Colombian Firm Energy Market: Discussion and Simulation
Peter Cramton(joint with Steven Stoft and Jeffrey West)
9 August 2006
Outline
• Discussion of issues• Simulation of market
– Purpose– Model 1
• Historical prices• Simulated units
– Outline of Model 2• Historical prices and output• Actual units
– Outline of Model 3• Full simulation of auction and investment decision
• Conclusion
Issues
Issues
• Reducing risk in early auction years
• Further protection from insufficient competition
• Long lead-time projects
• Why not have a higher strike price?
• Repowering bids
Reducing risk in early years
• Early years of auction– Ceiling and floor on firm energy payment to
existing suppliers– Spread between ceiling and floor expand
each year– Spread starts at 0 (transition years)– Increases to
• Ceiling = 2 CONE• Floor = .5 CONE
Insufficient competition rule
• Add additional requirement to assure competition from non-dominant players
• At qualification, quantity of new projects from small players (less than 15% firm-energy market share) > 50% of required new firm energy
• Otherwise insufficient competition:– Auction held– New entry paid clearing price– Existing capacity paid 1.1 CONE
Long lead-time projects
• 4-year planning period may be too short for large hydro projects (6-8 years to build)
• Allow large hydro projects to lock in auction price from 4-year ahead auction seven years (or less) ahead
• Large hydro project is price taker• Decides after auction a fraction of its firm energy to lock
in at 4-year ahead auction price• Total quantity of firm energy in years > 4 that load
purchases is limited by a percent of new firm energy required in that year based on planning projections:
Years ahead: 7 6 5Percent limit: 40 50 60
Strike price
• Why not have a very high strike price?(US$250 or more)– Benefits of call option are largely lost
• Load hedge• Mitigation of market power in spot energy market
– No reason to set strike price higher than marginal cost of an expensive thermal unit
Repowering bids
• Easily accommodated in auction
• Two types:– Quick switchovers (down time less than 1
year)• Repower bid is a new entry bid and a conditional
retirement
– Extended down time (more than 1 year)• Retirement followed by new entry bid 1 or more
years later
Simulation
Purpose
Purpose
• Assess supplier risk
• Consider variations of market design
• Evaluate alternative auction parameters
Model 1
Model 1Historical prices, simulated units
• Sample: October 1995 through May 2006
• Scarcity hours: spot price > strike price
• One long dry period: thirteen months– 30 Mar 1997 to 21 Apr 1998
• One short period of high prices(start of market)– 21 Nov 1995 to 24 Dec 1995
Scarcity hours by year
Energy Year Mean Standard Deviation1995 (Oct-Nov) 92 90.26 24.36
1996 419 118.33 47.701997 2371 110.66 42.451998 3296 93.74 31.011999 02000 02001 02002 17 63.92 2.932003 02004 2 85.18 0.672005 0
2006 (Dec-May) 0Note: A scarcity hour is any hour in which spot price exceeds strike price.
Number of Scarcity Hours
Spot Price During Scarcity Hours(Jan. 2006 US$ per MWh)
Scarcity hours by month and year
Energy Year December January February March April May June July August September October November Total1995 - - - - - - - - - - 92 921996 416 1 2 4191997 5 62 60 74 153 663 703 651 2,3711998 724 744 672 744 412 3,2961999 02000 02001 02002 16 1 172003 02004 1 1 22005 02006 - - - - - - -Total 1,140 744 689 750 412 62 60 74 153 664 704 745 6,197
400-599 scarcity hours600 or more scarcity hours
Almost every hour is a scarcity hour in long dry periods.
Thermal percent of load
Energy Year December January February March April May June July August September October November Average1995 - - - - - - - - - - 15.3 14.7 15.01996 19.8 22.0 14.3 11.9 10.5 10.7 9.6 8.6 11.5 14.7 13.8 13.2 13.41997 14.1 10.8 12.6 14.0 14.9 17.3 16.0 17.8 20.2 26.9 29.0 32.2 18.81998 29.2 36.2 36.7 35.6 37.8 21.1 16.2 12.3 11.2 12.6 17.0 18.8 23.71999 18.0 18.9 17.5 16.0 17.5 18.5 16.7 19.3 18.9 26.1 20.0 23.4 19.22000 29.3 28.3 31.1 30.4 28.3 22.8 22.7 18.8 25.0 28.9 29.3 26.1 26.82001 19.2 25.9 29.9 31.1 32.5 29.3 26.1 21.7 21.2 21.9 19.9 24.5 25.32002 19.3 24.6 31.0 25.0 25.2 19.7 18.0 17.2 21.4 21.6 24.0 21.2 22.42003 27.6 25.6 24.7 27.2 23.7 21.7 17.7 14.4 18.3 21.8 19.2 16.8 21.62004 18.2 18.9 22.0 24.6 21.2 16.4 20.6 15.7 16.0 20.0 15.4 10.6 18.32005 16.7 18.1 18.1 17.8 21.1 15.2 17.1 20.1 24.2 26.6 16.6 15.1 18.92006 21.0 15.9 16.1 23.3 17.9 17.7 - - - - - - 18.6
Average 21.1 22.3 23.1 23.4 22.8 19.1 18.1 16.6 18.8 22.1 20.0 19.7 20.6
400-599 scarcity hours600 or more scarcity hours
Thermal share of load much higher in dry periods.Hydro share is still large in long dry periods.
Model 1: Thermal unit
• Random time until failure
• Random time to repair
• Both exponentially distributed– Long-run availability: 70% and 95%– Mean time to repair: 10 hours and 40 hours
• 1000 simulations over entire time period– Calculate distribution of net firm energy
payment
Firm energy payment
• All amounts in January 2006 US dollars
• Auction not modeled so assume payment
• Firm energy payment = $10.86/MWh– Should be $13.045– Exact value not relevant
Net firm energy payment
• Net firm energy payment= Firm energy payment+ reward for over performance− penalty for under performance
• In hours where spot price > strike price,Reward or penalty
= (Qactual – Qobligation)(Pspot – Pstrike)
• Qobligation = supplier’s share of load
Energy rents
$0
Unit’s marginal cost
Strike price
Peak energy rent
Energy rent
Unit does not operate
Forward energy contract
Call option (FEM)
0
5
10
15
20
25
Jan
2006
US
$ pe
r M
Wh
of f
irm e
nerg
y
1995(Oct-Nov)
1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006(Dec-May)
p=70%, r=10 hrs p=70%, r=40 hrs Annual PER
p=95%, r=10 hrs p=95%, r=40 hrs Median +/- 1 Std. Dev.
Long-run availability = p, Mean time to repair = rThermal annual net firm energy payment (thermal follows load)
Model 1 results: thermal
• Net firm energy payment roughly constant• Some variation in dry periods
– Standard deviation is small compared to mean– Variation greatest for unreliable units with long mean
repair times
• Slight reduction in dry periods (about 10%)– Thermals under perform on average
• Over perform in low-load conditions• Under perform in high-load conditions• Small positive correlation between price and load
Alternative obligation:Thermal constant
• Idea: Make obligation more consistent with unit’s actual dispatch
• Give thermal a constant obligation during scarcity hours (obligation = LR availability)
• Hydro follows residual demand(load minus thermal obligation)
• Can still treat as one product– Load following is not scare– Service is priced at zero in competitive market
0
5
10
15
20
25
Jan
2006
US
$ pe
r M
Wh
of f
irm e
nerg
y
1995(Oct-Nov)
1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006(Dec-May)
p=70%, r=10 hrs p=70%, r=40 hrs Annual PER
p=95%, r=10 hrs p=95%, r=40 hrs Median +/- 1 Std. Dev.
Long-run availability = p, Mean time to repair = rThermal annual net firm energy payment (thermal follows load)
0
5
10
15
20
25
Jan
20
06
US
$ p
er
MW
h o
f fir
m e
ne
rgy
1995(Oct-Nov)
1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006(Dec-May)
p=70%, r=10 hrs p=70%, r=40 hrs Annual PER
p=95%, r=10 hrs p=95%, r=40 hrs Median +/- 1 Std. Dev.
Long-run availability = p, Mean time to repair = rThermal annual net firm energy payment (thermal constant)
Model 1: Hydro unit
• Actual quantity of firm energy in dry period is a random variable (normal distribution)
• Unit sells its mean firm energy in dry period (mean availability = 30% or 50%)
• Actual firm energy has standard deviation(sd = 10% or 15%)– Note: Probably too high. Will rerun with
empirically fitted distribution from hydrology data from 1950s.
0
5
10
15
20
25
Jan
20
06
US
$ p
er
MW
h o
f fir
m e
ne
rgy
1995(Oct-Nov)
1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006(Dec-May)
p=30%, sd=10% p=30%, sd=15% Annual PER
p=50%, sd=10% p=50%, sd=15% Median +/- 1 Std. Dev.
Dry period availability = p, Standard deviation = sdHydro annual net firm energy payment (thermal follows load)
0
5
10
15
20
25
Jan
20
06
US
$ p
er
MW
h o
f fir
m e
ne
rgy
1995(Oct-Nov)
1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006(Dec-May)
p=30%, sd=10% p=30%, sd=15% Annual PER
p=50%, sd=10% p=50%, sd=15% Median +/- 1 Std. Dev.
Dry period availability = p, Standard deviation = sdHydro annual net firm energy payment (thermal constant)
0
5
10
15
20
25
Jan
20
06
US
$ p
er
MW
h o
f fir
m e
ne
rgy
Thermal follows load Thermal constant
1996 1997 1998 1996 1997 1998
p=70%, r=10 hrs p=70%, r=40 hrs Annual PER
p=95%, r=10 hrs p=95%, r=40 hrs Median +/- 1 Std. Dev.
Long-run availability = p, Mean time to repair = rThermal annual net firm energy payment
0
5
10
15
20
25
Jan
20
06
US
$ p
er
MW
h o
f fir
m e
ne
rgy
Thermal follows load Thermal constant
1996 1997 1998 1996 1997 1998
p=30%, sd=10% p=30%, sd=15% Annual PER
p=50%, sd=10% p=50%, sd=15% Median +/- 1 Std. Dev.
Dry period availability = p, Standard deviation = sdHydro annual net firm energy payment
Thermal-constant alternative
• No impact on risk• Thermal: Higher mean in dry period• Hydro: Lower mean in dry period• Obligation better matches actual dispatch
– Units enter spot market with balanced position– No incentive to exercise market power– With load-following approach:
• Hydro increases slope of supply curve (increasing price in high-load hours)
• Thermal bids higher (increasing low-load price)
Model 2
Model 2Historical prices, output; Actual units• Assume each unit sells its firm energy
certification(either reference or maximum for hydro)
• Calculate net firm energy payment for each unit in each hour– Aggregate over month– Aggregate over year– Aggregate over company’s portfolio
• Provides some insight on supplier risk
Model 3
Model 3Full simulation of auction and investment
• Can ask new questions– How does acquired firm energy differ from
firm energy target?– What is the impact of increasing the slope of
the demand curve around the target?
• Stationary model– Three project types: baseload, peaker, hydro– Baseload and peaker: Capacity, FC, VC– Hydro: Capacity, FC, Firm Energy
Conclusion
Conclusion• Call option reduces market risk
– Load is hedged from high spot prices– Hedge is not too costly for suppliers to offer
• Physical asset covers obligation
• Call option reduces supplier risk– Get nearly constant payment, rather than highly
variable peak energy rents• Call option improves spot market
– Mitigates market power problem during scarcity– Better spot market improves forward energy market
• Spot energy prices are more stable and predictable
• Thermal-constant obligation is better than load-following obligation