Beruflich Dokumente
Kultur Dokumente
Screens
Ricardo J. Galarza
1
, Member, IEEE, Ibrahim Mqasqas
2
, and James C. David
3
Abstract--. The presence of congestion in power systems
creates new challenges for detecting local market power. Indexes
based on pure economics rarely apply under congestion. When
transmission constraints limit the transfer of energy into an area
(or load pocket), energy suppliers inside the load pocket face less
competition (and have a greater concentration) than they would
during unconstrained periods. Transmission constraints come
and go and with them come and go the instances where market
supply is concentrated in the hands of a few suppliers. This
dynamic environment calls for market power screens using tests
that can recognize and adapt to the current situation, whether
constrained by transmission or not. This paper explores and
proposes elements of design and methodologies to be used for
building local market power screens, considering that both
market conditions and transmission network conditions could
lead to market power. When appropriate, the paper describes
currently used tests, indicating pros and cons.
Index Terms Electricity markets, market power, congestion,
local market power screens.
I. INTRODUCTION
he goal of market power screens is twofold. First is the
use of clear and unambiguous metrics to detect situations of
market power, that is, where energy supply is concentrated in
the hands of very few suppliers. Second is the use of clear and
unambiguous metrics to detect the potential for or actual
exercise of market power. It is of great importance that such
metrics also differentiate between a scarcity condition and
market power.
The monitoring of supplier bids in both the day-ahead and
real-time markets has been well developed in NYISO as well
as other ISOs [6], [7], [10]. For example, mitigation of market
power in the NYISO relies on the examination of the behavior
(conduct) of resources by examining their offers (bids) to sell
energy in the day-ahead and real-time markets [3]. A
performance measure (clearing price impact test) is added as a
means to mitigate only when significant (subject to threshold)
impact is found.
1
Ricardo J. Galarza is with PSM Consulting, Guilderland, NY 12084, USA
(E-mail: rgalarza@psm-consulting.com).
2
Ibrahim Mqasqas (E-mail: imqasqas@nyiso.com) is with NYISO,
Schenectady NY 12303, USA
3
James C. David (E-mail: jdavid@nyiso.com) is with the NYISO,
Schenectady NY 12303, USA
However, the transmission system impact on market power
has not been as well developed. The issue at hand is actually -
much more complex than just monitoring static bids. In
reality, transmission system conditions leading to market
power is dynamic, being affected by numerous conditions
such as planned and forced outages, availability of suppliers,
load levels, and others. It is the presence of congestion which
typically causes inadequate supply competition and hence
presents conditions that could lead to potential market power.
As a result, the market can and should be monitored
dynamically; to test for uncompetitive outcomes as well as
conditions that could lead to uncompetitive outcomes. Screens
designed for that purpose should also adapt to the constant
changes in the market place. Market power (or conditions
leading to) might be present in one particular time interval and
not exist for another interval, even for the same transmission
constraint. The flagging for uncompetitive market outcomes
(prices) must be then based on both the actual market and
transmission system structural conditions; including
dynamically occurring transmission constraints and
corresponding (dynamically occurring) load pockets. For this
reason, it is important that the screen design reflects the
dynamic nature of the problem, so as to activate or
suspend mitigation based on the actual potential to exercise
market power.
The remainder of this paper is divided into three sections
that provide (i) an analysis of the design elements in building
screens; (ii) the presentation of specific screens that can be
used to detect market power; and (iii) conclusions.
II. ANALYSIS OF DESIGN ELEMENTS
This section discusses the elements that could be used in
building various screens and the expected characteristics of
these elements. Numerous examples are included to illustrate
different methodologies as potential candidates for providing
structural tests. The section also includes, when appropriate,
examples of similar screens developed (or being developed)
by other ISOs.
The design elements are as follows:
1. Definition of local: geographically and dynamically-
T
1710 142440178X/06/$20.002006IEEE PSCE2006
defined load pockets;
2. The use of generator shift factors;
3. Structural tests with basis in economics; and
4. Screen design alternatives.
A. Geographically (fixed) and dynamically-defined Load
Pockets
Partition of a control area into zones has been used by
some dispatch centers to handle generation control and
congestion management. A similar concept has been applied
in electricity markets for pricing purposes. As an example, the
NYISO uses a grid partitioning into a number of predefined
pricing zones for the purpose of load settlements in both day-
ahead and real-time markets. An LBMP zone is a
geographically and electrically contiguous area within the ISO
Control Area. Currently, there are eleven zones defined in the
NYISO control area as illustrated in Fig. 1. The zones are
separated by transmission interfaces that have historically
experienced congestion within the ISO Control Area.
NYCA Zones
A-K
A
B C
D
E F
G
H
I
J
K
IESO
(O)
HQ
(M)
ISO-NE
(N)
PJM
(P)
A B C
D
E F
G
H
I
K J
P
O
M
N
Fig. 1. NYISO LBMP zones
For example: zone A (West) is connected to zone B
(Genesee) through the inter-zone tie called Dysinger East,
zones I (Dunwoodie) and J (New York City) are connected
through the Dunwoodie South interface, and so on for the rest
of the LBMP zones. The assumption made at the time of
market start-up was that generators in each area will have very
similar nodal (LBMP) prices so that one single zonal price
will reflect closely the prices of individual generator nodes
included in the area of reference. Incorporated in this
assumption was the fact that congestion was assumed to take
place, most of the time, at the inter-zone interface level; hence
there was a congestion-caused separation of zones into
different LBMP prices. LBMP zones are geographically static,
and they also do not change with any market or system
operating condition.
The current mitigation scheme in New York City pre-
defines each load pocket and their interfaces with other parts
of the system. As shown in Fig. 2, all the load pocket ties are
considered together for the purpose of mitigation (i.e.: the
load pocket A interface is comprised of three transmission
facilities, 1, 2, and 3).
LOAD POCKET A G_1
G_3
G_2
G_4
G_5
INTERFACE A
LOAD POCKET B
RESOURCES IN LP A
G_1, G_2
RESOURCES IN LP B:
G_3, G_4, G_5
F
A
C
IL
IT
Y
3
F
A
C
IL
IT
Y
2
FACILITY 1
Fig. 2. Load Pockets and Interface Definition
Assume load pocket A is connected to the rest of the
system through Interface A. If the load pocket is a net
importer and the MW flow of Interface A is below its limit
(that is, each single facility is below its thermal limit), it is
assumed that there is workable competition and no mitigation
applies. However, if a single facility, say transmission facility
1, reaches its thermal limit; the interface as a whole is
considered to be at its limit (facilities 2 and 3 are considered
to be at their limits, even if that is not the case). As a result, no
more MWs can be imported, any increase in load should be
met by resources inside the load pocket, local market power is
assumed, and mitigation is enforced. In reality, the re-dispatch
of generators outside the load pocket might shift flows from
one transmission facility to another, relieving an overloaded
facility, increasing the ability to move energy into the load
pocket, and eliminating a binding interface constraint.
The assumption that only one facility needs to be at its limit
(binding) for the entire interface to bind may be overly
cautious. This practice can imply that congestion exists when
none actually exists. The market impacts in an already
congested area may be unacceptable.
In most ISOs, there has been a trend to use a pre-defined
set of zones for market mitigation purposes, [7], [13], [14].
They are usually called transmission-constrained zones in
which some form of mitigation or bid cap is implemented. The
zones may or may not be the same zones used in pricing.
Although its application is simple, the validity of this
approach should be re-examined; since it fails to recognize the
dynamics of congestion.
In reality, when a transmission constraint is binding, two
sets of resources can be described independently of the zones
they are located: resources that could alleviate the constraint
by increasing their MWoutputs (also referred as generators on
the constrained-side of a constraint) and resources that could
worsen the condition when increasing their MW outputs (also
referred as generators on the unconstrained-side of the
constraint) These two sets of generating resources do not
1711
generally follow any fixed (geographical or electrical)
boundaries as defined in local load pockets and/or LBMP
zones. In this case, it is more appropriate to talk about
dynamic load pockets, defined along the line of these two sets
of resources described before.
An example that illustrates the issue: the transmission
facility LEEDS_345-N.SCTLND_345-1 was binding for the
contingency MTN: SCB1 R391OR R94301 O/S LE in the
day-ahead market for 07/11/2003 at 9:00 AM, [5]. The set of
generators that could alleviate this constraint were distributed
across some of the same LBMP zones which also contained
resources that could worsen the constraint. Extensive review
of historical congestion data for the NYISO indicates that the
previous example regularly occurs in a congested system. The
example also shows that competitive analysis under the
presence of transmission constraints is more complex than
pre-defined local load pockets (or LBMP zones). Based on
this analysis, the following elements of design should be
evaluated, as a minimum, in this case:
1) Design Element 1 (DE-1)
Using historical data and currently available software,
study feasibility of using dynamically-defined load
pockets along constrained (unconstrained) side of a
transmission constraint for defining local market
power zone and corresponding structural test;
Analyze historical congestion patters, and perform
evaluation on most common constraints;
Determine congestion trend if appropriate
(concentration, number of hours binding, shadow
prices);
Compare with similar tests performed on the zonal
LBMP level.
.
B. Using Shift Factors
NERC defines a Generator Shift Factor (GSF) as follows:
a factor to be applied to a generators expected change in
output to determine the amount of flow contribution that
change in output will impose on an identified transmission
facility or Flowgate [12] . The importance of shift factors in
designing local market power screens is that they can be used
to determine the effective MW available for relieving or
worsening congestion on a particular transmission facility.
Effective MW depends mainly on each units operating
point (difference between operation point and upper operating
limit), the fraction of the resource MW production that
actually flows on the monitored line, and the direction of the
flow.
Shift factors values vary between -1 to 1 (or in percentage,
-100% to 100%) [1]. A positive GSF indicates that increasing
MW output from that generator will result in an increase of
the flow on the particular transmission facility (or Flowgate).
Conversely, a negative GSF indicates that increasing MW
output of the generator will result in a decrease of the MW
flow on the transmission facility. If the facility is constrained,
a positive GSF indicates generators that worsen the constraint
and a negative GSF identifies those generators relieving the
constraint.
Fig. 3 shows an example of how many generators with
positive and negative values there are for a particular
constraint. Notice that if only positive GSF are used to
examine generator bid behavior, then a small minority (about
20%) will be tested in such screen.
Distribution of GSF
-0.6
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
0 20 40 60 80 100 12
G
S
F
0
Fig. 3. Example of distribution of GSF for a given constraint
Historical analysis of GSF (together with the total MW)
can provide insight into: which set of generators are actually
being used to relieve constraints, dynamic zones defined along
the logic down-stream or up-stream of a constraint, the
electrical distance of the generator to a constraint (i.e.: GSF
close to zero are usually electrically distant from the
constraint), and the magnitude of a potential threshold to be
used in defining dynamic zones.
The use and computation of shift factors varies across
control areas. There is also not a single criterion about using
shift factors in market operations and/or congestion
management. Some ISOs, [13], divide the system or control
area into zones, defining zonal-averaged shift factors instead
of individual generator shift factors. Others do not use average
shift factors for a zone, but they utilize more than one
reference bus [14]. The NYISO uses a single reference bus for
its market operations software. Since the results of any
analysis using GSF are sensitive to the choice of the reference
bus, a special consideration should be given to this fact when
implementing a screen for a particular ISO. It is also of
importance to consider the fact that in transmission networks
the limiting constraints are very often on the flows that would
result in the event of a contingency rather than flows for
normal (pre-contingency) operating conditions.
The design elements on this case are as follows:
1712
1) Design Element 2 (DE-2)
Using historical data and currently available software
investigate GSF variation over time for a single
Flowgate,
Examine GSF variation for a single day/single
Flowgate for defining potential thresholds, and
Compute available effective MW for providing
congestion relief (MW x GSF) for a selected set of
commercially significant Flowgates and rank
generators accordingly.
C. Structural market power tests and congestion
Several structural tests based on economic principles are
examined in this Section to evaluate their performance in
detecting market power in congested systems.
1) Herfindahl-Hirshman Index (HHI)
Economic market power indicators such as Herfindahl-
Hirschman Index (HHI ) can be calculated as indicated in (1),
where n the number of the market participants, and q
i
is the
% of market share of participant i; defined as the output
(MW) of supplier i divided by the total market output (MW)
[2].
) 1 (
1
2
n
i
i
q HHI
HHI and market share tests could take on very different
values (at different locations) depending upon congestion on
the network; that is, HHI at load locations depends on which
supplier is able to compete to meet the load at its particular
location. We illustrate this fact with an example. Consider the
load pocket illustrated in Fig. 4; with four generators with a
total capacity of 600 MW, load of 900 MW, and import
capability of 750 MW. Clearly, there is not enough generation
within the load pocket to meet the load and imports are
required. Assuming the import is comprised of five generator
owners, each one of 150 MW, the market shares of G1, G2,
G3, and G4 are 14.8%, 14.8%, 7.4%, and 7.4% respectively,
and the overall HHI taking account of imports is 1,166.
Notice that the total market output in this unconstrained case
is 1350 MW (600MW +750 MW).
LOAD POCKET
F
A
C
I
L
I
T
Y
B
F
A
C
I
L
I
T
Y
A
G_1
G_3 G_2
G_4
INTERFACE CAPABILITY = 750 MW
G1 = 200 MW
G2 = 200 MW
G3 = 100 MW
G4 = 100 MW
LOAD= 900 MW
Fig. 4. Example of HHI application and load pockets
FERC considers a market with a result of less than 1,000 to
be competitive; a result of 1,000-1,800 to be a moderately
concentrated marketplace; and a result of 1,800 or greater to
be a highly concentrated marketplace [15]. The load pocket
of the example is competitive when using this commonly
accepted measure of market concentration.
However, assume now that the interface is constrained,
total market output in the constrained case is 600 MW, the
market shares of G1, G2, G3, and G4 are 33.3%, 33.3%,
16.7%, and 16.7% respectively, and the HHI value for the
load pocket alone (imports do not count) is 2,778; clearly no
longer a competitive load pocket. The transition from a
competitive case (HHI = 1,166) to a noncompetitive one (HHI
= 2,778) is impacted by congestion at the load pocket
interface. Notice that (as in many actual load pockets) there is
enough power to meet the load; rather, it is a structural
condition such as the transmission system being congested,
that is preventing competition within the load pocket. Fig. 5
shows the calculation of total HHI for the NYISO based on
generation dispatch.
HHI, DA Market (07/11/2003)
0
5000
10000
15000
20000
25000
30000
35000
40000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
HB
M
W
0
200
400
600
800
1000
1200
H
H
I
Dispatch (MW) BID (MW) HHI
Fig. 5. Example of Calculated HHI (total) for the NYISO
Very little insight, if anything is given by the HHI values in
this case. Moreover HHI varies around a value of 1,000 which
1713
is considered to be a competitive outcome; in reality, for the
very same day, and due to a transmission constraint there was
a portion of the zone Capital that was not competitive, with
HHI values of 2,800 and generators with market share greater
than 20%.
From the examples described before, it is obvious that HHI
and market share alone do not provide adequate detection of
local market power. In a very congested system, its
application is even more questionable.
D. Pivotal Supplier Test
Another structural test, based on economics, is the pivotal
supplier test. A simple definition of a pivotal supplier is the
following: a resource is defined to be pivotal when its output
is required to meet the load. The test recognizes that electricity
markets are particularly vulnerable to supplier market power,
and that loads (demand) have little or no price sensitivity
making the price elasticity of demand close to zero. Notice
that the pivotal supplier test is basically another way of testing
for economic withholding, that is, generators that unilaterally
withhold MW production may be able to increase prices,
because of the combination of zero or low demand elasticity
and/or the limited capacity (both generator capacity and
transmission system transfer capability) of other rival
resources to supply the required MW of demand [15]. Its
application could take on different forms as described with the
following examples.
Consider first the extreme case of only one generator inside
a load pocket with a total load of 100 MW; since the load
cannot be met without the generators MW output, the
generator in question will be always pivotal. Now assume
there are two generators with a capacity of 100 MW and 10
MW respectively, applying the definition of pivotal supplier,
the bigger generator is pivotal, the smaller one is not. Next,
assume both generators are equally sized (say, 50 MW each),
since both units are needed to meet the load, they are each
pivotal (also considered to be jointly-pivotal). Moreover,
assume the load pocket includes three equally-sized generators
(say, 34 MW each), then all three generators are pivotal, and
the load cannot be fully served with the loss of any one
generator. Lastly, assume the load pocket contains five
generators of 34 MW of capacity, any combination of three
generators will be jointly pivotal.
The first conclusion of this analysis is that there are choices
for considering pivotal suppliers: single or jointly pivotal (two
or three). Hence, market power screens based on single-
pivotal-supplier tests will probably be appropriate for extreme
cases of market power and certainly not be suitable for more
realistic market power situations. Another conclusion is that
the choice of two or three jointly pivotal suppliers and
combination of tests should be carefully analyzed for its
adoption as a structural test.
The examples described in the previous paragraph assume
that every single MW is available from each generator to
serve the load. In a congested system, even if there are enough
MWs available to serve the load, not every MW produced by
a generator might reach the load, but only a portion of it, as
specified by GSF. Since GSF determine the impact of each
supplier on the constraint congestion (either relieving or
worsening the constraint); it seems an obvious choice to
perform the pivotal supplier on that basis. However, even in
this case there could be choices of methodologies with their
own limitations and drawbacks. The first element to be
considered is the choice for determining the relevant supply
and demand to perform pivotal supplier analysis.
This is again illustrated with an example. Consider the
following two options for defining the load: Option 1) given
a transmission constraint define the load for pivotal supplier
analysis as that affected by the transmission constraint, and
Option 2) consider the analysis is independent of the load in
the sense that it is always being served, but rather focus on the
net demand for the residual transfer capability in each
Flowgate.
Regarding the supply the options could be: Option 1) for
any transmission constraint, there are generators that can
relieve or worsen the constraint, choose potential pivotal
supplier generators with negative GSF, and Option 2) based
on the definition of GSF given before, assume that potential
suppliers will maximize congestion by reducing (rising) the
output of generators with negative (positive) GSF, and that
rival generators will employ the opposite strategy.
In Option 1, although much more realistic that Option 2,
there is the problem that analysis could only be performed
after the fact. The transmission constraint, GSFs, and load
affected by the constraint need to be known before pivotal
supplier analysis can be completed.
Option 2, although currently being used, [7], is completely
dissociated with actual power system operations, since
generators not only need to know how and when to worsen
a constraint (self-dispatching themselves accordingly) but they
also would be ignoring dispatch instructions. It also assumes
that all other rival generators competing to serve the load act
unilaterally and simultaneously in an opposite strategy (so as
to relieve the constraint) also ignoring dispatch instructions.
The following example illustrates the issue. Fig. 6 shows a
three-area system; with generators dispatched in economic
order, a total load of 500 MW, and the transmission facility
W-S carrying a maximum flow of 200 MW. Table 1 in the
Appendix contains the generator data used in the example.
1714
Area
West
190 MW
Limit= 200 MW
200 MW
G
w2
= 300 MW
Load Area 3
320 MW
Load bus 2
180 MW
10 MW
G
w1
= 90 MW
Area
South
Area
East
G
s1
= 100 MW
G
s2
= 10 MW
Fig. 6. Three-area system for pivotal supply analysis
Applying the basic assumption behind Option 2, for a
generator to be pivotal, it should be able to cause the
transmission facility W-S to be congested (at its MW limit),
and also be able to sustain the congestion, even if any other
generator in the system able to relieve the constraint will do
so.
Assume the case of the potential pivotal supplier owns G
W2
and G
S2
, the strategy for this owner is such that all generators
with positive GSF with respect to facility W-S will increase
production and all generators with negative GSF will back-
down to their minimum. At the same time, an opposite
strategy is assumed for all others generators not owned by the
potential pivotal supplier. The strategies can be quantified (see
equations in Appendix) using GSF as follows: the total
(available) incremental change in line W-S MW flow due to
the potential pivotal supplier, IFP
w-s
, so as to worsen the
constraint will be:
MW IFP
s w
0 ) 0 333 . 0 ( 0 333 . 0