Automated classification of HVAC systems through analysis of system behaviour

Abstract

Technical monitoring applications support the operator in identifying potential improvements in plant operation and deriving recommendations for action. Today, the integration of building automation network data points into technical monitoring applications is complex and costly. The current state of the art is to manually integrate data points into dashboard applications. The decision which data point describes which HVAC component is made by engineers. The basis for the decision is usually identifiers or data point descriptions. An automated recognition of the structural information of the HVAC systems would help to automate the process and make the implementation of a technical monitoring in existing buildings and new buildings easier. One way to determine structural information from data points is to look at the behaviour of an HVAC system type. Common HVAC systems follow known regularities in their construction and behaviour. By analysing the system behaviour, conclusions can be drawn about the heating circuit type (flow temperature controlled, outdoor temperature controlled, etc.). For example, an outdoor temperature-controlled heating circuit will behave differently than a flow temperature controlled heating circuit. Likewise, by analysing the system behaviour, it is possible to predict which data point is assigned to which system component. If two data points increase by a similar value with a short time offset, it is highly probable that they represent the supply and return temperature of a heating circuit. If the flow and return are heated up, it is likely that a pump has started up beforehand, which was represented by a binary switching command. This paper describes an automated method to classify heating circuit data based on system behaviour. BACnet trend objects of heating circuits from different buildings, that are maintained by the building management of the city of Cologne, serve as data sets. This ensures that the developed method can be applied to a wide variety of heating circuit types. The automated classification of a heating circuit is intended to reduce the effort of manually assigning data points to specially created dashboards. Building operators can thus be supported in the creation and implementation of technical monitoring.