IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 20, NO.

1, FEBRUARY 2005 239

Capacity Subscription: Solving the Peak Demand

Challenge in Electricity Markets
Gerard L. Doorman

Abstract—In competitive power systems, investment in gener- Integration variable for .

ation capacity is left to market participants. As a consequence, Total installed capacity.
due to the special characteristics of electricity, power systems Sum of all ’s.
may become unacceptably unreliable during peak demand. One
way to solve this problem is self-rationing, a variant of priority Stochastic state (values between and
service, adapted to a competitive environment. With this concept, ).
consumers subscribe to their anticipated demand for capacity Value of where total demand equals in-
during system peak conditions. When these conditions occur, the stalled capacity.
individual consumer’s demand is limited to subscribed capacity.
This paper proposes capacity subscription, self-rationing in a
competitive market, as a solution to the challenge of matching I. INTRODUCTION
generation to peak demand. It is shown that this can be economi-
cally efficient. Some illustrations are given, and a framework for
market implementation is outlined. With capacity subscription,
consumers pay for their individual preferences for uninterrupted
A FTER an initial “the market will take care” attitude, gen-
eration adequacy during peak demand has become one of
the focal points in the discussions about restructuring. Charac-
supply, thus making the generation adequacy aspect of reliability
a private instead of a public good. teristically, peak demand is very volatile, and has a short dura-
tion. To make investment in generation profitable, prices would
Index Terms—Peaking capacity, power systems, restructuring,
self-rationing. have to be extremely high in these periods. The fact that de-
mand peaks occur infrequently results in high uncertainty and
high requirements to the rate of return of the investment. Many
NOMENCLATURE countries are presently looking for solutions to this fundamental
Lagrange multiplier of the installed capacity peak load problem.
constraint for the stochastic state in period One solution is the ICAP or installed capacity model used
. among others in the PJM market. The ICAP model imposes a
Cost of capacity. capacity requirement on the market, but the requirement is based
Consumer-type index (values between on traditional engineering criteria. Without this requirement,
and ). there is an increased probability of a capacity deficit resulting in
Consumer c’s optimal capacity. involuntary rationing of demand [1]. An alternative to imposing
Marginal cost of energy. a capacity requirement on the market is to make rationing
Demand when . more efficient and preferably market based. A specific form
Expected value of consumer surplus for con- for rationing, priority service, was proposed in [2]. The authors
sumer . argue that priority service can be seen as a rationing scheme in
Expected value of producer surplus. the case of deficient supply. The concept of self-rationing [3]
Cumulative density function of . can be seen as a specific implementation of priority service.
Cumulative consumer distribution function. This paper proposes a new concept, capacity subscription,
See text. as the solution to the potential shortage of peaking capacity
Price of capacity. in restructured markets. Capacity subscription is based on the
Period (value between 1 and ). same principles as self-rationing, but capacity subscription has
Fixed energy price when . a number of new elements.
(Spot) energy price when . i) While self-rationing was developed for a traditional
Consumer c’s marginal willingness to pay in market organization, capacity subscription is adapted to
period and state . a market-based environment.
Consumer c’s demand in period for price ii) It is shown that an optimal solution can be reached in a
p and state . market based on capacity subscription, by proving that op-
timal prices for energy and capacity equal their respective
Manuscript received May 26, 2004. This work was supported by the Norwe- marginal costs.
gian Research Council projects “Large Scale Power Exchange” and “Compe- iii) A framework for a practical market organization based on
tence Building—Capacity Shortage.” Paper no. TPWRS-00663-2003. capacity subscription is presented.
The author is with SINTEF Energy Research, 7465 Trondheim, Norway
(e-mail: [email protected]). Implementation in a market-based environment has two sig-
Digital Object Identifier 10.1109/TPWRS.2004.841230 nificant implications. Firstly, while self-rationing can be a way
0885-8950/$20.00 © 2005 IEEE

Authorized licensed use limited to: INTERNATIONAL ISLAMIC UNIVERSITY. Downloaded on July 30,2020 at 08:24:29 UTC from IEEE Xplore. Restrictions apply.
240 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 20, NO. 1, FEBRUARY 2005

to reduce costs in a traditional system (by reducing the need for when a capacity shortage occurs.2 In this case, the word fuse
peak generation capacity), capacity subscription can be a solu- becomes confusing, and we can call it a LLD. Although devised
tion to solve the problem of potential generation deficiency in a for a regulated environment, this scheme is intuitively attractive
competitive system, enabling a market solution without an ex- in competitive systems for a number of reasons:
ogenous capacity requirement and without unacceptable supply • Willingness to pay for capacity reflects consumers’ pref-
interruptions. Secondly, in a traditional context, the utility can erence for quality of supply. Consumers with a low pref-
set prices at whatever level that covers costs. In a competitive erence for quality of supply will obtain a small amount of
system, prices are driven down to marginal costs. Unless op- capacity, resulting in considerable curtailment when there
timal prices equal marginal costs, the market solution will not is a capacity shortage. On the other hand, consumers with
be efficient. a high preference for quality of supply will obtain more
It will be argued that, with the implementation of capacity capacity, and will not be curtailed.
subscription, a competitive market for capacity can be cre- • Capacity installation can be based on consumers’ real
ated where demand reflects consumers’ real preferences. The preferences, instead of surveys where consumers answer
proposed solution will effectively solve Stoft’s Second De- hypothetical questions.
mand-Side Flaw, the lack of real-time control of power flow • Quality of supply becomes a private good instead of a
to specific consumers [4]. public good, and demand and supply can be matched in
The paper is organized as follows: In Section II, self-rationing an optimal way.
is introduced, with reference to the literature for a traditional en- • Introduction of demand for capacity enables the creation
vironment. It is argued why the concept is useful in order to of a capacity market, where demand reflects the real func-
solve the peak load problem in competitive markets. As will tion of capacity, i.e., to ensure uninterrupted supply during
be explained, a remaining problem with self-rationing is “too all circumstances. If demand is high, the price of capacity
much load-relief,” i.e., once the rationing scheme is activated, will become high when there is a shortage, and producers
too much demand is rationed and idle capacity is available. In can invest in peaking capacity. On the other hand, if de-
Section IV, capacity subscription with optimal allocation of ex- mand is low, i.e., many consumers are satisfied with being
cess relief is analyzed, while Section V briefly discusses an curtailed during shortage periods, prices will stay low, and
alternative, random allocation. Section VI introduces the im- no new investments are made.
plementation of a market with capacity subscription and dis- • Although definitely more complicated than an en-
cusses the market equilibrium from a more practical point of ergy-only market, the transaction costs of self-rationing
view. Section VII gives some illustrations. Section VIII com- as proposed by Woo are probably lower than for full spot
pares capacity subscription with spot pricing, while conclusions pricing.
are drawn in Section IX. Before supporting these claims, some comments on [5] are
appropriate. The decisive difference compared with earlier con-
II. SELF-RATIONING IN COMPETITIVE POWER SYSTEMS tributions is the use of controllable LLDs that will not be acti-
vated before demand actually equals capacity. Consumers will
Self-rationing works as follows: each consumer subscribes to
learn to anticipate this, and obtain capacity according to their
a particular level of capacity some time before actual consump-
expected demand when system demand is high.
tion. Uncertainty is originally modeled through the stochastic
Despite its innovative contribution, there are two reasons why
temperature variable , where a high temperature implies high
inefficiencies remain in Woo’s model [6]. Firstly, marginal will-
demand. The consumer pays a capacity charge for the amount
ingness to pay generally differs between consumers at the time
he subscribes to, and an energy charge for actual consumption.
of consumption. This means that after activation of the LLDs,
If his usage exceeds subscribed capacity, a fuse curtails con-
some consumers will be willing to pay more for additional
sumption to the subscribed level. Consumers differ according to
consumption than others, leaving an economic inefficiency.
their willingness to pay for power or, put alternatively, quality
Secondly, some consumers will be limited by their consumption
of supply. In a regulated environment, the task of the utility is
even if system demand is less than capacity, because consumers
to find the optimal usage ( /kWh) and fuse ( /kW) prices, and
reach their subscribed capacity at different temperatures. In
the optimal installed capacity. Panzar and Sibley derive optimal
Woo’s scheme, this situation is avoided through the property
prices equal to marginal costs of production and capacity, re-
of weak separability, which entails that all consumers reach
spectively, and optimal installed capacity equal to the sum of all
the activation temperature at the same instant. As Doucet and
fuse sizes. They show that their scheme is ex post optimal under
Roland point out in [6], in that case, the possibility to control
the assumption that consumers are similar in the way in which
the LLDs is superfluous. However, the assumption of weak
temperature changes (uncertainty) affect their preferences.
separability is unrealistic and consequently too strong.
A problem with this analysis is that of “untimely curtail-
ments” when idle capacity exists, exemplified in [3] by the “in-
somniac” blowing his fuse at 4 A.M.1 Woo [5] proposes an in-
2The description is based on systems where a generation deficit may occur
novative solution, by letting the utility activate the fuses only
during infrequent high demand peaks due to especially hot or cold weather. In
systems with a flatter load profile over the year, such a deficit may occur any
1Basically the same criticism is given against demand charges in [9] (pp. time. This does not make a principal difference, but may alter the frequency
69–71), where it is stated that “Demand charges do not send good price sig- of load limiting device (LLD) activation and therefore consumers’ choice of
nals”. capacity levels.

Authorized licensed use limited to: INTERNATIONAL ISLAMIC UNIVERSITY. Downloaded on July 30,2020 at 08:24:29 UTC from IEEE Xplore. Restrictions apply.
DOORMAN: CAPACITY SUBSCRIPTION: SOLVING THE PEAK DEMAND CHALLENGE IN ELECTRICITY MARKETS 241

In [6], it is shown that the inefficiencies in Woo’s model The value of is given by
can be reduced by reducing the use of the LLDs. This scheme
reduces, but does not eliminate the problem of too much load
relief. Furthermore, it will not work in a competitive market,
because in this case it will not be possible to fix the price
above marginal cost. In Section IV, a model for a competitive
market will be derived. In (1), the first line represents the consumer surplus for values
of , i.e., when LLDs are not activated. The second line
III. OPTIMAL ALLOCATION OF EXCESS RELIEF represents the surplus for values of , with activated
LLDs for consumers with demand up to their subscribed ca-
In this section, a mathematical model for a market with ca- pacity. The third line represents the surplus for those consumers
pacity subscription and optimal allocation of excess load relief that buy power in excess of their subscribed capacity at the spot
is derived. Reserves and losses are neglected. They are relatively price . The last line is the cost of buying capacity.
easy to include, but the notation becomes more cumbersome. Provided all excess load relief is allocated, the effective
A “sticky” market price is assumed, i.e., it is assumed that system demand at is
the price reflects market conditions over a period of some dura-
tion, but cannot vary continually.
Uncertainty is modeled by the stochastic variable , which (2)
takes values between and . When demand is tempera-
ture dependent, temperature is the main source of uncertainty, The optimal subscribed capacity chosen by the consumer sat-
where the use of air conditioning implies a positive correlation isfies the first-order condition
between temperature and demand, and electrical heating a neg-
ative correlation.

A. The Consumer Problem

A formal discussion of the concept of utility in the current
context is given in [3] or alternatively in [7]. In the tradition (3)
of the literature, all variables are assumed continuous to ease
This states that in the optimum the price of capacity equals
the mathematical description. This may seem slightly odd with
the expected total cost of buying power in excess of subscribed
regard to consumer type and time interval, but is acceptable
capacity at the spot price in those periods that the LLDs are
given the large number of consumers and time steps (e.g.,
activated.
8760 h/year). It is assumed that all consumers have to subscribe
to some amount of capacity.
B. The Producer Problem
For some value of the stochastic variable , total demand
equals total installed capacity. is defined by Full allocation of empty capacity at implies that
demand equals available capacity in these states. Thus, producer
surplus is

Optimal allocation of excess relief implies that consumers

who prefer this possibility buy meters for hourly registration,
and are allowed to buy power in excess of their subscribed ca-
pacity, paying the spot price. The consumer’s problem is to buy
the amount of capacity at a per unit price , that maximizes
his expected surplus
(4)

Producer profits are thus made up of:

• the difference between price and cost, multiplied with total
demand in “normal” circumstances plus demand below
subscribed capacity under “activation” circumstances;
• the difference between spot price and cost multiplied with
demand above subscribed capacity;
• revenues from selling capacity minus the cost of installing
and maintaining capacity at the specific cost .
(1) The following constraints apply:

is assumed decreasing in and increasing in and . (5)

Authorized licensed use limited to: INTERNATIONAL ISLAMIC UNIVERSITY. Downloaded on July 30,2020 at 08:24:29 UTC from IEEE Xplore. Restrictions apply.
242 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 20, NO. 1, FEBRUARY 2005

C. Welfare-Optimal Solution With Optimal Allocation of be obtained in a market environment. Neither will the solution
Excess Capacity be optimal ex post, because the random allocation mechanism
The purpose of this section is to find optimal prices and in- naturally leaves consumers with different willingness to pay,
stalled capacity from a single-owner point of view. The ques- cf. [5, p. 76].
tion will then be if this solution is sustainable in a market-based In a practical context, the resulting loss of economic effi-
system. ciency must be compared with the cost of letting consumers buy
Based on the previous two sections, we can set up the La- power in excess of their subscribed capacity.
grangian for the problem
V. MARKET STRUCTURE AND FUNCTIONING
This section sketches a framework of a market design that
(6) supports capacity subscription and briefly discusses the be-
Here, the are the Lagrange multipliers of the constraints havior of the market participants.
(5). The derivatives of (6) are given in the Appendix. A market structure is assumed with a 24-h day-ahead spot
Solutions with or equal to zero are of no practical interest, exchange, where market participants place buy bids and sell of-
if they should exist. A solution with equal to zero does not fers with the help of price-volume relations. There will also be
make sense either, for then consumers would procure enough a real-time market to ensure the continuous balance between
capacity for any conceivable demand, and there would be no supply and demand.
need for rationing. Thus, an inner solution is sought, satisfying In addition to the daily spot exchange, there is a market for
the KKT-conditions capacity. Basically, this market has a longer horizon (season
or year), but buying and selling of capacity can in principle
be done more or less continuously, facilitating the market par-
ticipants’ changing needs over time. In this market, suppliers
The set of equations describing the KKT conditions can by bid their anticipated available capacity during system-peak as
inspection be shown to satisfy a price volume relation, while consumers do the same for their
anticipated need for capacity.
In a competitive electricity market, the system operator’s
main task is to control the continuous match between demand
This means that and are equal to marginal costs. An im- and supply (with the help of the real-time market) and to
plicit expression for can be derived from (9). Generally, operate the transmission network. In a market with capacity
the price for high-load conditions exceeds marginal subscription, the System Operator must know the total amount
cost with the amount necessary to limit demand to available of capacity sold in the market in (2)3. There are two different
supply. Optimal installed capacity equals the sum of subscribed ways to use capacity subscription. The first is related to 24-h
capacity. day-ahead (spot) market. When the forecast for the next day
In a properly designed market, prices equal marginal costs. indicates that demand will exceed system capacity ,
It has been shown that optimal prices for energy and capacity the system operator announces that LLDs will be activated,
equal their respective marginal costs, so a market based on ca- and demand will be limited to total subscribed capacity. All
pacity subscription can result in the optimal solution, based on parties can then take activation of the LLDs into account in
individual consumer’s preferences for uninterrupted supply. their spot market bids and offers. The second way is when de-
mand limitation is used in the real time market, when demand
IV. RANDOM ALLOCATION OF EXCESS RELIEF threatens to exceed supply on short notice. In this way, capacity
subscription can avoid both random rationing and extreme price
In the previous section, it was shown that optimal, price peaks in the real time market. To perform these functions in a
based allocation of excess load relief results in optimal prices timely manner, it must be assumed that:
for energy and capacity equal to their respective marginal costs.
• the system operator has frequent overview over total de-
However, a disadvantage of such a scheme is that consumers
mand and generation (e.g., every 5 min or more often);
need to be informed of spot prices during LLD activation, and
• LLD activation is fully automated and immediate.
that consumption in excess of subscribed capacity needs to be
Conceptually, there are three time horizons for the capacity
metered. A practical alternative to optimal allocation would be
market: firstly, in the short run (some hours to some days),
random allocation, where the system operator would allocate
available capacity is fixed and consumers’ willingness to pay
excess power to randomly selected consumers on a rotating
for capacity determines its price . Secondly, in the medium
basis. Assuming that there would be a number of LLD activa-
run (a few days to a few years), some more capacity can be
tion instances annually, activation could be spread as evenly as
made available: mothballed plants can be made operational,
possible between consumers, removing the need for advanced
maintenance can be rescheduled, or small-size gas turbines
options for buying capacity in excess of subscribed capacity.
or diesel engines can be installed. Within this time horizon,
Random allocation is explored further mathematically in
[8]. As could be expected, in this case, optimal prices do not 3Network congestion is disregarded here, but there are no principal obstacles
equal optimal costs, and the ex ante optimal solution will not to use self-rationing in a congested network.

Authorized licensed use limited to: INTERNATIONAL ISLAMIC UNIVERSITY. Downloaded on July 30,2020 at 08:24:29 UTC from IEEE Xplore. Restrictions apply.
DOORMAN: CAPACITY SUBSCRIPTION: SOLVING THE PEAK DEMAND CHALLENGE IN ELECTRICITY MARKETS 243

the price of capacity is determined both by the supply and

demand side. Thirdly, in the long run, it is possible to build
new optimally sized peaking plants, and the cost of these will
determine .
If there is excess capacity in the market, the probability of
is practically zero, and it does not make sense for a
consumer to buy capacity as long as its price is positive. In this
case will be zero, which is the rational outcome. However, if
there is a shortage of capacity, the price in the short and medium
term may exceed the long term balance price , because con-
sumers may be willing to pay a higher price to avoid being lim-
ited in their consumption. This gives a signal to the supply side
to provide more capacity in the long run at a cost , resulting in
a long-term price , as shown in Section IV-C.
Of course, with optimal allocation of excess capacity a con-
sumer can choose not to buy capacity, but to purchase energy at
the spot price when . However, in this case he
has to buy a real-time meter, which is expensive for small con-
sumers. Moreover, the consumer becomes exposed to a poten- Fig. 1. Example of two-consumer demand pattern.
tially unlimited spot price. Buying a certain amount of capacity
can be viewed as a forward contract that allows the holder to
purchase energy up to subscribed capacity at an agreed price .
In this way, capacity subscription combines the advantage of
spot-pricing, in this context price-rationing of excess demand,
with the advantage of tariffs, predictability.

VI. ILLUSTRATIONS
A. LLD Activation
The effects of LLD activation are illustrated with the fol-
lowing two-consumer, three-day example, shown in Fig. 1.
Each day is divided in six periods (instead of 24 for
readability reasons). The solid horizontal black line in each
pane shows each consumer’s and total subscribed capacity Fig. 2. Example supply and demand curves for power in excess of subscribed
respectively. capacity.
Day 1: Consumer 1 exceeds subscribed capacity in
period 5. Consumer 2 exceeds subscribed capacity in pe- following marginal willingness to pay for power in excess of
riod 3 and 5. In the latter period, total subscribed capacity their subscribed capacity
is exceeded, the LLDs are activated, and both consumers’
demand is limited.
Day 2: Total capacity is exceeded in period 4. Consumer 1
is unaffected, while consumer 2’s demand is limited. and that the marginal unit is a gas turbine with a marginal cost
Day 3: Consumer 2 exceeds subscribed capacity in period of 15 cents/kWh. This resulting supply and demand curves in
4, total capacity is also exceeded, and the LLDs are acti- excess of subscribed capacity are shown in Fig. 2.
vated. However, because consumer 1 uses well below sub- As a result, the excess capacity is allocated to consumer 3,
scribed capacity, there is excess load relief. In the case of who pays 25 cents/kWh for 1 unit of power in excess of sub-
optimal allocation, Consumer 2 would buy the empty ca- scribed capacity.
pacity if the price is lower than his willingness to pay.
C. How Much Capacity Should a Consumer Subscribe to?
B. Optimal Allocation of Excess Load Relief To illustrate this question, we can look at hourly metered load
By increasing the number of consumers to three, optimal for a specific consumer. We have the following data for a typ-
allocation of excess capacity can be illustrated. Assume these ical consumer: annual consumption 26390 kWh, peak demand
consumers have a capacity subscription of 4, 4, and 2, re- 12.4 kW, load factor 0.24 (2128 h).
spectively, so total subscription is 10. During a certain hour, We make the assumption that the total system experiences
their unrestricted demand is 3, 5, and 4. LLD activation results a shortage of capacity during the five days with highest spot
in a demand of 3, 4, and 2, totaling 9, and there is excess prices. Furthermore, we assume that LLD activation occurs
relief of 1. Assume further that these consumers have the from hour 7–10 and 16–19 on working days.

Authorized licensed use limited to: INTERNATIONAL ISLAMIC UNIVERSITY. Downloaded on July 30,2020 at 08:24:29 UTC from IEEE Xplore. Restrictions apply.
244 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 20, NO. 1, FEBRUARY 2005

Fig. 3. Example capacity and shedding costs.

When demand is limited, the consumer experiences a certain of enormous amounts of data. Capacity subscription re-
discomfort, which can be valued at some Value of Lost Load quires a communication solution, but not hourly or more
(VOLL). It is disputable if this is the correct value because con- frequent metering.
sumption is not completely shut off, only limited. In Fig. 3, the 5) Spot pricing might not reduce demand sufficiently, be-
effect of a value of VOLL of 0.6 and 2.5 /kWh is illustrated. cause consumers can and most certainly would protect
In the figure, the black straight line represents the capacity themselves through longer term (possibly purely finan-
cost, with an assumed capacity price of 25 /kW. The dark cial) contracts. Once they again face a fixed price, they
marked and the dashed line show total capacity costs with a will not bother about high spot prices. Although they
shedding cost of 0.6 /kWh and 2.5 /kWh respectively. The might profit from selling back when the spot price is
gray line shows the number of kWh that is rationed, using the high, most consumers probably would not do this.
right hand scale. 6) To make spot prices work to limit demand, there can be
With the lower VOLL value, total costs reach a minimum no price cap. Experience from existing markets shows that
of 150 with a subscribed capacity of 2 kW and 160 kWh of price caps will be imposed when spot prices become ex-
energy not served. However, with the higher value, the minimum treme, and this will be even more the case if final con-
is 243 for a subscribed capacity of 9 kW and only 7 kWh of sumers were directly exposed to the spot price.
energy not served.
VIII. CONCLUSION
VII. IS SPOT PRICING AN ALTERNATIVE? Capacity subscription in a competitive power system creates
Another alternative that also would avoid random rationing product differentiation and enables the creation of a true ca-
is real time spot pricing. Chao and Wilson [2] present the fol- pacity market, where “capacity during system-peak conditions”
lowing advantages of priority-service-type solutions over spot is traded. The essential feature of the solution is a centrally con-
pricing. trolled LLD, enabling the system operator to limit demand to
total subscribed capacity when generation capacity is insuffi-
1) Capacity subscription yields important information about cient. Based on their preferences for uninterrupted supply and
consumers’ preferences that can give a more efficient level the price of capacity, consumers buy their preferred amount of
of generation capacity. capacity, while generators can sell their available capacity. The
2) Capacity subscription offers consumers more certainty interesting feature of this solution is, that it guarantees that there
about their future bill than spot pricing. always will be “enough” generating capacity, avoiding random
3) A third, major point is that spot pricing may not be fea- rationing and thus solving one of the major problems of com-
sible. Due to transaction costs, the best one can hope for petitive power markets.
is that consumers adjust demand to prices based on a Because not all consumers will reach their subscribed capacity
day-ahead spot market. This does not take into account at the instant when the LLDs are activated, there will be some idle
short-term variations in real time that also may result in a capacity after activation, representing an economic inefficiency.
generation deficiency. With optimal, price based allocation, optimal prices are equal
There are other advantages as well: to marginal costs for energy and capacity, and the solution is
4) Continuous spot pricing (8760, 17 520 or even more pe- compatible with a competitive market. With random allocation,
riods per year) for all consumers necessitates the handling optimal prices are generally not equal to marginal prices and

Authorized licensed use limited to: INTERNATIONAL ISLAMIC UNIVERSITY. Downloaded on July 30,2020 at 08:24:29 UTC from IEEE Xplore. Restrictions apply.
DOORMAN: CAPACITY SUBSCRIPTION: SOLVING THE PEAK DEMAND CHALLENGE IN ELECTRICITY MARKETS 245

an optimal solution will not be reached. However, by using a

“rolling allocation” routine, the System Operator can divide
the excess capacity fairly between all consumers.
Consumers that put a high value on the possibility of being
able to buy electricity in excess of their subscribed capacity
during system-peaks, may invest in real time meters as an al- (8)
ternative to buying more capacity. Typically, this will be most
interesting for large consumers. Capacity subscription requires
a less expensive infrastructure than full spot pricing: the LLDs
and the necessary communication infrastructure, but no hourly
metering. It can be left to the market to develop solutions to
control demand efficiently at the consumer’s place. If capacity
(9)
prices are high, there will be a motivation to develop devices
that control demand with minimal loss of comfort, allowing con-
sumers to buy less capacity. On the other hand, if there is excess
capacity, prices will be low, and consumers will prefer to sub-
scribe on higher levels of capacity.
From a theoretical point of view, this paper has shown that
capacity subscription can be an efficient solution. These results
were illustrated with a framework for practical implementation,
and examples of some aspects were given. Further research is
needed in the following directions. (10)
• Field tests, including testing of potential LLD devices, and
communication with consumers.
• Development of a market design draft. REFERENCES
• Estimation of the amount of excess load relief in a prac- [1] M. A. Crew, C. S. Fernando, and P. R. Kleindorfer, “The theory of
peak-load pricing: A survey,” J. Regulatory Econom., vol. 8, no. 3, pp.
tical system, and evaluation of efficiency losses with 215–248, 1995.
random allocation. [2] C. Hung-Po and R. Wilson, “Priority service: Pricing, investment,
• Which groups to include in capacity subscription, and how and market organization,” The Amer. Econom. Rev., vol. 77, no. 5, pp.
899–916, 1987.
to treat groups who do not need to buy capacity. [3] J. C. Panzar and D. S. Sibley, “Public utility pricing under risk: The case
• The handling of transmission constraints. of self-rationing,” The Amer. Econom. Rev., vol. 68, no. 5, pp. 888–895,
1978.
[4] S. Stoft, Power System Economics. New York: IEEE Press, 2002.
APPENDIX [5] C.-K. Woo, “Efficient electricity pricing with self-rationing,” J. Regula-
DERIVATION OF THE LAGRANGIAN tory Econom., vol. 2, pp. 69–81, 1990.
[6] J. A. Doucet and M. Roland, “Efficient self-rationing of electricity re-
With optimal allocation given by (1) and by (4)), visited,” J. Regulatory Econom., vol. 5, pp. 91–100, 1993.
the derivatives of (6) become [7] S.-U. Lee, “Welfare-optimal pricing and capacity selection under an
ex ante maximum demand charge,” J. Regulatory Econom., vol. 5, pp.
317–335, 1993.
[8] G. L. Doorman, “Peaking Capacity in Restructured Power Systems,”
Ph.D. dissertation, Norwegian Univ. Sci. and Technol., Trondheim,
Norway, 2000.
[9] F. C. Schweppe, M. C. Caramanis, R. D. Tabors, and R. E. Bohn, Spot
Pricing of Electricity. Norwell, MA: Kluwer, 1988.

Gerard Doorman received the M.Sc. degree in

electrical engineering from the Norwegian Institute
of Technology (NTH), Trondheim, in 1981 and the
Ph.D. degree from the Norwegian University of
Science and Technology (former NTH) in 2000.
He began his career as a Research Scientist at
(7) the Norwegian Electric Power Institute. Later, he
worked with a power company on hydropower
optimization, and as a Consultant, among others,
with demand-side management. Since 1992, he has
been with SINTEF Energy Research, Trondheim,
working with hydropower optimization and power system restructuring issues,
presently as a Senior Research Scientist.

Authorized licensed use limited to: INTERNATIONAL ISLAMIC UNIVERSITY. Downloaded on July 30,2020 at 08:24:29 UTC from IEEE Xplore. Restrictions apply.