Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years
-
Upload
drake-garner -
Category
Documents
-
view
23 -
download
1
description
Transcript of Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years
![Page 1: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/1.jpg)
Using DOASA to investigate recent
trends in the NZ electricity market’s
management of dry years
Presented to EPOC Winter Workshop
July 2014
![Page 2: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/2.jpg)
Motivation• 2009 Ministerial Review• Electricity Industry Act 2010• To examine impact of measures on
management of dry years• We studied 2008, 2012 and 2013
![Page 3: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/3.jpg)
Measures undertaken• Sale of Whirinaki• Reserve energy scheme abolished• Customer Compensation Scheme• Scarcity pricing• Stress testing• Physical asset swaps between SOEs• Virtual asset swaps between SOEs• Electricity futures market
![Page 4: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/4.jpg)
DOASA• Variant of the SDDP algorithm • Aims to optimise the scheduling of hydro
and thermal generation resources in the face of uncertain future inflows
• Developed by Stochastic Optimization Limited
• A version is available as part of the Authority's suite of EMI tools
![Page 5: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/5.jpg)
Stagewise dependence of inflows (1)• Physical inflows exhibit stagewise
dependence• Inflows in one period partly depend on
inflows in earlier periods• Important in order to correctly account for
the probability of dry (and wet) sequences
![Page 6: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/6.jpg)
Stagewise dependence of inflows (2)• Some SDDP implementations use linear
Periodic Autoregressive (PAR) models to represent stagewise dependence
• DOASA approximates this by artificially spreading (increasing variance of) inflows
• Dependent Inflow Adjustment (DIA) feature automatically spreads inflows for each catchment consistent with accumulation over a specified number of weeks
![Page 7: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/7.jpg)
Water Values• DOASA yields marginal water values for
each hydro storage reservoir• Function of reservoir storage and date• Represent the expected value of holding
back a unit of water for release in some future period
• And hence the opportunity cost of instead releasing that unit of water now
• Balance risk of spill and running out of water
![Page 8: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/8.jpg)
Two approaches used• Compare the water values implied by
actual reservoir levels (under various hydro policies) with the prices that actually occurred
• Simulate operation with actual inflows (under various hydro policies) and compare simulated reservoir levels with the levels that actually occurred
![Page 9: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/9.jpg)
Implied water values approach• Water values depend on model inputs
(thermal prices and shortage prices)• Somewhat subjective• Tuning factors required to match market
prices• We chose to scale the demand• Manually spread inflows – 2x• Observed trends are interesting
![Page 10: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/10.jpg)
Simulated reservoir levelsapproach• Depend more on relativity between water
values and thermal/shortage prices• Actual prices less important• Less dependent on market dynamics• More objective• Model risk aversion by adding one or more
fictitious dry years when creating policies• DIA accumulation over 4 weeks
![Page 11: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/11.jpg)
Implied water values
![Page 12: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/12.jpg)
Water values vs spot price2008: Spreading factor 2; 110% Demand
01Jan 01Apr 01Jul 01Oct 01Jan
0
100
200
300
400
500
600
700
$/M
Wh
Lake Pukaki water v alue
Lake Pukaki water v alue smoothed
Benmore Av erage Daily Spot Price
Benmore Maximum Daily Spot Price
![Page 13: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/13.jpg)
SI water values vs SI storage2008: Spreading factor 2; 110% Demand
2525
50
50
50
50
75
75
75
75
100
100
100
100
150
150
150
150
200
200
200
200
300
300
300
400
400
400
% s
tora
ge
01Jan 01Apr 01Jul 01Oct 01Jan
0
25
50
75
100
Water v alue contours Actual storage Av erage storage
![Page 14: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/14.jpg)
Water values vs spot price2012: Spreading factor 2; 120% Demand
01Jan 01Apr 01Jul 01Oct 01Jan
0
100
200
300
400
500
600
$/M
Wh
Lake Pukaki water v alue
Lake Pukaki water v alue smoothed
Benmore Av erage Daily Spot Price
Benmore Maximum Daily Spot Price
![Page 15: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/15.jpg)
SI water values vs SI storage2012: Spreading factor 2; 120% Demand
50
50
50
75
75
75
75
100100
100
150
150
150
150
200
200
200
300
300
300
400
400
400
% s
tora
ge
01Jan 01Apr 01Jul 01Oct 01Jan
0
25
50
75
100
Water v alue contours Actual storage Av erage storage
![Page 16: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/16.jpg)
Water values vs spot price2013: Spreading factor 2; 120% Demand
01Jan 01Apr 01Jul 01Oct 01Jan
0
100
200
300
400
500
600
700
800
900
1000
$/M
Wh
Lake Pukaki water v alue
Lake Pukaki water v alue smoothed
Benmore Av erage Daily Spot Price
Benmore Maximum Daily Spot Price
![Page 17: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/17.jpg)
SI water values vs SI storage2013: Spreading factor 2; 120% Demand
50
50
50
75
75
75
75
100100
100
150
150
150
150
200
200
200
300
300
300
400
400
400
% s
tora
ge
01Jan 01Apr 01Jul 01Oct 01Jan
0
25
50
75
100
Water v alue contours Actual storage Av erage storage
![Page 18: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/18.jpg)
Best fit• 2008 – 110% demand• 2012 & 2013 – 120% demand• Implies market was more risk averse in
2012 and 2013 than in 2008
![Page 19: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/19.jpg)
Other observations• Spot prices move ahead of the movement
in model water values• Market is anticipating movements in
storage to some extent• Better information than the model?• Partly due to stagewise dependence?
![Page 20: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/20.jpg)
Simulated reservoir levels
![Page 21: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/21.jpg)
Simulation results2008: One fictitious dry year policy
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
% s
tora
ge
Lake Hawea
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
125
150
% s
tora
ge
Lakes Manapouri Te Anau
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
% s
tora
ge
Lake Pukaki
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
% s
tora
ge
Lake Taupo
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
125
% s
tora
ge
Lake Tekapo
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
% s
tora
ge
NZ
Actual storage Av erage storage Simulated storage
![Page 22: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/22.jpg)
Simulation results2008: Three fictitious dry year policy
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
% s
tora
ge
Lake Hawea
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
125
150
% s
tora
ge
Lakes Manapouri Te Anau
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
% s
tora
ge
Lake Pukaki
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
% s
tora
ge
Lake Taupo
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
125
% s
tora
ge
Lake Tekapo
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
% s
tora
ge
NZ
Actual storage Av erage storage Simulated storage
![Page 23: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/23.jpg)
Simulation results2012: One fictitious dry year policy
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
% s
tora
ge
Lake Hawea
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
125
150
% s
tora
ge
Lakes Manapouri Te Anau
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
% s
tora
ge
Lake Pukaki
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
% s
tora
ge
Lake Taupo
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
% s
tora
ge
Lake Tekapo
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
% s
tora
ge
NZ
Actual storage Av erage storage Simulated storage
![Page 24: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/24.jpg)
Simulation results2012: Three fictitious dry year policy
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
% s
tora
ge
Lake Hawea
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
125
150
% s
tora
ge
Lakes Manapouri Te Anau
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
% s
tora
ge
Lake Pukaki
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
% s
tora
ge
Lake Taupo
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
% s
tora
ge
Lake Tekapo
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
% s
tora
ge
NZ
Actual storage Av erage storage Simulated storage
![Page 25: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/25.jpg)
Simulation results2013: One fictitious dry year policy
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
% s
tora
ge
Lake Hawea
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
125
150
175
% s
tora
ge
Lakes Manapouri Te Anau
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
125
% s
tora
ge
Lake Pukaki
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
% s
tora
ge
Lake Taupo
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
% s
tora
ge
Lake Tekapo
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
125
% s
tora
ge
NZ
Actual storage Av erage storage Simulated storage
![Page 26: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/26.jpg)
Simulation results2013: Three fictitious dry year policy
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
% s
tora
ge
Lake Hawea
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
125
150
175
% s
tora
ge
Lakes Manapouri Te Anau
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
125
% s
tora
ge
Lake Pukaki
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
% s
tora
ge
Lake Taupo
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
% s
tora
ge
Lake Tekapo
01Jan 01Apr 01Jul 01Oct 01Jan0
25
50
75
100
125
% s
tora
ge
NZ
Actual storage Av erage storage Simulated storage
![Page 27: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/27.jpg)
Closest match(Total NZ storage)
• 2008 – One fictitious dry year policy• 2012 & 2013 – Three fictitious dry year
policy• Again implies market was more risk averse
in 2012 and 2013 than in 2008
![Page 28: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/28.jpg)
Other observations• Individual reservoir levels depart from the
simulated levels much more than the aggregated level …
![Page 29: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/29.jpg)
Lake Pukaki and Lake Tekapo (1)• Lake Pukaki and Lake Tekapo in 2008• Both lakes controlled by Meridian• Actual levels quite similar • SPECTRA? - lumped island reservoirs• Heuristic to split out reservoir policies• Model runs Lake Tekapo almost empty!• Tekapo directly supplies 185MW of plant• Pukaki directly supplies 1550MW of plant
![Page 30: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/30.jpg)
Lake Pukaki and Lake Tekapo (2)• Similar pattern in 2012• After Tekapo stations sold to Genesis• Not so obvious in 2013• Disrupted by canal outages?
![Page 31: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/31.jpg)
Lake Hawea• 2008 - Lake Hawea drawn down rapidly• Almost empty through July – August• Contact under-hedged?• 2012 – lower inflows April through August• Yet more gradual draw down• Contact better positioned?• Model holds Hawea high (valuable)• Only 13% of Clutha inflows are controllable• Compare: Waitaki 70%; Waikato 77%
![Page 32: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/32.jpg)
Lakes Manapouri/Te Anau• Model runs lakes lower than in reality• Operating rules allow temporary
excursions above main operating range• Model has hard limit at top of main
operating range• Hence model overstates risk of spill• Should model a higher limit to better
approximate the rules
![Page 33: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/33.jpg)
Some questions• Management of individual reservoirs under
the market deviates markedly from DOASA’s supposedly optimal strategy
• Does this imply a loss of welfare?• How material is it?• Would a greater range of hedge products
better align individual company incentives with total welfare maximisation?
• What sort of products?
![Page 34: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/34.jpg)
How much risk aversion is “optimal”?• Theoretically a risk neutral policy has the
lowest expected cost• This should be the case over a very large
number of Monte Carlo simulations (with inflow spreading)
• This approximates the population the policy is optimised for
• But …
![Page 35: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/35.jpg)
Expected cost of 2008 policies 400 Monte Carlo simulations
![Page 36: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/36.jpg)
Why did risk neutral policy not minimise cost?• Algorithm was terminated after 300
iterations so not completely converged?• 400 Monte Carlo simulations - not a large
enough statistical sample to approximate the entire population?
![Page 37: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/37.jpg)
What about historical simulations?
![Page 38: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/38.jpg)
Expected cost of 2008 policies Historical simulations
![Page 39: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/39.jpg)
Expected cost of 2012 policies Historical simulations
![Page 40: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/40.jpg)
Expected cost of 2013 policies Historical simulations
![Page 41: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/41.jpg)
Observations• Even “worse” than Monte Carlo• Significant load shedding cost for risk
neutral policy • Risk neutral policy clearly not optimal
![Page 42: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/42.jpg)
Why?• Inflow spreading technique not adequately
accounting for risk of extended dry sequences?
• 82 historical inflow sequences only a small sample of entire population
• Reduced set of 20 historical inflow years used to generate policies did not adequately capture some of the driest historical sequences in full 82 year set?
![Page 43: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/43.jpg)
So how much risk aversion is “optimal”?• 2008 & 2013 – load shedding avoided and
cost minimised for one fictitious dry year policy
• 2012 – load shedding avoided and cost minimised for two fictitious dry year policy
• But 2012 & 2013 simulations look like a three fictitious dry year policy (best match)
• Implication: market might be overly risk averse now?
![Page 44: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/44.jpg)
Conclusions• Market appears significantly more risk averse
since Authority’s security of supply measures• Market managed 2012 and 2013 dry year
events more effectively than in 2008• Maybe industry as a whole is now treating
hydro risk a little too conservatively?• Management of individual hydro reservoirs
appears to deviate significantly from welfare maximising strategy
• Perhaps new hedge products would help this
![Page 45: Using DOASA to investigate recent trends in the NZ electricity market’s management of dry years](https://reader038.fdocuments.net/reader038/viewer/2022110405/5681331e550346895d99ea14/html5/thumbnails/45.jpg)
Questions … ?