Systems Research Group – NetOS
The progress in LED devices and the growing acceptance of modified power distribution for lighting give rise to compelling cases for lighting system overhaul. For most buildings this will be a one-time opportunity. This opportunity allows one to include configurable control and monitoring at low marginal cost; the system can be a platform for innovation, not just for lighting, but in understanding and incorporating user behaviour in design and control of all building energy use.
There are a variety of advantages that 24V DC power distribution for LED lighting has over 120V or 240V AC. The power is more easily and efficiently converted to the constant current sources required by LEDs and the same can be said for powering processors and sensors co-located with luminaires. Where existing wiring separates lighting circuits from mains rings, lighting wiring may be reused (as long as the entire circuit is converted!) giving a low capacity communication channel which is now separated from the mains. Battery back-up is straightforward; all lighting is emergency lighting. Integration with renewables - particularly solar - is straightforward avoiding conversion to and from mains level AC. There is nothing magic about 24V and some will argue for higher voltages to reduce transmission losses, but the availability of automotive qualified components, the increased reliability of ceramic capacitors at lower voltages, and the efficiency of converstion arguments should be considered as well (and two batteries are smaller than four of half the capacity).
So where's the Computer Science?
To a first order, who cares? This is more about thinking at a systems level, escaping the notion that we should use new technology in the same we used the old - the "horseless carriage" approach. Take appart an LED lightbulb that plugs into the mains when it fails (and it will). Most of it is a power supply. Not surprisingly it is not a very good power supply - converting 240VAC into 3V or 6V DC in that tiny space is not easy. The power supply breaks, not the LED. The notion of light bulb will fade away.
To a second order, there is lots. Luminaires last a long time. Adding a processor and sensors to every light is entirely reasonable. Flexible communication, control, inference, correctness, privacy and an open platform. This was kicked off to some extent by work in the C-Aware project where we quickly realised we needed better contextual information on user behaviour in order to understand energy use, for example to be able to classify use as waste or to be able to find more efficient ways of realising the same outcome. As a trivial example, if no one enters a room for known periods of time does it need to be heated?
How are we investigating this? Essentially by building a pilot, learning as we go. This is a systems project so we use components when available and build when not. There is a pilot network gently rolling out in the William Gates Building (the home of the Computer Laboratory).
More to come....