Modelling Electricity Prices

1
Modelling Electricity Prices

• Dipl.-Ing. Dr. Josef H. Bogensperger
• Head of Risk Management
• VERBUND Austrian Power Trading AG

2
Welcome, may we introduce ourselves
VERBUND is ...

• ... the largest producer and distributor of
  electrical energy in Austria
• ... the operator of the super-regional
  high-voltage grid of Austria with important
  connections to the neighbour countries in Central
  Europe
• ... one of the biggest Austrian listed companies
• ... the most eco-friendly electricity generator
  in the European Union
• ... one of the most modern electricity companies
  in Europe
• ... a successful player at the European
  electricity market

3
VERBUND Key Figures 2005 Numbers
Share Chart

• Sales revenue 2,51 bn
• EBIT 526,5 mln
• Free Cash Flow 597,3 mln
• Moodys Rating A1
• Electricity sales 57,282 GWh
• Generation 29,011 GWh
• Hydro 24,788 GWh
• Thermal 4,223 GWh
• Employees 2,436

22
51
Shareholding Structure
27
4
VERBUND Key Figures 2005 Volumes
5
VERBUND-Austrian Power Trading AG

• Energy Turnover (APT trades in the name and on
  behalf of VERBUND)2004 86 TWh (69
  international, 79 thereof in Germany)
• Products Physical Energy, Energy
  Derivatives,Proofs of Origin, Emission-Certifica
  tes
• Target Groups Retail, Wholesale, Big
  (Industrial) End users
• Employees 120
• Owner 100 Verbund

6
VERBUND Organizational Chart
VERBUND HOLDING
VERBUND- Austrian Power Grid AG
VERBUND- Austrian Hydro Power AG
VERBUND- Austrian Power Trading AG
VERBUND-Austrian Power Sales GmbH
VERBUND-Austrian Thermal Power GmbH Co. KG
VERBUND Italia
Energia
APT SK
POWEO
APT CZ
Energie Klagenfurt GmbH
APT Hungária
VERBUND- BeteiligungsgmbH
APT Energa Hellas
APT Slovenija
VERBUND- Management Service GmbH
APT Deutschland
7
Table of Contents

• Price forward curve (PFC)
• Hourly price forward curve (HPFC)
• Simulation Results
• Scenario Backtesting

8
Specification of firm schedule

• Fixed schedule The customer defines an hourly
  fixed schedule, every deviation will be charged
  extra by actual prices

9
Hourly curve needs a two component model
Stochastic spot prices
Historical factors
yearly, quarterly PFC
hourly PFC
monthly PFC
Market data
Pricing
2nd component
1st component
10
Price forward curve (PFC)
price
month products
year products
quarter products
balance of month (BOM)
time
now

• load data of structured products is hourly ? we
  need hourly prices

11
Price forward curve (PFC)

• Prices for future delivery at various points in
  time
• Granularity of observable products reduces over
  time
• Reflects the actual market opinion
• Emerge from traded standard products
  (month/quarter/year, base/peak)

12
Price forward curve (PFC)

• We use all historical data from the European
  Energy Exchange (EEX) as model input.

• Assume the following price processes for
  historical data of future products
• Year............................
• Quarter in year (i)........
• Month in quarter (k).....

13
1st component receive monthly PFC
historical factors
quarterly PFC
monthly PFC
actual market data
includes specific times of history
14
Another method to receive quarterly PFC

• Assuming that the following correlation is high
• ? we can use actual quotes for the estimation of

gtgt
15
Arbitrage free monthly forward curve (MPFC)

• Monthly prices have to be arbitrage free compared
  to quarterly and yearly prices, therefore
• Settlement prices are arbitrage free, intraday
  prices in general not
• ? be careful using actual quotes, get them first
  arbitrage free

year i, month j in year i Number of
hours prices at t
16
2nd component receive HPFC

• Spot price simulation Sampling method
• Cluster similar historical data season, weekend,
  holidays,
• Random selection of a historic day of a
  corresponding class
• ? we get a sample of
  hourly prices for month
• Get hours arbitrage free For example, shift all
  hourly prices till
• is fulfilled.

17
2nd component receive HPFC
Possible improvement

• Modelling spot prices as stochastic process on a
  daily basis using
• Jump Diffusion or
• Regime Switching

18
How to display results
short
long
19
How to display results

• Price distribution for 1000 possible scenarios
• If quarters do not quote, we look at the
  summer/winter spread of the scenarios

20
Back testing models

• Back testing is the validation of the underlying
  simulation
• Regarding HPFC in APT back testing is solved
  using a database
• Historical and simulated price data are fed into
  a database
• Historical prices since the start up of the
  certain power exchange
• Simulation prices HPFC for 1000 scenarios
• Possibility of selection criteria on specific
  periods of time to find similarities
• Graphical query results indicate strengths and
  weaknesses of a model

21
Back testing models
22
Thank you for your attention!
Dipl.-Ing. Dr. Josef H. Bogensperger VERBUND-Austr
ian Power Trading AG Tel. 43 (0)
50313-53020 Mob. [phone] Fax. 43 (0)
50313-153020 email Josef.Bogensperger_at_verbund.at