Welcome to join the Swedish Wind Centre webinar: Two Models for Better Wind Power Forecasts
Two Models for Better Wind Power Forecasts
How can forecasts for wind power generation be improved? In this webinar, Tomas Landelius from SMHI and Olle Viotti from Svenska kraftnät present two models that can help address this challenge.
Introduction, Johan Arnqvist, PhD in meteorology, Uppsala University
What challenges and opportunities are at the horizon when it comes to forecasting of Variable Renewable Energy (VRE)? What is currently happening within the IEA? What is the role of VRE forecasting in the transformation of the electricity market and the energy system the latest couple of years?
Tomas Landelius, researcher, SMHI
Numerical weather prediction (NWP) has undergone a rapid transformation through recent advances in machine learning. Most current ML-based weather models are trained on reanalysis datasets and still rely on a traditional NWP analysis to provide their initial state. In the EU funded Weather Generator project, we take a different approach: developing a foundation model trained jointly on model data and observations, enabling the system to both learn and initialize the atmospheric state directly. This talk will present the architecture of the Weather Generator and discuss how such foundation models can be extended to application specific forecasting tasks, including high resolution prediction of wind power production.
Olle Viotti, researcher Svenska kraftnät
The rapid variations and inherent uncertainty of wind power pose challenges for the power system, creating a need for short-term forecasts with a horizon of approximately 30 minutes. At these timescales, both weather prediction data and time-series analysis of individual wind farms are insufficient. I will present a method for analyzing the spatio-temporal correlations between the production of different wind farms, which can be used to improve forecast precision during critical events such as front passages
Most welcome!
Swedish Wind Centre