Wind forecast uncertainty and impact on the electricity network.

Abstract

The objective of the thesis is to understand and characterize the wind power forecast uncertainties and develop a model to realistically mimic the typical error made on wind power forecast in the day-ahead planning of generation schedules. The aim is to take into account the wind power forecast error at the day-ahead time scale in the simulation runs of the Tractebel ENGIE software tool called Scanner. For this purpose, we conduct a review of the scientific literature available about existing models for wind forecast and representation of wind errors and their possi- ble integration in similar tools. Then, we characterize the errors made in the past by transmission system operators in Central West Europe for day-ahead wind power forecast. Next, we develop the model to mimic the typical day-ahead wind power forecast errors and write the Matlab code necessary for its implementation. Finally, we perform two tests of a typical case, without and with the introduction of the error. The researches show that different statistical laws can be used for the model. How- ever a trade-off must be done between the complexity of a law and the quality and reality of the results obtained by such a law. This is the reason why we chose a hyperbolic distribution to describe the error. We generate the error via the sampling of a hyperbolic distribution previously fitted and adjusted to mimic the typical error made in the past. Moreover we add some improvements to our model such as the correlation between the errors made at dif- ferent hours or the better wind power forecast for the first hours of the next day than for the last hours of the next day. We always keep in mind that we want to obtain a model as close as possible from the reality. Finally we observe different results concerning general costs, power production and an analyse of a critical period where large errors have been observed. We conclude that the results observed can be explained in a logical way due to the introduction of the forecast uncertainty.