Forecasting Electricity Spot using
Futures Contracts and Energy
Reservoir as External Signals
A Discrete Time Heteroscedastic Approach

 
Jonas Berg and Torbjörn Kronblad

Abstract

Due to the deregulation of the Nordic power market during the last few years
the conditions on the market has changed. The electricity price has become
increasingly volatile while at the same the time the Nordic power market
turnover has grown exponentially. Accurate forecasts of electricity prices are
useful in a variety of areas, such as planning production or creating trading
strategies.
In this thesis we model and forecast Elspot, which is the physical electricity
system price, traded at Nord Pool. We model on a weekly basis and use
discrete time models, both with and without respect to heteroscedasticity.
The linear ARMAX model is compared to ARMAX models where dierent
GARCH structures has been added. Calculated energy reservoirs and
Nord Pool futures contracts are tested as external signals for improving forecasts.
Integrating the physical and financial market in this way is an interesting
aspect.
As it turns out, using futures as external signals results in good forecasts,
especially at shorter forecast horizons. The addition of GARCH to the general
ARMAX model gave us some improvement, even though it was not very
large. Our best performing model was the ARMAX(2; 1; 1; 1)-EGARCH(1; 1)
model.