Market Performance: a Hindcasting PerspectiveEPOC Jul 2014Version 2.1

Grant Telfar, Meridian Energy

July 2014

Introduction & Background

• With market listing complete additional scrutiny of industry and company performance is likely:– As well as a desire to understand and analyse key

industry drivers and their impact.

• Hindcasting is an analytical approach that can be applied to examine how decision support tools, analysis, and key assumptions performed after the fact:– A priori we forecast a range of future outcomes.– Ex post we observe actual market outcomes.

• Meridian has undertaken five hindcast exercises in recent years – the first in 2008 and most recently in early 2013:– Hindcasting is seen by Meridian as part of good

internal management.

• In early 2013, Meridian undertook an updated hindcasting exercise:– In particular an examination of the 2012 year.– Seeking to isolate individual forecast assumptions,

their impacts, and to test their relevance.– Seeking to embed the hindcasting discipline within

Meridian to compliment its existing self-examination process and metrics.

2

• Over recent years the NZ electricity industry has been buffeted by unprecedented levels of dissatisfaction from the wider public, policy makers, consumer groups, and regulators.

• Dissatisfaction has been particularly focused on market and regulatory structures and intensified over the 2008-13 period due to:

− Winter 2008: the 4th ‘dry winter’ since 2001.− Retail tariffs: up 70+% since 2001.− ComCom: the Wolak report alleging $4.3B of

excess profits since 2001.− 2009 Ministerial Review: suggestion of poor

retail competition.− 2010 Electricity Industry Act: suggestion of over-

reliance on public conservation campaigns and mis-management of reservoir operation.

− 2012-3 the emergence of NZ Power, growing awareness of fuel poverty, and equity issues.

• Three years on from the implementation of the EIA there is a natural question as to how the ‘new’ wholesale market is performing:

− Especially in the light of the extreme dry of 2012

Updating Meridian’s Hindcasting Perspective

• It is cleanest to start with a metric that judges behaviour from the perspective of what is best for NZ:– This is the type of view that regulators typically

apply on behalf of consumers and tax-payers.– Once outcomes have been assessed against this

metric then additional questions of appropriate commercial performance may be posed.

• A range of potential benchmarks can be suggested for examining the NZ power system.

• Traditionally in the NZ context stochastic reservoir and power system models are applied:– Spectra– SDDP– DOASA– EMarket

• From a NZ inc perspective most of these models seek to balance the costs of excess thermal fuel burn against the costs of excess system shortage:– In the face of significantly uncertain hydro inflows.– To minimise the overall NZ fuel (offer) supply cost.

3

• Meridian has undertaken five hindcasting exercises in recent years

• Some common conclusions were reached:− Market outcomes are largely determined by

environmental factors – inflows and the unavoidable costs of generating and managing the system.

− The market is not a perfect reflection of a centrally controlled benchmark – but it is surprisingly close to one with reservoirs being managed in the best interests of NZ.

− A range of decision support models demonstrate a good ability to reflect market outcomes.

• We focus here on the most recent of these – the 2012 hindcast exercise:

− Focusing on wholesale market outcomes.− Covering the Jul-2009 to Dec-2012 period.

• As with previous hindcast exercises, we begin to answer the question of market performance (and Meridian’s ability to assess it) by first addressing the question of what is an appropriate metric to use in measuring the actions of the market.

2012 Hindcasting

• Implicit in the above assumptions, we are now only examining a key residual problem: the NZ hydro-thermal reservoir management problem:– All demand conditions, geothermal output, and

plant outages are assumed to be known with perfect foresight – rather than via the application of generic planning assumptions.

– Whereas hydro and wind conditions are assumed to be unknowable.

– Thermal and hydro offer behaviours are assumed to be broadly cost reflective (a modified SRMC).

• Beginning in Jul09:1. Optimise the use of water in storage in the face of

historical hydrological uncertainty – ie all weekly flows over the 1931-2012 period.

2. Run the model to simulate the 42 month Jul09 to Dec12 period for all hydrological sequences.

3. Examine the single inflow ‘sequence’ corresponding to the Jul09 to Dec12 period.

• Having done this we can now compare modelled benchmark results to actual market outcomes.

4

• We configure a number of decision support models to reflect the fundamental underlying costs of electricity supply and demand in NZ:

− Approach intended to be broadly consistent with the system short-run marginal cost (SRMC).

− NOT a re-litigation of the market’s operational decisions – rather we are using the models in the role of a proxy ‘regulatory benchmark’.

• The decision support models used are:− Spectra: legacy ECNZ stochastic DP – 2 node− LPcon: Meridian internal stochastic DP – 22 node− Emarket: Energy Link market simulation tool – 2

node & 22 node configuration

• We update each model for events of Jul09-Dec12:− Storage conditions as at Jul2009.− Actual inflows and wind conditions.− Market geothermal and co-generation output.− Observed thermal and geothermal availability

(planned & unplanned outages).− Observed hydro and wind availability (planned &

unplanned outages).− Observed market HVDC configuration/outages.− Observed market NI and SI demand.− Assumed thermal SRMC offers and fuel costs.

The Full Forecast Distribution

• Beginning in Jul09 each model forecasts a wide range of outcomes over the following 42 month period corresponding to 82 different historical hydrological sequences:– We can view this via a familiar distributional perspective.– These price and storage charts show a wide range of

possible outcomes (in blue).– Superimposed over the top of these charts are the actual

market outcomes (in red) as occurred in reality.

• We can observe that prices and storage have oscillated within the extreme bounds of the feasible forecast produced (in this case by Spectra) beginning in Jul-2009.– This holds true for both the ‘dry’ period of 2012 and the

‘wet’ period of 2009-11

• Now we examine the single hydrological forecast consistent with what occurred on the day …

5

-

200

400

600

800

1,000

1,200

1,400

1,600

1,800

Jun

09

Se

p0

9

De

c09

Ma

r10

Ma

y10

Au

g1

0

No

v10

Fe

b11

Ma

y11

Jul1

1

Oct

11

Jan

12

Ap

r12

Jun

12

Se

p1

2

De

c12

sto

red

en

erg

yy

[G

Wh

]

Pukaki Hindcast Storage

$-

$50

$100

$150

$200

$250

$-

$50

$100

$150

$200

$250

Jun

09

Se

p0

9

De

c09

Ma

r10

Ma

y10

Au

g1

0

No

v10

Fe

b11

Ma

y11

Jul1

1

Oct

11

Jan

12

Ap

r12

Jun

12

Se

p1

2

De

c12

ba

se

loa

d p

ric

e [

$/M

Wh

]

SI Hindcast SRMC Prices

Comparison: Storage

• The storage levels from all models show good alignment over an extraordinary 42 month period that has traversed both an extreme dry period and an extended extreme wet period:– There are some divergences between modelled outcomes

– but these are not large.– However late 2011 and winter 2012 show notable

differences with LPcon holding the Southern lakes first lower and then higher than Spectra and Emarket:• In the 2012 situation this is closer to market outcomes.

– To some extent in terms of practical use storage outcomes in all models are configurable and if storage/risks implied here are considered unacceptable then outcomes can be altered.

• Total NZ storage levels have show a very good level of alignment between the all modelled results:– There are divergences between modelled outcomes but

these are small (in general).– Note that in all models (to differing degrees) the handling:

• Of Hawea storage is not great.• Of Taupo storage is adequate.• Of the split between Tekapo and Pukaki is mixed but never

fully satisfactory.

• This broadly implies that there are no gross differences in the use of water between the models.

6

0 250 500 750 1,000 1,250 1,500 1,750 2,000 2,250 2,500 2,750 3,000 3,250 3,500 3,750

-250 500 750

1,000 1,250 1,500 1,750 2,000 2,250 2,500 2,750 3,000 3,250 3,500 3,750

Jun

09

Se

p0

9

De

c09

Ma

r10

Jun

10

Se

p1

0

De

c10

Ma

r11

Jun

11

Se

p11

De

c11

Ma

r12

Jun

12

Se

p1

2

sto

rag

e [G

Wh

]

NZ Hydro Storage:LPcon, Spectra & EMarket

ALL Storage Market ALL Storage LPcon ALL Storage Spectra

ALL Storage EMarket-LP ALL Storage EMarket-SP NZ Historical Average

Hyd Seq Beginning: 2009-2010

0

250

500

750

1,000

1,250

1,500

1,750

2,000

2,250

2,500

2,750

-

250

500

750

1,000

1,250

1,500

1,750

2,000

2,250

2,500

2,750

Jun

09

Se

p0

9

De

c09

Ma

r10

Jun

10

Se

p1

0

De

c10

Ma

r11

Jun

11

Se

p11

De

c11

Ma

r12

Jun

12

Se

p1

2

sto

rag

e [G

Wh

]

Pukaki & Tekapo Hydro Storage:LPcon, Spectra & EMarket

PKI+TEK Storage Market PKI+TEK Storage LPconPKI+TEK Storage Spectra PKI+TEK Storage EMarket-LPPKI+TEK Storage EMarket-SP PKI+TEK Historical Average

Hyd Seq Beginning: 2009-2010

Comparison: Generation

• Meridian total market generation:– Is delivered in a more volatile manner by LPcon and

EMarket at the weekly level than suggested by Spectra – perhaps reflecting the LP nature of the underlying analytical engines, differences in tributary flow modelling, or both:• Reflected in the standard deviations of weekly generation.

– There are minor differences between modelled generation levels (highest in LPcon and all higher than market) driven by spill differences:• Eg higher levels of spill encountered in LPcon – at both

Manapouri and Pukaki/Tekapo.

• Thermal generation trends are very similar between the models:– Thermal volumes are high in LPcon & EMarket-LPcon

reflecting higher levels of spill in all Southern catchments.– Thermal volumes are particularly high in EMarket-Spectra

due to Taupo spill effects (static reserves representation)

7

- 50 100 150 200 250

Market

Lpcon

Spectra

Emarket-Lpcon

Emarket-Spectra

week generation [GWh]

Meridian Weekly Generation:LPcon, Spectra & EMarket

StdDev

Average- 50 100 150 200 250

Market

Lpcon

Spectra

Emarket-Lpcon

Emarket-Spectra

week generation [GWh]

Weekly Thermal Generation:LPcon, Spectra & EMarket

Huntly

CCGT

OCGT

-

50

100

150

200

250

300

350

400

Jun

09

Se

p0

9

De

c09

Ma

r10

Ma

y10

Au

g1

0

No

v10

Fe

b11

Ma

y11

Jul1

1

Oct

11

Jan

12

Ap

r12

Jun

12

Se

p1

2

De

c12

ge

ne

rati

on

[GW

h]

Meridian Hydro Generation:LPcon, Spectra & EMarket

Man+Wind Market LPcon

Spectra EMarket-LPcon EMarket-Spectra

Hyd Seq Beginning: 2009-2010

-

50

100

150

200

250

300

350

400

Jun

09

Se

p0

9

De

c09

Ma

r10

Jun

10

Se

p1

0

De

c10

Ma

r11

Jun

11

Se

p11

De

c11

Ma

r12

Jun

12

Se

p1

2

ge

ne

rati

on

[GW

h]

Thermal Generation:LPcon, Spectra & EMarket

Market LPcon Spectra EMarket-LPcon EMarket-Spectra

Hyd Seq Beginning: 2009-2010

Comparison: Market Prices

• Average modelled NZ prices over the full 42 month period are similar, all in a $55-$65 range compared to $64/MWh in the market:

• The general shape of modelled prices from all models track together reasonably tightly through high priced dry periods and low priced wet periods:– Both the SI and the NI show good general alignment.– There are clearly short lived spot market events that

different models reflect quite different: eg Aug-2011– There are market events that none of the models reflect

well: eg Nov-2010– The largest divergence is the second half of 2012 where SI

Spectra and EMarket-Spectra prices increase to high levels.

• Note that price outcomes show significantly more sensitivity to changes in models and/or assumptions than do physical outcomes:– A range of reasonable sensitivities can drive a movement in

average prices of +/30%.

8

$-

$25

$50

$75

$100

$125

$150

$175

$200

$225

$-

$25

$50

$75

$100

$125

$150

$175

$200

$225

Jun

09

Se

p0

9

De

c09

Ma

r10

Jun

10

Se

p1

0

De

c10

Ma

r11

Jun

11

Se

p11

De

c11

Ma

r12

Jun

12

Se

p1

2

ba

se

loa

d p

ric

e [

$/M

Wh

]

NI Wholesale Market Prices:LPcon, Spectra & EMarket

HLY LPcon HLY2201 Market NI Spectra

HLY EMarket-LP HLY EMarket-SP

Hyd Seq Beginning: 2009-2010

$-

$25

$50

$75

$100

$125

$150

$175

$200

$225

$-

$25

$50

$75

$100

$125

$150

$175

$200

$225

Jun

09

Se

p0

9

De

c09

Ma

r10

Jun

10

Se

p1

0

De

c10

Ma

r11

Jun

11

Se

p11

De

c11

Ma

r12

Jun

12

Se

p1

2

ba

se

loa

d p

ric

e [

$/M

Wh

]

SI Wholesale Market Prices:LPcon, Spectra & EMarket

BEN LPcon BEN2201 Market SI SpectraBEN EMarket-LP BEN EMarket-SP

Hyd Seq Beginning: 2009-2010

NZ Average Price

LPconEMarket-

LPconEMarket-Spectra Spectra Market

Jul09-Dec12 64$ 54$ 68$ 62$ 64$ FY2010 63$ 48$ 57$ 57$ 54$ FY2011 53$ 37$ 54$ 51$ 49$ FY2012 80$ 72$ 75$ 75$ 91$

Comparison: Market Outcomes

• High level outcomes over the Jul-2009 to Dec-2012 period are broadly similar between all of the models:– There are variations in prices, spill, and thermal generation

with physical outcomes in general being better aligned than pricing outcomes.

– Storage levels are similar overall - but over the crucial 2012 period LPcon held Southern reservoirs at higher levels than delivered by the other models (but lower during late 2011).

– Modelled generation revenues are similar – driven by modelled prices.

• Through a period of starkly different market conditions the consistency between the models is good.

• In broad terms this means that all the models are reflecting very similar inputs, market drivers, and underlying economic rationales.

• While the consistency of model outcomes – both between different models and within the same model but using different input assumptions is broadly good:– Primal outcomes (physical – eg storage) are significantly

more robust than dual (eg price, revenue) outcomes.– Eg, changes to thermal offer/fuel price assumptions or

shortage assumptions or ... may yield storage outcomes that are only modestly different but present price outcomes that vary much more significantly.

9

LPcon, Spectra, and EMarket HindcastJul2009 to Dec2012

HydroSeq:2009-2012

LPconEMarket-

LPconEMarket-Spectra Spectra Market

Prices SI 60.1$ 53.3$ 64.3$ 61.4$ 61.3$

NI 66.7$ 54.7$ 69.7$ 62.3$ 65.4$

SI-S/FIR 2.9$ 2.3$ -$ -$ 1.4$

NI-S/FIR 16.2$ 5.8$ -$ -$ 3.9$

Generation Pukaki 22,821 23,849 24,024 23,430 22,552

Manapouri 16,563 16,977 16,466 16,698 16,575

MEL Wind 3,837 3,862 3,866 3,869 3,863

Taupo 14,591 14,707 12,013 14,258 14,230

Waikaremoana 2,049 1,972 1,864 1,998 1,935

TPD 4,733 4,599 4,601 4,629 4,762

Tekapo 3,260 3,560 3,678 3,488 3,581

Clutha 12,541 12,237 11,851 12,840 12,441

TPL NI Hydro 4,941 4,743 4,743 5,339 5,276

TPL SI Hydro 3,662 3,253 3,314 3,571 3,712

Todd Hydro 1,077 925 925 1,111 1,140

Other NI Hydro 531 1,033 1,033 799 480

Other SI Hydro 841 737 737 790 537

Geothermal 14,295 17,064 17,067 17,628 17,628

Co-Gen 4,899 4,605 4,605 4,899 4,899

Other Aux 10,015 5,016 5,047 6,537 6,537

Huntly 10,047 10,468 11,951 9,158 8,656

CCGTs 20,755 20,129 21,751 20,370 21,619

OCGTs 2,062 2,024 2,450 2,027 2,324

TOTAL Hydro 87,611 88,594 85,250 88,951 87,220

TOTAL Thermal 32,863 32,620 36,151 31,555 32,599

TOTAL Other 33,046 30,547 30,585 32,932 32,927

Dem Res NI 0- - 13- 6- -

SI 25- - 2- 80- 444-

HVDC S->N 6,192 7,009 6,478 7,100 5,731

Spill Pukaki 1,921 1,140 772 1,397 2,275

Tekapo 274 270 143 508 385

Clutha 1,149 1,198 1,610 934 1,050

Manapouri 1,710 390 1,511 320 320

Taupo 321 321 2,963 701 226

Storage Waitaki Avg 1,101 1,220 1,176 1,071 1,213

Tekapo Avg 545 507 511 550 564

Hawea Avg 219 215 219 125 170

Taupo Avg 373 401 259 468 357

Conclusions

• Comfort should be taken from the fact that:– Reservoir operation (both physical and pricing) is

being managed in a rational manner.– Market outcomes are close to what could be

achieved under central control.– A range of different reservoir management tools

demonstrate a good ability to shadow the market.

• However care should be taken in drawing strong conclusions about the goodness of outcomes without an appropriate context and sensitivities:– Price outcomes in particular show a large

sensitivity to changes in input assumptions.

• Three years on from the implementation of the EIA what conclusions can be drawn about the ‘new’ market arrangements:– Not a lot has changed – indeed similar behaviours

are seen in aggregate to those observed in previous hindcast exercises:• Thermal stations generate when they ‘should’.• Reservoirs are being managed ‘appropriately’.

– However market prices do appear more volatile with short lived events pushing prices up quickly.

– There is some small evidence of the market holding Waitaki reservoirs higher than the benchmark during 2012.

10

• Market outcomes have broadly matched benchmark modelled results:

− All models considered have matched high level market outcomes.

• Market outcomes through both an extreme dry period (2012) and a prolonged extreme wet period (2009-2010) have been largely driven by the unavoidable costs of generating and managing the system:

− Market results are still dominated by hydrology.− Physical reservoir management is all about

managing what inflows turn up, when, and with very limited storage capacity.

− Market price outcomes are the result of balancing escalating thermal costs against too much reservoir spill in a fashion that ensures security of supply is not compromised.

• The market has not been a perfect reflection of a centrally controlled benchmark:

− This is to be expected.− However the gap between market outcomes and

benchmark outcomes is not large.− Imperfections should be considered in the light of the

significant successes – particularly in market investment and the allocation of risk.

Additional Material

EPOC Jul 2014Version 1.0

Grant Telfar, Meridian Energy

July 2014

Hindcasting at Meridian

• From a NZ inc perspective these models seek to balance the costs of excess thermal fuel burn against the costs of excess system shortage:– In the face of significant hydro uncertainty.– To minimise the NZ fuel (offer) cost.

• The decision support models used here:− Spectra: legacy ECNZ stochastic DDP− LPcon: Meridian internal stochastic DDP− EMarket: Energy Link market simulation

tool – 2 node & 22 node configuration• LPcon is an in-house Meridian hydro-

thermal power system model:− 2 stage optimisation/simulation. − Weekly resolution and a 15 block LDC.− Stochastic DDP creating water-values.− Simple thermal offers/cost.− DC load flow – 22 regions used.− Dynamic risk and reserves.− Diurnal wind characteristics.

12

• Hindcasting is an analytical approach that can be applied to examine how decision support tools, analysis, and key assumptions performed after the fact:− A priori forecast vs ex post outcomes.− Hindcasting is seen by Meridian as part of

good internal management.• We seek to answer the question of market

performance by first addressing the question of what is an appropriate metric to use in measuring the actions of the market:− Starting with a metric that judges behaviour

from the perspective of what is best for NZ.• Traditionally in the NZ context stochastic

reservoir and power system models are applied:− Spectra− SDDP− DOASA− Emarket− ...

Comparison: Pukaki and Tekapo Storage

13

0

250

500

750

1,000

1,250

1,500

1,750

2,000

2,250

2,500

2,750

-

250

500

750

1,000

1,250

1,500

1,750

2,000

2,250

2,500

2,750

Jun

09

Se

p0

9

De

c09

Ma

r10

Jun

10

Se

p1

0

De

c10

Ma

r11

Jun

11

Se

p11

De

c11

Ma

r12

Jun

12

Se

p1

2

sto

rag

e [G

Wh

]

Pukaki & Tekapo Hydro Storage:LPcon, Spectra & EMarket

PKI+TEK Storage Market PKI+TEK Storage LPconPKI+TEK Storage Spectra PKI+TEK Storage EMarket-LPPKI+TEK Storage EMarket-SP PKI+TEK Historical Average

Hyd Seq Beginning: 2009-2010

Comparison: NZ Storage (No Manapouri)

14

0 250 500 750 1,000 1,250 1,500 1,750 2,000 2,250 2,500 2,750 3,000 3,250 3,500 3,750

-250 500 750

1,000 1,250 1,500 1,750 2,000 2,250 2,500 2,750 3,000 3,250 3,500 3,750

Jun

09

Se

p0

9

De

c09

Ma

r10

Jun

10

Se

p1

0

De

c10

Ma

r11

Jun

11

Se

p11

De

c11

Ma

r12

Jun

12

Se

p1

2

sto

rag

e [G

Wh

]

NZ Hydro Storage:LPcon, Spectra & EMarket

ALL Storage Market ALL Storage LPcon ALL Storage Spectra

ALL Storage EMarket-LP ALL Storage EMarket-SP NZ Historical Average

Hyd Seq Beginning: 2009-2010

Comparison: Meridian Generation

15

-

50

100

150

200

250

300

350

400

Jun

09

Se

p0

9

De

c09

Ma

r10

Ma

y10

Au

g1

0

No

v10

Fe

b11

Ma

y11

Jul1

1

Oct

11

Jan

12

Ap

r12

Jun

12

Se

p1

2

De

c12

ge

ne

rati

on

[GW

h]

Meridian Hydro Generation:LPcon, Spectra & EMarket

Man+Wind Market LPcon Spectra EMarket-LPcon EMarket-Spectra

Hyd Seq Beginning: 2009-2010

Comparison: Thermal Generation

16

-

50

100

150

200

250

300

350

400

Jun

09

Se

p0

9

De

c09

Ma

r10

Jun

10

Se

p1

0

De

c10

Ma

r11

Jun

11

Se

p11

De

c11

Ma

r12

Jun

12

Se

p1

2

ge

ne

rati

on

[GW

h]

Thermal Generation:LPcon, Spectra & EMarket

Market LPcon Spectra EMarket-LPcon EMarket-Spectra

Hyd Seq Beginning: 2009-2010

Comparison: NI Market Prices

17

$-

$25

$50

$75

$100

$125

$150

$175

$200

$225

$-

$25

$50

$75

$100

$125

$150

$175

$200

$225

Jun

09

Se

p0

9

De

c09

Ma

r10

Jun

10

Se

p1

0

De

c10

Ma

r11

Jun

11

Se

p11

De

c11

Ma

r12

Jun

12

Se

p1

2

ba

se

loa

d p

ric

e [

$/M

Wh

]

NI Wholesale Market Prices:LPcon, Spectra & EMarket

HLY LPcon HLY2201 Market NI Spectra

HLY EMarket-LP HLY EMarket-SP

Hyd Seq Beginning: 2009-2010

Comparison: SI Market Prices

18

$-

$25

$50

$75

$100

$125

$150

$175

$200

$225

$-

$25

$50

$75

$100

$125

$150

$175

$200

$225

Jun

09

Se

p0

9

De

c09

Ma

r10

Jun

10

Se

p1

0

De

c10

Ma

r11

Jun

11

Se

p11

De

c11

Ma

r12

Jun

12

Se

p1

2

ba

se

loa

d p

ric

e [

$/M

Wh

]

SI Wholesale Market Prices:LPcon, Spectra & EMarket

BEN LPcon BEN2201 Market SI SpectraBEN EMarket-LP BEN EMarket-SP

Hyd Seq Beginning: 2009-2010

Comparison: HVDC Transfers

19

-75

-25

25

75

125

Jun

09

Se

p0

9

De

c09

Ma

r10

Jun

10

Se

p1

0

De

c10

Ma

r11

Jun

11

Se

p11

De

c11

Ma

r12

Jun

12

Se

p1

2

en

erg

y tr

an

sfe

r [G

Wh

]

HVDC Transfers:LPcon, Spectra & EMarket

Market LPcon Spectra EMarket-LPcon EMarket-Spectra

Hyd Seq Beginning: 2009-2010

Conclusions

– Market outcomes are close to what could be achieved under central control.

– A range of different reservoir management tools demonstrate a good ability to shadow the market.

• However care should be taken in drawing strong conclusions about the goodness of outcomes without an appropriate context and sensitivities:– Price outcomes in particular show a large

sensitivity to changes in input assumptions.

• Three years on from EIA implementation not a lot seems to have changed – similar behaviours are seen to those observed in previous hindcast exercises:

• Thermal stations generate when they ‘should’.

• Reservoirs are being managed ‘appropriately’.

• Market prices appear reasonable.

20

• Market outcomes have broadly matched benchmark modelled results.

• Market outcomes through both an extreme dry period (2012) and a prolonged extreme wet period (2009-2010) have been largely driven by the unavoidable costs of generating and managing the system.

• The market has not been a perfect reflection of a centrally controlled benchmark:− However the gap between market outcomes

and benchmark outcomes is not large.− Imperfections should be considered in the

light of significant successes – particularly in market investment and the allocation of risk.

• Comfort should be taken from the fact that:− Reservoir operation (both physical and

pricing) is being managed in a rational manner.