I'm having a loft conversion done at present, and I'd like to take the
opportunity to upgrade my central heating control system, which is
currently pretty dumb (3-on 3-off per day controller for hot water and
heating, with thermostatic radiator valves). I'm looking for some
software that will let me run much more sophisticated control
algorithms. Some of the things I'd like to be able to do a
- control three heating zones plus hot water
- allow different control temperatures for each zone at different
times of day and days of the week
- correct for outside temperature, wind speed and direction etc.
- don't heat rooms when they are unoccupied
- don't heat rooms when the windows are open
- heat only as much water as is expected to be required for the time
of day (i.e. don't heat a full tank unless people are expected to be
bathing or the washing is being done)
- use oil, off-peak electricity and full-price electricity in the most
cost effective manner
- automatically adjust boiler operating temperature for optimum
efficiency (i.e. lower the boiler temperature when the heat demand is
lower)
- learn house warm-up time under various conditions and come on "just
in time" to achieve the required temperature
- easy to use and flexible manual override facilities via wall-mounted
control panels, a web based interface or by phone (e.g. hotter, cooler,
on, off, come on for 1 hour etc.)
- full logging of temperatures, boiler firing times and valve opening
times so that performance can be analysed and improved
- interface to temperature sensors, e.g. Dallas One-wire sensors
My vision is to run the control software on a small solid state single
board computer with the necessary I/O (interface to temperature and
other sensors, mains relays for zone valves and pumps, wall mounted
control panels, web server for network based contol etc.), with the
control program developed on a PC and downloaded to the SBC when the
algorithms need to change.
What software exists which can help implement this sort of control?
What languages are best for implementing the algoithms?
Who else is working in this field?
What is the best hardware?

Thanks for your help - Rowan

On Tue, 03 Jan 2006 15:18:53 -0800, rowan wrote:

What software exists which can help implement this sort of control?
What languages are best for implementing the algoithms?
Who else is working in this field?
What is the best hardware?


From a recent discussion on our local Linux Users' Group:

http://www.o2m8.com/

http://www.automatedhome.co.uk/

wrote:

Oh another one!

There is another thread running on this he

http://groups.google.com/group/uk.d-...4ea70c 6c5acb
http://tinyurl.com/9tvhd

I'm having a loft conversion done at present, and I'd like to take the

Does that mean you will be losing storage tanks etc?

opportunity to upgrade my central heating control system, which is
currently pretty dumb (3-on 3-off per day controller for hot water and
heating, with thermostatic radiator valves). I'm looking for some
software that will let me run much more sophisticated control


algorithms. Some of the things I'd like to be able to do a

- control three heating zones plus hot water
- allow different control temperatures for each zone at different
times of day and days of the week
- correct for outside temperature, wind speed and direction etc.
- learn house warm-up time under various conditions and come on "just
in time" to achieve the required temperature
- easy to use and flexible manual override facilities via wall-mounted
control panels, a web based interface or by phone (e.g. hotter, cooler,
on, off, come on for 1 hour etc.)

All this can be done with standard off the shelf control kit withotu
needing to get too cleaver (unless you want to!)

S Plan+ system with some programmable (optimising) thermostats would do
that. Add a external weather compensator etc.

- don't heat rooms when they are unoccupied
- don't heat rooms when the windows are open
- heat only as much water as is expected to be required for the time
of day (i.e. don't heat a full tank unless people are expected to be
bathing or the washing is being done)

- use oil, off-peak electricity and full-price electricity in the most
cost effective manner

Ignore electric while the oil tank is not empty then ;-)

- automatically adjust boiler operating temperature for optimum
efficiency (i.e. lower the boiler temperature when the heat demand is
lower)

This may depend on how the temperature is adjusted on your boiler - on
many it is a hard wired knob which is not that well suited to automated
control.

My vision is to run the control software on a small solid state single
board computer with the necessary I/O (interface to temperature and
other sensors, mains relays for zone valves and pumps, wall mounted
control panels, web server for network based contol etc.), with the
control program developed on a PC and downloaded to the SBC when the
algorithms need to change.

Small embedded PCs with flash file systems are ideal for this sort of
application and can be bought off the shelf. You could in reality do it
with just about anything (old micro computer of some description) since
you have very low computational requirements.

You can mix in some X10 automation stuff as required for some of the
applications, and a smart security system like:

http://www.tlc-direct.co.uk/Main_Ind...ain/index.html

will handle the GSM and Telephone interfaces to the home automation kit
to give you the non web based remote access.

What software exists which can help implement this sort of control?
What languages are best for implementing the algoithms?

Whatever you can: a) get a development system for that suits your chosen
target and b) that you feel comfortable developing in.

Who else is working in this field?

Have a look at some of the many X10 and home automation sites.

What is the best hardware?

Define best?

--
Cheers,

John.

/================================================== ===============\
| Internode Ltd - http://www.internode.co.uk |
|-----------------------------------------------------------------|
| John Rumm - john(at)internode(dot)co(dot)uk |
\================================================= ================/

"John Rumm" wrote :


mega snip
Have a look at some of the many X10 and home automation sites.

What is the best hardware?

Define best?


You need to be very careful with projects like this, especially when you
have loads of sensors and functions. They introduce very many corner cases
and single points of failure. You will need to do a lot of very methodical
testing. What happens when a sensor fails? Can you detect it and act
appropriately? In the system I am designing (using C, by the way) it is
designed to fail passive, reverting to the old control system.

Most SBC solutions use cooperative multitasking. This requires careful
coding in order not to break it.

Also, plan what to do if you decide to move house. A homebrew heating
control system is not likely to add to the value of the house. In my case I
shall be able to revert to a normal if somewhat outdated control system at
the drop of a hat.

Dallas 1-wire sensors are OK, but bear in mind that for any significant
distance you'll need twisted pair cabling. Also think carefully about the
interface to them. If interfacing directly to an SBC the timing is critical,
and you'll need interrupts turned off for much longer than is desirable on a
multitasking SBC. Use the 18S20's via an I2C interface chip instead.

I'd advise looking carefully at risks versus benefits wrt the complexity of
your project. Does taking wind speed and direction into account (for
example) result in measurable benefit and what is the risk in terms of
software complexity?

If you want to go ahead, get yourself an evaluation kit and start playing
and learning. This will help you understand the magnitude of the task ahead.

HTH,
Steve S

Steve S wrote:
"John Rumm" wrote :


mega snip
Have a look at some of the many X10 and home automation sites.

What is the best hardware?

Define best?


You need to be very careful with projects like this, especially when you
have loads of sensors and functions. They introduce very many corner cases
and single points of failure. You will need to do a lot of very methodical
testing. What happens when a sensor fails? Can you detect it and act
appropriately? In the system I am designing (using C, by the way) it is
designed to fail passive, reverting to the old control system.

Most SBC solutions use cooperative multitasking. This requires careful
coding in order not to break it.

Also, plan what to do if you decide to move house. A homebrew heating
control system is not likely to add to the value of the house. In my case I
shall be able to revert to a normal if somewhat outdated control system at
the drop of a hat.

Dallas 1-wire sensors are OK, but bear in mind that for any significant
distance you'll need twisted pair cabling. Also think carefully about the
interface to them. If interfacing directly to an SBC the timing is critical,
and you'll need interrupts turned off for much longer than is desirable on a
multitasking SBC. Use the 18S20's via an I2C interface chip instead.

I'd advise looking carefully at risks versus benefits wrt the complexity of
your project. Does taking wind speed and direction into account (for
example) result in measurable benefit and what is the risk in terms of
software complexity?

If you want to go ahead, get yourself an evaluation kit and start playing
and learning. This will help you understand the magnitude of the task ahead.

HTH,
Steve S

The majority of the OP's desired functions are available on commercial
heating control systems already. However theyre not cheap.

For the computer side of things, I would steer well clear of PCs, as
you need a system that will stay running for years at a time, and PCs
simply dont have the necessary hardware or software reliability. My
first choice would likely be a BBC B. If youre not familiar with them,
theyre 1982 32k 4MHz command line machines with inbuilt OS and BASIC,
and are stable as a rock. They also have lots of A./D I/O onboard, and
are easy to program, so are popular for scientific monitoring and
control projects. Despite being hopelessly out of date technically,
they still retain a niche market for these sort of apps. E-bay. Burn
your software to EPROM, and the EPROM plugs into the machine. Add
sideways RAM if you want more RAM, or get a BBC Master instead (128k).

And, as has been said, the system must fail to functional. If it
doesnt, youre buying yourself a whole lotta trouble downline, and
trouble costs money.


NT

wrote:My
first choice would likely be a BBC B. If youre not familiar with them,
theyre 1982 32k 4MHz command line machines with inbuilt OS and BASIC,
and are stable as a rock. They also have lots of A./D I/O onboard, ...

snip

I'd avoid onboard A/D like the plague. The distances involved are much too
great to use analogue signals reliably.

Steve S

Steve S wrote:
wrote:My
first choice would likely be a BBC B. If youre not familiar with them,
theyre 1982 32k 4MHz command line machines with inbuilt OS and BASIC,
and are stable as a rock. They also have lots of A./D I/O onboard, ...

snip

I'd avoid onboard A/D like the plague. The distances involved are much too
great to use analogue signals reliably.

Steve S

I would have to disagree, having routed analogue round houses and
commercial premises with unshielded untwisted cable, and had nice
consistently clean signals. There are various ways, but the simplest
for these slow moving signals would be to put a cap across the line at
the computer end.

NT

wrote:

For the computer side of things, I would steer well clear of PCs, as
you need a system that will stay running for years at a time, and PCs
simply dont have the necessary hardware or software reliability. My

I don't think that is true. Embedded PC hardware is designed for high
availability applications and should run flawlessly for long periods.
Remember also there is no reason you need to stick anything Microsoft on
top of it - you can even program for the bare platform if you really want.

So software reliability is down to you.

There are also plenty of real time and/or embedded OSs available that
can be used (anything from Linux, to VxWorks, Psos, OS/9 etc). To be
fair to MS, even embedded Win2k or WinXP is actually pretty solid in
most applications - remember most of the clunky UI related crud is
missing and you are back to the well designed (i.e. Dave Cutler and co)
kernel at the heart of it.

OS choice will depend on the sophistication of the application and the
need to support more complex items like TCP/IP stacks and industry
standard file systems.

first choice would likely be a BBC B. If youre not familiar with them,
theyre 1982 32k 4MHz command line machines with inbuilt OS and BASIC,
and are stable as a rock. They also have lots of A./D I/O onboard, and

Lots of them are failing now - just due to age alas. Many with PSU
faults attributable to failing capacitors. I am not sure I would want to
spec a modern design with a system that is no longer supported or
available.

The IO they had was good at the time - but perhaps less use than might
be expected in a modern context. Lack of TCP/IP, X10, Bluetooth, USB,
WiFi stacks and interface logic may prove a bigger hurdle.



--
Cheers,

John.

/================================================== ===============\
| Internode Ltd - http://www.internode.co.uk |
|-----------------------------------------------------------------|
| John Rumm - john(at)internode(dot)co(dot)uk |
\================================================= ================/

John Rumm wrote:
wrote:

For the computer side of things, I would steer well clear of PCs, as
you need a system that will stay running for years at a time, and PCs
simply dont have the necessary hardware or software reliability. My

I don't think that is true. Embedded PC hardware is designed for high
availability applications and should run flawlessly for long periods.

Maybe you're right. But OP needs to bear in mind the system needs to
run reliably for at least 30 years, and servers are not designed to
that high a reliability level. You'd need to source that PC very
carefully. Or perhaps have a backup available with no need for extra
setup work.


Remember also there is no reason you need to stick anything Microsoft on
top of it - you can even program for the bare platform if you really want.

So software reliability is down to you.

I took it as evident MS products would not feature in this one. Linux
is AFAIK not nearly stable enough to run for 30 years without
maintenance, or even 3. Perhaps there are PC OSes that are, and I just
dont know of them.


first choice would likely be a BBC B. If youre not familiar with them,
theyre 1982 32k 4MHz command line machines with inbuilt OS and BASIC,
and are stable as a rock. They also have lots of A./D I/O onboard, and

Lots of them are failing now - just due to age alas. Many with PSU
faults attributable to failing capacitors. I am not sure I would want to
spec a modern design with a system that is no longer supported or
available.

Todays PCs will be no longer supported or available when they go wrong,
so I'm not sure how much difference there. The 2 plusses with BBCs is
1. they have an ongoing niche market, making replacement easy and
minimal cost.
2. They need almost no setup: just plug the EPROM in and switch on.


The IO they had was good at the time - but perhaps less use than might
be expected in a modern context. Lack of TCP/IP, X10, Bluetooth, USB,
WiFi stacks and interface logic may prove a bigger hurdle.

I guess so, had forgotten about the web interface requirement.


NT

In article . com,
writes:
For the computer side of things, I would steer well clear of PCs, as
you need a system that will stay running for years at a time, and PCs
simply dont have the necessary hardware or software reliability. My

Um, they do, providing you aren't running anything Microsoft-based.
I run my system on an old 120MHz Pentium (which is still _way_ over
powered for the task) using Solaris x86, and it has never gone down
except when there's a power cut, which is roughly once every 400
days so far.

And, as has been said, the system must fail to functional. If it
doesnt, youre buying yourself a whole lotta trouble downline, and
trouble costs money.

Or fail-safe and alert you, which is what mine would do (hasn't
ever done so yet, except when deliberately engineered to do so
to test it works correctly). I also watchdog the system from the
burglar alarm, which will alert me if it fails, and can remotely
reboot it if required (again, never required as yet).

--
Andrew Gabriel

Andrew Gabriel wrote:
In article . com,
writes:

For the computer side of things, I would steer well clear of PCs, as
you need a system that will stay running for years at a time, and PCs
simply dont have the necessary hardware or software reliability. My

Um, they do, providing you aren't running anything Microsoft-based.
I run my system on an old 120MHz Pentium (which is still _way_ over
powered for the task) using Solaris x86, and it has never gone down
except when there's a power cut, which is roughly once every 400
days so far.

And, as has been said, the system must fail to functional. If it
doesnt, youre buying yourself a whole lotta trouble downline, and
trouble costs money.

Or fail-safe and alert you, which is what mine would do (hasn't
ever done so yet, except when deliberately engineered to do so
to test it works correctly). I also watchdog the system from the
burglar alarm, which will alert me if it fails, and can remotely
reboot it if required (again, never required as yet).


The point of 'fail to functional' is that the computer isnt relied on
at all, and if it dies you dont have a problem to solve. I would not
want to rely for the coming 30 years or more on the right person being
there at the right time, remembering the right things and having the
right kit. Imho its important that the system continue working,
presumably with a more basic control system. If not set up this way,
youre just building in a problem.

A way to achieve this is to use a simple hardware control system and
use the secondary PC system to tweak the first control system. If the
PC dies, no further tweaks, it keeps running, with more basic control.


NT

In message , Andrew Gabriel
writes
In article . com,
writes:
For the computer side of things, I would steer well clear of PCs, as
you need a system that will stay running for years at a time, and PCs
simply dont have the necessary hardware or software reliability. My

Um, they do, providing you aren't running anything Microsoft-based.
I run my system on an old 120MHz Pentium (which is still _way_ over
powered for the task) using Solaris x86, and it has never gone down
except when there's a power cut, which is roughly once every 400
days so far.

You're going to have to go out and buy a 2nd hand Prius

--
geoff

On 3 Jan 2006 15:18:53 -0800, wrote in message
.com:

I'm having a loft conversion done at present, and I'd like to take the
opportunity to upgrade my central heating control system,

lots of interesting questions snipped

What is the best hardware?

Good question. For example, it would (IMO) be inappropriate in most cases to
substitute a computer-operated low-water-shutoff when extremely reliable
electromechanical ones have long been available inexpensively and are
installed and understood by HVAC professionals.

So consider accomplishing as much as you can with conventional equipment and
use uP/plc/pc-control only where you have to. So, for example, use
conventional, commercially available communicating thermostats (Aprilaire,
RCS, etc) and have the setpoints varied under software control.

The arena in which software decision making may be most appropriate is in
realtime fuel-switching decisions. For example, in some places in US, the
cost of electricity depends on time of day, time of year and total usage.
Last month, the doubling of the price of natural gas over the last year in
my area (Central US) has led to resistance electrical heating being less
expensive than the most efficient possible heating with natural gas. It has
been less expensive than heating oil and propane for some time. Use of heat
pumps (instead of resistance) cuts the energy cost roughly in half again and
geothermal in quarter under some circumstances. Each has different advantage
and ranges of applicability, as well as and installation and maintenance
costs.

The software I use (CyberHouse by Savoy, now not available to consumer) has
an energy use module that can download energy cost from the web in real
time, monitor power consumption on nine circuits and control Cutler Hammer
circuit breakers. Savoy's efforts in this area were funded in part by US
Government grant, and the original installations were in Texas in about
1999.

(This question was crossposted to multiple groups, which may result in some
confusion.)

.... Marc
Marc_F_Hult
www.ECOntrol.org

Marc F Hult wrote:


The software I use (CyberHouse by Savoy, now not available to consumer) has
an energy use module that can download energy cost from the web in real
time, monitor power consumption on nine circuits and control Cutler Hammer
circuit breakers. Savoy's efforts in this area were funded in part by US
Government grant, and the original installations were in Texas in about
1999.



Marc,
Do you also have an automation panel or is your system all PC software
driven?

Gary

On Thu, 05 Jan 2006 21:14:12 -0500, ChainSmoker
wrote in message
:

Marc F Hult wrote:


The software I use (CyberHouse by Savoy, now not available to consumer)
has
an energy use module that can download energy cost from the web in real
time, monitor power consumption on nine circuits and control Cutler
Hammer
circuit breakers. Savoy's efforts in this area were funded in part by US
Government grant, and the original installations were in Texas in about
1999.



Marc,
Do you also have an automation panel or is your system all PC software
driven?

Gary

Neither, although I recently purchased an Elk M1G automation panel which is
not yet installed.

Since 1999, the system has consisted in autonomous devices that communicate
with and through pc-based software.

Subsystems (Security = Napco 9600; Thermostats =
Enerzone/Statnet/Aprilaire; IR = Nirvis Slinke; Lighting = X10 + DIY;
Analog and digital I/O with controller = Elk MM443S Magic Modules; others)
all work whether or not the PC is running.

So it is not "PC software driven" but, rather, 'PC software coordinated' in
what I describe as a 'federated system' and is also described as
"PC-centric". www.premisesystems.com and www.charmedquark.com are examples
of such systems. Www.homeseer.com has also matured from its original
X10tricity to encompass many devices. ( X10sic systems in general are
hometoys and aren't dependable in my opinion and experience.)

IMO, a system that depends on a one-off program written and maintained by
one mortal individual is fragile indeed -- no matter how well executed.

A 'federated system' is intrinsically more robust than a pc-based in
important respects and is usually more practical to maintain. The HVAC
repair folks have a thermostat that they understand and can service, the
alarm monitoring company interfaces with a security panel it is used to --
and so on.

.... Marc
Marc_F_Hult
www.ECOntrol.org