Delegating to a building-management system

The ability of computer science to learn from experience and anticipate events can be put to good use in building-management systems, says JAMES PALMER.In 1950 Alan Turing published a paper in Mind magazine stating that, in time, computers would be programmed to acquire abilities to rival the human brain. While computers are still a long way from successfully emulating human intelligence, computer science has come a long way in mimicking our ability to learn from experience and anticipate events. Couple this with the powerful memory of modern computers, and we can identify many building-services applications that would benefit from this ‘artificial intelligence’. Human limitations As a mere human being with limited capacity to remember, analyse and control the services in my building, I could be forgiven for needing assistance. How long would it take me, for instance, to spot the difference between heating and lighting that has been left on because Bob in accounts works late every other Thursday, or because the cleaners simply forgot to turn them off? Would I even be there to do anything about it? What I need is a set of eyes watching every aspect of my building 24 hours a day, learning what is normal and what is exceptional. Indeed, one of the difficulties I face is that, in a sense, technology has made my buildings harder to manage because I am bombarded with information. What I really need is management systems that highlight problems and unusual occurrences — management by exception, in fact. Today’s intelligent systems have the ability to log and archive hundreds of thousands of values. We can also run data-analysis software to help us understand the data logs. Data analysis is thus mainly concerned with what has been happening historically, rather than what is about to happen. A statistical analysis application called AutoTrack, developed by North Building Technologies, uses powerful statistical-analysis techniques on live data logs to monitor and learn the way a building or its services operate. Data processing occurs dynamically, minute-by-minute, hour-by-hour, month-by-month. AutoTrack is a like a self-adaptive stop-start routine, but is much more sophisticated. It can be applied to any parameter — temperature, humidity, light level, power — and to any building system. Once AutoTrack has learnt the weekly profile of a data channel, it can use the data intelligently, detecting, for example, when a value is about to exceed the limits of its usual operating range. Significantly this may provide opportunities to prevent out-of-limit events occurring. As you would expect, alert messages may be sent to supervisory or maintenance staff by email, SMS text message, or to a PC screen. Management by exception AutoTrack has the ability to learn that the ambient temperature is lower in the night than in the day. It will discover that the temperature is higher in summer than winter. Once it has worked it all out, it will automatically set alarm limits that are in tune with the building. AutoTrack can predict alarm events much earlier than conventional BMS alarm monitoring methodologies, which can only respond to variations by sending out alarms against pre-set alarm limits. This means, of course, that the event has already occurred, and the problem is already there. The traditional optimum start/stop feature of a HEVAC control system could be greatly enhanced if it could also learn the occupancy profile of the building. For example, staff levels may be lower during a holiday period, and people may take longer lunch breaks before Christmas. As soon as these patterns are picked up, the normal operating parameters for a building can automatically adjust to daily and seasonal fluctuations in its requirements. Similarly, it might normally take an hour for your shopping centre to reach optimum temperature, but if the controller could anticipate that on Fridays and Saturdays, the building will be packed with people, all of whom generate body heat, it might decide it does not need to start the heating quite so early. Evolution of PCs Since Turing’s prediction back in the 1950s, when we saw the first computer-based building-management systems coming into use, much has changed — to our advantage. In the 1970s, centralised computers were used to control HEVAC radially wired to sensors and actuators. The 1980s saw the advent of EMS, BEMS and BMS with distributed processing and dedicated control networks. All this is a far cry from today’s systems with their embedded controllers in air-conditioning units, flexible networks, web servers and Ethernet capability. It is true that the majority of work on artificial intelligence in buildings has been in the area of predicting occupancy levels. But technology is capable of doing so much more for us — if only we will let it. James Palmer is UK business-development manager with North Building Technologies Ltd, PO Box 2673, Brighton, Sussex BN1 3US./i>
Related links:
Related articles:

modbs tv logo

The electric economy

Universities drive down carbon emissions

Kevin Pocock of Mitsubishi Electric discusses the drivers for universities to achieve zero carbon in their built environment – and the challenges.

  • Calendar Icon
    11 November, 2020, 10:00 - 12 November, 2020, 16:00
  • Calendar Icon
    20 November, 2020, 19:00 - 23:59
    CSA Awards