Joseph H. Yan, Gary A. Stern, Peter B. Luh, and Feng Zhao

24 IEEE power & energy magazine 1540-7977/08/$25.00©2008 IEEE march/april 2008

CCURRENTLY, IN DEREGULATEDwholesale electricity markets (Pennsyl-vania-Jersey-Maryland Interconnection,New York Independent System Opera-tor, Independent System Operator NewEngland, Midwest Independent SystemOperator, Electric Reliability Council ofTexas, and California Independent Sys-tem Operator) in the United States (e.g.,the day-ahead and real-time energymarkets), market participants submitenergy and ancillary service bids fortheir supply or demand bids to an inde-pendent system operator (ISO), and theISO runs auctions to determine theselections of resources, thelevels of output from theselected resources, and themarket clearing prices(MCPs) or locational mar-ginal prices (LMPs). (Forsimplicity of presentation,MCP is used in the remain-ing part of this article toillustrate the key ideas.)Markets are then settledbased on the MCPs, i.e.,selected supply bidders arepaid and selected demandbidders are charged at theMCPs regardless of their bidprices. This “pay-at-MCP”settlement scheme has beenwidely accepted by the U.S. electricitymarkets as opposed to the “pay-as-bid”scheme, where the payments for select-ed bids are determined as their bidcosts. Under the pay-at-MCP payment

scheme, however, ISOs in the UnitedStates currently minimize the total “as-bid” cost in their auctions to determinethe bid selections and the generationlevels of selected bids. This “bid-costminimization” auction is inconsistentwith the pay-at-MCP settlement, i.e.,the minimized bid cost in the auctionprocess is not the same as the paymentcost made during the settlementprocess. As a result of this inconsisten-cy, the total payment cost could be sig-nificantly higher than the minimized“as-bid” auction cost. To eliminate thisinconsistency, an alternative auction

mechanism that directly minimizes thepayment cost (“payment-cost minimiza-tion”) has been discussed, and signifi-cant progress towards an implementablesolution has been made in recent years.In the following, a brief history of thedebate on auction objective functions in

ISO markets is first presented alongwith the flaws of the bid-cost minimiza-tion auction. Then, the challenges of thepayment-cost minimization auction arepresented, followed by a summary ofrecent progress in addressing thesechallenges. Lastly, several questionsand concerns regarding payment-costminimization are discussed.

What’s the Issue?Before deregulation, vertically integrat-ed local utilities managed the entireprocess of power generation, transmis-sion, and distribution to serve their cus-

tomers. Classical unitcommitment and economicdispatch models were run todetermine generation sched-ules. The objective of thesemodels was to select a set ofgeneration units and theirgeneration levels to minimizethe total production cost (fuelcosts plus variable operationsand management) while satis-fying the demand and otherrequirements (e.g., reserverequirements). Under fixeddemand (e.g., the forecastload), the minimization of pro-duction cost was then equiva-lent to the maximization of

social welfare, as illustrated in Figure 1.The unit commitment and economic

dispatch problems belong to a class ofmixed-integer programming problemswith the objective of minimizing produc-tion cost subject to system demand, unitcapacity, reserve requirements, and other

the

busi

ness

sce

ne

payment versus bid cost minimization in ISO markets

Digital Object Identifier 10.1109/MPE.2007.915189

figure 1. The supply and demand curves. Social welfareis defined as the summation of consumer surplus andproducer surplus. With fixed (or vertical) demand, themaximization of social welfare is equivalent to the mini-mization of the production cost.

VerticalDemand

Q*

Supply (Marginal Cost)

Price

Quantity

ProductionCost

ConsumerSurplus

ProducerSurplus

26 IEEE power & energy magazine march/april 2008

individual unit constraints. These prob-lems are generally considered to be NP-hard. However, due to their separablestructure, they can be efficiently solvedby using decomposition and mixed-inte-ger programming techniques, and thereare well-developed software packagesreadily available to solve them. (An opti-mization problem is separable and canbe decomposed into subproblems if theobjective function and the constraints areadditive in subproblem decision vari-ables.) The key idea of decomposition isto first relax system-wide demand andreserve constraints that couple differentunits and then decompose the probleminto individual unit subproblems to besolved iteratively. The mixed-integerprogramming techniques include branch-and-bound and cutting plane methods.

Since deregulation, ISOs have runcentralized forward energy auctions inorder to select generation bids to meetthe demand and other requirements. An

important question immediately fol-lows: What should be the appropriatesettlement scheme? In other words, howshould the payments be determined?Two options—”pay-as-bid” versus“pay-at-MCP”—have been extensivelydiscussed. The pay-as-bid approach set-tles the payment based on bid costs,while the pay-at-MCP approach usesthe MCP for settlement. Abundantresearch has been done on the settle-ment issue, and it has been concludedthat pay-as-bid “would do consumersmore harm than good” (Blue RibbonPanel Report, 2001). The primary rea-son for this conclusion is that under thepay-as-bid settlement scheme, marketparticipants would bid substantiallyhigher than their marginal costs (sincethere is no incentive for participants tobid their operating cost) to try toincrease their revenue and, thus, offsetand very likely exceed the expectedconsumer payment reduction. As a

result, currently all ISOs in the UnitedStates adopt the pay-at-MCP principle.

While the market settlement schemehas been extensively studied, littleattention has been paid to the auctionmechanism. Currently, ISOs minimizethe total bid cost in their auctions. Theobjective function for a simplified auc-tion can be formulated as

min{pi,(t)}

J,

with J ∫T∑

t=1

I∑

i=1

{Oi(pi(t), t) + Si(t)},

(1)

where pi(t) is the selected megawattlevel of bid i at hour t, Oi(.) is the bidcost curve of bid i, and Si(t) is the start-up cost of bid i at hour t. [Note that Si(t)is incurred if and only if bid i is turnedfrom “off” status at hour t − 1 to “on”status at hour t.] The problem is verysimilar to a traditional unit commitmentproblem, only with bid costs replacing

CAPE lets you reconfigure a network with a few quick clicks.

1.888.240.4044 (U.S. only) • 734.761.8612 • eii@electrocon.com • www.electrocon.com/pe

Try CAPE’s serious modeling for 60 days, free.

Serious models:• Detailed bus structures

• Vendor-specific relay models

• Current-limited wind generators

For serious work:• Automatic relay setting

• Stepped-event simulations

• Automated coordination checking

• Export settings to relay vendor software

More detailed models give you better results.

The last several articles in this column have examined electric markets in

other parts of the world. These markets are designed from very distinctive

viewpoints, from representing the customer to aggregating the sellers.

Markets should be designed according to the total set or rules from bid-

ding through settlement. Settlement is the function to resolve the differ-

ences between the bids submitted and the actual operation of the system

over time, including all charges incurred to implement the contracts. The

present article discusses a change in

the market design to harmonize the

sett lement process with the bid

selection process.

The market settlements rules are

implemented to generate appropri-

ate price signals. The price signals are

communicated to the market partici-

pants via a complex system of settle-

ments rules implemented as part of a

particular market design. The full set

of rules is between 50 and more than

100 for some markets. A particular

implementation of the detailed rule

set is best understood if the rules are categorized into subsets, which might

run into the high hundreds. The Electric Reliability Council of Texas Settle-

ment Protocols described 42 rules. The amended California IS0 marked

design MD02 describes about 180 settlement business rules.

This system of rules covers the complete requirements needed to

financially balance the market, taking as inputs product prices; all traded

product volumes; grandfathered contracts; reliability must run (RMR)

contracts; various market conditions such as transmission congestion,

outages, generation, and transmission losses; and imports and exports on

the boundary interfaces.

The price signals get propagated back to the market participants after

being run through the set of settlement rules and, depending on the specific

market events and conditions, the participants are being invoiced for the cor-

responding debits or credits incurred by them participating on the market.

The correctness of the complex system of rules is crucial to the success-

es of the market design. However, analytically proving this correctness

could be a daunting task given the complexity of a particular market

implementation. Therefore, a way must be devised to decisively prove that

the market implementation is designed to properly telegraph market sig-

nals to the participants, which is also economically easy to implement and

able to be applied to a wide variety of market designs. This article attempts

to resolve the bids with the settlement by an alternative formulation of the

auction process.

—Gerald Sheblé

Associate Editor, Business Scene

Pay as Bid or Pay as Settled—Why Is There a Difference?

©ARTVILLE

march/april 2008 IEEE power & energy magazine 29

production costs in the objective func-tion. As a result, the existing softwarepackages for unit commitment and eco-nomic dispatch can be readily adaptedto solve the bid-cost minimization prob-lem. However, a basic question aboutthe above auction objective (1) is: Whydo we minimize the “as-bid” cost inauctions while the “pay-as-bid” settle-ment scheme has already been rejected?

Actually, corresponding to the twosettlement schemes, there are also twooptions for the auction mechanism:“bid-cost minimization” and “payment-cost minimization.” Unlike the bid costin (1), the payment-cost minimizationauction minimizes the consumer pay-ment cost, i.e.,

min{MCP (t )},{pi(t )}

J,

with J ∫T∑

t=1

I∑

i=1

{MCP(t)pi(t) + Si (t)}.

(2)

It is, therefore, natural to expect a debateon the auction objectives similar to thaton the settlement scheme. However,there is only limited discussion of pay-ment-cost minimization, and the adop-tion of bid-cost minimization was carriedout without much justification. Webelieve the auction issue at least deservesthe same level of debate the settlementissue received. Furthermore, one focusfor the study should be the consumerbenefit since it is the major objective ofderegulation and has been used to justifythe pay-at-MCP settlement scheme.

Despite the lack of debate regardingthe auction mechanism, one justificationfor bid-cost minimization is that the auc-tion would maximize the social welfare.The question, however, is whether or notthe market participants’ bid costs reallyand fully reflect production costs. If theanswer is yes, then the bid-cost mini-mization should be adopted and nodebate on the auction mechanism is

needed. If the answer is no, then we needto think further about the current use ofthe bid-cost objective function. Unfortu-nately, many studies have provided theevidence that generation bid prices areoften above marginal cost. One studyconcluded that sellers either submittedbids at prices significantly above margin-al cost of their generation unit or with-held part of the available capacity frombeing bid or scheduled into the market.Another study pointed out that, in sum-mer 2000, the California wholesale elec-tricity expenditure was US$8.98 billion,up from US$2.04 billion in summer1999, and that 59% of this increase wasdue to market power. A third studyshowed that there was overwhelmingevidence of a significant market powereffect reflected in wholesale marketprices in California during summer2000. Many researchers concluded thatgenerators bid higher and that their bidswere unrelated to production costs.

Furthermore, many ISOs (e.g., PJM andNY ISO) allow the existence of virtualbids in their day-ahead energy marketsunder their current market rules. Thesevirtual bids are purely financial and havelittle to do with the production costs ofphysical generation resources. Conse-quently, minimizing the bid cost, includ-ing that of the virtual bids, leads nowhereclose to social welfare maximization. Allthe above results make the maximizationof social welfare under bid-cost mini-mization unreachable.

What’s Wrong with Bid-Cost Minimization?As discussed above, ISOs in the UnitedStates currently use the pay-at-MCPsettlement scheme but minimize the as-bid cost in the auction process. Thisinconsistency between payment costand minimized bid cost may lead todramatically high consumer payments,as illustrated by Example 1.

Example 1Consider an auction for one hour withfour supply bids from four units andsystem demand of 100 MWh. For sim-plicity, transmission constraints andreserve requirements are not consid-ered. Also, the startup costs of select-ed bids are assumed to be fullycompensated. The supply bid pricesand characteristics of the four unitsare summarized in Table 1.

Under bid-cost minimization, thecheap units, A and B, will be selectedfor a total capacity of 90 MW, and theremaining 10 MW (100 MW − 90MW) goes to unit C, with additionalbid cost of US$1,020 (10 × 100 +20), instead of unit D, with additionalbid cost of US$2,300 (10 × 30 +2,000). This leads to a minimized totalbid cost of US$2,370, and the MCP

i s s e t b y t h e u n i t C p r i c e(US$100/MWh). Since the pay-at-MCP settlement scheme is used, everyselected bid is paid at the MCP. Thisleads to a total payment cost ofUS$10,020, which is significantlyhigher than the minimized bid cost. Itcan be seen that the high payment costis caused by the selection of unit C(having a lower bid cost) that sets ahigh MCP. The above bid-cost mini-mization results are presented in Table 2.

Now consider the solution under apayment-cost minimization auction.The results are presented in Table 3.

It can be seen that the payment-cost minimization auction results indifferent unit selections than the bid-cost minimization solution. Unit D isselected since it sets a lower MCP andthus lowers the consumer paymentcost as opposed to the selection of unitC. The different selections of unitsunder the two auction schemes areillustrated in Figure 2.

It can be seen from Figure 2 that theselection of high-price unit C under bid-cost minimization is caused by ignoringthe system-wide cost impact of its highbid price (US$100/MWh). As a result,consumers pay a significantly higheramount in the settlement process. Thiscan also be used to explain the pricespikes caused by the selection of somesmall-sized and high-priced units in thebid-cost minimization markets. In con-trast, payment-cost minimization con-siders the actual payment cost, and thuswould be less likely to select thosehigh-priced units, thereby reducing theclearing price spikes. Further study ofthe bidding behaviors of market partici-pants has shown more interestingresults. One of our findings is that sup-pliers are more likely to bid high pricesfor their small units under bid-cost

Capacity (MW) Bid Price ($/MWh) Startup Cost ($)Unit A 45 10 0Unit B 45 20 0Unit C 12 100 20Unit D 80 30 2,000

System Demand = 100 MWh

table 1. Bids of a four-unit 1-h example.

minimization compared to payment-cost minimization. This can be used toexplain the notorious hockey-stick bid-ding behaviors observed in currentmarkets. For example, a supplier havinga portfolio of generation capacities maychoose to speculate on its small units,as illustrated in Figure 3.

Challenges ofPayment-Cost Minimization In spite of the advantage of payment-cost minimization for reducing con-sumer payments, many challengesremain to be addressed. The first chal-lenge is how to solve the problem inview of the inseparable structure of apayment-cost minimization problem.While the solution may seem obviousfor the simple Example 1, it is actuallyvery difficult to solve larger problemswith more units and more hours sincethe number of possible unit selections isan exponential function of the numberof units and hours. More importantly,unlike the bid-cost minimization prob-lems where MCPs are byproducts of theminimization, a salient feature of pay-ment-cost minimization is the involve-ment of MCPs in the auction objectivefunction (2). This makes the payment-cost minimization problems inseparablein individual bids, i.e., there exists cross-product terms between MCPs and bidvariables. As a result, existing approach-es for unit commitment and economicdispatch cannot be used to solve thepayment-cost minimization problems.Also, a comprehensive comparison ofthe two auction schemes needs to bedone. This greater complexity associ-ated with solving the problem mayalso explain the lack of debate overthe auction mechanism, as described

Bid-Cost Minimization Auction Solution Settlement At MCP = US$100/MWhEnergy Startup

Selected As-Bid Energy Startup Cost Subtotal Payment Payment Subtotal MWh Cost (US$) (US$) (US$) (US$) (US$) (US$)

Unit A 45 450 0 450 4,500 0 4,500Unit B 45 900 0 900 4,500 0 4,500Unit C 10 1,000 20 1,020 1,000 20 1,020Unit D 0 0 0 0 0 0 0Total 100 2,350 20 2,370 10,000 20 10,020

table 2. Solution of bid-cost minimization auction for Example 1.

32 IEEE power & energy magazine march/april 2008

previously. The above challenges havebeen addressed in our recent study andongoing research, as presented in thefollowing section.

Recent Progress andOngoing Researchon Payment-CostMinimization Auctions To obtain insights on solving payment-cost minimization, a simplified marketmodel with single block supply bids and

fixed demand over the planning horizonis considered first for study. Ancillaryservices and transmission constraints arenot included, and selected bids areassumed fully compensated for theirstartup costs. MCPs are defined as thehighest price of selected bids. TheseMCPs are part of the decision variablessince they explicitly occur in the pay-ment cost objective (2). As a result, thedefinition of MCP needs to be opera-tionalized in the problem formulation.

Our approach is to incorporate “MCP-bid price” inequality constraints (i.e.,MCP should be greater than or equal tobid prices with zero prices defined forthose bids that are not selected) into thepayment-cost minimization formulation.Since the problem cannot be separatedinto individual bid subproblems with theexistence of the cross-product termsbetween MCP and unit generation levelsin (2), traditional Lagrangian relaxationtechniques cannot be directly applied.

Settlement At MCP =Payment-Cost Minimization Auction Solution US$30/MWh

Energy Payment Startup Payment Selected MWh (US$) (US$) Subtotal (US$)

Unit A 45 1,350 0 1,350Unit B 45 1,350 0 1,350 Consistent with the Unit C 0 0 0 0 auction solutionUnit D 10 300 2,000 2,300Total 100 3,000 2,000 5,000

table 3. Solution of payment-cost minimization auction for Example 1.

Our method is to usean innovative “surro-gate subgradient”method within an aug-mented Lagrangianrelaxation frameworkto overcome the diffi-culties caused by theproblem inseparability.The key idea of themethod is that therelaxed problem doesnot need to be solvedoptimally as requiredby the traditional sub-gradient method.Rather, an approximatesolution to the relaxed problem is suffi-cient if the “surrogate optimization” con-dition is satisfied, implying that the“surrogate subgradient” forms an acuteangle with the direction toward theoptimal multiplier vector. The relaxedproblem is thus optimized with respectto a particular supply bid one at a timeuntil the condition is satisfied. In opti-

mizing a bid, other variables may have tobe adjusted to satisfy the surrogate opti-mization condition. An augmentedLagrangian technique that adds quadraticpenalty terms into the traditionalLagrangian is used to reduce solutionoscillation. Numerical testing resultsdemonstrate that the method is effectiveand near optimal and that, for a given set

of supply bids, theresulting payment costis significantly lowerthan that obtained byminimizing the totalbid cost. The aboveaugmented Lagrangianrelaxation and surro-gate optimizationframework has alsobeen extended to incor-porate market practicessuch as demand bidsand partial capacitycost compensation.

In addition to ener-gy, ancillary services

(e.g., regulation, spinning reserve, andnonspinning reserve) also play animportant role in power operation andsystem reliability and are often procuredthrough auctions. There are two ways toconduct ancillary service auctions: oneis to conduct the energy auction first andthen the ancillary services auction(sequential optimization); the other is to

figure 2. Different solutions under bid-cost minimization and pay-ment-cost minimization.

10

100

MWh

$/MWh

45 45

BA

C

20

10

30

MWh

$/MWh

45 45

BAD

2K

Bid Cost Min. Payment Cost Min.

Bid Cost Increase for the Last 10 MWhSystem Cost Increase for the Last 10 MWh

co-optimize the energy and ancil-lary services (simultaneous opti-mization). The latter has beenshown to yield better solutions inour study, and it is adopted in thepayment-cost minimization prob-lem. With ancillary services, thecosts that consumers have to payinclude those for both energy andancillary services. In addition,since ancillary services are com-peting against energy usage forthe same generation capacity andthe provision of some ancillary services(e.g., regulation and spinning reserve)requires the units to be “on,” the energyand ancillary services are coupled in acomplicated manner. The payment-costminimization for simultaneous auctionof energy and ancillary services hasbeen solved by extending the augmentedLagrangian and surrogate optimizationframework developed earlier by our uni-versity collaborators. We are currentlyconducting more numerical testing andpreparing a journal paper.

Since standard market design (SMD)and most ISOs have adopted LMPs, wehave also solved the LMP payment-costminimization problem with transmissioncapacity constraints. The considerationof transmission constraints complicatesthe problem by entailing power flow lim-itations and introducing LMPs. For sim-plicity, transmission loss is notconsidered and dc power flow is used.LMPs are defined by “economic dis-patch” for the selected supply bids. Tocharacterize LMPs that appear in the

payment-cost objective function,Karush-Kuhn-Tucker (KKT) con-ditions of economic dispatch areestablished and embedded as con-straints. The reformulated problemis difficult in view of the complexrole of LMPs and the violation ofconstraint qualifications caused bythe complementarity constraints ofKKT conditions. Our key idea is touse a regularization technique tomanage the violation of constraintqualifications and then extend the

surrogate optimization framework. Spe-cific techniques to satisfy the “surrogateoptimization condition” in the presenceof transmission capacity constraints aredeveloped. Numerical testing results ofsmall examples and the IEEE ReliabilityTest System with randomly generatedsupply bids demonstrate the effective-ness of the method.

The above results concentrate on thedevelopment of solution methodologiesfor payment-cost minimization prob-lems with various market setups. Testing

figure 3. Illustration of hockey-stick bidding.

MWhU1

U2

$/MWh

results show that with the same set of supply bids, payment-cost minimization leads to reduced consumer payments com-pared to bid-cost minimization. However, the cost savingsmay not be realized since market participants may bid differ-ently under the two auction mechanisms. This leads to theneed to investigate the strategic behaviors of power suppliers.In our recent effort, the bidding behaviors of suppliers underthe two auction mechanisms are studied by using a game the-oretic framework. Market participants are assumed to havecompeting objectives, and a supplier maximizes his/her profitby selecting an appropriate bidding strategy from finite strate-gy choices. Since each participant’s bidding decision dependson the other participants’ decisions, a matrix game is formedfor each auction method and a Nash equilibrium is used asthe solution concept. Simple two-supplier Nash games withcontinuous strategies are first analyzed to provide insights.General matrix Nash games are then solved by using theconcept of approximate Nash, with our auction algorithmsdeveloped earlier serving as the core. Testing results demon-strate that, in general, payment-cost minimization leads tosignificant payment reduction with relatively small increasesin production cost and bid cost as compared to bid-cost min-imization. Furthermore, testing examples demonstrate thathockey-stick bidding is less likely to occur under payment-cost minimization. Our ongoing research includes the studyof continuous games for the two auction methods and theinvestigation of broader economic implications.

DiscussionSince the payment-cost minimization auction was broughtinto discussion, many insightful questions and concerns havebeen raised. For example, our study shows that the consumersavings are obtained at the cost of reducing generation rev-enues. Therefore, an important question is: Would this causerevenue adequacy difficulties (i.e., generation companies donot make enough money from the energy markets to coverboth their production costs and capital costs) and thus removethe incentives to build new generation? While the payment-cost minimization auction would in general reduce generationcompanies’ revenues from the ISO day-ahead market com-pared to the bid-cost minimization auction, the answer to thisquestion actually lies beyond the day-ahead market itself. It isacknowledged that the problem of generation companies notgetting enough revenue from the market to recover their costsalready exists in today’s markets with bid-cost minimization.We therefore believe the source of the problem is that theshort-term energy market alone is not sufficient to cover boththe production costs and capital costs of generators or to bringlong-term incentives for new generation. As a result, the rev-enue shortage problem should be resolved within a contextincluding capacity markets, long-term contracts, etc.

Other concerns, such as environmental impacts of the pay-ment-cost minimization auction, have also been raised.Example 1 shows that payment-cost minimization selectsUnit D with low price and high startup cost (e.g., an outdated

march/april 2008 IEEE power & energy magazine 35

If you participate in the FTR (CRR/TCC) market,you know that developing competitivetransmission-rights bidding strategies andportfolios is critical to your success.

That’s where Nexant’s HEDGE® can help.As the industry’s only comprehensive studypackage for FTR (CRR/TCC) auctions, thissoftware provides you with:

• All the power and flexibility needed forsimulating and analyzing auction clearingin the ISO FTR markets

• Our acclaimed auction modeling andcalculation engine—the same engine thatis supplied to all ISOs to routinely con-duct and/or analyze their FTR auctions

HEDGE package capabilities include:

• All auction timeframes (month, season,year, etc.)

• Peak, off-peak, 24-hour, and multi-calen-dar-periods

• Bids and offers with single price or curve

• Point-to-point and flowgate rights

• Obligations and options

• DC and AC network models

• Wide range of price and sensitivity results

• Corresponding allocation calculations

• Up-to-date modeling of all new FTRproducts

Are you an FTR/CRR/ TCCmarket participant?

Software & Information Systemsenergysolutions@nexant.com

http://www.nexant.com/products/erms_hedge.html

HEDGEThe comprehensive multifunctionFTR/CRR/TCC auction package.

®

coal-burn generator) rather than thesmall Unit C with high price and lowstartup cost (e.g., new-tech jet-enginegenerator). The concern is that Unit Dselected under payment-cost minimiza-tion might cause more pollution thanUnit C. While the selection of Unit Dshows nothing wrong from an optimiza-

tion point of view, a response to the envi-ronmental concern may be to include theenvironmental cost in the objective func-tion, and this depends on the marketrules. Still, the auction objective shouldbe to minimize the total payment cost,including environmental payment costsif required by market rules.

The above questions are examplesof the ongoing discussions. They showthat the issues of the auction objectiveare being brought into wider discus-sions. We believe these discussions tobe valuable to various groups.

ConclusionsWhether ISOs should minimize thetotal as-bid cost or the total paymentcost in their auctions is a crucial deci-sion for both consumers and generationcompanies since the selection willaffect both the bids and the MCPs. Fur-ther, the MCPs have financial impactson forward transactions outside the ISOmarkets as well as on long-term invest-ment decisions. Illustrative exampleshave demonstrated that for the same setof bids, payment-cost minimizationleads to payment reductions comparedto the bid-cost minimization techniquethat is currently used by ISOs. Consid-ering that the total value of electricitypurchased in these systems is in thebillions of dollars annually, even a verysmall percentage of payment-costreduction will result in millions of dol-lars in annual savings for consumers.This article summarizes our recentdevelopments in the solution methodol-ogy of payment-cost minimization andthe economic analysis of the two auc-tion methods. Furthermore, topics suchas revenue adequacy implications arebrought into discussion. Generallyspeaking, the research on the appraisalof the two auction methods is still at theearly stages, and we believe that a com-prehensive study of the two auctionmethods is highly valuable for bothresearchers and industrial practitioners.We hope this article can initiate moreserious debate among researchers andstakeholders as to which objectiveshould be used in ISO markets.

For Further ReadingJ.H. Yan and G.A. Stern, “Simultaneousoptimal auction and unit commitment forderegulated electricity markets,” Elect.J., vol. 15, no. 9, pp. 72–80, Nov. 2002.

See videos of FLIR’s thermal security cameras at www.fl ir.com

Heat Can’t Hide

See tonight – and every night – with FLIR

Copper theft from remote, and often unattended, power stations has reached

epidemic proportions. But the cost of a metals theft goes beyond the loss

in materials – companies lose operating revenue, must repair damaged

infrastructure, and deal with stations being off-line while they repair the damage.

But with FLIR thermal security cameras, even an unattended station isn’t

unprotected. FLIR’s thermal cameras provide a 24/7, single-camera security

solution. They help power facility personnel see clearly at night, keeping facilities

secure and guarding against theft by seeing an intruder’s body heat.

FLIR helps you stay secure at night – because no one can hide their heat.

The World Leader in Thermal Security Cameras

1.877.773.3547

VSR-6 at $3,499.00

(continued on page 92)

36 IEEE power & energy magazine march/april 2008

92 IEEE power & energy magazine march/april 2008

cale

ndar

PES meetingsfor more information, www.ieee.org/power

TTHE IEEE POWER ENGINEERINGSociety’s (PES’s) Web site (http://www.ieee.org/power) features a meet-ings section, which includes calls forpapers and additional information abouteach of the PES-sponsored meetings.

April 2008Transformation and Distribution Con-ference and Exposition, 21–24 April2008, Chicago, Illinois, USA, contactDonald A. Preston, +1 504 466 4235, fax+1 504 466 4235, e-mail d.preston@ieee.org or Tommy Mayne, Lacombe,LA, +1 504 427 3390, fax +1 985 8828059, t.w.mayne@ieee.org, http://www.ieeet-d.org/ (sponsored by PES).

July 2008PES General Meeting, 20–24 July,Pittsburgh, Pennsylvania, USA, con-tact General Chair David J. Vaglia,davevaglia@ieee.org, Technical Pro-gram Chair Kalyan Sen, senkk@ieee.org (sponsored by PES).

August 2008T&D Latin America, 13–15 August,Bogota, Columbia, e-mail tydla2008@ieee.org, http://www.ieee.org.co/~tydla2008 (sponsored by PES).

October 2008International Conference on PowerTechnology (POWERCON), 12–15October, New Delhi, India, contact Dr.Subrata Mukhopadhyay, +91 1123383778, fax +91 11 26170541,

pesrrap@ieee.org, http://www.ewh.ieee.org/r10/delhi/piconf.htm (cospon-sored by PES).

March 2009PES Power Systems Conferenceand Exposition (PSCE) , 15–18March, Seattle, Washington, USA,contact General Chair Hardev Juj,hsjuj@bpa.gov or co-chair MaxEmrick, memrick@ci.tacoma. wa.us,http:/ /www.pscexpo.com/ 2009/(sponsored by PES).

July 2009PES General Meeting, 26–30 July,Calgary, Alberta, Canada, contact Gen-eral Chair W.O. (Bill) Kennedy,b7kennedy@shaw.ca or Technical Pro-gram Chair Om Malik, maliko@ieee.org (sponsored by PES). p&e

Digital Object Identifier 10.1109/MPE.2008.916899

the business scene (continued from page 36)

P.B. Luh, W.E. Blankson, Y. Chen,J.H. Yan, G.A. Stern, S-C. Chang, andF. Zhao, “Payment cost minimizationauction for deregulated electricity mar-kets using surrogate optimization,”IEEE Trans. Power Syst., vol. 21, no. 2,pp. 568–578, May 2006.

P.B. Luh, Y. Chen, W.E. Blankson, Y.Chen, J.H. Yan, G.A. Stern, W.E. Blank-son, and F. Zhao, “Payment cost mini-mization with demand bids and partialcapacity cost compensations for day-ahead electricity auction,” Electric PowerNetwork Efficiency and Security, L. Miliand J.A. Momoh, Eds., to be published.

G.A. Stern, J.H. Yan, P.B. Luh, andW.E. Blankson, “What objective func-tion should be used for optimal auc-tions in the ISO/RTO electricitymarket?” in Proc. IEEE Power Engi-neering Society General Meeting,Montreal, Canada, June 2006.

F. Zhao, P.B. Luh, J.H. Yan, G.A.Stern, and S-C. Chang, “Payment costminimization auction for deregulatedelectricity markets with transmissioncapacity constraints,” IEEE Trans.Power Syst., to be published.

F. Zhao, P.B. Luh, Y. Zhao, J.H.Yan, G.A. Stern, and S. Chang,

“Bid cost minimization vs. pay-ment cost minimization: a gametheoretic study of electricity mar-kets,” in Proc. PES General Meet-ing, Tampa, FL, 2007.

AuthorsJoseph H. Yan is with SouthernCalifornia Edison.

Gary A. Stern is with SouthernCalifornia Edison.

Peter B. Luh is with the Universityof Connecticut.

Feng Zhao is a Ph.D. student at theUniversity of Connecticut. p&e