Self-learning forecasting methods in the field test

05.12.2024

The Fraunhofer Institute for Energy Economics and Energy System Technology (IEE) wants to help market players to leverage and optimize flexibility options in the energy system.

The partners in the "KonSEnz" research project (continuously self-learning forecasting methods and services in smart energy markets and grids) are hoping to achieve reliable forecasts. In addition to the Fraunhofer IEE, the University of Kassel and wind turbine manufacturer Enercon from Aurich in Lower Saxony are also involved. The transmission system operators Amprion, Tennet and 50 Hertz are also associated partners.

"Our forecasting methods will do far better justice to the increasing dynamics on the generation and consumption side than the traditional methods," promises Dominik Beinert, co-project manager at Fraunhofer IEE. Grid operators, direct marketers and plant operators are set to benefit, as they will be provided with "powerful tools" with which they can, for example, make optimum use of flexibility in the energy system.

According to a press release from the Fraunhofer Institute, the training of forecasting models will no longer have to be initiated manually in future. Instead, the project partners are developing methods that incorporate changes on the generation and consumption side directly and independently into the forecasts.

Processes of continuous adaptive learning and the ongoing updating of the models ensure this. In contrast, previous methods could only deliver results with a time delay, in particular because the training of the models was discontinuous.

Available to all as open source in future

When developing the new method, the project partners refer to a number of use cases, such as the operation of PV systems with forecasts of generation and self-consumption. Another use case is the prediction of power flows between the high-voltage and extra-high-voltage grid, where continuous changes to the switching states in the grid must be taken into account. This should enable grid operators to reliably forecast the overloading of equipment.

The project partners have announced that the new methods will be designed as scalable, resilient microservices that can be seamlessly embedded in various control, management and operating systems thanks to numerous interfaces. The researchers want to publish their results as an open access publication and make the software modules available to the energy industry as open source.

Author: Fritz Wilhelm, E&M powernews