PSERC
Market Design and Gaming in
Competitive Electricity Markets
Shmuel S. Oren
University of California at Berkeley
Presented at the Iowa State University
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
PSERC
OBJECTIVE OF MARKET DESIGN
Develop a set of trading rules and procedures
so that when all market participants act
selfishly so as to maximize profit while
following the rules, the market outcome will
replicate the results of a benevolent central
planner with perfect information, or a
perfectly regulated monopoly
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
PSERC
POWER INDUSTRY RESTRUCTURING
Customers
Competitive
FERC
Regulated
State
Regulated
Demand
Management
Transmission
Micro-Grids
Retail
Providers
Distribution
Grid
Management
Power
Generation
Power Mkts.
& Reliability
SC
Grid
Reliability
ISO
LSE
February 4, 2005
PX
TO
All rights reserved to Shmuel Oren, UC Berkeley
SC
SC- -Scheduling
SchedulingCoordinator
Coordinator
PX
Power
Exchange
PX - Power Exchange
ISO
ISO- -Indep.
Indep.System
SystemOperator
Operator
TO
TO- -Transmission
TransmissionOwner
Owner
LSE
LSE- -Load
LoadServing
ServingEntity
Entity
Alternative Market Structures
CALIFORNIA
PJM
(before 2001)
SC
PX
SC
ISO
ISO
LSE
TO
U. K. (before 2001)
PX
LSE
February 4, 2005
TO
NORD POOL
SC
ISO
LSE
PX
PX
ISO
TO
LSE
All rights reserved to Shmuel Oren, UC Berkeley
TO
PSERC
The Basic Tradeoffs
„
„
PSERC
A decentralized electricity market is inherently incomplete.
The set of products traded does not capture all the
complexities of the system (e.g. ramp rate and reactive
power are not priced). Results in suboptimal dispatch and
uplifting of non-priced coordination costs, which are
susceptible to gaming.
A centralized market must use multidimensional bidding
schemes (e.g. energy, start up costs, ramp rate,
inflexibility etc.) and complex market clearing rules.
Susceptible to incentive compatibility problems (bidders
can game the market by revealing false information).
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
Market Design Components
Integrated Market Design
Market-Reliability Interface
Energy
Multiple settlements
Unit Commitment
Auction structure
Transmission
Congestion management
Property rights
Hedging
Ancillary services
Forward markets
Ensuring supply adequacy
Demand side participation
Retail Services
Reliability Differentiation
Exposure to price volatility
PBR for distribution companies
Capacity Expansion
Market Mitigation
Locational price signals
Planning Reserves
Transmission expansion
Monitoring
Automatic mitigation
Damage control caps
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
PSERC
PSERC
CONGESTION MANAGEMENT
All rights reserved to Shmuel Oren, UC Berkeley
FERC Division of Market Development, Dec. 19, 2001
Transmission Constraints in
ISO-NE
the
NWPA Contiguous U.S.
PSERC
Pacific DC Intertie
Northeast of Boston
NYISO
MAPP
CA/OR Interface
WY/ID Interface
Southwest MI
East NY
Central IA
SW CT Interface
PJM
ECAR
East KS/MO Interface
CA/MX
MAIN
West VA/PJM Interface
Southeast PA
RMPA
SPP
Central CA
Central MO
Southeast West VA
Northeast AZ
Southern CA
SERC
AZ/NM/SNV
FRCC
: Constraint and constrained flow direction
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
8
PSERC
Objectives of Congestion Pricing
„
„
„
„
„
Provide incentives to generators to account
for transmission constraints in their schedules
by direct assignment of congestion cost.
Provide price signals for efficient redispatch
due to transmission constraints
Create price signals for efficient use of the
transmission system
Create price signals for transmission
upgrades
Prevent gaming due to misallocation of
constraints cost
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
Locational Marginal Prices
$/MWH
$/MWH
S2
Export = Import
S1
p*
P1
p1
P2
Export
p2
Import
Power flow
D1
0
PSERC
q1 q1*
Region 1
0
q2* q2
Region 2
With transmission constraint
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
D2
General Definition of LMP
PSERC
The LMP is the cheapest cost of
delivering one additional MWh to a
location while respecting all limits
(including contingency limits)
Marginal value of transmission between
two locations = LMP difference
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
Alternative Measures of Congestion
PSERC
Cost and Corresponding Settlements
location 1
Export
location 2
Import
Redispatch
Social cost
Incremental cost
P2
E
A
P*
P1
Congestion rent
February 4, 2005
Export supply function
at location 1
B
Congestion
Relief
payment
C
D
Flow limit Unconstrained
Optimal export
Congestion relief
cost
Avoided cost function
at location 2
Export
Quantity
All rights reserved to Shmuel Oren, UC Berkeley
The case for socialization of
congestion relief cost
location 1
Net Supply
location 2
Net Demand
Price
LMP at 2
Shadow price
On line
Congestion
social cost
Congestion
Relief
payment
Export supply function
at location 1
Congestion
relief cost
LMP at 1
Avoided cost function
at location 2
Congestion
rent
February 4, 2005
Constrained Scheduled
flow
flow
All rights reserved to Shmuel Oren, UC Berkeley
Quantity
PSERC
PSERC
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
The DEC Game
Redispatch
Social cost
Price
P2
Export supply function
at location 1
A
B
P*
P1
PSERC
C
D
E
Congestion
Relief
payment
F
Congestion relief
cost
Avoided cost function
at location 2
Flow limit
Scheduled flow
Unconstrained optimal
export
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
Quantity
ENRON’s “Death Star” strategy
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
PSERC
The DEC Game on Path 26 in
California
BEFORE
February 4, 2005
AFTER
All rights reserved to Shmuel Oren, UC Berkeley
PSERC
NORTH2001
LOAD 20,700 MW
GEN 22,000 MW
1200 MW
720 MW
0
58
3750 MW
2200 MW
WEST2001
LOAD 3,700 MW
GEN 5,300 MW
M
W
0
31
M
W
LOAD, GENERATION
AND TRANSFERS BETWEEN
ERCOT REGIONS
ON PEAK SUMMER 2001
February 4, 2005
SOUTH2001
LOAD 33,000 MW
GEN 45,200 MW
All rights reserved to Shmuel Oren, UC Berkeley
PSERC
SIGNIFICANT CONSTRAINTS
SUMMER 2001 PSERC
BASED UPON THERMAL LIMITS ONLY
Will Vary Based Upon System Conditions &
GenerationFebruary
Dispatch
4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
ERCOT Initial Design
„
„
„
Zonal markets linked by CSCs (4 zones 4 CSCs)
QSEs submit balanced bilateral zonal schedules and zonal portfolio
balancing energy offers with resource specific premiums
All schedules accepted. Congestion relieved in two steps
„
„
„
PSERC
Interzonal (CSC) congestion relieved with counterflow from zonal
balancing energy offer stack (INCs and DECs).
Intrazonal (OC) congestion relieved (if possible) by dispatching local
balancing energy resources (redispatched resources paid market clearing
premiums if competitive solution exists, otherwise they get OOME
payments
Counterflow cost initially socialized with two “trigger points”
„
„
CSC congestion charges and TCRs to be activated 6 months after
cumulative counter flow cost reaches $20M over 12 month.
Direct assignment of OC congestion cost to be initiated 6 moths after cost
reaches $20M over 12 month period.
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
Observed Scheduling Patterns at PSERC
ERCOT During Aug. 2001
• Consistent overscheduling of generation in
the South Zone which resulted in significant
resource imbalance credits to QSEs from
ERCOT.
• Consistent overscheduling of load in the
North Zone which resulted in significant load
imbalance credits to QSEs from ERCOT.
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
ERCOT Daily Load Imbalance by Zone,
August 2001
PSERC
30%
North
South
West
25%
20%
15%
10%
5%
0%
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
-5%
-10%
-15%
-20%
-25%
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
23
24
25
26
27
28
29
30
31
ERCOT Daily Resource Imbalance by Zone,
August 2001
PSERC
20%
North
South
West
15%
10%
5%
0%
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
-5%
-10%
-15%
-20%
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
22
23
24
25
26
27
28
29
30
31
PSERC
QSE B Daily Load Imbalance, August 2001
350%
All Zones
North Zone
Percent Over (Under) Scheduled
300%
250%
200%
150%
100%
50%
0%
1
2
3
4
5
6
7
8
9
10 11
12
13 14
15 16 17
18 19
20
21 22
-50%
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
23 24
25
26 27
28 29
30
31
PSERC
Daily Costs for Local Congestion, CSC Congestion, and Total BENA
$14,000,000
$18,049,376
$14,486,872
$12,000,000
Local
BENA
CSC
$10,000,000
$8,000,000
$6,000,000
$4,000,000
$2,000,000
$0
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
($2,000,000)
August 2001
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
PSERC
Findings
• Six QSEs received more than $46 million in LI
credits for the 15 day period.
– August 14 and 15 accounted for 44% of LI credits to the six
QSEs.
– Three of the six QSEs received 91% of LI credits for the 15
day period.
• Eight QSEs received $48 million in RI credits
for the 15 day period.
– August 14 and 15 accounted for 38% of RI credits to the
eight QSEs.
– Three of eight QSEs received 95% of RI credits for the 15
day period.
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
Epilog
„
„
„
„
„
PSERC
Interzonal Congestion charges were activated in February 2002
Interzonal congestion uplift was virtually eliminated and the
DEC game between zones was suppressed.
The PUCT filed suite against the QSEs that collected BENA
settlements through over and under scheduling (for protocol
violation). An out of Court settlement was reached with several
of the QSEs (not including ENRON) who agreed to refund a total
of about $25 million dollars. The settlement is pending
commission approval.
Local congestion cost is still socialized and uplift cost is
escalating
New rule will institute a “Texas Nodal” system including direct
assignment of all congestion rents by 2006
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
Uplifted Local Congestion
Cost in ERCOT
PSERC
Local Congestion Costs (Million$)
$60
$50
$40
$30
$20
$10
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
Feb-03
Dec-02
Oct-02
Aug-02
Jun-02
Apr-02
Feb-02
Dec-01
Oct-01
Aug-01
$0
PSERC
Pricing Congestion the Right Way
„
„
„
Charge scheduled transactions the locational
price differences between injection and
withdrawal point OR
Charge scheduled transactions a congestion
fee that equals to their induced flow on
congested lines times the “shadow prices” on
these lines
Pay redispatched INCs at import location P2
and sell back energy to DECs at export
location P1
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
Congestion Rent is the Only Incentive
Compatible Congestion Charge
Redispatch
Social cost
Price
P2
Shadow
price
On line
P1
Export supply function
at location 1
A
C
B
D
E
Congestion
Relief
payment
Congestion relief
cost
F
Avoided cost function
at location 2
Flow limit
Congestion Rent
February 4, 2005
Scheduled flow
Quantity
Unconstrained optimal
export
All rights reserved to Shmuel Oren, UC Berkeley
PSERC
PSERC
ENERGY MARKETS
All rights reserved to Shmuel Oren, UC Berkeley
Economic Withholding by
Pivotal Supplier
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
PSERC
“Hockey Stick” Bidding in RT
Ballanacing Market
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
PSERC
The Texas Ice Storm
2/25-27/2003
„
„
„
„
PSERC
ERCOT had to procure 100% of the
balancing offer stack during several
hours
Market clearing price was set by 1MW
offered on a regular basis at $990 by a
QSE that offered the rest of its
balancing energy at $200/MW or less.
$17 million changed hands in few hours
One REP declared bankruptcy.
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
Price-Setting Bid Curve at
ERCOT, 2/25/2003
Bid Price ($/MWh)
$1,000
PSERC
$990
$800
$600
$400
$200
$149
$0
0
50
100
MW
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
150
ERCOT MCPE by Interval,
2/25/2003
PSERC
$1,000
$800
$600
$400
$200
$0
0:15
February 4, 2005
3:15
6:15
9:15
12:15
15:15
All rights reserved to Shmuel Oren, UC Berkeley
18:15
21:15
Why Mitigate
„
„
„
„
„
PSERC
Market power cannot be eliminated completely.
Because of the nature of electricity some suppliers
will have locational market power at some times
New entry takes time and while it happens suppliers
can extract unacceptable rents
Demand inelasticity can result in occasional market
failure (markets do not clear or suppliers can name
any price they want)
ISOs are bound by strict rules in acquiring balancing
energy and ancillary services that do not allow them
to respond to price
Demand inelasticity and ISO rigidity enables
predatory bidding strategies that are designed to
exploit market failure conditions
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
Alternative Approaches to
Automatic Market Mitigation
„
„
„
„
„
„
PSERC
Damage control offer caps
Exclusion from price setting
Out of merit (OOM) dispatch and RMR
dispatch
Must offer requirements
Mitigation of offers based on conduct and
impact test of individual bidders (NYISO)
Mitigation based on market failure tests (bid
sufficiency, pivotal bidder, hockey stick)
(ERCOT)
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
Hockey Stick Mitigation
in Texas
PSERC
(Adopted by the PUCT May 2003)
„
„
„
„
If balancing bid stack is exhausted, set
clearing price based on 95% of offer
stack
Mitigated MCPE = 95% price point X
1.5
Offers at prices above mitigated MCPE
paid as bid
Sunshine policy: disclose identity of
offerers above $300/MWh (“Sunshine is
the best disinfectant”)
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
Retrospective Impact Analysis of
Hockey Stick Mitigation in Texas PSERC
July 1, 2002 through March 31, 2003
Bid Curves above $900:
Average 90%, 95%, 99%, and 100% Points
$1,000
Houston Zone
North Zone
South Zone
West Zone
Price ($/MWh)
$800
$600
$400
$200
$0
0%
20%
40%
60%
80%
Portion of stack
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
100%
PSERC
ANCILLARY SERVICES
All rights reserved to Shmuel Oren, UC Berkeley
Ancillary Services
„
„
Automatic Generation Control (AGC)
Reserves with varying different
response time
„
„
„
„
„
Spinning Reserves
Non-spinning Reserves
Replacement Reserves
Voltage Support
Black Start Capability
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
PSERC
Procurement Auction Features PS
ERC
„
„
Key feature - downward substitutability
Basic Approaches:
„
Sequential auctions (substitution through
rebidding)
„
„
„
„
Auction order: Regulation, Spin, Nonspin, Replacement
Offer prices: Single, Simultaneous multiple, Sequential
Uniform market clearing price in each auction
Simultaneous auctions (substitution by ISO)
„
„
„
February 4, 2005
Objective function: Min social cost or Min procurement
cost
Settlement rules: MCP based on offer type, MCP based
on use, Marginal value, Pay as Bid.
Pricing to buyers: Marginal cost, MCP based on type
All rights reserved to Shmuel Oren, UC Berkeley
Example
PSERC
Two type example
„
Regulation 600 MW at $10 /MW
Spin
200 MW at $ 5 /MW
100 MW at $15 /MW
300 MW at $20 /MW
dRG=500
dSP=500
20
15
10
5
600
Regulation
February 4, 2005
700
200
500
Spin
All rights reserved to Shmuel Oren, UC Berkeley
Sequential Auction
PSERC
dRG=500
dSP=500
20
15
10
5
600
700
200
Regulation
500
Spin
Total Procurement Cost
= 500*10 + 500*20
= $15,000
Revealed Social Cost
February 4, 2005
= $10,500
All rights reserved to Shmuel Oren, UC Berkeley
Simultaneous Auction
Minimum Social Cost Selection
dRG=500
20
15
10
5
PSERC
dSP=500
pRG = 20
pRG = 15
pSP = 20
600
Regulation
700
200
500
Spin
Settlement Rule
Procurement Cost ($)
Misrepresent quality
Highest Accepted Bid 16,500
Incentive compatible
Marginal Value
20,000
Pay as Bid
10,500 (Soc. Cost) Misrepresent cost
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
Rational Buyer Auctions
(Min procurement cost)
dRG=500
PSERC
dSP=500
–
20
15
10
5
+
600
Regulation
700
200
500
Spin
Procurement
Cost ($) Social Cost ($)
Saves (black):4500
14,000Spends (red): 2000 11,000
Difference: 2500
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
Price Reversals in New England Under
PSERC
Sequential Auction
1.1. Operating Reserves Market Flaws: TMSR, TMNSR,
TMOR.
During June and July the ISO identified design flaws in the operating
reserve markets of Ten Minute Spinning Reserve (TMSR), Ten Minute
Non-Spinning Reserve (TMNSR), and Thirty Minute Operating Reserve
(TMOR). The operating reserve design flaws resulted in markets that
are not workably competitive. In a competitive market, higher quality
goods should command higher prices. In the operating reserves
markets prices did not appear to be reflective of costs and lesser
quality reserves received higher prices. For example, Thirty Minute
Operating Reserve (“TMOR”) is less useful than Ten Minute NonSpinning Reserve (“TMNSR”), Ten Minute Spinning Reserve (“TMSR”),
or Energy. However, at times, TMOR prices exceeded the prices for all
three of the other products.
[From ISO New England Monthly Market Report, June 1999]
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
Price Reversals in California Under
PSERC
Rational Buyer Procurement
„
„
Price reversals have been reported in the CAISO ancillary
service market under the rational buyer auction
On March 20, 2000 in HE 19, there was a spike in the
Replacement Reserve price. The following are published
prices for NP15:
Reg-Up: P1 = 18.44 $/MW
Spin: P2 = 35.97 $/MW
Non-Spin: P3 = 18.00 $/MW
Replacement: P4 = 198.98 $/MW
Note that P4 > P3, P2,P1. Also P2 > P1.
„
A similar pattern in AS prices was observed on Feb 29,2000
in HE18 when replacement reserve prices were 122$/MW
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
NP15 A/S Price Reversal
Frequency in 2000
PSERC
70%
60%
50%
40%
30%
20%
10%
0%
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
If Reg = 4, Spin = 3, Non-spin = 2, Repl. Res = 1,
a price reversal occurs if Price_i > Price_j for i < j where i,j denote the type of A/S
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley
Price Reversals in ERCOT AS Markets
PSERC
Under Sequential Auction
Responsive Reserves
Average Daily Prices
Regulation Up
Average Daily Prices
$/M W
1000
$/MW
Max Price
Min Price
Ave Price
$60
$50
Max Price
Min Price
Ave Price
800
$40
600
$30
400
$20
$10
200
$0
0
10
20
30
40
0
50
0
Days Between August 1 - September 21
10
20
Non Spinning Reserves
Average Daily Prices
$/MW
1000
Max Price
Min Price
Ave Price
800
600
400
200
0
0
February 4, 2005
30
40
Days between August 1 - September 21, 2001
10
20
30
40
Days
between
August
- September
All rights
reserved
to 1Shmuel
Oren, 21,
UC2001
Berkeley
50
50
Conclusions
„
„
„
„
„
PSERC
If gaming is possible it will happen
Market design must be proactive in anticipating
perverse incentives and gaming opportunities
Structural solutions and incentive mechanisms are
preferred but not always possible
Ex- post market power mitigation too cumbersome
and ineffective
On line ex-ante intervention is needed to mitigate
market failure and regulate behavior but distortions
should be minimized.
February 4, 2005
All rights reserved to Shmuel Oren, UC Berkeley