Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10 5/3/83; site cubsvax.UUCP
Path: utzoo!watmath!clyde!floyd!cmcl2!rocky2!cubsvax!peters
From: peters@cubsvax.UUCP
Newsgroups: net.misc,net.physics
Subject: Re: Why don't thermostats work?
Message-ID: <164@cubsvax.UUCP>
Date: Mon, 6-Feb-84 22:09:18 EST
Article-I.D.: cubsvax.164
Posted: Mon Feb  6 22:09:18 1984
Date-Received: Thu, 9-Feb-84 14:13:34 EST
References: <877@ihuxl.UUCP>, <194@heurikon.UUCP>
Organization: Columbia Univ Biology, New York City
Lines: 49

heurikon!jeff's commentary on this question is excellent... I just want
to amplify a bit.  (By the way, I used to work for Owens/Corning Fiberglas,
and though I didn't work directly in energy conservation, some of my friends
were experts in "HVAC" [afficionados will know the acronym], and I picked
up some things informally... I also had to use PID controllers (see below)
to regulate processes.)

Jeff's comment about adding an integrating function to compensate residual 
errors is correct.  A proportional controller will in general settle down
or oscillate about a temperature different from the setpoint;  the integrator
amounts to an automatic reset function.  Even this, however, works well only in
steady-state conditions;  in this case, that would mean constant outdoor 
temperatures, essentially.   By the way, such controllers are called "PI" 
controllers -- for Proportional Integrating.  For non-steady state systems, 
such as when the temperature varies outdoors, or in a chemical process, a ton of
cold reactant is added to the kettle halfway through the process, a third 
function is added, which responds to the rate of change of the error, and adds 
an extra boost of heat if all of a sudden, say, someone opens the door on a cold
winter day.  "PID" controllers, where the D stands for "Differentiating," 
incorporate this as well as the P & I functions, and let me tell you they are
a son-of-a-you-know-what to tune to a process!

Home thermostats are really only on-off sytems -- not even Proportional! --
and these tend to oscillate around the set point quite severely.  What the 
anticipator does is to heat up the bimetallic element while the couple is
calling for heat, to compensate for the time-lag  involved with the room air
diffusing into the thermocouple box.  A guy I worked with wrote his Ph. D.
thesis about modelling a home furnace/thermostat system.  As earler articles
have pointed out, it's *very* complicated.  I believe there were ten or 
fifteen terms in his model.

Now, a few more comments about how to make it better.  Industrial heating
systems work differently.  In, say, an office building, in, say, the winter,
the periphery of the building (that means near the windows, for all you
hackers) is always being heated.  The room air coming from the ceiling vents
is switched between heated and chilled air to either "buck" or augment the
peripheral heating system.  Without that, it would always be very cold near
the windows, due to radiative heating (i. e., of the cold walls by warm bodies).
These systems, unfortunately, are also difficult to "tune," or "balance," and
maintenance people don't usually know enough to do it.

Eventually, thermostats will have a "learn" cycle, in which they record
temperature changes, etc., and adjust their own parameters, perhaps on the
fly.  In additiion, if they have access to outside temperatures, together
with the information that heat flux is proportional to (T[in] -T[out]),
they should be able to do very well indeed.

{philabs,cmcl2!rocky2}!cubsvax!peters            Peter S. Shenkin 
Dept of Biol. Sci.;  Columbia Univ.;  New York, N. Y.  10027;  212-280-5517