X-Store

The integration of renewable energy generation into electricity and heating grids requires the coordinated management of energy storage systems. This determines the mode (charging, discharging or doing nothing) for each storage system at any given time. Concepts from the engineering discipline of control engineering make such decisions possible - e.g. based on models of system behaviour.

In the sub-project X-Store, model predictive control (MPC) was designed for existing heat storage systems and tested under real-time conditions. The scientific focus lay on the adaptation of tensor-based methods for the design of decentralised learning controls using the example of storage tanks. The application-oriented goal was to optimise the operation of storage components with different dynamic properties that are integrated into heterogeneous networks. This means that the control components of a smart grid, which are often referred to as intelligent, can not only be systematically designed, but also continuously improved during operation with the help of measured data.

Results

In the project, the applicability of the control concepts was demonstrated at the Technology Centre Energy Campus (TEC) of the Competence Centre for Renewable Energies and Energy Efficiency (CC4E) in Hamburg Bergedorf.

A storage-centred, model-predictive control system was implemented at the TEC and operating parameters were determined that led to energy savings of up to 40% compared to the previous conventional heating control system while increasing room comfort. In addition, the provision of flexibility, which is important for sector coupling, is simplified in principle by switching to model-based control. The characterisation of the storage behaviour carried out contributed to the retrospective evaluation of the design of the heating system and can help in the planning of future components.