Research project DestinE: Destination earth machine learning demonstrator – energy systems
Development of a machine learning demonstrator for faster and interactive simulation of European energy systems
As we move away from fossil fuels and electrify heating and transport, the demand for solutions to integrate weather-dependent renewable energy into a reliable power supply is increasing across Europe. A key prerequisite for this is the consideration of climate and weather-based uncertainties in the modelling of energy systems. However, existing climate information systems often have only low regional accuracy, making it difficult to predict changes on both short and long timescales. Against this backdrop, in 2021 the European Union launched the ambitious Destination Earth (DestinE) initiative, which focuses on creating digital twins of our planet. To enable a seamless fusion of real-time observations and high-resolution forecasts and projections, two different digital replicas of the highly complex Earth system are to be created during the project: one for weather-related and geophysical extremes, and the other for adaptation to climate change. Both serve the goal of improving our knowledge of the Earth's climate and its interaction with other sectors concerning human action and life, such as agriculture, water supply or energy. This should help to better predict disasters and ultimately support policymakers in managing climate-related changes.
Research project DestinE: Destination earth machine learning demonstrator – energy systems | |
|---|
Duration | May 2025 to October 2026 |
Funded by | European Commission/European Centre for Medium-Range Weather Forecasts (ECMWF) |
Project participants | - DLR Institute of Networked Energy Systems
- Renewables Grid Initiative
- Aarhus University
|
Within the overall DestinE undertaking, experts from various scientific disciplines are implementing several sub-projects. The sub-project “Destination Earth Machine Learning Demonstrator – Energy Systems” aims to replace time-consuming load flow calculations with machine learning (ML) approaches. This should enable accurate and cost-effective forecasts for critical situations in the European electricity system and facilitate the consideration of uncertainties in the design of future energy systems. To achieve this, an innovative machine learning demonstrator is being developed, based on a sophisticated neural network that explicitly considers the laws of physics and utilises climate data as well as other relevant input data such as expected energy demand and potential energy generation from renewable sources. Based on this data, the trained ML demonstrator should deliver predictions on the cost-optimal use of power plants and storage facilities, and on load flows.
The Institute of Networked Energy Systems aims to support transmission system operators with improved information and tools for decision-making in grid planning and security of supply analysis within the sub-project “Machine Learning Demonstrator – Energy Systems”. The developed approach is suitable for both local and continental energy systems. This work is based on the DestinE Use Case Energy Systems, for which DLR had integrated climate and weather information into the modelling of energy systems. Building on these results, the ML demonstrator aims to validate and improve the integration process of climate and weather information in order to develop more accurate and efficient models of energy systems, taking uncertainties into account and thus improving the resilience of the systems.
For the modelling of power systems, the Institute will use meteorological information in a semi-operational workflow within the demonstrator, replacing the usual but complex and less efficient linear optimal power flow (LOPF) approach. For this purpose, the ML application draws on time series from DestinE’s Climate Adaptation Digital Twin as well as other open datasets. The later applicability of the demonstrator is to be ensured by involving stakeholders from energy and climate research in the development process in a participatory approach, taking into account interests relating to the planning of network expansions and improvements as well as the safe operation of networks and power plants in the model development.
More information:
