Generative Downscaling of Offshore Wind Fields

European offshore wind capacity is projected to increase tenfold in coming decades, posing challenges for power grid stability due to the variability of wind energy. As wind energy becomes a larger part of the power mix, accurate short-term wind forecasting is essential for effective grid integration, wind farm operation, energy trading, and turbine control.

Advanced machine learning models will be developed to produce more accurate forecasts, and we are building short-term wind forecasting methods for European offshore regions, offering improved accuracy and addressing a critical need in wind energy for enhanced short-term forecasting. This innovation can help stabilize grids, reduce energy costs, and provide improved wind forecast models for the industry.

As part of this project, the MSc candidate will investigate generative AI methods for the downscaling of offshore wind fields. The thesis will focus on developing and evaluating generative machine learning algorithms capable of transforming coarse-resolution wind information into high-resolution wind fields, with potential applications in offshore wind forecasting and energy systems.

The student will be supervised by PhD candidate Francesco Pinto}and Prof. Angela Meyer.

The Energy, Weather \& AI Lab is affiliated with both Bern University of Applied Sciences (Switzerland) and Delft University of Technology (The Netherlands). The student will therefore have the opportunity to visit and work at either institution, depending on their preference.

An important goal of the thesis is to provide the student with the opportunity to contribute to research that can lead to a publication in a high-impact scientific journal.

The interview will involve a discussion of the following two papers:

• Lopez-Gomez et al. - Dynamical-generative downscaling of climate model

ensembles

• Pinto et al. - Skillful forecasting of offshore winds from satellite scatterometer constellations

• Good knowledge of deep learning and PyTorch (necessary)
• Fluency in English (necessary)
• Interest in the intersection between energy, weather and AI (desired)
• Prior knowledge of physics and/or meteorology is a plus, but not necessary