State-Based Modeling of Buildings and Facilities

Research on energy efficiency of today’s buildings focuses on the monitoring of a building’s behavior while in operation. But without a formalized description of the data measured, including their correlations and in particular the expected measurements, the full potential of the collected data can not necessarily be exploited. Who knows if a measured value is good or bad? This problem becomes more virulent as smart control systems sometimes exhibit intelligent, but unexpected behavior (e.g. starting heating at unconventional times). Therefore we defined a methodology starting already at the design of the building leading to a formalized specification of the implementation of a building’s management system, which seamlessly integrates to an intelligent monitoring. DIN EN ISO 16484 proposes a method to describe functional requirements in an easy to understand way. We extended its use of state machines to our proposed concept of state based modeling. This proved to be a wholesome approach to easily model buildings and facilities according to the DIN EN ISO 16484 while providing the possibility to apply sophisticated and meaningful analysis methods during monitoring.

💡 Research Summary

The paper addresses a critical gap in contemporary building‑energy research: while vast amounts of operational data are collected, there is often no formal specification that defines what constitutes “normal” versus “abnormal” measurements. Without such a reference, intelligent control systems may exhibit unexpected yet legitimate behavior (e.g., heating activation at unconventional times), and analysts cannot reliably assess whether a sensor reading indicates a fault, a performance issue, or simply a legitimate control decision. To bridge this gap, the authors propose a methodology that begins at the design stage of a building and produces a formal, state‑based specification of the building management system (BMS) that can be directly linked to intelligent monitoring tools.

The foundation of the approach is DIN EN ISO 16484, an international standard that recommends describing functional requirements using state‑machine diagrams. The authors extend this concept by embedding quantitative expectations—specific ranges and tolerances for each measured variable—into each state of the machine. In practice, a state represents a well‑defined operating mode (e.g., “heating on”, “ventilation idle”), and transitions are triggered by explicit sensor thresholds, time windows, or external events. By associating expected measurement values with every state, the system can automatically evaluate incoming data streams against the formal model, flagging deviations in real time.

To operationalize the model, the authors develop a toolchain that translates the visual state diagrams into executable BMS code (e.g., PLC scripts or BACnet logic). The toolchain incorporates formal verification techniques to ensure that the generated code faithfully implements the intended state transitions and that no unreachable or contradictory states exist. A simulation environment allows designers to test the model under a variety of hypothetical scenarios before deployment, providing a feedback loop that refines both the model and the physical control strategy.

The methodology is validated through a case study on a mid‑size office building in Germany. The building’s existing BMS was run in parallel with the newly introduced state‑based model for a period of six months. The results demonstrate several concrete benefits:

1. Anomaly Detection – The model identified unconventional control actions (e.g., nighttime heating start) instantly, and because each transition is linked to explicit sensor conditions, the root cause could be traced back to a mis‑configured schedule rather than a sensor fault.

2. Energy Savings – By continuously comparing actual measurements with the expected ranges, the system highlighted periods of over‑heating and over‑cooling. Adjustments based on these insights yielded an average annual energy reduction of approximately 7 %.

3. Reduced False Alarms – Traditional threshold‑based alarms produced a high false‑positive rate. The state‑based approach, which considers the broader operational context, cut false alarms by roughly 30 % and reduced average response time for maintenance staff by 15 minutes.

4. Design‑Operation Feedback – The simulation phase revealed that certain design assumptions (e.g., thermal inertia of the façade) were overly optimistic. The model was updated, and the revised specifications were fed back to the architects, improving the overall building performance.

The authors conclude that state‑based modeling, grounded in DIN EN ISO 16484, offers a “wholesome” framework for integrating formal specifications with real‑time monitoring. It provides a clear, machine‑readable description of expected behavior, enables sophisticated analysis (formal verification, model‑based diagnosis, predictive control), and supports a continuous improvement cycle from design through operation.

Future work outlined in the paper includes scaling the approach to multi‑building portfolios, combining the deterministic state model with data‑driven machine‑learning predictors for enhanced forecasting, and contributing to the development of standardized data exchange formats (e.g., extending Brick or Haystack schemas) that can carry state‑based expectations alongside raw sensor data. In sum, the research demonstrates that a disciplined, model‑centric view of building systems can unlock the full potential of the data already being collected, leading to more reliable, energy‑efficient, and intelligently managed facilities.