Some of what you suggest is relatively easy to implement with Sense as one (probably the main) component in your “Energy Control Cloud” using something like IFTTT as an integrator but I understand your query regarding the intelligence aspects.
Sense ML is based necessarily on training from the electrical signal dataset and direct integration with other sensors & systems & datasets would, I believe, take the Sense team “off message”, away from the core capabilities.
A good example of this is Smart Plugs. Many (including myself) have argued that Sense should make its own Smart Plug. Why not? At first take it seems like an obvious next step but the fact is other companies make them and they have been integrated into Sense in such a way that for a small company it makes little sense to be distracted by making “little Senses”. Moving on from that “obvious” product, it’s easy to extrapolate to it being even more of a challenge to directly integrate other pre-existing datasets. Meanwhile the integration of Smart Plugs; HUEs and so on is happening gradually and with each step Sense’s capabilities expand.
As @RyanAtSense points out, Sense seems to be constantly re-thinking the Sense UI (and backend) methods in a way to enhance the integrations and where possible incorporating native intelligence. Case in point: the newly-released “Home Details” addition –
You will be reassured to know that Sense added this (it seems) based on feedback from the Community … a lot of us talked about something like that. The device list details will feedback to the ML and help with the detection and alerting possibilities. I think they will also carefully consider the Heating/Cooling/Water device categories AND consequently the interplay of these with overall energy use and their control & integration. Something I pointed out in response to the v27 release was along the lines of what I believe you are thinking. Essentially it’s taking the device disaggregation and then an intelligent re-aggregation of the energy signatures (& use) to optimize control. With the right external hooks (via IFTTT or whatever) it will enable more intelligent suggestions/controls/alerts. When you throw solar and EVs and batteries into the mix, btw, it starts to become really fun.
After some time on the Community and using Sense I start to understand that the arc of a product like this is (necessarily) a slow progressive build of features. What wasn’t immediately clear to me was that this is actually an obvious outcome of a ML-based system (or humans for that matter): Sense needs more data to learn and needs to grow to get more data; conversely, Sense (Humans) cannot outgrow the data (food).
I found this AskSense particularly enlightening (around 6:09)