Development of data fusion methods to support energy operational optimization and predictive maintenance of heating systems in multi-family buildings

In the project "ADAPTeR", digital methods are being developed to operate heating systems in multi-family buildings more energy-efficiently and reliably. To achieve this, measurement data, maintenance information, and technical specifications are combined in a digital asset data sheet and analyzed using innovative data fusion and AI methods. The consortium, consisting of Green Fusion, B&O Service, HAW Hamburg, and Fraunhofer ISE, enables condition-based and predictive maintenance, reducing energy consumption, CO₂ emissions, and costs, while providing clear decision support for technical staff.

Initial Situation

Heating systems in multi-family buildings are primarily maintained on an interval basis. Operational failures often remain undetected for long periods due to a lack of significant measurement data and structured documentation. As a result, only symptoms are typically addressed, not the underlying causes of energy-related malfunctions. Complex systems, such as hybrid plants with heat pumps and fossil backups, cannot be tested in all operational states in practice. Thus, disruptions in renewable systems often go unnoticed, while operations switch unnoticed to inefficient fossil generators—resulting in increased energy consumption, CO₂ emissions, and costs.

Objective

ADAPTeR aims to shift the management of heating systems in multi-family buildings from interval-based to condition-oriented and predictive strategies. The core goal is to create a digital asset data sheet as a central access point, where measurement data, maintenance information, and technical specifications are provided in a machine-readable format. Building on this, data fusion and AI methods will identify malfunctions early, reveal energy optimization potentials, and determine the optimal time for inspections and maintenance. Multicriteria assessments will consider energy, CO₂, costs, labor, and resource usage.

Approach

The project will develop a digital asset logbook with a web interface that consolidates technical data, maintenance history, and continuous measurement data into a standardized data model. Interfaces to field buses and protocols, as well as mobile measurement technology, will connect the systems. Data fusion algorithms, innovative models based on a semi-Markov approach, and AI methods will analyze the plant's condition and derive key figures, failure probabilities, and maintenance recommendations. The expertise of the practical partners will be utilized to enable targeted assessments and the derivation and prioritization of suitable measures. Practical tests in real residential buildings will validate the solutions and further develop them into user-friendly tools for energy-efficient and reliable plant operations.

The “ADAPTeR” project is funded (Fkz.: 03N1119A) by the Federal Ministry of Economic Affairs and Energy (BMWE).