The ARCHE project set out to streamline the design and commissioning of self‑optimising control systems for building energy management. Central to the effort was the creation of a unified methodology that integrates model‑based design, simulation, and real‑time testing of distributed energy systems. An expert system was developed to support semi‑automatic design of control architectures, leveraging semantic technologies to describe the structure of building energy networks. This system feeds into a toolchain that links the design environment with the simulation platform SimulationX, which was extended to accommodate the new self‑optimising components. Within this environment, model‑predictive control (MPC) and reinforcement learning (RL) techniques were applied to generate and evaluate control strategies before deployment.

The project validated the developed components on four demonstrator buildings: the Musterhaus, Blankenburger Straße, Brehmestraße, and a Bürogebäude. For each site, detailed simulation models were constructed and used to test the self‑optimising controllers in a virtual environment. Subsequent hardware‑in‑the‑loop (HiL) and software‑in‑the‑loop (SiL) tests confirmed that the controllers could be transferred to real‑time operation without loss of performance. Parallel to the control development, a monitoring framework was established to capture energy consumption data. The team identified several optimisation opportunities in data handling, such as adopting ISO 8601 timestamps, avoiding changes to scaling factors during operation, and using generic, human‑readable data point names. These guidelines improved the reliability of long‑term monitoring and facilitated automated analysis of system performance.

Load forecasting was addressed through the evaluation of multiple regression algorithms. The forecasts were then used to optimise operation schedules, reducing peak demand and improving the utilisation of renewable generation. The combined use of forecasting and optimisation demonstrated that the self‑optimising controllers could adapt to varying supply and demand conditions, thereby enhancing the overall energy efficiency of the buildings. While the report does not provide explicit numerical performance figures, it emphasises that the integration of MPC, RL, and semantic modelling significantly reduced design costs and increased the tangible benefits of deploying renewable, CO₂‑free energy sources in building operations.

The consortium comprised six partners: Fraunhofer IIS/EAS, FASA AG, ifm Software GmbH, ESI GmbH, Geo‑En GmbH, and EASD GmbH. Fraunhofer IIS/EAS led the design of the sEMS methods and the expert system, while FASA AG handled user‑side validation and measurement verification. ifm Software represented the building automation perspective, managing the Musterhaus demonstrator and bridging the expert system with the automation software. Geo‑En was responsible for energy system planning and monitoring of the Blankenburger Straße and Brehmestraße sites. ESI GmbH provided simulation expertise, supplying the SimulationX environment and integrating it into the toolchain. EASD GmbH contributed modelling and optimisation expertise, analysing the control design processes and producing the final guidelines for data handling. The project report, version 0.3 dated 1 February 2023, reflects the culmination of these collaborative efforts. Although the funding source is not explicitly stated in the excerpt, the project aligns with German research initiatives aimed at reducing market barriers for renewable energy deployment in buildings. The ARCHE project therefore delivers a comprehensive, technically robust framework that simplifies the deployment of self‑optimising controllers, thereby advancing the adoption of environmentally friendly energy systems in the built environment.