You are on page 1of 8

520

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 18, NO. 2, MAY 2003

Demand-Side View of Electricity Markets
Daniel S. Kirschen, Senior Member, IEEE

Invited Paper

Abstract—This tutorial paper discusses some aspects of electricity markets from the perspective of the demand-side. It argues
that increasing the short-run price elasticity of the demand for
electrical energy would improve the operation of these markets.
It shows, however, that enhancing this elasticity is not an easy task.
The tools that consumers and retailers of electrical energy need to
participate more actively and effectively in electricity markets are
discussed. The paper also describes how consumers of electricity
can take part in the provision of power system security.
Index Terms—Contracts, load management, load modeling,
load shedding, power demand, power industry, power system
economics, power system security.

I. INTRODUCTION

R

EDUCING the price paid by consumers for electrical
energy is invariably the first reason given for introducing
competitive electricity markets. In most instances, however,
consumers have very little influence on the design of these
electricity markets. Committees composed of representatives
from generators, transmission, and distribution companies,
retailers, and regulators take most decisions. There is an
obvious and understandable reason for this situation: most
consumers, with the possible exception of the largest ones, do
not have the financial incentive and the expertise required to
contribute effectively to such a complex and time-consuming
task. Possibly as a consequence of this lack of representation,
most electricity markets do not treat consumers as a genuine
demand side capable of making rational decisions but simply
as a load that needs to be served under all conditions. Over
the last ten years, wholesale electricity markets have become
quite sophisticated and have been fairly successful. However,
active participation in these markets by the demand-side remains minimal. Retail electricity markets have been much less
successful than wholesale markets and rarely give consumers
the opportunity to buy electrical energy at spot prices. Recent
experience in California made painfully clear, that introducing
competition on the supply side while shielding the demand
from liberalized prices seriously distorts the market [1].
This paper discusses the unusual economic characteristics of
the demand for electrical energy and the issues that must be
addressed if the demand side is to participate actively in the
market. It also outlines the tools and techniques that have been
Manuscript received November 7, 2002.
The author is with the University of Manchester Institute of Science and Technology, Manchester, U.K. (e-mail: d.kirschen@umist.ac.uk).
Digital Object Identifier 10.1109/TPWRS.2003.810692

or should be developed to help consumers take advantage of
the opportunities offered by competitive markets. A more active
participation of the demand side would make electricity markets
more efficient and more competitive. It would also promote a
more optimal allocation of economic resources.
II. ECONOMIC CHARACTERISTICS OF
DEMAND FOR ELECTRICITY

THE

The development of electricity markets is based on the
premise that electrical energy can be treated as a commodity. In
some respect, this assumption is undeniable: electrical energy
is an undifferentiated good that can be traded in quantity
because it is easily measured. Microeconomic theory therefore
suggests that consumers of electricity, like consumers of all
other commodities, will increase their demand up to the point
where the marginal benefit they derive from the electricity is
equal to the price they have to pay. For example, a manufacturer
will not produce widgets if the cost of the electrical energy
required to produce these widgets makes their sale unprofitable.
The owner of a fashion boutique will increase the lighting
level only up to the point where the additional cost translates
into additional profits by attracting more customers. Finally, at
home during a cold winter evening, there comes a point where
most people will put on some extra clothes rather than turning
up the thermostat and face a very large electricity bill.
If these industrial, commercial, and residential customers pay
a flat rate per killowatt hour for the electricity they consume,
they are insulated from the spot price for electricity and their demand is affected only by the cycle of their activities. Averaged
over a few weeks or months, their demand reflects their willingness to pay this flat rate. But what happens when the price
of electrical energy fluctuates more rapidly? Empirical evidence
suggests that demand does decrease in response to a short-term
price increase, but that this effect is relatively small [2], [3]. In
other words, the short run price elasticity of the demand for electricity is small. On a price versus quantity diagram, the slope of
the demand curve is therefore very steep. Determining the shape
of the demand curve with any kind of accuracy is practically
impossible, especially for a commodity like electrical energy. It
is interesting, however, to compare the range of prices for electrical energy sold on a competitive market such as the Electricity
Pool of England and Wales ( Table I) with a measure of the value
that consumers place on the availability of electrical energy. One
such measure is the value of lost load (VOLL), which is obtained
through surveys of consumers and represents the average price

0885-8950/03$17.00 © 2003 IEEE

1 illustrates the origin of these price spikes. Fig. electricity is indispensable in manufacturing and is regarded as essential to the quality of life by most individuals in industrialized societies. the new market equilibrium ( . Two economic and social factors explain this weak elasticity. Most industrial consumers therefore will not reduce their production drastically to avoid a small increase in their electricity costs. The intersection of these two curves determines the market equilibrium (i. This convenience has become so ingrained that it is fair to say that very few people carry out a cost/benefit analysis each time they turn on the light. the savings might be more than offset by the loss of profit. For the same period as the data shown in Table I. In the short term. (a) Market equilibria for a “normal” commodity.e. [7]. Similarly. the market price ( ) and the quantity transacted ( ). A disruption in the supply of the commodity (due for example to a breakdown at one of the production plants) is modeled by a leftward shift of the supply curve. While most of the capacity is offered within a relatively narrow price band. Fig. [5]. First. (b) Typical supply and demand curves for electrical energy. The demand curve represents the amount that consumers will purchase as a function of the price of the commodity. the price for the most expensive generation increases (a) (b) (c) Fig. At the same time. 1(a) shows the supply and demand curves for a “normal” commodity. For a normal commodity. ) is such that the change in price that results from this disruption is relatively small because it is accompanied by a change in the quantity purchased. The second factor explaining this weak elasticity is partly historical. 1(b) shows a typical supply curve in an electricity market [6]. it can be shown that this curve is equal to the aggregated marginal cost curve of all the suppliers.. electricity has been marketed as a commodity that is easy to use and always available. the VOLL used in England and Wales was set at 2. Fig. most residential consumers will probably not reduce their comfort and convenience to cut their electricity bill by a few percents. Since the early days of commercial electricity generation over a century ago. (c) Supply and demand curve following a major generation outage. It also represents only a relatively minor fraction of the cost of living for most households. the cost of electrical energy makes up only a small portion of the total cost of producing most industrial goods. This very low elasticity of the demand has two consequences that are undesirable for the good operation of electricity markets: it causes the large price spikes that have been observed in many spot electricity markets and it facilitates the exercise of market power by generating companies [1]. The supply curve indicates the amount of the commodity that suppliers are willing to sell as a function of the market price for this commodity.768 £/MWh. 1. [4]. .KIRSCHEN: DEMAND-SIDE VIEW OF ELECTRICITY MARKETS 521 TABLE I POOL SELLING PRICE ON THE ELECTRICITY POOL OF AND WALES (IN £/MWh) ENGLAND per megawatt hour that consumers would be willing to pay to avoid being disconnected without notice. In a perfectly competitive market.

consumption would resume at its usual level. ] over which the price oscillates. Such a contract again isolates the consumer from the vagaries of the spot market. most people would reduce the lighting level in their house and might even consider turning off their television. Medium-sized consumers are sometimes given this option at a later date. As soon as the price returned to a more normal level. The potential savings easily justify the purchase and installation costs of such metering systems for very large consumers. In its simplest form. In fact. buy more efficient appliances or move their factory to a region with lower average electricity prices. and therefore. it is important to consider the obstacles that will need to be overcome if this goal is to be achieved [12]–[14]. this can be an intermediate product. Such provisions encourage the consumers to be more flexible in their use of electricity and give the retailer some flexibility in managing the elasticity of its aggregated demand. The theory behind the spot pricing of electricity was developed by Schweppe et al. VOL. In the long run. This form of demand shifting is possible only if the consumer is able to store one of the factors of production. It must be emphasized that the preceding discussion pertains to the short run price elasticity of demand for electrical energy. On the other hand. The seller of the contract would be an aggregator or electricity retailer. but also a form of storage. residential consumers in some countries take advantage of lower nighttime tariffs by waiting until later in the evening to wash clothes or heat hot water. As the supply curve shifts to the left. for most other types of loads (heating. the contract might give the retailer the right to ask the consumer to reduce its load up to a certain extent during periods of high prices. Many economists will indeed argue that price spikes caused by shortage conditions are healthy because they highlight the true state of the market and encourage economically efficient behavior [9]–[11]. this elasticity is much higher because consumers have considerably more options. the price for peak load conditions ( ) increases abruptly because the demand curve has a very steep slope. every half-hour or every 15 min. NO. Enhancing demand response to price signals therefore requires not only some communication and control electronics. the customer should be kept informed in advance of the price effective for each period. In most jurisdictions. it could negotiate a more sophisticated contract with the retailer. industrial processes). depending on the market. For example. The very large number of smaller customers is a second reason for not allowing them to participate directly in the market. air conditioning. Even if we assume that all consumers are buying electrical energy on the spot market and that they are instantaneously informed of its price. The organization of the retail electricity market presents the first obstacle.522 IEEE TRANSACTIONS ON POWER SYSTEMS. 1(c) illustrates what happens following the outage of a significant amount of “cheap” generation capacity. The one on the left corresponds to light load conditions and the one on the right to peak load conditions. 2. The intersection with the supply curve indicates the range [ . Fig. electrical energy. Residential and small commercial consumers are usually not allowed to participate directly. [15]. a manufacturer may decide to delay the completion of a particularly energy-intensive step of a production process until the night shift if she expects that the price of electrical energy will be lower at that time. For example. Two practical reasons account for this discrimination [16]. they can heat with gas rather than electricity. The consumption of each participant in the market must therefore be recorded for each period and these recorded values must be downloaded regularly to the market settlement system. Shifting the supply curve to the left does not have to be the result of unplanned generation outages. In the long run. Consumers often equate price spikes and a high average price. the largest consumers are usually either obliged or given the opportunity to participate directly in the wholesale electricity market right from its inception. Generating companies wishing to exert market power can achieve the same effect by withholding capacity. MAY 2003 sharply. INCREASING THE PRICE ELASTICITY OF THE DEMAND Since increasing the price elasticity of demand would improve the behavior of electricity markets. most of the electrical energy that is not consumed during a high-price period would simply be shifted to a period with a lower spot price [3]. there is not a single demand curve. The very low elasticity of the demand ensures that the resultant price increase during peak hours will be large enough to offset the reduction in volume sold. Lighting and entertainment loads are exceptions because there is no justification for making up unsatisfied demand: very few people would feel the need to turn on extra lights to compensate for having had to sit or work in relative darkness earlier in the day. reduces the elasticity of the demand. It is only when consumers are exposed to variable prices that they will consider taking short run actions to maximize the value they get from the electrical energy they consume. The premium compensates the retailer for the risk associated with buying electricity at a variable price and selling it at a fixed price. the importance of electricity in daily life represents a second barrier to enhancing the elasticity of demand [17]. occasional price spikes can thus contribute to keeping the average price of electricity lower. III. Depending on the application. the demand curve shifts from left to right depending on the time of day and day of the week. If the buyer feels that this premium is excessive. the recurring communications cost) will often outweigh the potential savings for smaller consumers. In addition. Since the demand for electricity is cyclical. the spot price is set every hour. or dirty clothes. a contract guarantees a fixed price per kilowatt hour to the consumer in exchange for a premium above the expected average spot price for electrical energy. [19]. Two representative curves are shown. If the spot price was particularly high during a particular evening. Similarly. these costs (in particular. they have no incentive to modify their behavior. Such a price spike would result in a net reduction in the amount of electrical energy consumed for lighting and entertainment on that day. heat. As long as consumers purchase electricity on the basis of a tariff that has a fixed rate per megawatt hour. 18. Unless such a . Consumers who cannot or do not wish to buy electricity on the spot market have the option to enter into contracts that isolate them from the variations in the spot price. On the other hand. For example. First. Educating the public on this distinction would help electricity markets mature [8]. [18].

transmission congestion can have a sudden and hard to predict effect. this lower average spot price has an influence on the price of long term contracts. Most industrial consumers would have difficulties shifting significant amounts of load at short notice. it should still be possible to purchase electrical energy on a flat per kilowatt hour tariff. Unfortunately.. On the other hand. this operating profit must be sufficient to cover the additional investment costs required by the storage facility. If the cost of implementing load control measures exceeds the potential reductions in electricity costs. A. electrical energy. failures leading to withdrawal of capacity and price spikes) and others are not always publicly announced in advance (e. In such markets. if a consumer feels that the potential savings might be significant. The optimal storage and consumption strategy depends on the configuration of the production process. predicting price spikes (i. On the other hand. Abating price spikes has more than a short-term effect because it also lowers the average spot price of electrical energy. using storage will return an operating profit only if the difference between the price during charge and the price during discharge exceeds the losses associated with storage. when competition is less than perfect. In addition. Price Forecasting If a consumer is to buy electricity on the spot market. When the configuration of the storage and production process is not simple (i.. • use of storage does not affect other production costs. [22].. substantial benefits will accrue if the demand of a relatively small number of consumers becomes even moderately price responsive. Forecasting prices accurately is extremely complex because of the number of influential factors and the lack of information on some of these factors. The machine used for the first manufacturing step has a capacity that can be adjusted between 0 and 300 widgets per hour. the price profile is determined centrally one day ahead of real time. It remains to be demonstrated that these techniques are sufficiently robust to support actual trading decisions. The resulting surplus or deficit therefore has to be socialized (i. prices determined so long in advance often do not reflect the true cost of production.. TOOLS FOR CONSUMERS OF ELECTRICAL ENERGY The introduction of a competitive electricity market does not oblige consumers to change the way in which they use electricity. It should also be noted that the adoption of the same forecast by a large number of consumers would reduce the accuracy of this forecast because most of these consumers would adjust their demand on the basis of the same data and thus smooth the variations in prices. • the production requirements are specified in terms of a total quantity to be produced over a time horizon. Market designs that encourage demand side participation are therefore unlikely to come into force without pressure from consumers. This machine . Even under these conditions. On the load side. it is essential that it predict as accurately as possible the evolution of these spot prices over the time horizon used to plan production. some generators have the ability to influence prices to suit their own objectives. Larger and more frequent opportunities for adjusting demand in response to prices can probably be found if production can be scheduled to take advantage of low prices or avoid heavy consumption during periods of high prices. or an intermediate product and has access to a forecast of electricity prices. B.e. The first and most obvious benefit of this increased overall demand elasticity is an immediate and substantial reduction in the magnitude and number of price spikes.KIRSCHEN: DEMAND-SIDE VIEW OF ELECTRICITY MARKETS storage facility is a natural part of the process. when pricing is locational. A few papers proposing techniques for short term price forecasting have been published [27]–[30]. the biggest opportunities for rescheduling demand) is extremely difficult. The following sections describe the functionality of the tools that consumers might need if they decide to participate actively in the market for electrical energy.. some decision support tools are useful. Since the price of electrical energy derives from the market equilibrium. the consumers obviously do not have to forecast prices and can plan their consumption of electrical energy with confidence.e. the investment required may outweigh the potential savings. 523 A reasonable accuracy is achievable when forecasting the long-term average value and expected volatility of prices to evaluate hedging contracts [23]–[26]. Example 1: Consider an industrial plant whose production target is 1200 widgets over a 12-h period.g. A complete model of the industrial process will usually need to be developed and optimized using a numerical optimization package. Production Optimization A consumer who has the ability to store heat. economic and special factors that are used in load forecasting should also be taken into account when forecasting prices. The process used to manufacture these widgets requires two steps. Such a strategy is optimal (or quasioptimal) when the following conditions hold: • a single form of storage is available. As some authors have pointed out [1]. meteorological. A straightforward strategy consists in filling the store during periods of low prices and emptying it during periods of high prices. • quality of the stored item does not deteriorate over time. all the temporal. fashioning an optimal production strategy becomes more complex very rapidly. [7]. An increase in the price responsiveness of the demand is thus not in the best interest of generating company. charged to all participants in the form of an additional fee). when the above conditions are not satisfied). Finally. In addition. In some competitive electricity markets. it is influenced by both load and generation factors [21]. The generation side is considerably more troublesome because some events occur at random (e. A discussion of the factors that make some customers more likely to respond to price signals can be found in [20]. planned outages for maintenance). • amount stored does not decay over time. IV.e. could optimize its production schedule to take advantage of the periods of lower electricity prices. Ultimately. due to the uncertainty on the actual demand and on the available means of production.g. Many economists dislike this approach and the trend among new market implementations has been toward decentralized price discovery with shorter delays between gate closures and real time.

Fig. and cost of material movement.S. Widgets can be stored between the two manufacturing steps.S. A simple calculation shows that this simplest of production strategies would have a total energy cost of U. Electricity retailers will indeed try to get their business by offering them various types of contracts [7]. are risk averse or consume so little electrical energy that considering other types of contracts is not worthwhile. The net saving depends on the fraction of the total cost represented by the energy cost and the additional costs due to storage. Under such a contract. 18.$ 500. 3. loss of productivity. producing a saving of U. Fig. the price of electricity under this type of contract has two components. To reduce electrical energy costs. Many of the concepts underlying these contracts have been applied already as part of demand side management programs in traditional vertically integrated utilities. Tools for Selecting the Best Contract The previous section suggests that most consumers of electrical energy are unlikely to find it profitable to buy electrical energy exclusively on the spot market.S. The horizontal axis shows the fraction of the total cost made up by the nonoptimized energy cost. This does not mean that they will not be able to take advantage of a liberalized electricity market. only a few basic types will be discussed here. 2. 2 shows the prices that are forecasted for a 12-h period. While the cost of such a contract may be high. It is also the “default” type of contract offered to smalland medium-size customers. These additional costs include such items as losses. In this example. This production schedule would have a total energy cost of U. Fig. risk means the probability that the total cost will deviate from the expected value. The retailer therefore assumes all the price and quantity risks and is therefore likely to ask a substantial premium over the average price of electricity it expects to have to pay on the wholesale market. it completely isolates the consumer from the vagaries of the spot market. It has been suggested that changing .$ 245. 3 shows how this net saving varies for this example and three values of the relative additional costs. [31]–[33]. Consumers thus need to determine which contract best suits their needs. 1) Flat Rate Contracts: This is the simplest and oldest form of contract for the supply of electrical energy. VOL. Since a very wide variety of contracts is possible. MAY 2003 Fig.$ 255. the machine used for the first manufacturing step could be run at full capacity during the four hours with the lowest price for electrical energy (hours 1 to 4) and idled during the rest of the time. The parameter x represents the relative increase in nonenergy cost caused by the storage. C. the retailer undertakes to provide any amount of electrical energy to the consumer at any time of the day at an agreed and fixed rate per kilowatt hour. but also the risk involved. consumes 10 kWh per widget. In this case. The amount of electrical energy that is required for the second manufacturing step is negligible.524 IEEE TRANSACTIONS ON POWER SYSTEMS. NO. breakage. As Hirst has repeatedly pointed out [10]–[14]. If the need to store widgets between manufacturing steps increases the production cost. The output of the machine used for the second manufacturing step cannot be easily adjusted and should therefore be held constant. If both manufacturing steps are run at a constant rate. the net saving will be less than this value. Price profile for the 12-h period considered in the example. This evaluation should take into consideration not only the total cost of electrical energy that is expected to result from a particular contract. this parameter is assumed to be independent of the amount of time widgets are stored. 2. This type of contract is thus best suited to consumers who have an unpredictable demand. The first component covers the average cost of supplying the energy while the second represents an insurance premium against the risk of price spikes. 100 widgets will be produced and 1000 kWh will be consumed at each hour. Net achievable saving for the example.

The previous section discussed some of the contracts that retailers and consumers sign to share this risk. While some interruptible contracts allow the system operator to disconnect loads at very short notice in the event of an emergency. The separation of generation and system op- . As compared to the flat rate contract.e.e. can function only if the power system remains secure in the face of unpredictable contingencies. DEMAND CONTRIBUTION TO SYSTEM SECURITY So far. these customers will have to be classified into groups that have similar load profiles. some of the risk is shifted from the retailer to the consumer because the consumer has an incentive to consume during periods when the rates are lower. Because of the random nature of the consumers’ load. from a consumer’s perspective these two random variables are independent. This reserve allows the prompt restoration of the balance between generation and load in the event of the failure of a generating unit. 4) Spot Market Linked Contracts: In this type of contract. Ideally. Since it is clearly impossible for a retailer to forecast the demand of each of its customers. the rate is fixed for the duration of the contract but depends on the time of the day. Furthermore. The retailer must identify these groups and determine their load profiles and the dependence of these profiles on meteorological and temporal factors [34]. The cap limits the risk associated with high prices while the collar restricts the windfall reaped during periods of low prices. VI. a compensation that includes a payment proportional to the number of interruptions may be desirable. Once retail competition is in place. this perfect match usually cannot be achieved and the retailer is forced to buy or sell the imbalance on the spot market. 2) Time of Use Contracts: In this type of contract. this paper has considered only the trading of electrical energy. and independent companies usually compete to perform this role. All retailers will indeed not have the same mix off consumers: new entrants are likely to focus their marketing efforts on the larger industrial and commercial consumers while the incumbent supplier is likely to keep a substantial fraction of the residential customers. A cap and a collar are sometimes placed on the rates paid by the consumer. To this end. Electricity markets. The premium asked by the retailer is thus likely to be lower than for a flat rate contract. the load to be supplied by an individual retailer must be predicted from the bottom up (i. however. To manage this risk. Instead. generators. In exchange for paying a small premium. Retailers will also reduce their exposure to price risks by signing long-term flat rate or time of use contracts with generators for the bulk of the energy that they have to supply to consumers. the retailer must forecast as precisely as possible the demand of its customers. To reduce this risk. In many cases. This is a more complex problem than the traditional load forecasting problem that power system operators face. a single company no longer supplies the entire load of a region or even of a distribution feeder. It also prevents the system from collapsing following a fault on the transmission system. Such forecasts are difficult to obtain and are often wrong. In its simplest form. most of this reserve has traditionally been provided by the supply side. its annual energy consumption can also help a retailer focus its marketing efforts toward more profitable customers. the handling of these imbalances represents a risk for the retailer. [25]. To evaluate fixed rate contracts. each of the competing retailers supplies part of this load. its load factor. 3) Interruptible Contracts: Contracts specifying fixed rates leave the retailer exposed to large risks during price spikes in the wholesale market. Evaluating contracts linked to spot prices requires a prediction of the statistical distribution of these prices for each period over the duration of the contract [24]. While the total load of a region can be forecast using traditional “top down” methods. the retailer thus acts simply as an intermediary between the consumer and the spot market. To achieve security. a retailer would like to match the power it sells to consumers exactly with the power it buys using long-term contracts.KIRSCHEN: DEMAND-SIDE VIEW OF ELECTRICITY MARKETS the default to a different type of contract could significantly increase the responsiveness of the demand to price signals. Being able to predict the profile of a potential customer on the basis of extrinsic factors such as its type of industrial activity. the consumer avoids the rather large transaction costs associated with participating in the wholesale market. This type of contract therefore gives to the consumer control over the level of risk that it is willing to accept. V. the probability distribution of its load for each of the periods defined in the contract). it is easy to calculate the expected value of the total cost of electrical energy over the duration of the contract. Distribution companies.. retailers would like to be able to reduce the amount of energy that they have to deliver during such periods. Consumers whose demand is naturally higher at times of low prices or who can shift their consumption to these periods are likely to benefit from this type of contract. Since the spot price for electrical energy tends to be volatile. Fortunately. the rate paid by the consumer is linked to the spot price of electrical energy. It is also possible to compute the probability that this cost will exceed the expected value by more than a given percentage. they are likely to offer better rates to consumers who agree to reduce their consumption for short periods at the request of the supplier. It must be emphasized that the cost to a consumer of being interrupted is usually not directly proportional to the energy not served because of the losses and inconvenience associated with shutting down processes.. a consumer needs to know the statistics of its load (i. Computing the expected value and the standard deviation of the total cost is more complicated in this case because the calculation involves two random variables: the price of electricity and the consumer’s demand. Using that information. Risk management is at the center of a retailer’s function because it purchases electrical energy in bulk at variable prices and sells it to consumers at fixed rates. a substantial operational reserve must be maintained at all time. by aggregating the forecast for each consumer). each retailer’s share varies as they gain or loose customers. TOOLS FOR RETAILERS OF ELECTRICAL ENERGY Retailers buy electrical energy on the wholesale market and sell it to consumers whose load is too small to justify taking part 525 directly in the wholesale market.

“Industrial power demand response analysis for one-part real-time pricing. IEEE. and G. (2002. p. and N. Bohn. 65–75. Hughes. A. Washington. vol. Conf. vol.. and J.” IEEE Trans.” IEEE Trans. no. Sci. 2002. [Online]. Exposito. Hirst.. p. but would also help these markets operate more efficiently and satisfactorily. no. M. Comput. Caramanis. L. pp. Conf.html. Rep. Feb. Univ. R. J. Ramos.. Heffner. An attempt to quantify this benefit can be found in [7]. A. Lecture Notes in Computer Science. [17] G. pp. 2002. 5. M. Santos.lbl. CONCLUSION Enhancing the ability of the demand for electricity to respond to price signals could benefit not only the consumers who choose to participate actively in electricity markets. Santos.. ser. A.” The Elect. 1998. pp. Univ. Faruqui. Santos. vol. 2001. [2] J. no./Feb. 3. L. “Measuring and modeling of industrial demand response to alternative prices of the electricity. 18. Strbac. G. S. In some cases. 2000. 66–73. A. Mar.) Do enabling technologies affect customer performance in price-responsive load programs?. C. 13. Edison Electric Institute Rep. Dominguez.” The Elect. Apr. Szkuta. Hirst and B. pp. . and T. vol. vol. Cappers. R.) Demand responsive programs—An emerging resource for competitive electricity markets?. Faruqui and S. no. J.lbl. A Comparison of Two Techniques for Next-Day Electricity Price Forecasting. [10] E. “Market power and price volatility in restructured markets for electricity.526 IEEE TRANSACTIONS ON POWER SYSTEMS. 4. Boisvert. pp. Contreras. O’Sheasy. pp.ehirst. no. pp. 851–857. E.. This helps reduce the market power of generating units that are flexible or are located at critical locations in the transmission network might be able to exercise in the market for providing security services. Political Econ. D. 100. In exchange. providing security from the demand side should therefore be cost effective because it affects the consumer only during emergencies. vol.” J. .” The Elect. 2002. [18] J. 2412. Deb. 2000. [Online]. Boisvert. 14. 15.. 1.” in Proc.. [7] R. LBNL-50328. Aug. [20] B. 1998. consumers can participate in electricity markets by taking advantage of the new types of contracts being offered by electricity retailers. [28] F. vol. [19] D. no.C. 2. 14. Newbery. G.” IEEE Trans. Angelus. Available: www. 13. “RTP in California: R&D issues and needs. This increase in system security has a value that is shared by all consumers. Aug. 139. Available: http://eetd. Santos. 2002. no. 52–59. J. Feb. M. N.” The Elect. J. Faruqui. Unfortunately installing the necessary equipment and modifying the production schedules to take advantages of lower electricity prices will often not be cost effective.html. [12] E. Power System Economics. pp. J. S. J. 11. Goldman. 2002. 15. they would receive an option fee and an exercise fee proportional to the energy not supplied if their load is disconnected. Program on Workable Energy Regulation. R. REFERENCES [1] D. D. 2002. MA: Kluwer.C. J. J. July 2002. “What makes a customer price responsive?. a consumer’s opportunity cost during normal operation is zero. “Electricity price short-term forecasting using artificial neural networks. In practice. M. R. J. L. Power Syst. Bunn. pp. “Factoring the elasticity of demand in electricity prices. working paper PWP-081. Apr. [Online].” Report prepared for the California Energy Commission. 24–29. Breipohl. 929–953. Lawrence Berkeley National Laboratory..) Understanding competitive pricing and market power in wholesale electricity markets. Kintner-Meyer. 163–169. Syst. pp. Lane. Niemeyer. 5. 15. May 2002. J. 41–51. J. pp. pp. vol.ehirst. 2000. [Online]. reduce the likelihood of involuntary emergency load disconnections. 612–617.” Public Utilities Fortnightly.. Lora. 1999. Feb. Pro[5] gram on Workable Energy Regulation. [8] A. Ramos. Conejo. May 2001. T.. Syst.. 1. J. Kirby. (2001. May 2000. pp. The trouble with electricity markets (and some solutions). “Forecasting long-term electric price volatility for valuation of real power options. and A.” The Elect. Kirschen. vol. Cappers. Power Syst. Available: www. M. [27] B. Electricity Market Price Forecasting: Neural Networks Versus Weighted-Distance k Nearest Neighbors. “Electricity price forecasting in deregulated markets. T. Conf. M. S. Winter Meeting. and A. Exposito. 2002. “Forecasting loads and prices in competitive power markets.. A. [21] D. “The value of dynamic pricing in mass markets. Lecture Notes in Computer Science. pp. Power Syst. Heffner and C. 963–966. 34–41. “Mitigating price spikes in wholesale markets through market-based pricing in retail markets. Albert. Power Syst. “How to buy low and sell high. Jan. vol. “How to incorporate volatility and risk in electricity price forecasting. J. 2002 IEEE Power Eng. pp.gov/EA/EMP/.” Proc. 13.org. Mauldin. R.” in Proc. 101. R. Dillon. Spot Pricing of Electricity. vol.” The Elect.” The Elect. Hirst. Cumperayot. California Energy Institute. 342–348. 6. Hawaii Int. J. 384–390.com/publications.” The Elect.ucei. “Price-responsive demand: Key to Competitive Electricity Markets. 4.” in Proc. . Lora. Finally demand responsiveness during periods of high prices tends to reduce the load.ucei. 17. Aug. Schweppe. Goldman. May 2000.-L. Nogales. J. Borenstin. [13] E. R. [24] R. A. 159–164. Available: http://www. “The benefits of customer participation in wholesale electricity markets. E. (2002. Another disadvantage of demand-side participation in the provision of security is that it is harder to estimate accurately the size of the load reduction that will actually occur in the event of an emergency. 13–23. J. C. Aug. [30] A. Available: http://www. George. Washington. 15. no. and R. Edison Electric Institute Rep. Yusta and J. no. [9] S. 15. no. P. In theory. Session 15. the provision of security services by the demand side is likely to be limited to large consumers because of the cost involved in installing remote switching. [16] C. Eakin. Hsue. Tabors. G. vol. working paper PWP-067. 14. Caves. Green and D. P. and B. ser. Norwell. 1988. 88. Roos and I. [Online]. large consumers may also be able to compete in the market for the provision of power system security resources. Power Syst.) Retail-load participation in competitive wholesale electricity markets. Jan.” IEEE Trans. “Price-responsive demand in wholesale markets: Why is so little [14] happening.. Demand-side participation in the provision of system security does not benefit only the consumers who choose to participate. however. no. California Energy Institute. Sanabria.gov/EA/EMP/. Soc. 2.. Apr. June) Barriers to price-responsive demand in wholesale electricity markets. VII. NO. the amount of these option and exercise fees will depend on the location of the consumer.. [22] A. Paper 3. Sacramento. N. Neenan. [26] A. A. Mount. and M. and D. (2001.. and therefore. D. May 2002. Nevertheless. Sci. 1992. 1999.. J.” in Proc. 45–55. P. [25] V. and P. “Electricity price forecasting models. A.org.. C. pp. 4. “Forecasting next-day electricity prices by time series models. [6] R. Stoft. L. no. R. “Competition in the British electricity spot market. Mendes. 32–41. “The financial and physical insurance benefits of price-responsive [11] demand. 25–37. CA. pp. vol.com/publications. 3. . [15] F..” The Elect. 2001.. [3] A. (1999. [Online]. [29] A. New York: Wiley. If the demand side is used to handle emergencies in the transmission system. and R. Report LBNL-48 374 Lawrence Berkeley National Laboratory. Brown. 99. Apr. It also increases the options available to the system operator for maintaining security. 321–330. Unlike a generator that must reduce its production of electrical energy to provide reserve. K. T. pp. G. vol. 2453. [31] M. Available: http://eetd. vol. pp. [4] S. MAY 2003 eration that accompanies the introduction of competition gives a chance to the demand side to take a more active part in the provision of security for the system [35]–[38].. J. M. Espinola. vol. Hawaii Int. VOL. Consumers who do not mind being totally or partially disconnected during emergencies should be given the opportunity to offer their load as a contribution to the reserve. 3. Neenan. [23] T. vol.

Ahmed. “Application of data mining to electric load profiling. 122–128. Available: www. Feb. and Ph. [36] R. Rajamara. Nov. 527 [38] E. 1997. and Siemens on the development of application programs. Empros. Gedra and P. Gedra. Kirschen. pp. Power Syst.. 1994. pp. vol.” IEEE Trans. [37] M. Aug.html. Belgium. “Optional forward contracts for electricity markets.KIRSCHEN: DEMAND-SIDE VIEW OF ELECTRICITY MARKETS [32] T. Sarlashkar. 8. D. 1177–1184. in 1979 and the M. [34] B. From 1985 to 1994. S. V.” IEEE Trans. respectively.” IEEE Trans.. P. “Using utility information to calibrate customer demand management behavior models. and F.D. J. vol. 1998. Varaiya.Sc.ehirst. Hirst. U. W. 1766–1773. and R. Daniel S. Manchester. 13. 1999.D. Power Syst. pp. L. Manchester Institute of Science and Technology.. He is currently a Reader at the University of Manchester Institute of Science and Technology. Allan.” IEEE Trans. 317–322.K. [35] G.” IEEE Trans. W. Power Syst. . Fahrioglu and F. May 2001. “Markets and pricing for interruptible electric power. N. 9.com/publications. Strbac. L. Aug. Alvarado. degrees in electrical engineering from the University of Wisconsin-Madison in 1980 and 1985. Kirschen (M’86–SM’91) received the electrical and mechanical engineer’s degree from the Free University of Brussels. vol. vol. dissertation. 1096–1102. pp. vol. [33] T. 16. Pitt. [Online]. Price-responsive demand as reliability resources. Alvarado..” Ph. 1993. Univ. Power Syst. Power Syst. 12. S.. he worked for Control Data Corporation. pp. “A method for computing the value of corrective security. “The effect of elasticity on demand security prices for the Poolco and the bilateral contract models.