Ellen Grumert


Wind modeling

Abstract

The demand for environmentally friendly energy sources is huge today. The
wind power industry has been growing extremely fast the last decades. The
research for new methods and models in the wind power area is an ongoing
process. One important issue is to predict the wind in a good way. The wind
power industry needs the predictions to plan where to place new wind power
stations and to maintain already existing ones. The electricity market
needs the predictions in order to plan how much of the electricity that 
can be taken from wind power and how to price the electricity.

In this paper the primitive barotropic model is tested as a wind prediction
model. A particle filter is used to calibrate the model. The advantage with
using a particle filter is that a lot of useful information is included,
i.e. the distribution of the particles and hence the distribution of the 
wind forecasts. An important improvement with the model compared to the one 
point prediction models like time series processes as AR-models is that a 
spatial dependency is included. The model is tested out with given data and 
simulated data. This first simplified model shows that there is a big 
potential in this way of solving the problem.