Restructured Power System

UNIT-III
LOCATIONAL MARGINAL PRICES AND FINANCIAL TRANSMISION
RIGHTS

LOCATIONAL MARGINAL PRICES(LMP)

INTRODUCTION

The Locational Marginal Pricing (LMP) mechanism is one of the most commonly
employed tools for market settlement in the deregulated power system environment. The
Locational Marginal Price (LMP) at a bus signifies the cost of supplying the next increment
of load at that bus. The LMP is the sum of supplying energy marginal cost, cost of losses
due to the increment and transmission congestion cost, if any, arising from the increment.
and congestion, if any, arising from that increment. The LMP is the true indicator of
marginal pricing of energy. The calculation of LMPs implicitly involves congestion
management. Compared to other approaches of congestion management, the LMP approach
has found very wide acceptance throughout the world due to its inherent efficiency in the
network capacity allocation.

Many of the successfully running power markets like PJM, NYISO, ISO-NE, CAISO,
ERCOT, MISO and NEMCO have already implemented LMP mechanism in their systems,
whereas other markets are now evolving towards locational marginal pricing. The LMP
mechanism was first invented by Dr. William Hogan in 1992 [1], and introduced at
Pennsylvania-New Jersey-Maryland (PJM) ISO. However, the basis of the LMP mechanism
is the theory of spot pricing proposed by Schweppe et al.[2]. The distinguished feature of the
LMP mechanism is that the entire course of power scheduling (pool as well as bilateral
transactions) is done centrally, recognizing system conditions and constraints arising thereof.
The underlying principle of locational marginal pricing is that the energy price varies from
one location to another location in the presence of congestion and loss in the system.

FUNDAMENTALS OF LOCATIONAL MARGINAL PRICING

Locational marginal pricing is a centralized process of market clearing, where it is the

responsibility of the Independent System Operator (ISO) to determine the power dispatch
schedules as well as the energy prices. Unlike system uniform pricing (i.e., unconstrained
bidding) approach, network limits have to be considered while scheduling generators, loads
and bilateral transactions. Since network constraints are considered in the market clearing
process, it is not possible to determine the market equilibrium simply by the intersection of a
cumulative supply curve and a cumulative demand curve. Instead, the power dispatch
schedules and energy prices are calculated through an optimization approach consisting of
network and power flow related constraints.

An LMP market may be a single settlement or two settlement market. In case of a single
settlement market, scheduling is done only in day-ahead, whereas both day-ahead and real-
time scheduling is done for a two settlement market. A real-time market is essentially within

E.Muthukumaran AP/EEE Page 1

 Restructured Power System

the hour market. The real-time scheduling and settlement are further done in different time
blocks. For example, in PJM, real-time scheduling is done in 5- minute time blocks. The
real-time scheduling starts with the state-estimation solution at the beginning of each time
block. The state-estimation solution gives the actual injection by each generator and actual
withdrawal by each load at the current point of time. It should be noted that the above time
frames of settlement are not strict and change depending upon the market needs.

Apart from the offers from generators and bids from the loads, bids are also invited
from the bilateral transactions under locational marginal pricing. These bids are called up to
network usage charge bids. An up to network usage charge bid indicates the price that the
concerned entity is ready to pay at maximum for the scheduled quantity of its transaction.
Similar to a load bid curve, a transaction bid curve is also downward sloping. Apart from
bids and offers, self-scheduling is also allowed in some markets like PJM. Self-scheduling
means showing price-unresponsiveness or price-insensitivity. A self-scheduled generation,
load or transaction is alternatively called as an inelastic generation, load or transaction,
respectively. The power dispatch schedules are obtained by maximizing social welfare
function within the network limits. The energy price at a location is defined as the rate
decrease of optimal social welfare with respect to the fixed load increase at that location.
These prices are the locational marginal prices. Generators are paid and loads have to pay
the LMPs at the corresponding locations. For a bilateral transaction, the market player has to
pay a network usage charge depending upon the LMP difference between its sink and source
locations.

The bid and offer curves submitted by market participants may be piecewise or purely
linear curves. In case all the bid and offer curves are purely linear, the social welfare
function W(P) can be formulated as follows:

...........................................................................(1)
where,
......................................................................................(2)

ηd......................Number of load bids

ηg......................Number of generation offers
η.b.....................Number of transaction bids
Pd......................Column vector containing load bid variables
Pg......................Column vector containing generation offer variables
Pb......................Column vector containing transaction bid variables
........... Bid price as a function of quantity for the kth load bid
........... Offer price as a function of quantity for the kth generation offer
............ Bid price as a function of quantity for the kth transaction bid

For zero slope bid and offer curves, the expression of welfare function is simply

E.Muthukumaran AP/EEE Page 2

 Restructured Power System

reduced to the following,

......................(3)

In the absence of price-sensitive loads and transaction, the objective function can be
alternatively defined as minimization of generation cost C(Pg), where,

..............................................................................................(4)

As before, for zero slope offer curves, the expression of generation cost is reduced to
the following:

............................................................................................................(5)

It should be noted that when the objective function of the market clearing problem is
defined as minimization of generation cost, the definition of LMP is also changed as the rate
of increase of optimal generation cost with respect to the fixed load increase at a location.

Now, consider the case that the generators, loads and transactions submit piecewise
linear offer and bid curves, respectively. Under such a situation, the social welfare function
is to be formed by considering each linear segment as a separate offer or bid. For example,
consider the generation offer curve shown in Fig. 1. This offer curve is equivalent to the set
of three individual offers as following:

Figure 1: Generation offer curve

E.Muthukumaran AP/EEE Page 3

 Restructured Power System

Offer 1: Offered amount = 70 MW,

Offer 2: Offered amount = 70 MW,
Offer 3: Offered amount = 50 MW,

It should be noted that from the point of view of scheduling, the social welfare function
is nothing more than a simple mathematical function that has been devised to reach efficient
market equilibrium. Market equilibrium is said to be efficient if the following conditions
hold.

 Each selected generation is paid at least its offer price.

 The offer price associated with an unselected generation is higher than the
energy price at corresponding location.
 Each selected load or transaction has to pay at most its bid price.
 The bid price associated with an unselected load is lower than the energy
price at corresponding location.
 The bid price associated with an unselected transaction is lower than the LMP
difference between the corresponding locations.
 It is not possible to increase the output of a generator at a certain location,
and simultaneously increase the consumptions at some other location/s without
causing fall of at least one load node energy price below the gener¬ation node energy
price or creating network infeasibility, and vice versa.
 It is not possible to increase the quantity of a transaction without causing
network infeasibility or making its network usage charge negative.

ILLUSTRATIVE EXAMPLE

First five of the above conditions are explained with an example. Consider that the offer
price associated with a generation offer is $20/MWh and the offer gets fully selected in the
energy market. In case the market equilibrium is efficient, the LMP at this generator’s
location can never be less than $20/MWh. Similarly, let the offer price associated with a un-
selected generation- offer be $30/MWh. Under efficient market equilibrium, the LMP at the
corresponding location must be less than $30/MWh. The opposite conditions hold for a load
or transaction bid.

For the illustration of sixth and seventh conditions for efficient market equilibrium,
consider the sample system shown in Fig. 2. Fig. 3 shows one possible equilibrium state
satisfying the other necessary efficiency conditions. It can be seen that it is still possible to
increase the output of the generator at Node 1 and the consumption at Node 2 with neither
causing network infeasibility nor making the price at Node 2 to fall below the price at Node
1. Therefore, this market equilibrium is inefficient. The efficient equilibrium state is shown
in Fig. 4. The condition on bilateral transaction scheduling can be illustrated in a similar
way.

E.Muthukumaran AP/EEE Page 4

 Restructured Power System

Figure 2: Sample system

Figure 3: Inefficient market equilibrium

Figure 4: Efficient market equilibrium

A generation offer or load-bid may be real or virtual. A real load bid or generation offer
is always backed up by a real load or generation asset, respectively, whereas there is no
supporting physical load or generation asset, respectively, for a virtual bid or offer. For
E.Muthukumaran AP/EEE Page 5
 Restructured Power System

example, an entity can submit a generation offer at a location where it does not have any
generating asset. Virtual bids and offers are allowed only under two-settlement market
structure. Any virtual generation or load accepted in a day-ahead market must be counter-
balanced by a same sized load or generation, respectively, in the real time market. The
motivations behind encouraging virtual bids and offers are to quickly demolish the
difference between day-ahead and real-time prices, and to increase the market
competitiveness.

ZONAL AND NODAL LMP

Locational marginal pricing is done either on nodal basis or zonal basis [3]. For a nodal
market, a location is defined by a node or a load zone or a generation hub and the impact of
each nodal injection on line flows is individually considered in the power dispatch problem.
The dispatch solution has to respect the flow limit of each line. As far as the definitions of
load zone and generation hub are concerned, they are fundamentally similar. A load zone or
a generation hub is basically defined by a set of nodes and a load or generation distribution
vector, respectively. In order to calculate the line flows, the total load on a load zone is
distributed over its constituent nodes by means of the distribution vector assigned to it. The
total generation over a hub is distributed in the same fashion. However, for a load zone, the
total load on it is also same as the sum of the total load on each of its constituent nodes. This
is not true for a generation hub. Additionally, the scheduling and metering for a load zone is
always done in the aggregation level by assuming that the total load on this load zone
naturally remains distributed in a fixed ratio. In case of a generation hub, the scheduling and
metering, in the end, is always done at the nodal level. It is the responsibility of the entity,
requesting a transaction from this hub, to maintain the injection pattern specified by the
generation distribution vector. Each individual load within a load zone has to pay the load
zone price only, whereas each individual generation within a hub is paid the LMP of the
corresponding node. The objectives behind defining load zones are to reduce the spatial
price variation and to ease the financial settlement. On the contrary, the objective behind
defining hubs is to maintain consistency with the previous (i.e., before nodal pricing)
bilateral contractual agreements.

In case of a zonal market, the system is divided into multiple transmission zones and
each of these transmission zones defines a particular location. A transmission zone is
basically a set of nodes and lines that is modeled like a single node in the dispatch problem.
Only the inter-zonal line flow limits are considered during the power scheduling. The inter-
zonal line flows are determined by the zonal sensitivity factors. The main technical
challenge with zonal pricing is to appropriately divide the system into transmission zones so
that zonal sensitivity factors may exist. For this, the branch-node sensitivity factors relating
any inter-zonal line should be almost same for all the nodes constituting a transmission zone.

ADDITIONAL ASPECTS OF LMP MECHANISM

The advantage of locational marginal pricing is multi kinds. First, this kind of energy
pricing can generate useful price signals to identify suitable locations for building new
generators, load centers and transmission facilities. For example, if the LMP at a certain

E.Muthukumaran AP/EEE Page 6

 Restructured Power System

location continuously shows a higher value, it gives an initial indication that this location
may be a suitable place for building new generators. At the same time, penalizing each
bilateral contract with a location dependent network usage charge applies a strong financial
force on it, making it implicitly care for the congestion in the transmission network. The
pool prices also recognize the system limits. Unlike system pricing approach, the possibility
of administrative transmission load relief (TLR) thus becomes very small. The need for TLR
arises only during very rare occasions when the self-scheduled generations, loads and
transactions load the system beyond its capacity such that no solution of the dispatch
problem exists. Moreover, due to a joint optimization based dispatch calculation,
transmission capacities are allocated in optimal way to the energy market participants as per
their needs. The less the bid price of a generation bid the more the amount of it would get
selected. Similarly, the more the bid price of a load or transaction bid the more the amount
of it would get selected. The capacity allocation occurs in optimal way due to measuring the
relative needs of all the market part at the same time by a single and non-profit making
entity. However, this feature is not available in the physical transmission right methodology.

LMP FORMULATION AND IMPLEMENTATION

Calculation of LMP in a market environment involves solving an optimization problem.

LMP is the by-product of the solution to this optimization problem. In this section, we shall
discuss various models existing in literature to calculate LMP.

The LMP models existing in literature can be broadly categorized into two classes,
namely, the ACOPF model and the DCOPF model. The ACOPF model uses the AC
formulation to represent power flow and thus can accurately represent the line flows and
loss. On the contrary, the DCOPF models are based upon DC power flow approximation
and, therefore, are less accurate. The DCOPF model can be lossless or loss compensated. In
case of a lossless model, the network loss is ignored at the time of LMP calculation. Even
though ACOPF provides the most accurate LMP, it is quite complex to model its constraints,
and cannot be used for real time applications. Sometimes, obtaining converged solution for
very large systems becomes difficult. However, DCOPF is not plagued by the convergence
problem due to its simplicity. Solution of DCOPF can be obtained with a single run of
Linear Programming (LP) because of the linearity assumption. However, due to these
assumptions, the LMP obtained is not very accurate. The choice between ACOPF and
DCOPF is a trade-off between accuracy and speed of convergence. Some advantages of
DCOPF over ACOPF are as follows:
1. Simplicity
2. Speed of convergence
3. Availability of LMP components

Among the two major classes of LMP models, the ACOPF model is prone to lack
financial consistency in general. On the contrary, financial consistency can be obtained in a
DCOPF model. Moreover, without significant loss of accuracy, the DCOPF models are
much simpler that the ACOPF models. As a result most of the LMP markets employ
DCOPF models for the purpose of market clearing. Currently, NYISO is the only market
that employs an ACOPF model for calculating locational marginal prices.

E.Muthukumaran AP/EEE Page 7

 Restructured Power System

ACOPF

The ACOPF model involves solving a linear/non-linear objective function with linear
and non-linear equality and in-equality constraints. The objective function can be cost
minimization or social welfare maximization. Unlike DCOPF, it is required to solve the non-
linear power flow equations, and hence a non-linear optimization solver is required to obtain
a solution. For large systems, the dimension is quite high and it is generally performed as an
offline study. The mathematical formulation is as follows:

...........................................................................................(6)
.....................................................................(7)
....................................................................(8)
..............................................................(9)
..........................................................(10)
...........................................................(11)
..................................................................(12)

(7) and (8) are the real and reactive power balance equations for all buses in the system.
(9) models the line flow limit in-equality constraints. (10) and (11) enforce generations
limits on real and reactive power for all generators. Voltage limits are enforced using (12).
Additional constraints pertaining to taps and shunts may also be added to the optimization
problem. For a practical system, the problem size is quite large and hence obtaining a
converged solution sometimes becomes difficult.

The LMP at a bus is the lagrangian multiplier corresponding to the equality constraint.
Unlike in DCOPF, where the LMP components are readily available, additional de-
composition techniques are required to obtain LMP components [4-11]. Moreover, there
exists no unique solution to this decomposition. Perhaps, this is one the driving factors for
DCOPF to be preferred over ACOPF.

DCOPF

Nomenclature

The following is the list of symbols used in the DCOPF models to follow.

n.............................................Number of buses/nodes
M.............................................Number of Transmission Lines
Ci(Pgi).....................................Cost curve of generator located at bus i
ηg............................................Number of Generator offers
LF............................................Column vector containing loss factors
LF.............................................Matrix containing line-bus loss factors

E.Muthukumaran AP/EEE Page 8

 Restructured Power System

λ..............................................Lagrangian multiplier of equality constraint

μ..........................Column vector containing Lagrangian multiplier of in-equality constraints
SF............................................Sensitivity matrix relating Power injections and line flows
L..............................................Lagrangian function
Ploss.........................................Total system power loss
...................................... Power loss occurring on line k
..................................... Column vector containing power loss occurring on all lines
max
flow ...................................Column vector containing line power flow limits
Pgi.......................................... Power generation at node i
Pdi..........................................Power demand at node i
Pg............................................Column vector containing Power generation at all buses
Pd ...........................................Column vector containing Power demand at all buses
....................................Lower limit on Pgi
o..................................Upper limit on Pgi
D.............................................Loss distribution vector
K.............................................Loss distribution matrix
FND........................................Column vector of fictitious nodal demand
offset......................................Offset term required to compensate for over-estimation of
losses
offsetL............................... ......Column vector containing offset for all lines
1x ..........................................x * 1 vector of all ones
Rk .........................................Resistance of line k
Xk .........................................Reactance of line k
flowk ....................................Power flow on line k

Lossless Model

This is the most simplistic form of DCOPF, where the transmission line losses are
completely neglected. The solution of lossless DCOPF is the starting point for some of the
approaches that incorporate losses since it provides a good initial estimate. The formulation
of lossless DCOPF is given below.

..........................................................................................(13)

.........................................................................................(14)
.......................................................(15)
.......................................................................................(16)

Equation (13) is the objective function. (14) is the global power balance equation. (15)
ensures that the transmission line flows don't exceed the limits. (16) is the constraint on
individual generator limits.

E.Muthukumaran AP/EEE Page 9

 Restructured Power System

The Lagrangian for the above optimization problem can be written as follows.

.
...............................................................(17)

For the sake of simplicity, the flow constraints are considered for only one direction of
power flow. To obtain the LMP at any bus i, the partial derivative of the Lagrangian w.r.t.
the demand at that bus i is evaluated, given as follows:

...................................................................(18)

As seen from equation (18), the LMP obtained from lossless DCOPF consists of two
components. The first term (?) is referred to as the energy component (?_e), since it is the
Lagrangian multiplier of the global power balance equation. The second term is known as
the congestion component (?_c) as it is a function of the Lagrangian multiplier of the in-
equality constraint that corresponds to line power flows. It must be noted here that in the
absence of congestion (µ=0), the LMP at all buses will turn out to be same. This holds true
only in case of DCOPF. When we use ACOPF, the LMP at all buses will be different
because of the presence of transmission losses, even in the absence of congestion.
The features of LMP are better understood using an example, which is discussed next.

Example 1: A sample 5 bus system given in Figure 5. The system has five
dispatchable generators. The line data for the system is given in Table 1. The marginal
costs of all generators are given in Table 2. The loads are inelastic and are 300 MW
each.

Line Data A-B A-D A-E B-C C-D D-E

R (%) 0.281 0.304 0.064 0.108 0.297 0.297

X (%) 2.81 3.04 0.64 1.08 2.97 2.97

Limit
999 999 999 999 999 240
(MW)

Table 1: Line data of 5 bus system

E.Muthukumaran AP/EEE Page 10

 Restructured Power System

Figure 5: Sample 5 bus system

Generator Marginal Cost ($/MW)

G1 14

G2 15

G3 30

G4 35

G5 10

Table 2: Marginal costs of generators

The results for the lossless case, when bus 4 is taken as slack, are given in Tables 3-5.

Generator Dispatch(MW)

G1 110

G2 100

G3 0

G4 116.0757

G5 10 573.9243

E.Muthukumaran AP/EEE Page 11

 Restructured Power System

Table 3: Dispatch for lossless case

Line Power flow(MW) μ

A-B 379.7505 0

A-D 164.1738 0

A-E -333.9243 0

B-C 79.7505 0

C-D -220.2495 0

D-E -240 -52.0344

Table 4: Power flow for lossless case

LMP Components ($/MW)

Bus LMP ($/MW)
Energy Congestion

A 35 -19.1744 15.8256

B 35 -11.3202 23.6798

C 35 -8.3015 26.6985

D 35 0 35

E 35 -25 10

Table 5: LMP at various buses

The following observations about LMP can be made from the results:

1. The energy component will always be equal to the LMP at the reference/slack
bus.
2. The energy component at the reference/slack bus also consists of its
congestion component and loss component (if any). The energy component has been
named so just for representation and does not indicate the cost of
supplying only energy.
3. Even if the reference bus is changed, the energy and congestion components
will change, but the total LMP will remain the same.
4. In the absence of congestion, LMP at all buses will be same, and equal to that
of the marginal generator.

E.Muthukumaran AP/EEE Page 12

 Restructured Power System

5. The availability of LMP components is an inherent feature of DCOPF. D e-

composition of LMP into its components is required for the settlement of risk-
hedging tools like financial transmission right (FTRs) [12]. FTRs are basically risk
hedging instruments whose objective is to guard forward contracts from the uncertain
congestion charges.

The lossless model always yields a reference node independent solution. This is one the
desired features of LMP calculation because there shouldn't be inconsistency in financial
settlements. Financial consistency mainly implies non-negativity of net congestion
collection or adequacy of net congestion collection to make full payments towards a set of
simultaneously feasible FTRs; whichever is of interest. Some desirable features of LMP
calculation are the financial consistency, refer¬ence node independency, accurate modeling
of power flow, and the model simplic¬ity. The reference node independency of LMP
calculation is desired in order to obtain an undisputed solution of each quantity of interest.
Both, reference node dependent and independent loss-compensated models are available in
the literature for DCOPF formulations.

Even though the lossless DCOPF model is simple to formulate, it has certain
disadvantages. For large systems, the losses are not negligible and hence can't be neglected
in the dispatch because the mismatch will be quite large. This has forced researchers to
improve upon the lossless DCOPF model such that losses are considered in the dispatch.

Model 1

In this model, the losses are estimated from pre-calculated loss factors. The lossless
DCOPF model is a good starting point to evaluate loss factors. These loss factors can be
evaluated using AC power flow (ACPF) or DC approximation. The calculation of these loss
factors is dealt with in the next section. The mathematical modelling of this model is as
given below.

.........................................................................................(19)

...................................................................................(20)

..........................................................(21)
-flowmax≤SF x [Pg - Pd] ≤flowmax.......................................................................(22)
Pgmin≤ Pgi ≤ Pgmax...............................................................................................(23)

The difference between this model and the earlier model is in considering the system
losses as part of the dispatch. Hence, the total dispatch will account for the losses occurring
in the system as well. However, the loss distribution is not specified in this model. If we

E.Muthukumaran AP/EEE Page 13

 Restructured Power System

look at equation (22) closely, there is no mention of loss being compensated at individual
bus level. This signifies that all losses are being consumed at the slack/reference node itself,
instead of being distributed across the network . Even though the dispatch accounts for
transmission losses, there is no effect on line flows. This is a major drawback of this model.
Further, the solution of the model changes if the slack bus is changed. Hence, in 2004,
Eugene Litvinov et.al. proposed a vector loss distribution (VLD) model that accounts for
losses at individual bus level, which we will discuss next

Model 2 [12]

In this model, the losses are distributed as additional loads to buses using a loss
distribution vector (D). The problem formulation for this model is as follows.

..........................................................................................(24)

such that .......................................................................(25)

.............................................................(26)
-flowmax ≤ SF x [Pg - Pd -D x Ploss] ≤ flowmax.......................................................(27)

Pgimin≤ Pgi ≤ Pgimax. .............................................................................................(28)

The loss factors in this model are calculated on the lossless snapshot using the fast-
decoupled approach (a variant of calculating loss factors using ACPF). The difference
between this model and Model 1 can be seen when we compare equation (27) and (22). The
D vector assigns the total system losses as additional load to various buses. This affects the
line power flow, which is desired because losses occur on transmission lines and are spread
over the transmission network. The Lagrangian for this model is as follows.

. .............................................(29)

Because of the introduction of the additional constraint that considers losses, the
definition of LMP in this model is different. Apart from the energy and congestion
component, an additional component called loss component (?1) is introduced. The LMP at
any bus i can be defined as follows:

.......................................................(30)

E.Muthukumaran AP/EEE Page 14

 Restructured Power System

Ploss is a control variable in the optimization problem. Hence, the derivative of the
Lagrangian function is necessary to obtain the three components of LMP.

. ...................................................................................(31)

From the above two equations, LMP can be defined as:

.....................................(32)

Thus, the LMP obtained from DCOPF inherently consists of the individual components.
As such, no separate technique is required to obtain these components. Hence, LMP can be
written as a sum of the three components as follows:
. .........................................................................................(33)
. ................................................................................................................(34)
. ....................................................................................................(35)

. ...............................................................(36)

Even with the introduction of the loss distribution vector, this model has certain
drawbacks.

1. The choice of D is an open research problem since there can be many

solutions. The authors overcome this drawback by choosing D proportional to the
demand at a bus.
2. The slack dependency of Model 1 is converted to Dvector dependency of this
model. If both are chosen to be same, both models produce identical results.

With the inclusion of network loss in the LMP calculation, an additional term called the
loss component is added to the LMP. The LMP decomposition should be made in such a
way that the loss component vanishes when the system is lossless and that the congestion
component vanishes when the sys¬tem is uncongested. The implication of reference node
independency for a lossy model of LMP calculation is threefold:

1. The power dispatch solution should be independent of the choice of reference

node.
2. The locational marginal prices should be independent of the choice of refer-
ence node.
3. The congestion components of locational marginal prices should be indepen-
dent of the choice of reference node.

E.Muthukumaran AP/EEE Page 15

 Restructured Power System

Model 3 [16]

In 2007, Fangxing Li proposed an iterative technique that uses DC loss factors to obtain
LMP. The mathematical formulation of this model is given below.

. ....................................................................................(37)
such that :

. ............................................................................(38)

. .....................................................(39)
-flowmax ≤ SF x [Pg - Pd -FND] ≤ flowma.........................................................(40)
Pgimin≤ Pgi ≤ Pgimax. ........................................................................................(41)

This model introduces the Fictitious Nodal Demand (FND) vector that distributes losses
to different buses. The losses occurring on a line are halved and added as additional loads to
the buses connecting the line. This is a reasonable assumption and provides good accuracy in
terms of power flow. The expression for FND for a bus i is given in (42).

.........................................................................(42)

Where, l is the index of lines connected to bus i. Since this is an iterative technique, the
procedure of calculation of LMP is understood easily using the flowchart given in Figure 6.
The loss, offset and FND are estimated from previous iteration, and passed as parameter to
the next iteration. Since this is an iterative technique, these values will converge and the
difference between two iterations will be less than the tolerance. This is different from the
formulation of

Model 2. The Lagrangian function and LMP calculation is done as follows.

.....................................................(43)
The LMP is given by:

E.Muthukumaran AP/EEE Page 16

 Restructured Power System

...................................................(44)

Figure 6: Flow chart of iterative technique proposed by Fanxing Li

Model 4 [17]

Sarkar et.al proposed a Matrix Loss distribution (MLD) method to distribute losses to
various nodes. This method provides additional design flexibility in distributing losses to
more than two nodes and in different proportions. Its mathematical modelling is as given
below.

.........................................................................................(45)

..............................................................................(46)
.................................................................(47)
..................................(48)
Pgimin≤ Pgi ≤ Pgimax...............................................................................................(49)

E.Muthukumaran AP/EEE Page 17

 Restructured Power System

The loss occurring on each line is a control variable in this model. Hence, the partial
derivative of Lagrangian function w.r.t these controls will exists and appear in the
expression for LMP. This is similar to the calculation of LMP discussed in Model 2.
Model 5 [18]

In 2010, a truly reference independent iterative model was proposed by Furong Li et.al.
This model obtained loss factors from the converged ACPF solution by making use of
Newton Raphson Jacobian sensitivities. The difference between this model and Model 4 is
in the calculation of loss factors. The solution of LP is passed on to NRLF to obtain the loss
factors, offset and D. These are then passed as parameters to the next iteration. The
mathematical model is given below.

...........................................................................................(50)

......................................................................................(51)

................................................................(52)

-flowmax ≤ SF x [Pg - Pd -D x Ploss]≤ flowmax.........................................................(53)

Pgimin≤ Pgi ≤ Pgimax..................................................................................................(54)

In (53), D is calculated using (55).

................................................................................................................(55)
The LMP decomposition of this model is similar to the one discussed in Model 2

Loss Factors

DC loss factors

In some of the methods discussed earlier, loss factors are estimated using DC
approximation. Since voltages are assumed to be 1 p.u., the losses can approximated as:

.........................................................................................(56)

The loss factor (LF) at a bus i is defined as the partial derivate of the above expression
w.r.t. power injection at bus i(Ploss). Hence,

E.Muthukumaran AP/EEE Page 18

 Restructured Power System

...........................................................................(57)
Transmission line flow can be written as:

................................................................................(58)

LF can be further expanded as:

......................................(59)

(59) is the expression for loss factor using DC approximation. The offset is required
since product of LF and Pi over estimates the losses. In [16], it has been proved that this
product produces double the system losses. And hence, the offset needs to be negation of
(53), which is the actual loss occurring in the system.

AC Loss Factors

The loss factors using ACPF makes use of the NRLF Jacobian to obtain sensitivities.
Additional sensitivity of Ploss with load angle (d) and voltage magnitude (V) needs to be
obtained.

.....................................................................(60)

Where, S is the inverse of NRLF jacobian (J). From the above equation, we can write
LF as:

............................................................................(61)

The offset in this case is calculated as follows.

...........................................................................(62)

E.Muthukumaran AP/EEE Page 19

 Restructured Power System

SUMMARY

Use of LMP for energy pricing can generate useful price signals to identify suitable
locations for building new generators, load centers and transmission facilities. The LMP
mechanism promotes the optimal and efficient use of system resources like generation and
transmission. However, even though the LMP mechanism is economically very efficient,
caution must be exercised before its implementation. The LMP is very volatile as real time
approaches and safeguard mechanisms like FTRs are required to hedge against this
uncertainty. A certain level of maturity on the part of the market and its participants is
required for these mechanisms to be effectively implemented and remain in sync. Not all
market participants may understand the intricacies involved in the LMP mechanism. It is a
crucial link to various other mechanisms involved in a Centralized dispatch market
environment.

REFERENCES

1. W. W. Hogan, Contract networks for electric power transmission, Journal of Regulatory

Economics , vol. 4, pp. 211-242, 1992.
2. F. C. Scheweppe, M. C. Caramanis, R. D. Tabors, and R. E. Bohn, Spot pricing of
electricity, Norwell, MA: Kluwer, 1988.
3. J. E. Price, Market-based price differentials in zonal and LMP market designs, IEEE
Transactions on Power Systems , vol. 22, no. 4, pp. 1486-1494, November 2007.
4. L. Chen, H. Suzuki, T. Wachi, and Y. Shimura, Components of nodal prices for electric
power systems, IEEE Transactions on Power Systems , vol. 17, no. 1, pp. 41-49, February
2002.
5. J. D. Finney, H. A. Othman, and W. L. Rutz, Evaluating transmission congestion
constraints in system planning, IEEE Transactions on Power Systems , vol. 12, no. 3, pp.
1143-1150, August 1997.
6. T. Wu, Z. Alaywan, and A. D. Papalexopoulos, Locational marginal price calculations
using the distributed-slack power-flow formulation, IEEE Transactions on Power Systems ,
vol. 20, no. 2, pp. 1188-1190, May 2005.
7. X. Cheng and T. J. Overbye, An energy reference bus independent LMP decomposition
algorithm, IEEE Transactions on Power Systems , vol. 21, no. 3, pp. 1041-1049, August
2006.
8. T. Orfanogianni and G. Gross, A general formulation for LMP evaluation, IEEE
Transactions on Power Systems , vol. 22, no. 3, pp. 1163-1173, August 2007
9. L. Wang and M. Mazumdar, Using a system model to decompose the effects of
influential factors on locational marginal prices, IEEE Transactions on Power Systems , vol.
22, no. 4, pp. 1456-1465, November 2007.
10. E. Litvinov, T. Zheng, G. Rosenwald, and P. Shamsollahi, Marginal loss modeling in
LMP calculation, IEEE Transactions on Power Systems , vol. 19, no. 2, pp. 880-888, May

E.Muthukumaran AP/EEE Page 20

 Restructured Power System

2004.
11. Y. Fu and Z. Li, Different models and properties on LMP calculations, IEEE/Power
Eng. Soc. General Meeting , June 18-22, 2006.
12. E. Litvinov, LMP implementation in New England, in IEEE/Power Eng. Soc. General
Meeting, June 18-22, 2006.
13. Z. Li and H. Daneshi, Some observations on market clearing price and locational
marginal price, IEEE/Power Eng. Soc. General Meeting , June 12-16, 2005.
14. B. C. Lesieutre and I. A. Hiskens, Convexity of the set of feasible injections and
revenue adequacy in FTR markets , IEEE Transactions on Power Systems , vol. 20, no. 4,
pp. 1790-1798, November 2005.
15. G. B. Shrestha and P. A. J. Fonseka, Congestion-driven transmission expansion in
competitive power markets, IEEE Transactions on Power Systems , vol. 19, no. 3, pp. 1658-
1665, August 2004.
16. F. Li and R. Bo, DCOPF-based LMP simulation: Algorithm, comparison with ACOPF,
and sensitivity, IEEE Transactions on Power Systems , vol. 22, no. 4, pp. 1475 –1485, Nov.
2007.
17. V. Sarkar and S. Khaparde, DCOPF-based marginal loss pricing with enhanced power
flow accuracy by using matrix loss distribution, IEEE Transactions on Power Systems , vol.
24, no. 3, pp. 1435 –1445, Aug. 2009.
18. Z. Hu, H. Cheng, Z. Yan, and F. Li, “An iterative LMP calculation method considering
loss distributions,” IEEE Trans. Power Syst., vol. 25, no. 3, pp. 1469 –1477, Aug. 2010.
19. A. J. Wood and B. F. Wollenberg, Power Generation, Operation and Control. John
Wiley & Sons, 1996.

E.Muthukumaran AP/EEE Page 21

 Restructured Power System

FINANCIAL TRANSMISSION RIGHTS FTR

Stochastic locational prices create a demand by risk-averse market players for locational
price hedging instruments. One of such instruments is a Financial Transmission Right FTR. An
FTR gives the holder a right (or/and obligation) to a share of congestion rents received by the
System Operator during transmission congestion. The allocation of FTRs typically occurs as an
auction, where the benefit of the buyer or seller is maximized. The auction determines the
amount of FTRs allocated to market players and market clearing prices. The design of the
auction is decided by the System Operator and depends on the market structure. However, FTRs
may also be allocated to transmission service customers who pay the embedded costs of the
transmission system.
FTRs are typically longer term and may have duration from months to years. They can take
different forms such as point-to-point FTRs and flowgate FTRs, both of the obligation and
option type.
Point-to-point Financial Transmission Right
Point-to-point FTRs entitle (or obligate) the holder to the difference in locational prices times the
contractual volume. The mathematical formulation for the payoff is:
FTR = QPij(Pj -Pi)

where:

Pj is the bus price at location j,

Piis the bus price at location i

QPij is the quantity of active power scheduled on the path from ito j.

An FTR obligation may be viewed as an injection of quantity of active power QPijat bus i and a
withdrawal of the same QPij at bus j. If the contractual volume matches the actual traded volume
between two locations, an FTR is a perfect hedge against volatile locational prices.

The difference between the congestion rent and payments to FTR holders may be positive,
resulting in a surplus to the SO. The surplus is redistributed to FTR holders and transmission
service customers. On the contrary, if payments to FTR holders exceed the congestion rent, the
SO proportionally reduces payments to FTR holders or requires that the transmission owners
make up the deficit.
Point-to-point FTRs obligations are considered to be the most feasible hedging instrument in
practice. However, for point-to-point FTRs options the computational demands are more
substantial. Nonetheless, it is still technically possible, as they have been introduced in PJM in
2003.

E.Muthukumaran AP/EEE Page 22

 Restructured Power System

FLOWGATE FINANCIAL TRANSMISSION RIGHTS

Flowgate Financial Transmission Rights are constraint-by-constraint hedges that give the
right to collect payments based on the shadow price associated with a particular transmission
constraint called a flowgates. Flowgates FTR, often abbreviated as FGR, is based on a
decentralized market design. This approach results from a belief that the point-to-point FTR
markets may work inefficiently in practice and therefore, are not able to provide effective
hedging. The idea behind flowgates is that since electricity flows along many parallel paths, it
may be natural to associate the congestion payments with the actual electricity flows. Key
assumptions include a power system with relatively few, known in advance, commercially
significant flowgates resulting in a limited set of FTGs, as well as relatively stable power transfer
distribution factors PTDFs that decompose a transaction into the flows over these flowgates.

The payoff from the FGRs can be negative, zero or positive and is determined by taking
the associated flowgate shadow price times the amount of FTG, totaling them for all lines k
affected by the transaction between buses m and n (9). ηkrepresents the shadow price associated
with flowgatek, and Qk is the contracted volume of FGR.

FGRη−=Σηk.PTDFk,m-nQ(k)

In the AC transmission network, the power flows are determined by the line impedances,
and therefore cannot be effectively controlled. A given transaction may thus affect more than one
flowgate (transmission constraint). Parties that want to be fully hedged against congestion should
purchase a mix of flowgate FGRs that matches the distribution of flows from its transaction.

In practice, the assumptions of a limited set of commercially significant flowgates and

relatively unchanging PTDFs might be difficult to ensure in a dynamic power system, where
unanticipated transmission constraints may become binding. Furthermore, the flowgate approach
assumes that all constraints that are binding in the dispatch have been previously designated as
flow gates, and therefore the SO have made FGRs available for them. If some constraints are
designated, but not become binding, there is no mechanism that could allow market parties to
purchase a perfect hedge. Not charging for the non-predicted constraints and socializing costs
instead can however be a solution here.

Hedging
Hedge is a transaction that may consist of cash instruments or derivatives and is used to
offset risks associated with certain positions by establishing opposing positions. A hedge could
offset an exposure to changes in financial prices of other contracts or business risks. In general,
hedge is defined as a position or combination of positions, which could reduce some types of
risk. Hedge often indicates partially offsetting a long position in one security with a short or
short equivalent position in a related security. An example of a hedge is the purchase or sale of

E.Muthukumaran AP/EEE Page 23

 Restructured Power System

futures contracts to reduce the interest rate risk by offsetting cash market transactions at a later
date.
Hedging is the process of reducing uncertainty of future price movements. Hedging is an
operation undertaken by a trader or dealer who wishes to protect an open position, especially a
sale or a purchase of a commodity currency, security, etc. that is likely to fluctuate in price over
the period that the position remains open. This process could be implemented by a purchase or a
sale of derivative securities, such as options, to reduce or neutralize all or part of the risk of
holding another security. For example, undertaking forward sales or purchases in thefutures
market, or taking out an option which would limit the exposure to price fluctuations is regarded
as hedging.
Hedger is a participant who would enter the market with the specific intent of protecting
an existing or anticipated physical market exposure from unexpected or adverse price
fluctuations. For example, a generation company may contract to sell a large quantity of energy
for delivery over the next six months. Electricity products would depend on raw materials such
as natural gas that could fluctuate in price. If the generation company would not have sufficient
raw material in stock, an open position could result. The generation company may hedge the
open position by purchasing the required natural gas on a futures contract, and if the company
has to be paid for the natural gas in a foreign currency, the company’s currency needs could be
hedged by other contracts.
Two-way hedge and one-way hedge are hedging contracts which are dependent on the
contract’s strike price and spot price. In a two-way hedge contract, buyer and seller would agree
on a payment from one to the other when the strike price is either higher or lower than the spot
price. For example, when a power generating participant would agree with a retailer for a two-
way hedge contract, the generating party would pay the retailer the difference between the spot
price and the strike price when the spot price would get to be higher than the strike price, and
there tailer would pay the generating party when the spot price drops below the strike price. In a
one-way hedge, only the generating party would pay the retailer when the spot price is higher
than the strike price, and nothing would be paid to generating parties when the opposite occurs.
In this type of contract, a retailer would pay an annual fee. The energy that is traded by means
other than hedging contracts would be through the spot price market where the spot price is
calculated by balancing supply with demand using a bidding system. Even though the volume of
trades in the spot price market is low compared to other types of markets, spot price is usually
used as an index to strike price in hedging contracts, and market participants would devote a
considerable attention to spot price for redirecting their movements.
Market participants can use insurance packages offered by insurance companies to hedge
risks such as hedging the expensive cost of replacement power when an outage takes place,
hedging counter party risks or hedging extreme price spikes due to unexpected weather
conditions or due to similar reasons. Weather-related derivatives are implemented in electric
industry restructuring as risk hedging tools to hedge risks associated with unpredictable weather
conditions. Weather derivatives are the only contracts that offer energy providers a procedure to

E.Muthukumaran AP/EEE Page 24

 Restructured Power System

hedge volumetric risk. As an example, a weather derivative can be used in the form of an options
contract with a strike point set at a specified number of degree-days (DDs) or set based on
maximum minimum temperatures.

E.Muthukumaran AP/EEE Page 25