Physics-informed control systems combined with thermal energy storage enable efficient use of renewable energy at the district level by optimally coordinating generation, consumption, and storage.

Thermodynamic Control of Modular Energy Networks

Photovoltaics are already a cost-effective and low-carbon source of energy. However, their output varies over the course of the day and across seasons. As a result, future energy systems will be characterized by periods of near-abundant availability and times when electricity must be used efficiently as a scarce resource. Demand-side flexibility can play a key role in addressing this challenge. One major consumer in this context is space heating. While new buildings achieve low energy demand through advanced insulation, comprehensive retrofitting of existing buildings is often prohibitively complex. Climate-friendly heating solutions that do not require major structural modifications are therefore urgently needed.
Intelligent control systems are a central component in this context. They determine when and where energy should be generated, used, transported, or stored. A key advantage is that they rely only on low-cost sensors and minimal additional hardware, making them easy to deploy, cost-effective, and suitable for retrofitting. Such approaches are particularly promising at the district level, where medium- to large-scale thermal storage can be implemented and even small efficiency gains quickly become economically viable at scale.
 

Challenges and Methods

Documentation on existing buildings and heating networks is often incomplete or unavailable. As a result, self-learning approaches are of great importance. Typically, large training datasets are not available; instead, relevant data must be collected during operation, making efficient data usage essential. At the same time, it is neither practical nor desirable to learn fundamental physical relationships purely from data. Physics-informed learning methods address this by incorporating known physical principles as prior knowledge, while only learning the unknown aspects from data.
In addition, such systems are distributed networks with a potentially large and time-varying number of components. Scalability is therefore a key requirement. Modular approaches are particularly well suited to this setting and can be systematically analyzed using passivity theory. Long-term storage, operating over extended time horizons, introduces numerical challenges. Structure-preserving integration methods and differential geometric formulations provide a promising framework to address those challenges.

D²HeaTEC Project

The joint project D²HeaTEC (Decarbonizing District Heating with Techno-Economic Control) develops and evaluates an integrated control framework that coordinates both technical and economic aspects of decentralized, renewable heat supply. Two residential districts in Cologne serve as testbeds, enabling a direct comparison between natural gas-based heating systems and electric heat pumps.
The thermal systems are modeled and techno-economically controlled. The system model includes representations of generation units, distribution networks in both districts, and a modular apartment model. While generation and distribution are already monitored in real time, additional devices such as radiator actuators, heat cost allocators, and indoor air sensors will be installed in participating apartments specifically for this project. All data is aggregated in a pseudonymized form.
The control framework has access to these aggregated data and thus addresses a fundamental limitation of conventional systems: so far, residents, landlords, and network operators at best optimize within their respective local scope. The proposed approach overcomes this fragmentation.

Collaboration

If you are interested in research collaborations, theses, internships, or student assistant positions, feel free to get in touch. In addition to the listed opportunities, there are always further exciting and current topics available.
 

• Exercise Signals and Systems (SuS)
• Control Engineering Lab (LR)