Re: A research team has managed to coordinate 100 domestic air conditioners to stabilise the power grid in real time, as if they were a flexible power plant.

On 6/14/2025 2:53 PM, Carlos E.R. wrote:

That's not what they are doing. They are changing the load in tiny parts of the cycle, 50 or 60 times per second. Changing the load waveform.

No, they are deferring the start of the compressor and/or blower.
As home thermostats are not precision devices -- AND ONLY MEASURE THE
TEMPERATURE AT ONE LOCATION IN THE HOUSE'S VOLUME -- an occupant
would likely never notice if the plant turned on at its "intended"
setpoint or some delta above.
The question is whether or not the control algorithm allows the
local plant to be operated with greater *symmetric* hysteresis
or if it just lets the cooling cycle be delayed without regard
to the actual temperature.
It's just silly to rely on antiquated control approaches when
it's so easy to have smarts "at the edge".  E.g., how smart is
it to allow the thermostat to call for cooling (or heat)
one minute before the setpoint temperature is SCHEDULED
(by the user's scheduling decisions!) to go up (or down) by
an amount that likely exceeds the hysteresis in the system?
E.g., house maintaining 68F for heat and at 1 minute prior to bedtime,
the thermostat notices it is now 67.9F and calls for heat... when
the setpoint will be lowered to 62 a minute later?  (and, the
changing of the setpoint likely causing the heating cycle to be
prematurely aborted:  "Oh, the house is now at 67.93 degrees
and the setpoint is 62 so let's shutdown the plant instead of
waiting for it to attain the OLD setpoint of 68F...")
Imagine the stresses on the equipment by all this ineffective
cycling (esp the compressor!)
Ideally, you want to notice how the system behaves and tune
your controls to maximize *it's* ability to satisfy the
user's stated needs while reaping some rewards for the provider
(and, ultimately, for the user by way of reduced capital
investments)