Page 39 - BACnet_Europe-Journal_43
P. 39
Technology Technik
can be used to avoid “spikes” in building energy Cloud-based
consumption, by shutting down some appliances Internet AI platform
and extending the operation time of others.
Management BMS (Building
Another form of Artificial Intelligence is level LAN Management System)
represented by processes that autonomously
determine which actions are appropriate in a
given situation. They emulate human behavior Automation Controller
whereby different solutions are tried in order to level Controller (manufacturer B)
determine the best way forwards in a hitherto (manufacturer A)
unknown situation, and conclusions drawn
retrospectively. The learning task becomes more
challenging when feedback is given much later
and hinges upon events in the relatively distant
past. This is true in a human context, and equally Field level
true in computer environments. "communicative“component
„smart devices“
The best-known example in this category is classical component
“Reinforcement Learning”. Consider the issue
of determining the optimal start and stop times Fig. 2: linking building automation with a (cloud-based) AI platform 1 Prof. Dr. Michael Krödel
of heating to achieve a comfortable temperature Abb. 2: Verbindung von Gebäudeautomation mit (cloud-basierter) AI-Plattform
when the building opens. At the simplest level,
the learning algorithm receives the value from
the room temperature sensor and can act on the forecasting, e.g. for meeting rooms, a broader range of features, so this currently
actuator on the radiator. By a process of trial and canteens, pantries, transit areas, toilets and represents common practice.
error, the algorithm can determine the necessary parking spaces as well as the provision of
lead time. However, this simple example ignores information in the short term (for building The AI platform is built on a Smart Building
the fact that, for instance, the speed of heating occupants) and in the long term (for facility infrastructure, and all technical systems
also depends on the outside temperature, so managers, e.g. in form of advice on building should ideally be connected to a BMS (Building
the reading from an exterior temperature sensor restructuring). Management System). The BMS must be able to
L
needs to be considered. Instead of providing a oad management: forward-looking govern the building facility and room automation
pre-set target temperature, the algorithm may operation of electrical systems in order to systems.
be given evaluations (good / OK / cold) during the avoid (costly) peak loads.
P
day and must learn in response to this feedback. recautionary maintenance and optimized Requirements for Building Infrastructure
servicing: analysis of failure probability,
In addition, the algorithm can be provided with timely maintenance and consequential AI-based processes require high computing
an additional rating every month based on avoidance of technical failures. power and are therefore best run on a dedicated
E
the overall energy cost: encouraging efficient mployee-oriented value-added services: AI platform. This can be cloud-based or run on
behavior and discouraging inefficient responses. mobile devices can – for instance – be used a server in the company’s own data center. Due
A “stable” response that balances comfort and to generate space utilization forecasts, view to the higher computing power of cloud-based
efficiency can be established, but exploration canteen usage intensity, request parking server farms and the currently greater scope of
should continue to accommodate changes in space availability and preferred workspace services offered by cloud-based AI frameworks,
behavior and the environment. location or select individual meals. this option is shown in Fig. 2.
Compensation of skilled-staff shortages:
It can be seen that these three approaches are making effective use of facility maintenance This figure also shows that a smart building must
complementary. The learning method should be staff in managing the building’s technical be in place as a basis for the use of AI-based
chosen depending on the task in hand – each systems. processes. It is important that all building services
has its merits. ocus on meaningful sensor data: generate in this smart building are connected to a building
F
as much data as possible from as few management system (BMS) for communication
Concrete Applications sensors as possible – reducing redundancy, purposes. In many cases, building automation
cutting investment and operating costs. is still implemented in such a way that room-
Many diverse AI-based applications are available side and system-side automation are operated
in the field of building automation. They can be Demands upon System Architecture separately from each other as respective
broadly categorized as follows: island systems. This must be converted into a
Optimized facility management: needs-based An AI platform is indispensable for the holistically networked installation.
control of heating plants, circulating pumps, introduction of intelligent learning processes
lighting etc. (as opposed to control on the such as those described above. This can be It must also be ensured that a BMS can
basis of simple parameters or by timer). either cloud-based or server-based. Cloud- also control room and system automation.
Optimized utilization of spaces and based server farms offer more processing In many cases, the specific behavior of, for
infrastructure: capacity analysis and power, and cloud-based AI frameworks offer example, room-side bus systems and system-
BACnet Europe Journal 43 09/25 39
