march 19th 2009 ewec 2009 – marseille tryggvi jónsson enfor a/s & dtu informatics

10
ENFOR Tryggvi Jónsson, ENFOR A/S & DTU Informatics Forecasting Day-ahead Electricity Prices and Regulation Costs in Markets With Significant Wind Power Penetration March 19th 2009 EWEC 2009 – Marseille Tryggvi Jónsson ENFOR A/S & DTU Informatics Pierre Pinson DTU Informatics Henrik Madsen DTU Informatics March 19th 2009 EWEC 2009, Marseille

Upload: leila-hebert

Post on 31-Dec-2015

21 views

Category:

Documents


0 download

DESCRIPTION

Forecasting Day-ahead Electricity Prices and Regulation Costs in Markets With Significant Wind Power Penetration. March 19th 2009 EWEC 2009 – Marseille Tryggvi Jónsson ENFOR A/S & DTU Informatics Pierre Pinson DTU Informatics Henrik Madsen DTU Informatics. Intro. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: March 19th 2009 EWEC 2009 – Marseille Tryggvi Jónsson ENFOR A/S & DTU Informatics

ENFOR

Tryggvi Jónsson, ENFOR A/S & DTU Informatics

Forecasting Day-ahead Electricity Prices and Regulation Costs in Markets With Significant Wind Power Penetration

March 19th 2009

EWEC 2009 – MarseilleTryggvi Jónsson ENFOR A/S & DTU Informatics

Pierre Pinson DTU Informatics

Henrik Madsen DTU Informatics

March 19th 2009 EWEC 2009, Marseille

Page 2: March 19th 2009 EWEC 2009 – Marseille Tryggvi Jónsson ENFOR A/S & DTU Informatics

ENFOR

Tryggvi Jónsson, ENFOR A/S & DTU Informatics

Intro The nature of wind power places it inside the equilibrium at

almost all times.

Wind power enters the day-ahead electricity supply function as a stochastic quantity with uncertainty attached to it.

The most rapid changes in the supply function are mostly owed to wind power.

Case study: DK-1 area of Nord Pool’s Elspot

Nord Pool hydro dominated – DK-1 heavily penetrated by wind

March 19th 2009 EWEC 2009, Marseille

Page 3: March 19th 2009 EWEC 2009 – Marseille Tryggvi Jónsson ENFOR A/S & DTU Informatics

ENFOR

Tryggvi Jónsson, ENFOR A/S & DTU Informatics

On the market power of wind energy (forecasts) - I

March 19th 2009 EWEC 2009, Marseille

Page 4: March 19th 2009 EWEC 2009 – Marseille Tryggvi Jónsson ENFOR A/S & DTU Informatics

ENFOR

Tryggvi Jónsson, ENFOR A/S & DTU Informatics

On the market power of wind energy (forecasts) – II

March 19th 2009 EWEC 2009, Marseille

Page 5: March 19th 2009 EWEC 2009 – Marseille Tryggvi Jónsson ENFOR A/S & DTU Informatics

ENFOR

Tryggvi Jónsson, ENFOR A/S & DTU Informatics

Model for forecasting day-ahead prices Propose a three layer approach to forecast the day-ahead price

consisting of:

i. A mapping of forecasted wind power penetration and time to the average prices

ii. Dynamical weighting of recently observed prices and prediction error together with the output from the first layer.

iii. Uncertainty estimation conditioned upon the forecasted wind power penetration and the forecasted mean price

New observations considered as soon as they become available

Older observations discounted either exponentially or by a rolling window

March 19th 2009 EWEC 2009, Marseille

Page 6: March 19th 2009 EWEC 2009 – Marseille Tryggvi Jónsson ENFOR A/S & DTU Informatics

ENFOR

Tryggvi Jónsson, ENFOR A/S & DTU Informatics

Forecasting properties

Root Mean Square Error Forecasts on January 12th 2007

March 19th 2009 EWEC 2009, Marseille

Page 7: March 19th 2009 EWEC 2009 – Marseille Tryggvi Jónsson ENFOR A/S & DTU Informatics

ENFOR

Tryggvi Jónsson, ENFOR A/S & DTU Informatics

Forecasting regulation costs Imbalances on the producers side are priced by a two price

model at Elspot

Implies no regulation costs for producers bringing the system back to balance

Forecasting of regulation costs is done in two steps

i. Prediction of the sign of the system imbalance

ii. Probabilistic forecasting of the penalty

March 19th 2009 EWEC 2009, Marseille

Down regulation hoursUp regulation hours

Cost

Imbalance

Page 8: March 19th 2009 EWEC 2009 – Marseille Tryggvi Jónsson ENFOR A/S & DTU Informatics

ENFOR

Tryggvi Jónsson, ENFOR A/S & DTU Informatics

Predicting the system’s imbalance sign Predict the imbalance sign by binary classification for each

“direction”

Do not believe in linearly separable non-overlapping classes

Explanatory variables are

i. Forecasted wind power penetration

ii. Spot price forecast

iii. Time variables

iv. Import/Export forecasts

March 19th 2009 EWEC 2009, Marseille

Down regulation Up regulation

Hit rate 75.5% 76.4%

Bias 1.17% -0.20%

Page 9: March 19th 2009 EWEC 2009 – Marseille Tryggvi Jónsson ENFOR A/S & DTU Informatics

ENFOR

Tryggvi Jónsson, ENFOR A/S & DTU Informatics

Predicting the imbalance penalties Predict price intervals with certain probabilities directly

Predictions conditioned upon the forecasted sign

March 19th 2009 EWEC 2009, Marseille

Down regulation Up regulation

Page 10: March 19th 2009 EWEC 2009 – Marseille Tryggvi Jónsson ENFOR A/S & DTU Informatics

ENFOR

Tryggvi Jónsson, ENFOR A/S & DTU Informatics

Summary & Final Remarks The impact of wind power forecasts on the day-ahead prices is

substantial and nonlinear

Same applies for regulation costs

Wind power forecasts therefore play an important role in price forecasting

More intelligent trading with benefits for both producers and the system as a whole

More efficient risk hedging as well

Methods are operationally available and results indicate that they can be tailored to other markets

March 19th 2009 EWEC 2009, Marseille