Page 38 - BACnet_Europe-Journal_43
P. 38
Technology Technik
Artificial Intelligence in the field of Building
Automation
Künstliche Intelligenz in der Gebäude-
automation
The term “AI – Artificial Intelligence” is Similarly, cleaning schedules can be based not classic building automation with flexible IT-based
increasingly associated with buildings only on the current actual values in terms of the management systems. These offer unrestricted
and building automation. The question is: intensity of use of kitchens, canteens and toilets programming using modern IT languages and
what is it, where do its tangible benefits and other areas, but can be based on predictions tools, easy integration with other IT systems
lie in this field, and how does the building drawn from an analysis of usage patterns in the such as workspace/room reservation systems
infrastructure need to be adapted to realize previous days and weeks. This kind of forward- or data banks, and data visualization for facility
those benefits? looking building management can be applied in managers and for “ordinary” users.
Der Begriff „KI – Künstliche Intelligenz“ almost every area of building services, leading to
wird in zunehmendem Maße mit Gebäu- increased energy efficiency, reduced operating The growing assimilation of sensor-generated
den und Gebäudeautomation in Verbindung costs, improved space utilization and other data into the IT-based management level opens
gebracht. Die Frage ist: Was ist künstliche advantages. the way for more advanced data processing
Intelligenz, wo liegen ihre konkreten Vor- solutions to come into play – such as AI tools.
teile für die Gebäudeautomation, und wie All this – and much more – is possible when This is the pre-condition for the implementation
muss die Gebäudeinfrastruktur angepasst data on building system status and conditions of any prognosis-based form of building
werden, um diese Vorteile zu nutzen? is intelligently evaluated. This requires intensive management. The sophisticated processing
processing of large amounts of data, with many of sensor-generated data makes the Smart
Today’s building automation systems in the variables to be considered. Artificial Intelligence Building into a “Cognitive Building”. (Fig. 1)
main operate ‘statically’ in response to fixed (AI) offers many new, tailor-made solutions
time programs or simple control parameters. which are eminently suited to efficient building AI-Learning Process
Room temperature control is based on a preset management.
temperature that is the same throughout the day. The first step in any Artificial Intelligence process
Lighting is operated manually, with switches, or “Building Automation”, “Smart Building” is system learning. This can take three forms.
on the basis of simple presence switches. None and “Cognitive Building” nsupervised Learning
U
S
of this is truly ‘intelligent’. The new dimension upervised Learning
R
that AI can add into the building automation Initially, “Building Automation” was comparatively einforcement Learning
environment is to use autonomous analysis “unintelligent”. Systems were programmed
of the data as a basis for optimized operation. to follow a set of simple rules, allowing for “Unsupervised Learning” is used when
Thus, the heating and cooling dynamic of rooms, quick system start-up and subsequent ease of large quantities of data must be processed
weather forecasts, predicted room occupancy maintenance. and categorized. This grouping enables the
during the course of the day can all be factored recognition of deviations from norms and
into the operation of the heating. The “Smart Building” typically builds on this interdependencies. For example, sensor
data from identical circulation pumps can be
grouped. If data from one pump or group of
pumps deviates from the norm, there may be a
defect, and a human engineer can be sent to
investigate.
“Supervised Learning” often makes use of neural
networks. They consist of entry and exit nodes
as well as further nodes in the intermediate
layers. Mathematically weighted relationships
exist between the diverse nodes (neurons). In
order to optimize these relationships, the neural
network is subjected to a training phase with
known input and output patterns. In the field
of building automation, for example, a neural
network can “learn” the current consumption
profiles of different appliances and which
Fig. 1: Building Automation – Smart Building – Cognitive Building (source: IBM)
Abb. 1: Gebäudeautomation – Smart Building – Cognitive Building (Quelle: IBM) appliances are active when. This information
38 BACnet Europe Journal 43 09/25