Repository navigation
Cost-aware optimization layer for source selection and opportunistic preheating/precooling #257
Replies: 5 comments 12 replies
|
You've pretty much summed up a lot of my thoughts for a "holy grail" of Home Assistant HVAC control. |
|
Then lets put our minds together and build it. Ive been looking around and i think roommind is the perfect base. We do need to focus on getting a strong base (all basic functionalities working without bugs) and then the sky is the limit with this thing! |
|
Yes I think the same, roommind has the ability to do amazing things, initally I would like to see the abilty to control the heating/cooling based on excess energy being available eg available solar when its a cold day. pre-emptive heating |
|
oh hey, I was about to push a PR with the time of use / preheat and precool feature. I've just been testing it / making sure it works as intended. It's what I was already doing before I found roommind, and I wanted to add the heat decay (that's how I found roommind). I had already made it with the idea in mind that it may not be a fixed schedule because of microgrids, multi-fuel micro-grids, and certain jurisdictions where peaks are announced a few hours prior (looking at you Australia) |
|
has this been added? |
Uh oh!
There was an error while loading. Please reload this page.
Description
I would like to open a broader roadmap discussion around cost-aware optimization.
The project is already moving in a very strong direction with prediction, efficiency, and comfort. The next powerful step, in my opinion, is to add a true cost-aware layer on top of the climate logic.
Core idea
A system can be thermally efficient without being financially optimal.
In real homes, that matters a lot, especially where users have combinations such as:
• gas heating plus electric HVAC
• solar production
• low export compensation
• dynamic or variable electricity pricing
• batteries
• the ability to use the house itself as a thermal battery
The goal would be to let the controller optimize not only “how efficiently can I reach comfort?” but also “what is the lowest-cost way to reach and maintain comfort?”
This lines up very naturally with a broader central-brain architecture where later phases explicitly consider cost optimization and full energy orchestration.  
Why this matters
There are many moments where:
• electricity is cheap enough that AC heating beats gas
• gas is cheaper than electricity even when the AC COP looks decent
• solar surplus / battery makes electric heating/cooling effectively free or near-free
• very cheap or negative-price electricity makes preheating or precooling worthwhile
My own motivation comes from wanting to combine predictive climate control with broader home energy optimization and source selection. That is also very much aligned with a central brain approach where demand, source choice, and device execution are separated.  
Possible design direction
I could imagine three pillars:
• comfort
• energy efficiency
• cost efficiency
And optionally a mode or feature set that enables cost-aware behavior.
Potential parameters:
• electricity price input
• gas price input
• estimated HVAC COP behavior (Energy Meter)
• estimated Boiler gas consumption behavior (P1 meter) -> can be learned by running a calibration algorithm where you don't use hot water to see boiler cost/usage.
• solar surplus input (Energy Meter)
• house import/export input (P1 Meter)
• battery state / priority
• external EMS signal such as:
• surplus available
• low-cost window active
• avoid load now
Use case
Use cases
1. Gas vs electric source choice
Use the lowest-cost source, not only the theoretically most efficient one.
2. Solar surplus utilization
Heat or cool slightly earlier when solar would otherwise be exported cheaply.
3. Dynamic tariff exploitation
Preheat or precool during cheap windows and rely more on thermal inertia later.
4. House as thermal battery
Use predictive control to store cheap or free energy in the building mass when it makes sense.
Why I think this could be huge
If combined well with the learning and prediction side, this could become one of the strongest Home Assistant climate-control approaches available, because it would optimize for the thing most users ultimately care about: cost without sacrificing comfort.
I know this is broader and less immediately implementable than the more concrete control tickets, but I think it is worth discussing now so the architecture stays compatible with it.
I am also a developer and would be happy to help brainstorm proof-of-concepts.
Alternatives considered
No response
All reactions