||Optimal trading strategies for a wind-storage power system under market conditions
||Andersen, Philip Hvidthøft Delff (Mathematical Statistics, Department of Informatics and Mathematical Modeling, Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark)
||Madsen, Henrik (Mathematical Statistics, Department of Informatics and Mathematical Modeling, Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark)
Pinson, Pierre (Mathematical Statistics, Department of Informatics and Mathematical Modeling, Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark)
||Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark
||In this thesis, a model of a system consisting of electric power production on
wind turbines combined with a storage device is developed. By use of Monte
Carlo simulation, the operation of the system is optimised with respect to two
different objective functions. One strategy is to maximise the expected revenue
for the whole delivery period, the other is to minimise the expected regulation
costs. Moreover, two different markets are considered, with different horizons
and duration of the delivery periods.
A passive operation strategy for the electrical energy storage is defined, and
hence the delivery to the power net becomes a function of only the production
and the issued contract at the market. Since the production is assumed to be
uncontrollable, only the contract is left to optimise.
Three different models of the electrical storage devices are being used. Hence,
effects of all limited capacity, charging and discharging efficiencies, and limitations
on charging and discharging speeds can be observed.
For running the Monte Carlo Simulation, a non-linear estimate of the distribution
of the future production for each relevant horizon by use of adaptive
quantile regression with point forecasts as explanatory variable. From this, the
interdependence of future production at different horizons are estimated. From
these two estimates, the scenarios are simulated, and based on these the optimisation
problems are solved. The simulations are run throughout all in all more
than one year of data.
The results of the optimisation strategies are not as good as expected, assumed
because of too poor estimates of the distributions of the production. However,
by use of simulations, a potential gain of the method can be estimated. This
gain is expected to be realistic if a good model for prediction of the distributions
is found. The results are very depending on optimisation strategy and storage
model, and the obtained revenues are between -2% and 19% compared to when
using the point predictions as contracts and the respective storage devices.
Finally, different approaches to improve the method are discussed.
Creation date: 2009-09-02
Update date: 2009-10-21