Using neural networks to evaluate the uncertainties in linear optimisation energy system models
Duration: October 2019 until September 2022
Funded by: Federal Ministry for Economic Affairs and Climate Action
Project Participants: Institut für Technische Thermodynamik Institut für Vernetzte Energiesysteme Forschungszentrum Jülich GmbH Technische Uni Berlin – Fakultät für Mathematik und Naturwissenschaften GAMS Software GmbH, Frechen Konrad-Zuse-Zentrum für Informationstechnik, Berlin
Project Manager at the Institute of Networked Energy Systems: Jan Buschmann
Project Description: Incidents such as grid bottlenecks and power station failures have to be compensated for if energy supply is to be stabilised and its reliability guaranteed. To make the necessary preparations, the impact of these “uncertainties” are being studied in model-based simulations so that power stations have a strategy in place to cope with bottlenecks of this kind. The scientists working on the UNSEEN research project are devising a method for calculating and evaluating an unprecedented number of energy scenarios to ensure that these simulations are meaningful and reliable.
To produce reliable simulation results, for example to advise policymakers, a large number of scenarios need to be created using various different parameters. This simulates as many potential impacts as possible and allows transmission network operators to make long-term preparations for such incidents. Outages can thus be avoided and enough power station capacity made available to respond to and offset bottlenecks. Researchers are currently facing the challenge, however, that it takes a huge amount of time to run simulations with energy system models of the size needed here. At present, therefore, there is only little scope for studying the abovementioned uncertainties in sufficient depth and for determining the resulting potential for optimisation and the relevant solutions.
The UNSEEN research project is responding to this situation by incorporating artificial intelligence (AI) into its investigations. Using AI significantly cuts processing time for the simulation runs, enabling millions of scenarios to be calculated in virtually no time at all. First, AI supplies forecasts for eliciting potential optimisation measures. These forecasts are then used as the basis for an algorithm that calculates more detailed results for optimisation.
The project team at the Institute of Networked Energy Systems is focusing on studying and assessing the impact calculated in this way on the energy system. The scientists are considering grid-related situations such as the grid bottlenecks and power station failures mentioned above so that they can investigate and evaluate these uncertainties with the help of selected indicators.