Congestion Zones

Full locational pricing at every node in the network is a natural consequence
of the basic economics of a competitive electricity market. However, it
has been common around the world to assert, usually without apparent need
for much further justification, that nodal pricing would be too complicated
and aggregation into single price zones, with socialization of the attendant
costs, would be simpler and solve all manner of problems. On first impression,
the argument appears correct. On closer examination, however, we find
the opposite to be true, once we consider the incentives created by aggregation
combined with the flexibility allowed by market choices. The debate continues,
but the negative evidence is accumulating.

For example, the first region in the United States to abandon a zonal
pricing model after it failed in practice was PJM, from its experience
in 1997 when its zonal pricing system prompted actions which caused severe
reliability problems. Given this experience, PJM adopted a nodal pricing
system that has worked well since March 1998.1 Subsequently, the original
one-zone congestion pricing system adopted for the New England independent
system operator (ISONE) created inefficient incentives for locating new
generation.2 To counter these price incentives, New England proposed a
number of limitations and conditions on new generation construction. Following
the Commission’s rejection of the resulting barriers to entry for new
generation in New England, there developed a debate over the preferred
model for managing and pricing transmission congestion.3 One zone was
not enough, but perhaps a few would do? In the end, New England proposed
go all the way to a nodal pricing system.4

A similar zonal congestion management market design created similar problems
in California, which prompted the Commission to reject a number of ad
hoc market adjustments and call for fundamental reform of the zonal congestion
management system. “The problem facing the [California] ISO is that the
existing congestion management approach is fundamentally flawed and needs
to be overhauled or replaced.”5 As a further example, the zonal pricing
system in Alberta, Canada, apparently produced a related set of incentives
that failed to give generators the price signal to locate consistent with
the needs of reliability: “Most of the electricity generation sources
are located in the northern part of the province and ever-increasing amounts
of electricity are being transported to southern Alberta to meet growth,
… [t]his is causing a constraint in getting electricity into southern
Alberta and impacting overall security of the high-voltage transmission
system.”6 As a result, Alberta has proposed a central generation procurement
process under the transmission operator to provide a means to get generation
built in the right place. Hence we have the ironic result of a supposed
simplification of the market under zonal pricing that seems headed towards
replacing central procurement by the monopoly utility with central procurement
by the monopoly transmission provider. This is hardly a true simplification,
nor is it consistent with the original intent to move towards a competitive
market and away from monopoly procurement.

Fact: A single transmission constraint in an electric network
can produce different prices at every node. Simply put, the different
nodal prices arise because every location has a different effect on
the constraint. This feature of electric networks is caused by the physics
of parallel flows. Unfortunately, if you are not an electrical engineer,
you probably have very bad intuition about the implications of this
fact. You are not alone.

Fiction: We could avoid the complications of dealing directly
with nodal pricing by aggregating nodes with similar prices into a few
zones. The result would provide a foundation for a simpler competitive
market structure.

The accumulating evidence reveals the flaws in this seductive simplification
argument.7 In reality, the simplification creates unexpected problems.
These problems in turn cause the system operator to intervene in the market
by imposing non-market solutions and socializing the costs. In the end,
the truly simple system turns out to be a market that uses nodal pricing
in conjunction with a bid-based, security-constrained, economic dispatch
for the real network, administered by an independent system operator.
Purchases and sales in the balancing spot market would be at the nodal
prices. Bilateral transactions would be charged for transmission congestion
at the difference in the nodal prices at source and destination. Transmission
congestion contracts would provide price certainty for those who pay in
advance for these financial “firm” transmission rights up to the capacity
of the grid. The system would be efficient and internally consistent.

Note that the problem with zonal price aggregation and poor incentives
does not extend to the use of market hubs within the framework of nodal
pricing.8 The hub-and-spoke model fits quite naturally within the nodal
pricing framework and has been operating successfully in PJM, producing
a liquid forward market at the PJM “Western Hub.”9 Market hubs can and
do provide virtually all the benefits of simplification often attributed
to zonal price aggregation. The difference between a hub-and-spoke model
and zonal price aggregation is simple: zonal aggregation gives you the
hubs without the spokes. The spokes capture the difference between the
nodal price and the market hub price. Zonal price aggregation assumes
these differences can be ignored, and then socializes the cost when than
cannot. And just as the wheel would not support the hub without the spokes,
the missing spokes in the zonal model lead to a collapse of price incentives
followed by the inevitable requirement for operator intervention.

In some cases, of course, the arguments offered for zonal price aggregation
may be true. The differences in nodal prices may be small, most of the
time, and the occasional excursions would not be commercially significant.
Or, to be more precise, the occasional excursions would not be significant
as long as the system operator did not socialize the costs. Under these
circumstances, there is a clear business opportunity. The RTO need not
and should not do anything different. Within this framework, an entrepreneur
would be free and able to set up a business that provided the aggregation
service, charging participants for the claimed benefits and providing
a revenue stream to compensate for the small risks involved.

When viewed from this perspective, the arguments in favor of zonal price
aggregation should not be seen as applying to the RTO. As we have learned,
when the RTO follows this path, trouble soon appears. Rather, the arguments
for zonal aggregation should be seen as either wrong or right. If wrong,
they should be ignored. If right, they should lead to a successful business.
But zonal price aggregation is usually a bad market design for an RTO.

Flowgates and Decentralized Congestion Management

The essential market ingredients outlined above include a coordinated
spot market integrated with system operations to provide balancing services
and congestion management. In principle, an alternative to central coordination
would be a system of decentralized congestion management that used the
same basic information as the system operator but could be handled directly
by the market participants.

The most prominent example of such a decentralized congestion management
model is the so-called “flowgate” approach. This is interesting as both
a theoretical argument10 and because it is the procedure embraced by NERC
as a principal market alternative to its administrative Transmision Loading
Relief (TLR) procedures.11 The details can be complicated, but the basic
idea is simple. The argument begins with the recognition that the contract
path model is flawed. Power does not flow over a single path from source
to sink, and it is this fact that causes the problems that lead to the
need for TLR in the first place. If a single contract path is not good
enough, perhaps many paths would be better. Since power flows along many
parallel paths, there is a natural inclination to develop a new approach
to transmission services that would identify the key links or “flowgates”
over which the power may actually flow, and to define transmission rights
according to the capacities at these flowgates. This is a tempting idea
with analogies in markets for other commodities and echoes in the many
efforts in the electricity industry for MW-mile proposals, the General
Agreement on Parallel Paths (GAPP), and related efforts that could go
under the heading of transmission services built on link-based rights.

For any given total set of power injections and withdrawals, it is possible
to compute the total flows across each line in the transmission network.
Under certain simplifying assumptions, it would be possible further to
decompose the flows on the lines and allocate an appropriate share of
the flows to individual transactions that make up the total loads. If
we also knew the capacity on each line, then presumably it would be possible
to match the flows against the capacities and define transmission services.
Transmission users would be expected to obtain rights to use the individual
lines, perhaps from the transmission line owner or from others who owned
these capacity rights.

In principle, these rights on each line might be seen as supporting a
decentralized market. Associated with each line would be a set of capacity
allocations to (many) capacity right holders who trade with the (many)
users of the system who must match their allocated flows with corresponding
physical capacity rights. Within this framework there are at least two
interesting objectives. First, that the trading rules should lead to an
efficient market equilibrium for a short period; and second, that the
allocated transmission capacity rights would be useful for supporting
the competitive market for geographically dispersed buyers and sellers
of power.

As a matter of principle, it is likely that the first objective could
be met. There should be some system of tradable property rights that would
be sought by users of the system, and in so doing would lead to an efficient
short-run dispatch of the system. This would seem to be nothing more than
an application of the principles of competitive markets with well-defined
property rights and low transactions costs. There is a general belief
that this short-run efficiency would be available in principle: “Efficient
short-run prices are consistent with economic dispatch, and, in principle,
short-run equilibrium in a competitive market would reproduce both these
prices and the associated power flows.”12 The problem has always been
with the natural definitions of the “physical” rights: these are cumbersome
to trade and enforce. The property rights are hard to define, and the
transaction costs of trading would not be low.

The second objective is perhaps more important. Presumably the allocated
transmission capacity rights would extend over many short-run periods,
for example, even only a few days, weeks or months of hourly dispatch
periods.65 Presumably a natural characteristic that would be expected
of these physical rights would be that a seller of power with a known
cost of power production could enter into an agreement with a distant
buyer to deliver a known quantity of power at a fixed price, including
the out-of-pocket cost for transmission using the transmission right.
Many other contracts could be envisioned, but this minimal possibility
would seem to be essential; and it is broadly taken for granted that this
capability will exist in the open-access transmission regime. However,
any approach that defines tradable physical capacity rights based on flows
on individual lines faces obstacles that appear to make it impossible
to meet this minimal test.

There are many variants of such link-based transmission rights that one
can imagine, and the industry has been struggling with these ideas for
years. Here the flowgate argument follows the outline above. The system
operators and others demur on the grounds that the electric system is
more complicated and there are simply too many lines and possible constraints
to manage in a decentralized environment. The proponents argue that it
is not necessary to consider all the lines and all the possible constraints.
Rather they propose to consider only a few critical constraints, the flowgates,
and to focus decentralized trading on these. The assertion is that the
commercially significant congestion can be represented by a system with:

• Few flowgates or constraints.

• Known capacity limits at the flowgates.

• Known power transfer distribution factors (PTDF) that decompose a
  transaction into the flows over the flowgates.

Under these simplifying assumptions, the decentralized model might work
in practice. The RTO would identify the flowgates. The capacity rights
would be allocated or auctioned somehow to the market participants. Similarly,
the RTO would publish the PTDF table that would allow individual market
participants to compute the effect of their transactions on the flowgates.
The participants would then purchase the corresponding flowgate capacity
rights in the market. This trading of capacity rights would take place
in decentralized forward markets. Transactions that had assembled all
the capacity rights needed would then be scheduled without further congestion
charges. Real-time operations would be handled somehow, typically not
specified as part of the flowgate model.

There is some experience with this flowgate model. However, the experience
is limited and what experience we do have is not good. In particular,
these simplifying assumptions and the corresponding flowgate model for
decentralized congestion management were applied as part of the NERC Pilot
Project for Market Redispatch in 1999, to create a decentralized alternative
to administrative TLR curtailments. In the end, and despite the substantial
turmoil created by the TLR system, the result was that apparently there
were no successful applications of any decentralized trades under this
approach.14 By contrast and at the same time, the centralized coordinated
market in PJM regularly provided successful market alternatives to administrative
TLR curtailments. Perhaps the flowgate problems will be ironed out as
the NERC experiment continues,15 but the experience reinforces the need
to look more closely at the flowgate model.

Despite the appeal of a move away from the contract path model and closer
to the actual underlying reality of the transmission network, these generic
methods built on flowgate rights must confront the problems inherent in
the simplifications. Are there only a few flowgates? Are the capacity
limits known in advance? Are the PTDF impacts stable and known in advance
of real-time?

Those who demur in accepting the flowgate model as a method for organizing
the use of the transmission system would answer in the negative for each
of these three questions. First, there are many potential constraints,
so it would be necessary to obtain many capacity rights on flowgates.
The number of rights that would have to be acquired in a complete version
of a flowgate model generally would not be determined simply by the amount
of power that flows in the actual dispatch. Under current practice, the
system operators typically adhere to “(n-1) contingency” constraints on
power flows through the grid. This means that the allowed power loads
at every location in the transmission system must be such that in the
event one of series of possible contingencies occurs, the instantaneous
redistribution of the power flows that results will still meet minimum
standards for thermal limits on lines and will still avoid voltage collapse
throughout the system. We can think of the terminology as coming from
the notion that one of the “n” lines in the system may drop out of service,
and the system must still work with the (n-1) lines remaining. The actual
contingencies monitored can be more diverse, but this interpretation conveys
the basic idea of an (n-1) contingency-constrained power flow.

Hence, a single line may have a normal limit of 100 MW and an emergency
limit of 115 MW.16 The actual flow on the line at a particular moment
might be only 90 MW, and the corresponding dispatch might appear to be
unconstrained. However, this dispatch may actually be constrained because
of the need to protect against a contingency. For example, the binding
contingency might be the loss of some other line. In the event of the
contingency, the flows for the current pattern of generation and load
would redistribute instantly to cause 115 MW to flow on the line in question,
hitting the emergency limit. No more power could be dispatched than for
the 90 MW flow without potentially violating this emergency limit. The
90 MW flow, therefore, is constrained by the dispatch rules in anticipation
of the contingency. The corresponding prices would reflect these contingency
constraints.17

Depending on conditions, any one of many possible contingencies could
determine the current limits on the transmission system. During any given
hour, therefore, the actual flow may be, and often is, limited by the
impacts that would occur in the event that the contingency came to pass.
Hence, the contingencies don’t just limit the system when they occur;
they are anticipated and can limit the system all the time. In other words,
analysis of the power flows during contingencies is not just an exception
to the rule; it is the rule. The binding constraints on transmission generally
are on the level of flows or voltage in post-contingency conditions, and
flows in the actual dispatch are limited to ensure that the system could
sustain a contingency. Operation of a complete flowgate model, therefore,
would require a trader to acquire the rights on each link sufficient to
cover its flows on that line in each post-contingency situation.

A sometime argument is that this problem is not serious because the actual
dispatch will have only a few of the potential constraints actually binding.
Typically this is true, but it does not avoid the difficulty for the simple
reason that we don’t know in advance which constraints will be binding.
Were it otherwise the system operator would not have to monitor all the
constraints that are typically considered. In fact, the large list of
potential constraints monitored by the system operator is already a select
group identified as the important subset from the thousands or millions
of possible constraints that could be defined given the large number of
lines and the large number of contingencies. The mere fact that the system
operator as identified the constraints would arguably be enough to require
an associated flowgate capacity right in order to ensure that the resulting
transaction would be feasible.

The accumulating experience in PJM is well documented and amply illustrates
the point. In one outside study intended to support the development of
a zonal model and decentralized congestion management through something
like a flowgate model, a set of 28 constraints were identified as important
and analyzed for the variations in the equivalent of a PTDF table. While
28 may seem a large number and difficult to deal with in assembling the
capacity rights to use the transmission system, it turned out not to be
large enough. In the event, the first six months of operation of locational
pricing in PJM found 43 constraints actually binding. Most importantly,
none of these actual constraints were in the list of 28 supposedly easy-to-identify
flowgates.18 This suggests the magnitude of the difficulties faced when
predicting which constraints will be binding.

The obstacle of too many constraints to specify a complete flowgate model
might be overcome if it were still possible to identify in advance how
much capacity there is at each flowgate. This is an old problem with the
uncomfortable reality that for many of the constraints it is not possible
to specify the limiting value without also knowing the pattern of the
loads. For example, interface constraints for voltage protection are routinely
described as a range of maximum values on real power flows, with the actual
value being set and changed regularly during real time operations. The
PJM Eastern Reactive Transfer Limit is reset at least every 15 minutes
and can vary over a range of 4000 MW to 7000 MW, depending on system conditions.20
This is essentially the same problem as defining the available transmission
capacity. As the New York Power Pool (NYPP) observed in a typical comment
heard from system operators:

“The primary responsibility of the NYPP system operator is and must
be to maintain the reliability of the bulk power system. The operator
must have the flexibility to decide, for example, what level of transmission
reserve capacity should be retained under various conditions and facilities’
loadings to meet contingencies as they may arise. Thus, actual transmission
availability, or, more correctly, available transmission transfer capability,
may be less than the thermal limits of the facilities, and the difference
may change as conditions change. The Commission should make certain that
all participants understand and accept these factors. 21

In addition to recognizing that the capacity limits are not always known
in advance, the other reality is the lack of truly stable and known PTDF
tables. The flows over the lines and voltages at the buses will depend
on all the other receipts and deliveries on the grid. Thus, the flow over
a particular flowgate that can be attributed to a particular transaction
will be changing all the time, so it will be difficult to know how much
of a flowgate capacity right is required or how much would be used. There
are many causes of this ex ante ambiguity in the PTDFs. First, the PTDFs
are a function of the entire configuration of the grid. With any line
out of service, there are different PTDFs, and the configuration of the
grid is changing all the time. Even with the same configurations on the
wires, there are many electrical devices, such as phase angle regulators,
whose very purpose is to change the apparent impedance of lines as a function
of changing loads and, therefore, to change the PTDFs throughout the system.
Furthermore, there are inherent nonlinearities in the flows and constraints,
especially the ubiquitous so-called “nomogram” constraints that attempt
to approximate even more complex interactions in the system. It is simply
not true that the real system conforms to the simplified textbook approximation
of the pure DC-load model that is useful for illustrating the effects
of network flows, but that is at best only a linearized local approximation
of the real system that can be used to guide the dispatch. It is for these
reasons that PJM updates both the load flow estimate and calculation of
its equivalent of PTDF tables every five minutes.22 In reality, the PTDFs
needed for a complete flowgate model would be anything but known in advance.

These criticisms of the flowgate model sit at the foundation of the argument
that it would not be an appropriate model for operating the power system.
However, as with the arguments above for zonal congestion management,
the criticisms may be less applicable to a commercial model that would
serve as an entrepreneurial business. Suppose that the criticisms are
correct, but the commercial significance is small. The RTO could operate
the coordinated dispatch and define financial transmission rights as outlined
above for the real system rather than the flowgate approximation. Although
it is not possible to identify all the components of the flowgate model
in advance, it is possible to determine in advance if a particular load
flow would be feasible. Hence, despite the complexity of the grid, a set
of simultaneously feasible point-to-point financial transmission rights
can be defined. For a given configuration of the grid, the RTO can guarantee
the point-to-point FTRs without using its powers to tax the participants
and socialize the costs.

Under the simplifying assumptions of the flowgate model, it would be
possible to decompose these point-to-point financial transmission rights
into their component flowgates, implied flow capacities on flowgates,
and the associated PTDFs. If the approximation errors of the flowgate
model are not large, then it would be possible for a new business to provide
the service of organizing trading of flowgate rights that could be reconfigured
to create new FTRs. The differences in flows and capacities might be small,
most of the time, and the occasional excursions would not be commercially
significant. Or, to be more precise, the occasional excursions would not
be significant as long as the system operator did not socialize the costs.
Under these circumstances, there is a clear business opportunity. The
RTO need not and should not do anything different than outline above as
part of the essential market design. Within this framework, an entrepreneur
would be free and able to set up a business that provided the flowgate
service, charging participants for the claimed benefits and providing
a revenue stream to compensate for the small risks involved. In effect,
the business could take the financial risk that the reconfigured FTRs
might not be feasible in the real network, but if the flowgate assumptions
are valid this risk would be small.

When viewed from this perspective, the arguments in favor of the flowgate
approach should not be seen as applying to the RTO. When the RTO follows
this path, trouble is likely to appear because the real system is more
complicated. Rather, the arguments for the flowgate approximation should
be seen as either wrong or right. If wrong, they should be ignored. If
right, they should lead to a successful business. But the flowgate model
is likely to be a problematic market design for an RTO.

Footnotes

1 William W. Hogan, “Restructuring the Electricity market: Institutions
for Network Systems,” Harvard-Japan Project on Energy and the Environment,
Center for Business and Government, Harvard University, April 1999, pp.
37-44.

2 The use of zones for collecting transmission fixed charges is not the
issue here. The focus is on managing transmission congestion. For a critique
of the previously proposed one-zone congestion pricing system, see Peter
Cramton and Robert Wilson, “A Review of ISO New England’s Proposed Market
Rules,” Market Design, Inc., September 9, 1998.

3 Federal Energy Regulatory Commission, New England Power Pool Ruling,
Docket No. ER98-3853-000, with October 29, 1998.

4 ISO New England, “Congestion Management System and a Multi-Settlement
System for the New England Power Pool,” FERC Docket EL00-62-000, ER00-2052-000,
Washington DC, March 31, 2000. The proposal includes full nodal pricing
for generation and, for a transition period, zonal aggregation for loads.

5 Federal Energy Regulatory Commission, “Order Accepting for Filing in
Part and Rejecting in Part Proposed Tariff Amendment and Directing Reevaluation
of Approach to Addressing Intrazonal Congestion,” Docket ER00-555-000,
90 FERC 61, 000, Washington DC, January 7, 2000, p. 9. See also Federal
Energy Regulatory Commission, “Order Denying Requests for Clarifications
and Rehearing,” 91 FERC 61, 026, Docket ER00-555-001, Washington DC, April
12, 2000, p. 4.

6 “Alberta Transmission Czar Wants More Generation,” Electricity Daily,
Vol. 14, No. 77, April 21, 2000, p. 3.

7 William W. Hogan, “Nodes and Zones in Electricity Markets: Seeking
Simplified Congestion Pricing” in Hung-po Chao and Hilliard G. Huntington
(eds.), Defining Competitive Electricity Markets, Kluwer Academic Publishers,
1998, pp. 33-62. Steve Stoft, “Transmission Pricing in Zones: Simple or
Complex?”, The Electricity Journal, Vol. 10, No. 1, January/February 1997,
pp. 24-31.

8 William W. Hogan, “Restructuring the Electricity Market: Institutions
for Network Systems,” Center for Business and Government, Harvard University,
April 1999, p. 52, available from the author’s web page.

9 “The New York Mercantile Exchange will launch an electricity futures
contract March 19 at the PJM western hub, one of the most liquid markets
in the Eastern grid. … The PJM hub already features an active and growing
over-the-counter forwards market. A liquid hub can have a downside [for
the futures contract] given that players are content trading in the OTC,
said one Northeast broker.” Power Markets Week, February 8, 1999, p. 14.

10 Hung Po Chao and Stephen Peck, “A Market Mechanism for Electric Power
Transmission,” Journal of Regulatory Economics, Vol. 10, No. 1, 1996,
pp. 25-59. Steven Stoft, “Congestion Pricing with Fewer Prices than Zones,”
Electricity Journal, Vol. 11, No. 4, May 1998, pp. 23-31.

11 Congestion Management Working Group of the NERC Market Interface Committee,
“Comparison of System Redispatch Methods for Congestion Management,” September
1999.

12 W. Hogan, Contract Networks for Electric Power Transmission,” Journal
of Regulatory Economics, Vol. 4, 1992, p. 214.

13 This is apart from the problems encountered with changes of the grid
capacity or configuration. Link-based rights have other substantial problems
for dealing with system expansion.

14 Congestion Management Working Group of the NERC Market Interface Committee,
“Final Report on the NERC Market Redispatch Pilot,” November 29, 1999,
filed with FERC on December 1, 1999.

15 NERC, “Market Redispatch Pilot Project Summer 2000 Procedure,” March
31, 2000.

16 Expressing the limits in terms of MW and real power is shorthand for
ease of explanation. Thermal limits are actually in terms of MVA for real
and reactive power.

17 Jacqueline Boucher, Benoit Ghilain, and Yves Smeers, “Security-Constrained
Dispatch Gives Financially and Economically Significant Nodal Prices,”
Electricity Journal, November 1998, pp. 53-59.

18 Richard D. Tabors, “Transmission Pricing in PJM: Allowing the Economics
of the Market to Work,” Tabors Caramanis & Associates, February 24, 1999,
p. 31. This is a careful study that is among the rare instances with easily
available and documented assumptions. See the PJM web page for the record
of actual constraints.

19 See the PJM web page spreadsheet report on historical transmission
limits, “Historical_TX_Constraints.xls.” Over the period January 1998
to April 2000, there were 610 constraint-days recorded, with the same
constraint appearing on more than one day. Based on the “Monitor” and
“Contingency” names, there were 161 unique constraints.

20 Andy Ott, PJM, personal communication.

21 Comments of Member Systems of the New York Power Pool, “Request for
Comments Regarding Real-Time Docket Information Networks,” No. RM95-9-000,
Federal Energy Regulatory Commission, July 5, 1995, p. 9-10.

22 Andy Ott, PJM, personal communication.