Hi,
Years ago I made a cheap and dirty "setback" thermostat that tried to model the heating load (as a simple passive network). The goal was to be able to determine how far in advance the setpoint would have to be changed in order to achieve the desired temperature *at* the time specified.
And, do so dynamically (i.e., look at current conditions and adjust accordingly).
At the time, I just used a crude model: plant was an ideal source (sitting behind a switch) feeding the thermal mass of the building/air (C) through some transmission losses (R) with another leakage path across the thermal mass.
This was pretty cool for oil fired hot water but a waste of instruction fetches once I moved to GFA. ("R" is too small and there is very little L).
And, I didn't have access to as much real-time (as well as predictive!) meteorological data at the time :>
So, as with most engineering tasks -- complexity grows to exceed the available expertice (this is a corollary to software growing to use all available memory :> ) -- I'm making another pass at the problem with *more* available data.
"Do nothing simply if a way can be found to make it wonderful and complex"
But, can't get *too* carried away since you still don't have that many inputs to drive the model.
Indoor and outdoor temperature are easy. As is windspeed. I've no idea how to measure and compensate for solar radiation. I can get water content in any of several forms (RH, dew point, etc.). And, if necessary, many "measurement points" inside and outside the house (though obviously T is the cheapest to get; wind speed doesn't make much sense indoors! :> )
So, how to embelish a model without going hog-wild? And, to be able to accommodate different HVAC sources/sinks?
E.g., building is still a big C. R loss/gains through the exterior walls. The walls themselves probably introduce another lag (masonary).
I expect humidity only contributes to "comfort level" (though might it also affect C?) -- though it would also play a role in determining whether to use evapoprative cooler or ACbrrr...
This is largely a "toy". A "thought experiment" to ponder how various things affect the living space.
Yes, there are fancy modeling systems that you can use (e.g., TRNSYS) but that turns it into a "job". And, I think most of those systems are designed for predicting HVAC loads not *controlling* them (?).
--don