conference, MELECON 2006. IEEE; 2006.
[8] Rudkevich E, Hausman E, Tabors RD, Bagnall J, Kopel C. Loss hedging rights: a final
piece in the LMP puzzle. In: Proc. 38th Hawaii int. conf. system sciences, Waikoloa,
HI; 2005.
[9] Babic AB, Saric AT, Rfnkovic A. Transmission expansion planning based on
Locational Marginal Prices and ellipsoidal approximation of uncertainties. Int J
Electr Power Energy Syst 2013;53:175–83.
[10] Pillay A, Prabhakar Karthikeyan S, Kothari D. Congestion management in power
systems – a review. Int J Electr Power Energy Syst 2015;70:83–90.
[11] Volodin DV, Vaskovskaya TA. Clustering approach for determination of congestion
zones on nodal electricity markets in long term periods. 2015 IEEE Eindhoven
PowerTech. IEEE; 2015. p. 1–6.
[12] Twomey P, Green R, Neuhoff K, Newbery D. A review of the monitoring of market
power; 2006.
[13] Baughman ML, Siddiqi SN, Zarnikau JW. Advanced pricing in electrical systems.
Part I: theory. IEEE Trans Power Syst 1997;12(1):489–95.
[14] Baughman ML, Siddiqi SN, Zarnikau JW. Advanced pricing in electrical systems.
Part II: implications. IEEE Trans Power Syst 1997;12(1):496–502.
[15] Chen L, Suzuki H, Wachi T, Shimura Y. Components of nodal prices for electric
power systems. IEEE Trans Power Syst 2002;17(1):41–9.
[16] Wu T, Alaywan Z, Papalexopoulos AD. Locational marginal price calculations using
the distributed-slack power-flow formulation. IEEE Trans Power Syst
2005;20(2):1188–90.
[17] Rivier M, Pérez-Arriaga I. Computation and decomposition of spot prices for
transmission pricing. In: 11th PSC conference; 1993. p. 371–8.
[18] Cardell J. Marginal loss pricing for hours with transmission congestion. IEEE Trans
Power Syst 2007;22(4):1466–74.
[19] Orfanogianni T, Gross G. A general formulation of LMP evaluation. IEEE Trans
Power Syst 2007;2(3):1163–73.
[20] Hu Z, Cheng H, Yan Z, Li F. An iterative LMP calculation method considering loss
distributions. IEEE Trans Power Syst 2010;25(3):1469–77.
[21] Li F, Litvinov E, Zheng O. A combined LMP model considering reference in-
dependent decomposition and fictitious nodal demand. IEEE PES general meeting.
IEEE; 2010. p. 1–5.
[22] Li F. Fully reference-independent LMP decomposition using reference-independent
loss factors. Electr Power Syst Res 2011;81(11):1995–2004.
[23] Cheng X, Overbye TJ. An energy reference bus independent LMP decomposition
algorithm. IEEE Trans Power Syst 2006;21(3):1041–9.
[24] Li F, Bo R. DCOPF-based LMP simulation: algorithm, comparison with ACOPF, and
sensitivity, vol. 22; 2007. p. 1475–85.
[25] Xie K, Song Y-H, Stonham J. Decomposition model and interior point methods for
optimal spot pricing of electricity in deregulation environments. IEEE Trans Power
Syst 2000;15(1):39–50.
[26] Wang Q, Zhang G, McCalley JD, Zheng T, Litvinov E. Risk-based locational mar-
ginal pricing and congestion management. IEEE Trans Power Syst
2014;29(5):2518–28.
[27] Hajiabadi ME, Mashhadi HR. LMP decomposition: a novel approach for structural
market power monitoring. Electr Power Syst Res 2013;99:30–7.
[28] Rios-Zalapa R, Ruiz Casillas ME, Wan J, Cheung K. A dual-space approach for
pricing congestion in electricity markets. In: Energy market (EEM), 2010 7th in-
ternational conference on the European; 2010.
[29] Cheverez-gonzalez D, DeMarco C. Mutually orthogonal LMP decompositions: con-
gestion decomposes, losses do not. 2009 6th international conference on the
European energy market. IEEE; 2009. p. 1–6.
[30] Cheverez-gonzalez D, Lin J. Mutually orthogonal LMP decomposition: analysis of
PJM network by null space approach. IEEE PES general meeting. IEEE; 2010. p. 1–8.
[31] Sarkar V, Khaparde SA. Optimal LMP decomposition for the ACOPF calculation.
IEEE Trans Power Syst 2011;26(3):1714–23.
[32] Naren Bharatwaj V, Abhyankar AR. Max-min fair Financial Transmission Rights
payment-based AC optimal power flow locational marginal price decomposition.
IET Gen, Transm, Distrib 2014;8(10):1724–32.
[33] Vaskovskaya T, Guha Thakurta P, Bialek J. Identifying congestion zones with
weighted decomposition of locational marginal prices. In: PowerTech 2017,
Manchester, U.K.; 2017.
[34] Zimmerman RD, Murillo-Sanchez CE, Thomas RJ. MATPOWER: steady-state op-
erations, planning, and analysis tools for power systems research and education.
IEEE Trans Power Syst 2011;26(1):12–9.
[35] Li F, Bo R. Small test systems for power system economic studies. In: IEEE PES
general meeting; 2010. p. 1–4.
[36] Vaskovskaya T, Guha Thakurta P, Bialek J. 2746-node Polish energy system data of
transmission and voltage constraints contribution to the formation of LMP; 2017.
http://dx.doi.org/10.5281/zenodo.1001757.

499