Zephyr/Prediktor is the oldest commercial tool for the short-term prediction of wind power world-wide. It uses WAsP theory to downscale the wind fields from Numerical Weather Prediction input to the actual wind farm site in question. In order to remove the last remaining errors, a Model Output Statistics (MOS) module is used.

Purpose

The proposed Master thesis would investigate and improve the MOS module, both from a scientific point of view and from a numerical point of view. The Prediktor system is implemented in Java, therefore good knowledge of Java would be quite beneficial.

Description

There are three main fields to work on:

• An investigation of the convergence of the MOS parameters: What is the correct time window to use data from? 
• An investigation in the most effective MOS parameters: Which parameters give the qualitatively best result, and diminish the error best? Are there more meteorological fields which should be used for the prediction (e.g. atmospheric stability)? 
• An improvement of the numerical implementation of the MOS. At the moment, the MOS is implemented as a very generic, but quite heavy simplex optimization algorithm. This could be improved dramatically.
  

Qualifications

good programming skills, some basic understanding of wind flows over terrain, preferably knowledge of WAsP, and willingness to work at Risø

Project form

Thesis Final year project

Subjects

Computer science

Duration