FederatedForecasts

The research project "FederatedForecasts", a collaboration between Karlsruhe University of Applied Sciences (HKA) and the consulting company FORRS Partners GmbH in cooperation with four leading direct marketers, is exploring new ways in the field of federated machine learning (FML) to optimize renewable energy production forecasts. This project aims to explore the possibilities of FML to develop more accurate forecasting models for energy generation without compromising the data sovereignty of the companies involved. The co-operation partners from the energy sector include the companies BayWa r.e. AG, Trianel GmbH and Vattenfall Europe Sales GmbH.

In this research initiative, HKA, with project leaders Prof. Dr Andreas Wagner and Prof. Dr Christian Zirpins from the Faculty of Computer Science and Business Informatics at HKA, is taking the lead in developing the FML algorithms and setting up a robust IT infrastructure. FORRS contributes its extensive knowledge in system integration and modelling. The direct marketers contribute valuable practical experience and industry knowledge to strengthen the theoretical model.

The "FederatedForecasts" project aims to improve the accuracy of forecasting models, although the results and scope of the improvements still need to be researched. The project not only promises potential economic benefits for the companies involved, but could also have far-reaching effects on the efficiency of the entire electricity market. It also provides an important learning and research platform for students and professionals working on the latest developments in AI technologies and energy management.

The core of "FederatedForecasts" is the application of federated learning, which makes it possible to develop joint models based on decentralised data. Prof Dr Andreas Wagner, Vice Dean for Research at the Faculty of Economics at HKA, explains: "This technology offers the potential to strengthen collaboration between different market participants by sharing valuable insights without revealing sensitive data."

Sep. 01, 2023 - Aug. 31, 2027