Abstract: At the European Energy Exchange (EEX) a huge variety of energy derivatives are traded for many European countries as well as for Japan, e.g. Japanese Power Kansai Area Futures or Tokyo Area Power Futures. In the past years the increasing infeed from renewable energies became responsible for a large part of the variation in electricity prices at the EEX. In particular, the intraday market is highly dynamic with outstanding liquidity and steadily increasing traded volumes. Hence, risk management for the intraday market is challenging and an important topic for all market participants.
In this talk we look at several new statistical approaches for modeling and forecasting electricity prices. The target of the analyses are two price indices used for the intraday market at the EEX, ID3 and IDFull, which both represent weighted average prices over different time periods before delivery. In particular, we setup a Bayesian approach for forecasting the price spread between the intraday and the day-ahead market using predictive distributions. This way we can produce forecasts for the price trends on the intraday market and, in addition, assess the quality of these forecasts using predictive probabilities. Finally, we also investigate whether the forecasting quality can be further improved using artificial intelligence, in particular combined learning.
A large variety of information about historical prices as well as current weather forecasts are incorporated in the analyses as covariates or exogenous stochastic processes. The dependency of prices on renewable energies can be represented in flexible ways, which allows to vary price spikes both in direction and size according to the generation of wind and solar power.
This talk is based on several joint work with Daniel Lingohr, Daniel Nickelsen and Sebastian Uhl.