Professional Documents
Culture Documents
Abstract—In mixed pool/bilateral electricity markets, partic- the pool price and has no direct physical transmission implica-
ipants can sign forward bilateral contracts several months in tions [11], [12].
advance of its delivery. In addition, generators may sell to and Bilateral contracts are used in electricity markets to hedge
loads may buy from the pool at the spot price through the day-
ahead or balancing markets. Forward bilateral contracts have the
against pool price volatility; however, if improperly chosen, a
advantage of price predictability in comparison with the uncertain contract may actually worsen the benefit since, the eventual
spot price. However, the risk is that such a contract commits random spot price at market-clearing time, may end up being
the partners to a price that may be disadvantageous compared either too high or too low compared to the contract price.
to the spot price. Here, we propose a systematic negotiation Beyond spot price volatility [13], there exist other sources of
scheme through which a generator and load can reach a mutually risk associated with engaging in a bilateral contract, principally,
beneficial and risk tolerable forward bilateral contract, either
physical or financial. Under this approach, the generator and load random generator, and network-forced outages [14], load fore-
respond rationally to a stream of bilateral bids/counter-bids and cast errors, and fuel price uncertainty [15]. The negotiation of a
offers/counter-offers considering their respective benefits while bilateral contract will therefore converge only if both sides can
accounting for the risks incurred by the prediction uncertainty in find a pool/bilateral mix that provides an acceptable compro-
the pool spot price and other market parameters over the length mise between risk and benefit. The literature contains a wealth
of the contract. Each negotiating party can choose its own defi-
nition of risk which can be influenced by regret, value-at-risk or
of approaches to deal with risk/benefit tradeoff, typically as part
dispersion from the mean. Numerical tests show that this flexible of portfolio management [16], [17] and [18]. There is also a vast
negotiating approach can be readily put into practice. body of literature dealing with the broader issues of negotiation
Index Terms—Contract negotiation, forward bilateral contracts, and decision-making under uncertainty [19], [20].
market uncertainty, pool/bilateral electricity markets, regret, risk/ In the area of electricity markets, the use of bilateral contracts
benefit analysis, value-at-risk. to manage the risks associated with nodal price fluctuations due
to transmission congestion is analyzed in [21] and [22]. In [23],
the authors use stochastic optimization to determine the optimal
I. INTRODUCTION price and quantity of retail electricity contracts. The risk is then
N THE restructured power industry, load-serving entities measured as a term in the objective function that the retailer mini-
I and generation companies trade through both bilateral con-
tracts and the power pool [1]–[4] and [5].
mizes to limit its exposure to the volatile spot market. Schmutz et
al. [24] determine the value-at-risk from a portfolio of contracts
A bilateral contract is an agreement between two parties to using fuzzy set theory and scenario construction; Bjorgan et al.
exchange electric power under a set of specified conditions such [25] also begin with a portfolio of potential contracts and develop
as MW amount, time of delivery, duration, and price [6]. Bilat- a technique based on Pareto-optimality to manage the tradeoff
eral contracts can take the form of futures or forward contracts, between expected value and standard deviation of the portfolio
where the former are generally traded in an exchange [7], and profit; Marzano et al. [26] express the portfolio management
can be traded continuously up until their time of delivery. In problem as a dual dynamic program optimization which allows
contrast, forward contracts are typically negotiated directly be- for price dynamics over time. The pool/bilateral tradeoff problem
tween the load and generator with the terms of the contract re- has been addressed in terms of game-theory [27] and stochastic
maining fixed until the time of delivery [8]. optimization [28]. The optimal pricing of forward contract prices
Furthermore, a bilateral contract can be either physical or fi- has also been studied in [29] through an equilibrium model. In
nancial; the former meaning that all the power transacted bi- [29], an optimum forward price is determined by minimizing a
laterally must be self-generated and self-consumed at a pair global objective function of the expected profits and variances
of specified network buses [9], [10]. In comparison a contract of the generators and retailers, while meeting the condition of
is financial if the power transacted need not be self-generated zero net supply of forward contracts. The authors conclude that
and self-consumed but could be transferred up to the short-term forward contract prices will generally exceed expected spot
market-clearing time to another entity such as the pool [5]. Such prices when the demand expected value or volatility are high, a
a contract can be interpreted as a contract for difference that result that is tested by simulated and real data.
simply guarantees the difference between the contract price and The main innovation of this work is to propose a scheme
for a generator/load pair to negotiate a mutually beneficial and
risk-tolerable bilateral forward contract, or to conclude that it
Manuscript received March 7, 2006; revised October 6, 2006. This work was
supported by the Natural Sciences and Engineering Research Council, Canada,
is not in their mutual interest to enter into such an agreement.
and the Fonds québécois de la recherche sur la nature et les technologies. Paper In contrast to [29], we do not seek a common forward price that
no. TPWRS-00131-2006. meets a market equilibrium condition based on the optimization
The authors are with the Department of Electrical and Computer Engineering,
McGill University, Montreal, QC H3A 2A7, Canada (e-mail: sameh.elkhatib@
of a global objective function. In this work, during each bilat-
mail.mcgill.ca; francisco.galiana@mcgill.ca). eral negotiation session, we allow the parties wide flexibility to
Digital Object Identifier 10.1109/TPWRS.2007.894858 choose their own risk/benefit measures, under the assumption
0885-8950/$25.00 © 2007 IEEE
554 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 22, NO. 2, MAY 2007
that, in general, each negotiating party may have a unique in- III. BILATERAL CONTRACT CHARACTERIZATION
terpretation of what risk and benefit imply. The focus of our Whether physical or financial, a bilateral contract is typically
paper is therefore not on predicting how the market might be- negotiated weeks or months prior to its delivery and includes the
have under a common risk aversion rule and optimum forward following basic specifications: 1) starting date and time ( ); 2)
positions, but rather on analyzing bilateral negotiation schemes ending date and time ( ); 3) constant megawatt (MW) amount
between two specified participants, each of which is free to se- over the length of the contract, GD; 4) constant price in dollar
lect its risk aversion and benefit preferences. per megawatt hour ($/MWh) over the length of the contract, ;
The scope of this paper is restricted to forward bilateral con- and 5) range of hours when the contract is to be delivered.
tracts, either physical or financial, between a generator and load In a more general form, the MW amount GD and price
pair. The model assumed by the negotiation process is that the could be time-varying over the contract duration. This gener-
contract is to be delivered in a market where no single partici- alization is, however, not pursued in this paper which aims at a
pant can affect the spot price and where transmission congestion simpler negotiation that can be implemented by the two parties
and losses as well as their associated risks are not considered. independently without the need for complex calculations or the
The sources of risk that affect the negotiated contract are uncer- intervention of a mediator.
tainty in the spot price, fuel costs, load revenue factor and de-
mand. Risk can be measured by any of several well-established IV. UNCERTAIN MARKET PARAMETERS
approaches, namely, standard deviation [25], value-at-risk [13],
[30], [31], or regret [32]–[35]. The uncertain market parameters considered here are the
power plant efficiency and fuel cost coefficients , the load
revenue factor , and the spot prices and load demand levels at
II. BASIS OF RISK/BENEFIT ASSESSMENT APPROACH periodic time intervals, respectively,
and . The time interval index is
There is no unique way to define and assess risk and benefit. defined by sub-dividing the contract length into nc
Not only are the tolerance threshold to risk and its relation to intervals of equal duration (typically one hour).
benefit subjective but so are their definitions, in other words, Each of the above scalar parameters can be a random variable
the quantities that are deemed to constitute risk and benefit may with a discrete probability density function (pdf) defined over
vary from one person to another. a finite range. The spot price pdf’s estimated by the generator
In this paper we consider three approaches to assess the risk and load may be different. In addition, the pdf’s of the time-
of a projected bilateral contract: 1) based on the notion of regret, dependent parameters, and , may vary over the contract
under which the negotiating partner is influenced by the prob- duration to model growing uncertainty with time.
ability that the profit will not be sufficiently close to the best
possible (ideal) value, in other words what it could have earned V. GENERATOR RISK AND BENEFIT
if the market parameters had been predicted exactly at the time
of contract negotiation; 2) based on the notion of value-at-risk, A. Generator Profit
under which the negotiating partner is influenced by the min- Consider a generator with capacity and cost function
imum profit it can earn with a given probability; and 3) based , being the generator MW output at time . The gen-
on the notion of dispersion of the profit under which the negoti- erator must decide whether or not to sign a bilateral contract
ating partner is influenced by the probability that the profit will GD at a fixed price considering that at every time interval
deviate excessively from its mean value. , the generator could also sell power to the pool
The benefit of a bilateral agreement can also be measured through the short-term market. At time , at the uncertain spot
in various ways. One approach is through the expected rate of price , the amount sold to the pool depends on whether the
return, in other words the expected value of the ratio of profit contract is physical or financial as shown next.
to cost, a measure that can be applied in conjunction with any 1) A1-Generation Dispatch With Physical Contracts: Since
of the three risk measures. However, when risk is measured via for this type of contract the generator is obliged to self-generate
regret, the benefit can also be defined by the ratio of the expected its bilateral commitment, the level GD imposes a hard constraint
value of the actual and ideal profits. on the lower generation limit
Whether or not one of the above risk/benefit measures stands
out over any other is a subject of research and even controversy (1)
in decision-making [36], [37]. Decisions based on one risk mea- The generation dispatch at market-clearing under a physical
sure will in general not agree with the conclusions reached when contract is denoted by . If, as an example,
using a different measure. The results section illustrates the type the generator incremental cost takes the form with
of discrepancies that can occur under various choices of risk , then
measures. As this paper is not intended to debate what consti-
if
tutes a reasonable risk preference, the proposed bilateral nego-
if
tiation approach has been designed to be implemented under
if
any of the risk/benefit definitions proposed in the literature. We
(2)
also note that during a negotiation different risk/benefit mea-
sures (including a mix of the above three) may be used by each 2) A2-Generation Dispatch With Financial Contracts: Since
party according to how it is influenced by market uncertainty. with a financial contract the generator is not obliged to self-gen-
EL KHATIB AND GALIANA: NEGOTIATING BILATERAL CONTRACTS IN ELECTRICITY MARKETS 555
erate its bilateral commitment, the amount GD does not impose B. Generator Ideal Profit
a constraint on the lower generation limit The ideal profit is the maximum that the generator can earn
by adjusting its bilateral contract amount GD under the ideal
(3) assumption that the uncertain market parameters and are
exactly known at contract negotiation time. The ideal profit is
The generation dispatch at market-clearing under a financial needed when the negotiating generator bases its risk on the no-
contract is denoted by . Again if, as an example, the tion of regret, in other words, if the generator is swayed by the
generator incremental cost takes the form with extra profit it could have earned had it exactly predicted the un-
then certain market parameters at contract negotiation time.
Denoting the generic profit of a generator with either physical
or financial contracts by , the ideal generator
if
profit can be found numerically for any and , as explained in
if (4)
Section VIII, from
if
(8)
D. Generator Risk
The risk to a generator of embarking on a bilateral agreement
(6) can be measured in several ways. Here, we examine three alter-
natives all of which fit into the proposed negotiation scheme.
D1-Generator Risk Based on Regret: The difference be-
Observe that the second expression for the generator profit tween the ideal profit and the actual profit
with financial contracts in (6) has the same form as the gen- defines regret. The generator risk is defined
erator profit with physical contracts in (5), the distinction as the probability that the regret exceeds a fraction of the
being in the dispatched value of . In particular, in (6), for absolute value of the ideal profit, where is a self-specified
some time intervals, the profit due to the pool component, measure of the generator tolerance to such regret (the greater
may be negative, something that cannot the tolerance the higher )—see (9), shown at the bottom of
occur in (5) with physical contracts. the page.
(9)
556 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 22, NO. 2, MAY 2007
(10)
(11)
EL KHATIB AND GALIANA: NEGOTIATING BILATERAL CONTRACTS IN ELECTRICITY MARKETS 557
D. Load Risk
1) D1-Load Risk Based on Regret: Load regret is the differ-
ence between the ideal profit and the actual
profit . The load risk is the probability that
the regret exceeds a self-specified fraction of the absolute
value of the ideal profit—see (18), shown at the bottom of the
page.
2) D2-Load Risk Based on Value-at-Risk: Here, risk is the
probability that the profit will be too close to its minimum—see
(19), shown at the bottom of the page.
3) D3-Load Risk Based on Dispersion From the Mean: Risk
here is the probability that the load profit will deviate below its
mean value by an excessive amount—see (20), shown at the
bottom of the page, where is the mean value of Fig. 1. Generator and load risk for the initial proposed price of 35 $/MWh.
the profit. As in the generator case, a similar risk measure can
be defined as the ratio of the standard deviation to the absolute
mean approach for and GD contains two basic elements: 1) each
participant chooses its definition of risk and benefit as well as
the extreme limits beyond which it is not ready to compromise
(21) (these limits are affected by the type of risk measure used but
they all are typically less than one) and 2) the participants en-
gage in an iterative negotiation process under which one partic-
E. Load Benefit ipant offers or bids a contract price and a corresponding range
If the risk is measured in terms of regret, the benefit can be of GD. The receiving party conducts a risk/benefit analysis for
the expected value of the ratio of the actual and ideal profits the current which leads: a) to a new counter offer or bid; b) to
an agreement on a specific value of ; or c) to conclude
that there is no possibility of agreement. These steps are now
(22) described in more detail.
Alternatively, for any of the three risk definitions, the load A. Risk/Benefit Analysis
benefit can be expressed by the rate of return, namely the ex- For a proposed contract price the risk/benefit analysis by
pected value of the ratio of the profit to its payment to the pool either generator or load consists of two steps:
and to the generator supplying GD 1) Compute the respective risk and benefit versus the bilateral
contract amount GD over its feasible range. The various
substeps are detailed in Section VIII.
(23) 2) Identify the acceptable set of GD that meets the risk/benefit
criteria (e.g., Fig. 1). If no acceptable GD range is found,
the analysis rejects the proposed price .
(18)
(19)
(20)
558 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 22, NO. 2, MAY 2007
Fig. 2. Generator and load risk for the counter-proposed price of 43 $/MWh. Fig. 4. Generator benefit for three bilateral prices.
Fig. 7. Generator and load risk for the bilateral price of 34 $/MWh.
Fig. 9. Generator and load risk for the bilateral price of 38 $/MWh.
Fig. 8. Generator and load risk for the bilateral price of 42 $/MWh.
Fig. 10. Generator benefit for three bilateral prices.
ACKNOWLEDGMENT
The authors thank A. J. Conejo from the University of Castilla
La Mancha for his valuable comments and C. G. Quilez from the
University of Sevilla for her help in the numerical testing of the
Fig. 12. Generator and load risk based on regret for the agreed-to bilateral price
negotiation scheme.
of 38 $/MWh found using dispersion from the mean as the risk measure.
REFERENCES
[1] F. D. Galiana, I. Kockar, and P. C. Franco, “Combined pool/bilat-
eral dispatch. I. Performance of trading strategies,” IEEE Trans. Power
in Fig. 12, the risk would be 35% for the generator and 15% for Syst., vol. 17, no. 1, pp. 92–99, Feb. 2002.
the load. [2] X. Ma, D. I. Sun, and K. W. Cheung, “Evolution toward standard
Some notable observations on the negotiation process are as market design,” IEEE Trans. Power Syst., vol. 18, no. 2, pp. 460–469,
May 2003.
follows. [3] A. L. Ott, “Experience with PJM market operation, system design, and
a) Increasing the uncertainty range of any parameter usually implementation,” IEEE Trans. Power Syst., vol. 18, no. 2, pp. 528–534,
increases risk and decreases the probability of a negoti- May 2003.
[4] K. W. Cheung, P. Shamsollahi, D. Sun, J. Milligan, and M. Potishnak,
ated agreement. On the other hand, under perfect knowl- “Energy and ancillary service dispatch for the interim ISO New Eng-
edge of the input parameters, the parties cannot come to land electricity market,” IEEE Trans. Power Syst., vol. 15, no. 3, pp.
an agreement since, without uncertainty, each party would 968–964, Aug. 2000.
[5] Market Operations Document NE Market [Online]. Available: http://
know precisely whether a profit is possible or not. www.iso-ne.com/rules_proceds/isone_mnls/index.html
b) Unimodal nonuniform distributions of the uncertain pa- [6] I. Kockar, “Combined Pool/Bilateral Operation in Electricity Markets,”
rameters tend to decrease risk and increase the probability Ph.D. dissertation, Dept. Elect. Comput. Eng., McGill University,
Montreal, QC, Canada, 2003.
of a negotiated agreement. [7] J. C. Hull, Options, Futures, and Other Derivatives. Englewood
c) The presence of several simultaneous sources of uncer- Cliffs, NJ: Prentice-Hall, 1997, pp. 1–15.
tainty such as spot price and fuel costs increases risk and [8] ——, Introduction to Futures and Options Markets. Englewood
Cliffs, NJ: Prentice-Hall, 1998, pp. 1–15.
decreases the probability of a negotiated agreement. [9] I. Kockar and F. D. Galiana, “Combined pool/bilateral dispatch. II. Cur-
d) If no agreement is possible, other than halting the negotia- tailment of firm and non-firm contracts,” IEEE Trans. Power Syst., vol.
tion, one or both parties may opt to increase their tolerance 17, no. 4, pp. 1184–1190, Nov. 2002.
[10] L. R. Clarke, “New electricity trading arrangements in England and
to risk by increasing or by increasing the maximum tol- Wales,” in IEEE Transmission and Distribution Conf. and Exhib. 2002,
erable risk. Asia Pacific, Oct. 2002, vol. 2, pp. 1470–1472.
e) The effect of time-varying spot prices and demand levels [11] L. H. Fink, F. D. Galiana, and M. Ilic, Power Systems Restructuring:
Engineering and Economics. New York: Springer, 1998.
is also to increase risk compared to the case where these [12] M-28 Market Rule 1 Accounting Document NE Market [Online].
parameters are uniform across the length of the contract. Available: http://www.iso-ne.com/rules_proceds/isone_mnls/index.
html
X. CONCLUSIONS [13] R. Dahlgren and C. C. Liu, “Risk assessment in energy trading,” IEEE
Trans. Power Syst., vol. 18, no. 2, pp. 503–511, May 2003.
In mixed pool/bilateral markets, the spot price, system de- [14] D. Das and B. Wollenberg, “Risk assessment of generators bidding
mand and fuel costs at pool market-clearing are uncertain at the in day-ahead market,” IEEE Trans. Power Syst., vol. 20, no. 1, pp.
416–424, Feb. 2005.
time of negotiating a bilateral contract, typically several months [15] G. Gutierrez-Alcaraz and G. B. Sheble, “GenCos’ participation in the
in advance. Thus, although bilateral contracts are useful hedging unbundled energy market,” in Proc. IEEE PES Power Systems Conf.
instruments against the volatility of the spot market, the bilateral and Expo., New York, 2004, vol. 2, pp. 966–970.
[16] M. Denton, A. Palmer, R. Masiello, and P. Skantze, “Managing market
contract amount and price must be systematically negotiated in risk in energy,” IEEE Trans. Power Syst., vol. 18, no. 2, pp. 494–502,
order for both buyer and seller to meet their acceptable stan- May 2003.
dards of risk and benefit. [17] T. Aven, Foundations of Risk Analysis-A Knowledge and Decision-Ori-
ented Perspective. Chichester, U.K.: Wiley, 2003.
We have developed an iterative approach through which a [18] N. D. Pearson, Risk Budgeting-Portfolio Problem Solving with
generator and load can negotiate a bilateral contract described Value-at-Risk. New York: Wiley, 2002.
by its time of delivery, duration, amount, and price. The ap- [19] J. W. Pratt, H. Raiffa, and R. Schlaifer, “The foundations of decision
under uncertainty: An elementary exposition,” J. Amer. Statist. Assoc.,
proach is very flexible, allowing the negotiating parties to self- vol. 59, no. 306, pp. 353–375, Jun. 1964.
specify their risk/benefit measures, as well as the corresponding [20] A. Rubenstein, “Perfect equilibrium in a bargaining model,” Econo-
acceptable thresholds. Risk can be defined based on the notions metrica, vol. 50, pp. 97–110, 1982.
[21] T. S. Chung, S. H. Zhang, C. W. Yu, and K. P. Wong, “Electricity
of regret, value-at-risk, or dispersion from the mean, while ben- market risk management using forward bilateral contracts,” IEE Proc.
efit can be described in several ways including expected profit or in Generation Transmission and Distribution, pp. 588–594, Sep. 2003.
562 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 22, NO. 2, MAY 2007
[22] E. Tanlapco, J. Lawarree, and C. C. Liu, “Hedging with futures con- [34] D. Bell, “Regret in decision making under uncertainty,” Econom. J.,
tracts in a deregulated electricity industry,” IEEE Trans. Power Syst., vol. 30, pp. 961–981, Sep. 1982.
vol. 17, no. 3, pp. 577–582, Aug. 2002. [35] G. Loomes and R. Sugden, “Regret theory: And alternative theory of
[23] S. Gabriel, A. conejo, M. Plazas, and S. Balakrishnan, “Optimal price rational choice under uncertainty,” Econom. J., vol. 92, pp. 805–824,
and quantity determination for retail electric power contracts,” IEEE Dec. 1982.
Trans. Power Syst., vol. 21, no. 1, pp. 180–187, Feb. 2006. [36] ——, “Regret theory and information: A reply,” Econom. J., vol. 94,
[24] A. Shmutz, E. Gnansounou, and G. Sarlos, “Economic performance pp. 649–650, Sep. 1984.
of contracts in electricity markets: A fuzzy and multiple criteria ap- [37] K. Keasey, “Regret theory: And information: A note,” Econom. J.l, vol.
proach,” IEEE Trans. Power Syst., vol. 17, no. 4, pp. 966–973, Nov. 94, pp. 645–648, Sep. 1984.
2002.
[25] R. Bjorgan, C. C. Liu, and J. Lawarree, “Financial risk management in
a competitive electicity market,” IEEE Trans. Power Syst., vol. 14, no.
4, pp. 1285–1291, Nov. 1999. Sameh El Khatib (S’03) received the B.Eng. and
[26] L. G. B. Marzano, A. C. G. MeIo, and R. C. Souza, “An approach M.Eng. degrees in electrical engineering in 2002 and
for portfolio optimization of energy contracts in the Brazilian electric 2005, respectively, from McGill University, Mon-
sector,” in Proc. IEEE Power Tech Conf., Bologna, Italy, Jun. 2003. treal, QC, Canada, where he is currently working
[27] R. W. Ferrero, S. M. Shahidepour, and V. C. Ramesh, “Transaction toward the Ph.D. degree.
analysis in deregulated power systems using game theory,” IEEE Trans. His research interests include power system eco-
Power Syst., vol. 12, no. 3, pp. 1340–1347, Aug. 1997. nomics, portfolio management, and congestion man-
[28] C. C. Liu, H. Song, J. Lawarree, and R. Dahlgren, “New methods for agement.
electric energy contract decision making,” in Proc. IEEE Int. Conf.
Elec. Utility Deregulation and Restructuring and Power Technologies,
Apr. 2000, pp. 125–129.
[29] H. Bessembinder and M. L. Lemon, “Electricity pricing and op-
timal hedging in electricity forward markets,” J. Finance, vol. 3, pp.
1347–1382, Jun. 2002.
[30] K. Dowd, Beyond Value at Risk- The New Science of Risk Manage- Francisco D. Galiana (F’93) received the B.Eng.
ment. Chichester, U.K.: Wiley, 1998. (Hons.) degree from McGill University, Montreal,
[31] P. Jorion, Value at Risk-The New Benchmark for Controlling Market QC, Canada, in 1966 and the M.S. and Ph.D. degrees
Risk. New York: McGraw-Hill, 1997. from the Massachusetts Institute of Technology,
[32] V. Miranda and L. M. Proenca, “Why risk analysis outperforms prob- Cambridge, in 1968 and 1971, respectively.
abilistic choice as the effective decision support paradigm for power He spent some years at the Brown Broveri
system planning,” IEEE Trans. Power Syst., vol. 13, no. 2, pp. 643–648, Research Center, Baden, Switzerland, and held a
May 1998. faculty position at the University of Michigan, Ann
[33] B. G. Gorenstin, N. M. Campodonico, J. P. Costa, and M. V. F. Pereira, Arbor. He joined the Department of Electrical and
“Power system expansion planning under uncertainty,” IEEE Trans. Computer Engineering, McGill University, in 1977,
Power Syst., vol. 8, no. 1, pp. 129–136, Feb. 1993. where he is currently a Full Professor.