Data-Driven Identification and Operational Optimization of Energy-Flexible Thermal Supply Systems

The accelerating expansion of renewable energies in Europe and the rest of the world leads to the challenge of adapting energy demand to the increasingly volatile renewable electricity generation. To contribute to this goal, thermal supply systems can be used for demand response by utilizing their dedicated and inherent thermal energy storages and optimizing the loading and unloading operation. In this paper we present a method for the data driven system identification and operational optimization of industrial thermal supply systems. This method aims at developing optimization problems for the control of thermal supply system and identifying system dynamics specifically for this purpose. The method includes parameter identification and validation of individual components, such as compression chillers and thermal storages, using the system's measurement data. From the identified component dynamics an overall optimization model is formulated and tested, taking into account the interactions between single components. Finally, the optimization model is deployed, and the optimized control signals are passed onto the real systems components. For permanent application of the optimization, monitoring und maintenance of the model are integrated as a last and recurring step of the method. This includes updating model parameters in case of increasing mismatch between model and reality over time. These steps are successfully implemented in the real-world example of the cooling supply system of a climatized room at the ETA Research Factory at the Technical University of Darmstadt. In this example the room's thermal capacity is used to vary its cooling supply depending on time-variable electricity prices. The model is deployed as a model predictive control loop, regularly updating its optimal trajectory using electricity prices of a representative period from 2023. Just by optimizing the room's temperature control a reduction in energy cost of 10.06% compared the conventional operation was achieved. This results in a reduction of electricity related carbon emissions of 4.93%.