"Smart room-by-room heating control for homes"

Imperial MSc Computing Science students do a 3-month individual project over summer. Below is another proposal I and my Ph.D. supervisor have just submitted. Of course, there are no guarantees that any students will be interested...

Update: Dr Knottenbelt and I are planning to supervise at least two students on this project. And it looks like an industrial partner may (very kindly) be able to provide us with some PIR occupancy data for doing room occupancy prediction.

"Smart room-by-room heating control for homes"

On average, heating a home for a year costs about £600 and produces about 2.5 tonnes of CO2. Yet existing heating controls are really, really dumb and hence waste a lot of energy heating empty rooms. Smarter controls have been shown to reduce the energy required to heat a home by up to 30% and can potentially make a home more comfortable. The basic hardware setup includes:

  • Networked electronic radiator valve actuators to replace existing mechanical thermostatic radiator valve (TRV) heads. These provide room-by-room control of heating. Existing units are low cost (<£30 per room), user-installable and battery powered (e.g. the iTemp i30 (pictured below), the ELV TRV)
  • A networked boiler controller. Instead of controlling the boiler using a single timer and single thermostat, allow each room to call for heat from the boiler.
  • A "bridge" to connect the system to the home's TCP/IP network, allowing users to modify their heating schedule from a smart phone or tv or from outside the home.

Suitable software would enable features such as:

  • Room-by-room temperature schedules (e.g. heat the kitchen for breakfast and dinner; heat the living room for the evening etc)
  • Use the boiler as efficiently as possible (minimise the water temperature returning to the boiler to try to keep the boiler in "condensing mode")
  • Allow users to easily modify heating schedules from the home or over the network (e.g. from a smart phone). e.g. If you make a last-minute decision to go out for a pint then remotely tell the heating system to delay the start of the evening's heating.
  • Weather compensation (including both locally-installed external temperature sensors and downloaded forecasts to decide whether to bother putting the heating on in the morning)
  • The system should have a (user-configurable) degree of "intelligence". For example, the system could attempt to learn when particular rooms are occupied and the target temperature of preference for each occupant. Occupancy information could come from a variety of sources including: smart phone GPS (for house occupancy, not room occupancy), smart electricity meter disaggregation (if the TV is on then someone is probably in the living room), security system movement detectors, on-line calendar etc.
  • multi-user (most houses have more than one person living in them! each will have different preferences and schedules)
  • (more ideas listed here)

Commercial solutions do exist but they tend to be prohibitively expensive and/or use proprietary APIs and protocols (which means they can't be integrated into smart home systems). An open-source heating project has recently been started called OpenTRV ("TRV" stands for "thermostatic radiator valve").

Aim of the MSc project

The aim of this MSc project would be to contribute a useful "module" to the OpenTRV project. Exactly what you work on will be up to you. Possible projects include:

  • Automatically learn the thermal properties of the house and/or occupancy patterns (the temperature data might have to be simulated given that the project will run during summer!)
  • In conjunction with the "Automatically learn the thermal properties of the house and/or occupancy patterns" item, consider the simplest possible TRV UI (buttons, lights, LCD display, sound, tactile feedback, etc) and hardware (occupancy detection, window/door-open sensors, light levels, relative humidity, noise levels, etc) that could robustly deduce heating patterns with at least some way of manually forcing the valve off (other than frost protection) for holidays, and on, at least for a while, for visitors or temporary alternations in routine. Model how much energy an automatic 'setback' of a degree or so when a room is unoccupied, and what impact it could have on user comfort given the thermal mass and rad size of a room, and conversely how much it could improve convenience by needing little manual interaction. Do with a (supplied) regular occupancy pattern (eg child's bedroom) and a more sporadic or unusual one, such as maybe a student's or a night-shift worker's.
  • Model the building's physics to propose the most efficient heating schedule given a set of user-defined scheduling constraints
  • Hack / reverse-engineer an existing product so the OpenTRV project can interact with it. For example, add a ZigBee radio transceiver to the iTemp TRV either by completely replacing the "brains" with something like a Texas Instruments CS2530 ZigBee system-on-a-chip (a single chip with both the ZigBee radio, a small processor, 8KB RAM and 32-256KB flash) or by hacking the existing firmware and adding a radio transceiver module like the RFM12b (this will require some soldering)
  • Design, implement and test a secure "mesh networking" protocol to run on battery-powered radiator valves using low cost radio transceivers like the RFM12b. Needs to consume as little energy as possible to maximise battery life; but also needs to be responsive, reliable and resilient to failure of individual nodes. (Or just use ZigBee!)
  • Weigh pros and cons of simple star network using, for example, RFM12b modules vs mesh network (eg extra complexity and power drain for radios on more).
  • Research, design, implement and test algorithms to use the home's heat source as efficiently as possible. The heat source might be a modulating boiler, non-modulating boiler, heat pump, sunlight entering through south-facing windows etc. The "radiators" might be normal radiators or underfloor heating.
  • Design, implement and test the entire heating control system!
  • Review what already is already implemented and develop your own robust protocol to minimise radio on time for RX and/or TX (and thus energy consumption) for various components of the system such as TRVs, boiler interlock, any home automation gateway. Try to do this with and without the benefit of a local real-time clock, and maybe look at real microcontrollers such as PICAXE, JeeNode, etc to test the protocols and energy efficiency and timing accuracy. Look at the various trade-offs possible in battery life, system reliability, user annoyance and effort, etc...
  • Explore whether adding randomness deliberately to any of the algorithms may improve behaviour/performance, eg reduce collisions, "deadly embraces", etc, and explore what may already be in use in lower levels of any systems that you look at, eg in the radio protocols.

Some of these things are covered by patents so a patent search will be necessary during the literature review.

We can provide funding to buy hardware. And we have some basic tools to assist with reverse-engineering existing products and soldering. Some of the tasks above might sound like they are "electrical engineering" projects rather than CS projects but the electrical engineering involved in, say, adding a new radio to an existing product should be pretty minimal (and is something we can help with). The real work is the programming and whole system algorithm design and robustness.

Further reading


followed a link from the Navitron forum suplied by danakajack.

This would have been my perfect solution instead of the old fashioned solution I used on my property last year, separate 5-core wired room stats, (3 independent bedrooms, living room, 2linked bathrooms), or wired tank stat, (hot water), all calling for 'heat' via 3 2-channel Horstamnn Centaur C27 controllers).
This sounds much more elegant.

Thanks for the support!

A quick update: we have two great MSc students assigned to the heating project. They'll be working on their projects from June until Sept.

also known as mespilus on the navitron forum
Did you get to EcoBuild?
I found a stand of the 'Open therm Association'
and spoke to Jean-Francois Canard
of www.overkiz.com.
He is jf.canard@overkiz.com
They/he seemed to be concentrating on updating standard domestic 3-way valve based heating/hot water systems,
with smart radiators/valves calling for heat,
but also in conjuction with say CO2 monitors to check a human (/CO2 excreting entity) was in the room.

They may be a source of free/reduced price equipment.

Good luck.

Great, thanks for the tip!


Just came across this blog. This is what I'm about to embark on in my home. Is there any output from this project which is shareable?

Many thanks


Hi Dave,

There are some details of the student's work in this blog post.

Also, if you haven't done so already then it might be worth checking out the OpenTRV project.

Also, I have a list of existing commercial 'smart' TRV systems (although the list is a little old now!).

Best of luck with your project, Thanks, Jack

Add new comment