Guest talk: Irene Schicker "Post-processing of numerical weather predictions with focus on renewable energy"
When
Where
Event language(s)
Post-processing of numerical weather predictions with focus on renewable energy
Irene Schicker
Austrian national weather service ZAMG
Abstract: With the increasing feed-in rates of electricity from renewable energy sources, especially highly weather-driven sources such as wind and solar, weather and power forecasts that can cover the associated uncertainties are becoming more and more important. The same holds for the increase in extreme weather events or meteorologically induced events, calling for improved forecasting in all areas, from nowcasting to day-ahead and sub-seasonal forecasts. In weather forecasting, the ensembles of numerical weather prediction models generate the uncertainties. However, these
models are computationally expensive, time-consuming and require further post-processing models to fit site-optimized but also large-scale forecasts for wind energy. Here, post-processing is needed to break down the forecasts for single wind farms. In recent years, these post-processing models rely more and more on machine learning methods. Still, methods are based on statistical still are able to compete. In this talk an overview on the methods implemented at the Austrian met service with special emphasis on machine learning such as Unet based downscaling or Graph networks for targeted forecsats and their skills and drawbacks for improved predictions for renewable energy systems will be discussed.
Bio: Irene Schicker received her Masters' degree in Meteorology at the University of Innsbruck in remote sensing and glaciology. In her PhD studies she focused on very high resolution numerical weather prediction while working on a project level in different climate, renewables, and forecasting improvement projects. Since 2014 she is a researcher (since 2017 post-doc/senior researcher) at the Austrian national weather service ZAMG with focus on post-processing, both statistical and machine learning. In the past 5 years her focus was mainly on post-processing, forecast improvements across the time scales (nowcasting to sub-seasonal), and meteorologically driven extreme event detection for renewable energy systems using all available data types.
- Published:
- Updated: