Hard code room air conditioning capacities

[jbrossman]

In future iterations of ResStock, we should consider modeling room air conditioners differently in both baseline and upgrades. We should move from modeling them to meet a certain percentage of load, to modeling them at a certain capacity instead and seeing how much of the load is being met. The reasoning for this change is to more accurately show unmet hours, and to get closer to reality in which a person buys 1 room ac, they don't think about the cooling load of their home and then buy a certain size room ac.

Relevant details or previous work on this topic:

• looked at RECS 2020 data which has number of room air conditioners that people buy, very strong correlation with number of bedrooms. 0-1 bedroom buys 1 room ac, 2-5 bedrooms is 2 room ac, and 6+ bedrooms is 3 room ac
• joe created test branch (https://github.com/NREL/resstock/tree/hardsize-room-acs) where some initial work was done to change room ac to hard sized capacity in both baseline and upgrade scenarios
• multiple instances where unmet hours is being used as a thermal comfort metric instead of evaluating HVAC performance, so results are not making sense

[shorowit]

I don't know that I agree with this proposal. The fraction load served is a proxy for how much of home's floor area is served by the room AC. If you live in, say, a 800 sqft 1-br home in Miami and had a single 1-ton room AC in the bedroom, it might be possible to meet the entire cooling load year round according to our model, but in reality the room AC only throws the air so far and will only condition a small fraction of the home's floor area (i.e., the room it's in and a bit of overflow).

Also, real operation of room ACs can be difficult to model because they can be highly dependent on how much occupants are home and how much they turn them on. Central systems are left on all the time but room ACs are often not. That might be a reason to reduce the fraction load served even further.

[jbrossman]

Thanks for your perspective Scott, and I agree with many of your points. As you said above, modeling actual operation of room air conditioners is highly dependent upon many variables, some of them we can and cannot capture in our modeling. You know this better than anyone.

The original point of this was to better capture unmet hours, which I've seen people use as more of a thermal comfort metric instead of HVAC performance. It's an interesting dynamic, and hard to explain to non-modelers, that we have a single zone model, but that there are zero to few unmet hours for the whole home when the home is only a certain percent conditioned. I don't know of the right answer in this moment, how to accurately model room ac usage while also keeping a metric that both building scientists need but that our users are using as a thermal comfort metric.

This issue is more of a placeholder to reconsider in the future when we may have core funding available to make these types of changes. I've been thinking a lot about room ac modeling as part of this upcoming SDR, and Andrew encouraged me to make an issue for future reference if this comes up again.

[shorowit]

I agree that the reported unmet hours (or indoor temperatures) for partial space conditioning equipment are not ideal. But I'm pretty doubtful that there is any approach (short of multi-zone modeling, which brings its own complexities and is still not perfect) that can address the issue. Certainly we wouldn't want an approach that improves unmet hour reporting at the expense of accuracy of energy use predictions, which is what I think the above proposal would do.

(Also, is it really an unmet hour if the occupant is only trying to condition a portion of the home?)