After some significant hurdles I finally installed a heat pump mini-split in my 6,000 cubic foot apartment (i.e. small space). Appropriately perhaps I chose a Mitsubishi MxZ-FH06NA for the task, the unit is rated at 6kBTU cooling (and 8kBTU heating).
I have a dedicated Sense (the Main CTs) clamped to the 208V supply phases for the outdoor unit. The indoor unit, as for most small mini-splits, is powered directly by the outdoor unit.
Sense capture has been up-and-running from the initial startup of the unit.
Meanwhile I have an iQAir AirVisual temperature+humidity+CO2+PMx particle detector INSIDE the space sampling at 5 minute intervals and a PurpleAir monitor OUTSIDE (in the shade) near enough to the outdoor unit to capture the micro-climate of the outdoor unit but not so close that it will be influenced by the heat (or cold) thrown off by the unit.
I will be posting some ongoing analysis here as I tool up to build some coherent, and hopefully useful, charts. In the meantime, here are some initial thoughts based on the first few days of data:
A mini-split presents Sense with a really complex waveform! (I am reminded of my OLED TV signature)
The ramping power is everywhere for this system. Look at the startup load for example … 7 seconds to go from zero to the initial 223W! Note: When you look at the startup signature in anything other than the full resolution and fully expanded you would think the unit simply draws 223W, boom. At full resolution you might ask yourself how Sense would determine “When exactly the unit is ON?” … remember that Sense, when doing native device detection (disaggregation) is seeing a much more complex waveform than this – the aggregation of all the devices in my house.
Sense registers almost no current spikes for any startups but there are some interesting peaks at the ends of cycles.
Looking in Signals you can see that the unit is using a single phase (120V) to power certain components. Seems like in “idle” it’s consuming around 13W (on one phase), probably for the controller boards. This should all be fairly easy to correlate with the detailed specs but it does make me think: An installer empowered with detailed Sense data, maybe even including per-phase loads at high resolution, could quickly determine component-related startup issues. A dedicated Sense in some sort of capture mode in theory has plenty of free processor cycles because it doesn’t need to be doing any disaggregation or device integration (smart plugs and so on) processing.
Here’s the first few days from the very first system activation.
The standard commissioning procedure (which I carried out) sets the thermostat to 75F and runs for 30 minutes.
Apologies for the rough rendering. I’ll work out a more coherent charting method soon.
TEAL = Sense
BLUE = inside temperature, left vertical scale
RED = inside humidity, left vertical scale
YELLOW = inside CO2, right vertical scale
ON cycle begins …
ON cycle ends …