I used to be a Nest thermostat user. I never understood why it was called a “learning” thermostat. It seemed to learn that if I turned the temperature up to 22C during the winter then I also wanted to do that when it’s 22C outside. I’d have to manually reporgram the schedules it “learned” because they were rubbish. So I recently (2 months ago) switched to Tado. Being able to control the individual areas of the house is a bonus but it’s still not anywhere as smart as I would have expected. And this is the thing I don’t understand about these solutions - it’s cold, make it hotter. It’s hot, don’t make it hotter. Seems really simple.
So along with Tado I have various devices around my house reporting the temperature in various rooms. These are mainly z-wave multisensors and trisensors but I also have a number of devices that just happen to report the temperature and along with the multiple HAss components that tell me the weather outside I also have a Netatmo weather station. So I have lots of data to get started with.
The purpose of this post is just to see if anyone is already doing this before I start figuring out the best way. The easy way would be through HAss automations making use of the history and Bayes components. The hard way would be to create a model of when the heating was turned on and off compared to the various temperatures and other data (people in the house at the time, etc) and to query the model.
Let me know if you’ve got any pointers.
Steve
