The X‑Store subproject, part of the larger X‑Energy initiative, was carried out at the Hamburg University of Applied Sciences (HAW Hamburg) from 4 April 2018 to 30 June 2023 under the “Research at Universities of Applied Sciences” programme (grant code 13FH1I05IA). The project was led by Prof. Dr. Gerwald Lichtenberg, with project coordination handled by Mike Blicker and innovation management by Dr. Oliver Arendt. The research team comprised M.Sc. Björn Brunner, M.Sc. Kathrin Weihe and M.Sc. Marius Block. Funding was provided through the FH‑Impuls initiative, which supports high‑quality research at applied‑science institutions.
Technically, X‑Store focused on developing tensor‑based, model‑predictive control (MPC) methods for decentralized, learning‑enabled regulation of energy and heat networks, with a particular emphasis on storage components. The core scientific contribution was the adaptation of tensor‑based hierarchical design techniques to create storage‑centric MPC schemes. A grey‑box model of the demonstrator was built and parameterised in MATLAB/Simulink, enabling realistic simulation and validation of the control logic. The control strategy incorporated a nonlinear cost function and was implemented on the building management system of the new demonstrator, the Technology Center Energy Campus (TEC) of HAW Hamburg.
The project began with the Energiebunker Wilhelmsburg as the first demonstrator. During 2018–2021, the existing measurement infrastructure was analysed, new flow sensors were installed, and a grey‑box model was developed. In 2022, the demonstrator was switched to the TEC, where historical data were accessed via InfluxDB and the building automation system was integrated. After preliminary tests revealed a suboptimal configuration of the heating system that impaired the thermal storage, compensation strategies were devised and incorporated into the MPC framework.
Long‑term experiments conducted in the 2022/2023 heating season demonstrated that the storage‑based MPC could reduce energy consumption by up to 40 % compared with the conventional heating control scheme. This significant savings was achieved through more efficient utilisation of the thermal storage, better alignment of heat generation with demand, and continuous optimisation based on measured data. The experimental results also provided valuable insights into the design of future heating systems, especially when integrating new heat sources such as hydrogen or other alternative fuels.
The project’s work packages (A1–A3, B1–B3, C1–C3, D1–D3) covered demonstrator selection, infrastructure integration, model development, control design, implementation, testing, and evaluation. The switch of demonstrators required re‑establishing data access, re‑parameterising the model, and adapting the control script to the new building automation interface. After initial debugging, the control system was deployed for the start of the heating period, followed by a series of long‑term trials that informed subsequent optimisation and documentation.
In summary, X‑Store delivered a validated, storage‑centric model‑predictive control architecture that achieved a 40 % reduction in heating energy use on a real‑world campus demonstrator. The project demonstrated how tensor‑based hierarchical design can be applied to decentralized energy networks, providing a blueprint for future smart‑grid deployments. Collaboration across HAW Hamburg, the TEC, and the local energy utility ensured that the research was grounded in practical infrastructure, while the FH‑Impuls funding enabled the integration of advanced control theory with real‑time building management systems.