Get documents about "PSerc - Power Systems Engineering Research Center
Document Sample
Day Ahead and Real Time
Spot Markets
Shmuel S. Oren
University of California at Berkeley
Presented at the PUCT Public Workshop on
Elements of Market Design
Austin, Texas, November 1, 2002
The SMD NOPR Prescription
• Transmission Provider will operate the
transmission system using spot markets
• To manage congestion
• To balance generation and load
• Spot Markets support bilateral and spot
transactions
November 1, 2002 Shmuel Oren, UC Berkeley
...
Electric System and Market Time Line
PHYSICAL DELIVERY REAL TIME
Priority Service Contracts
Day Hour Minutes Seconds Cycles
100
Voltage Control/Dynamics
Price Menus
Fixed Fee
Product line and
Automatic
Service
Price Menue Design,
Control
Operation
Subscription Managmnt Transient Standby
and Billing.
<10
Frequency control
SYSTEM OPERATION
Settlement
Insurance
Ex-post
Service Options Design Load Balancing
Shmuel Oren, UC Berkeley 15-5 5-1
Premium Calculation (Spinning Reserves)
Service Reliabilty Forcst
Financial Settlmnt
Computation
Economic Dispatch
Trading Rules, (Congestion managment)
Scheduling
Spot Market
Bidding and
Bidding Protocols,
Trading
Online Comm. Network,
Optimal Bidding Software,
Bid Selection Software,
Online Financial Settlmnt. Unit Commitment
(scheduled congestion
managment)
Standardized Contracts,
Contracts Contracts
Future
Clearing House Service,
EFAs
Calls and Puts
On Line Trading Tech.,
EDI Capability, Hedging
Maintenance
Standards and
and Valuation Software.
Protocols
Forward
Years/Months
Contingency Analysis
CFDs
Hedging Instruments
Valuation Software
Fuel Inventory Mang.
Fuel Procurement
ASSET PROVISION
Trading Rooms,
Negotiation, OTC Trading,
Electronic Buletin
Boards Contract Bidding
Long Term Contracts
Design Principles
and Limits
Planning / Building
Load Forecasting
Investment
and Analysis
Portfolio Analysis,
Decades/Years
CAPM, Brokerage Risk and Decision Analysis
November 1, 2002
and Financing
Institutions
Capacity planning Tools
(Over/Under, Uplan)
Technology
Instruments
Supporting
Technology
Supporting
Tools and
Physical
Tools and
System
Market
Means
Form
The
SMD Spot Market
Characteristics
• Operated in two time frames
• Day-ahead
• Real-time
• Prices and schedules are based on
participants’ bids
• Bidding is voluntary, except when required
for market power mitigation
• Prices are market clearing
November 1, 2002 Shmuel Oren, UC Berkeley
Specific Characteristics of
Proposed Day-Ahead Market
• Physically feasible security constrained bid based
dispatch of energy and reserves
• Voluntary central unit commitment based on multi-
dimensional offers specifying costs and constraints of
each resource
• Self-commitment allowed
• Virtual bidding (identified as such) allowed
• Bilateral transactions allowed (do not have to submit
bids) but are subject to congestion charges based on
day ahead LMP
• CRRs in the form of PTP contracts, PTP options or
flowgate rights settled based on day ahead LMP.
• Two settlements: DA transactions settled at DA LMP
whereas RT deviations from DA schedules settled at RT
LMP.
November 1, 2002 Shmuel Oren, UC Berkeley
Important Observations
• Congestion management and transmission
pricing are independent from the organization of
spot markets
• It is possible to have LMP-based congestion
management and transmission pricing with only
a single settlement RT spot market (Victoria
Pool)
• It is possible to have a mandatory day ahead
market with central unit commitment and no
locational pricing (original UK system)
November 1, 2002 Shmuel Oren, UC Berkeley
Where is ERCOT?
• The balancing market (especially after relaxing
the balanced schedule requirement) qualifies as
a RT single settlement spot market but it is not
fully efficient because of limited information
exchange (portfolio bidding, no binding resource
plan)
• Ancillary service market is a limited DA market
with the RPRS procurement representing a
simplified unit commitment for covering
expected imbalances.
November 1, 2002 Shmuel Oren, UC Berkeley
Arguments for a Day Ahead Market
• Operator needs to know in advance what are the balancing needs
• With relaxed balanced schedules early knowledge of balancing
needs will enable more efficient deployment of resources (efficient
procurement of RPRS)
• DA market enables differential pricing of planned and unplanned
imbalances so as to reflect the higher cost of responding to
unplanned imbalances
• Provides an additional trading forum for market participants who
have predictable imbalances to cover their positions.
• Mitigates market power in RT
• Facilitates demand side response from load that cannot respond in
real time
• Enhances price discovery and provide a better (less volatile) bench
mark for contract settlements than RT prices.
November 1, 2002 Shmuel Oren, UC Berkeley
Scope of the DA market
• All DA transactions are bilateral, with
possible private exchanges (ERCOT, NETA)
• Voluntary RTO operated DA market.
• Mandatory RTO operated DA market. All
transactions must submit bids. No
guaranteed dispatch. Bilateral contracts are
financial hedges (CFD)
November 1, 2002 Shmuel Oren, UC Berkeley
Physical feasibility
• DA transactions can violate known
constraints (California PX, ERCOT step 1).
Constraints satisfied through RT balancing
market.
• DA transactions must satisfy security
constrained dispatch (PJM, NY)
November 1, 2002 Shmuel Oren, UC Berkeley
Settlements
• Single settlement: DA settled at RT prices
(Victoria pool)
• Two settlement: DA clearing prices
financially binding. Deviations from DA
schedules settled at RT prices
November 1, 2002 Shmuel Oren, UC Berkeley
Virtual Bidding
• Virtual DA bids are used to arbitrage DA
prices against RT prices
• Virtual bids must be declared
• Virtual bids participate in setting DA LMP
(PJM). Enables gaming of FTR values (virtual
bids can be used to congest or decongest
PTP)
• Virtual bids excluded from setting DA LMP
November 1, 2002 Shmuel Oren, UC Berkeley
Unit Commitment Considerations
• Central optimization can improve economic
efficiency
• Cost not one dimensional. Marginal cost affected
by schedule (e.g. startup)
• Without constraint information dispatch may be
infeasible resulting in deviations
• Constraints information can be gamed
• Optimal solution is not perfect and may not be
unique. May result in inequity
November 1, 2002 Shmuel Oren, UC Berkeley
Unit Commitment Alternatives
• Self commitment. Energy only bids in DA market
(CalPX)
• Self-commitment with energy only bids. Strips
and minimum daily revenue offers allowed.
(Spanish market)
• Central unit commitment of all resources
(including bilateral transactions) all bilateral day
ahead trading is financial (NY, old UK system)
• Voluntary unit commitment. Bilateral
transactions do not have to bid and self
committed energy bids accepted
November 1, 2002 Shmuel Oren, UC Berkeley
Unit Commitment Alternatives
• Unit commitment applied only to “net short”
(CAISO MD02)
• Technical parameters fixed for long time periods
• Treatment of ancillary services
• Joint optimization of energy and reserves (NY)
• Separate AS market (PJM - regulation)
• Treatment of transmission constraints
• Transmission constraints included in UC (NY)
• Loacational vs. systemwide reserves
• UC ignores transmission constraints
November 1, 2002 Shmuel Oren, UC Berkeley
Unit Commitment Problem
Minimize [ St at e T ransit io n Cost + Energ y co st ]
{ St at e, Out put } Time Resource
subject to: Dem and con st raint
Spinning reserve require ment
Int e rt emp o ral con st raint s ( e. g . Ramp in g )
Co mm it ment / availab ilit y con st raint s
(Transmission constraints)
November 1, 2002 Shmuel Oren, UC Berkeley
UNIT COMMITMENT PROBLEM
Transition Cost Commitmen t state {0,1} Ge neratio n cost
T I
Minimize [Sit (x (i,t 1) ,x it ) Cit (pit )]
Resource i
Ti me t
{xit ,p it }
t 1 i 1
Ge neratio n l evel
I
subject to p it Dt , t 1,...,T Demand constra int
i 1
I
pi
i 1
max
ui (x it ) Rt , t 1,...,T Spinning reserv e requirement
Resource fra cti on avai la ble given state
min max
pit ui (xit ) pit pit ui (x it ),i 1,...,I,t 1,...,T
Intertemporal
constraints
xi Xi , i 1,..., I Where xi (xi1 , xi 2 ,..., xiT ) Commitment
constraints
November 1, 2002 Shmuel Oren, UC Berkeley
DAY AHEAD BIDDING IN THE
UK SYSTEM
OFFER DATA PROVIDED TO GRID OPERATOR BY EACH GENERATOR, BY
10:00 AM ON EACH DAY:
• OFFERED AVAILABILITY (MAX MW FOR EVERY MINUTE)
• PRICES (SUPPLY CURVE)
• OPERATING CHARACTERISTICS:
• Run-up rates
• Run-down rates
• Synchronizing generation
• Minimum output
• Other
• DECLARED INFLEXIBILITY:
• Max On/Off between daily peaks
• Fixed output level at specified time intervals
• Minimum output level at specified time interval
November 1, 2002 Shmuel Oren, UC Berkeley
Description of the CALECO
System Used in Pool Simulations
U nit N ame M a x L o ad Mi n Lo a d S tartu p $ Min Up Min Down Fuel
Col-S tm 1000 25 0 100000 12 0 48 Coal
Stm1 750 50 15000 24 48 Gas
Stm5 330 50 15000 6 3 Gas
Stm6 3 30 50 15000 6 3 G as
Stm7 3 40 85 15000 6 3 G as
QF 1000 1000 N/A N/A N/A QF
P o n d Hy d ro 1 5 00 1 0 1 1 P o n d Hy d ro ( L im i te d )
ROR Hydro 9 00 0 0 1 1 ROR Hydro
Nuke 2000 0 200000 1 1 Nuke
CT s 1 0 00 0 0 1 1 D istil la te
Econ01 500 0 0 1 1 Transaction at $17.5/MWH
Econ02 5 00 0 0 1 1 Transaction at $30/MWH
L oad assum ption s: M a xi m u m lo a d = 9 7 4 9 M W
Minim um load = 4990 MW
T o ta l l o a d = 1 1 7 8 8 6 1 M W H
Load factor = 74%
168 hours
November 1, 2002 Shmuel Oren, UC Berkeley
LOAD NET OF BASE LOADED
RESOURCES AND HYDRO
100 00
hav y
Peak -S e H dro
900 0
R v y
un-of-Ri er H dro
800 0
700 0
600 0
LOA D /GEN ER A TION ( MW)
500 0
400 0
COAL-STEAM
300 0
QF
200 0
100 0 U
N KE
0
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160
H O URS
November 1, 2002 Shmuel Oren, UC Berkeley
November 1, 2002 Shmuel Oren, UC Berkeley
Profit Volatility in Power Pools
with Central Unit Commitment
DEVIATION FROM AVERAGE PROFIT ON DIFFERENT RUNS OF UNIT COMMITMENT
100
5
6
80 7
8
9
Deviation From Average (K$)
60 11
Average Profit/Cost (K$)
40
708 739 31 177 177 332 116 2,638 20,317
20
0
-20
-40
Econ01 Col-Stm Econ02 Stm6 Stm5 Stm1 Stm7 QF TOT Cost
Dispatchable Resources
November 1, 2002 Shmuel Oren, UC Berkeley
Incentive compatibility problems
• Optimum flat and dispatch decisions sensitive to small
changes in parameters
• Incentives for bidders to try manipulating the dispatch
and prices through declaration of constraints (such
manipulation was prevalent in the UK before NETA)
• Optimizing unit commitment with misrepresented
generator characteristics undermines efficiency objective
• Unit commitment does not account for all the variables
that affect a generator’s decision to run (e.g. long term
fuel supply contracts)
• Day ahead unit commitment does not capture
optimization over longer planning horizons (e.g. hydro
scheduling)
November 1, 2002 Shmuel Oren, UC Berkeley
Gaming Ramp Constraints
OffPeak Peak
Demand 1000MW 3000MW
Generator A Bids 1000MW at $10/MWh 1000 MW at $10/MWh
Generator B Bids 2000 MW at $15/MWh 2000 MW at $15/MWh
Generator C Bids 2000 MW at $25/MWh 2000 MW at $25/MWh
Least cost dispatch with no intertemporal constraints
Clearing Price Generator A Generator B Generator C
OffPeak $10/MWh 1000MW 0 0
Peak $15/MWh 1000MW 2000MW 0
Financials Soc. Cst=$600,000 Prft.=$60,000 Prft.=0 Prft.=0
Least cost dispatch with zero ramp rate constraint by Generator B
Clearing Price Generator A Generator B Generator C
OffPeak $15/MWh 0 1000MW 0
Peak $25/MWh 1000MW 1000MW 1000MW
Financials Soc. Cst=$780,000 Prft.=$180,000 Prft.=$120,000 Prft.=0
•Clearing prices in each time period are not the correct marginal costs
•Uniform price settlements based on the clearing price in each period
create perverse incentive to impose ramp constraints
November 1, 2002 Shmuel Oren, UC Berkeley
Conclusions
• Day ahead markets could be beneficial
• The devil is in the details
• Central unit commitment creates gaming
opportunities
• Voluntary DA market with self-
commitment is a good compromise
• Virtual bidding is a good idea but should
not be allowed to set DA LMP.
November 1, 2002 Shmuel Oren, UC Berkeley
November 1, 2002 Shmuel Oren, UC Berkeley
