Add Xiaomi Smart Purifier Elite (zhimi.airp.meb1)

[cristianchelu]

Works pretty well.

Only quirk is an unknown country_code command spammed by the MCU; device works fine despite this.

[08:20:07.959][I][app:185]: ESPHome version 2025.9.1 compiled on Oct 11 2025, 07:42:42
[08:20:07.960][I][app:187]: Project dhewg.esphome-miot version zhimi.airp.meb1
[08:20:09.798][W][miot:399]: Unknown command 'country_code'
[08:20:15.329][W][miot:399]: Unknown command 'country_code'
[08:20:20.751][W][miot:399]: Unknown command 'country_code'

Sending set_properties 9 11 17749 shuts it up with this trace:

[08:21:59.728][W][miot:399]: Unknown command 'country_code'
# `set_properties 9 11 17749` from HA
[08:22:00.684][W][miot:257]: Received property value without component: 9 11 17749
[08:22:05.908][W][miot:399]: Unknown command 'country_code\x00EU'
[08:22:06.003][W][miot:399]: Unknown command 'country_code'

and adding a select entity fails at compilation

select:
  - platform: miot
    id: country_code
    name: country-code
    miot_siid: 9
    miot_piid: 11
    options:
      "17230": CN
      "17749": EU
      "19282": KR
      "21591": TW
      "21835": UK
      "21843": US

Compiling .pioenvs/hallway-purifier/src/main.cpp.o
src/main.cpp: In function 'void setup()':
src/main.cpp:1163:36: error: narrowing conversion of '17230' from 'int' to 'unsigned char' [-Wnarrowing]
 1163 |   country_code->set_select_mappings({17230, 17749, 19282, 21591, 21835, 21843});
      |   ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
*** [.pioenvs/hallway-purifier/src/main.cpp.o] Error 1
========================= [FAILED] Took 12.77 seconds =========================

---

Also added template power/energy sensors based on kill-a-watt readings -- I can move this to the wiki if you don't want it in the standard config.

[dhewg]

Sorry, just noticed this now, I fail at notifications...

Feels like ages since I did anything with esphome, but thinking out loud:

• you want to calculate the current and accumulated power usage
• active_power has an update interval of 1s, hence you need the time /1 speed polling with the matching interval
• where does that fancy active_power formula come from?
• does the kWh sensor match the actual wall readings? I would expect some seconds to hour (60*60=3600) conversion?

[dhewg]

If you remove the quotes from the country numbers, and change the two uint8_t to uint32_t in components/miot/select/miot_select.h, does the country thing work then?

[cristianchelu]

you want to calculate the current and accumulated power usage

Correct. Like having a smart power plug attached, without the extra hardware.

active_power has an update interval of 1s, hence you need the time /1 speed polling with the matching interval

The more detailed the fan speed graph is, the better the total energy estimates get, more on that below. 1 Hz polling might be overkill or not, but I didn't dwell too much on this number.

where does that fancy active_power formula come from?

I made a spreadsheet with 1 minute average power measured at 100rpm intervals, and fitted a cubic function over the plot for the fan; then reorganized the terms for more efficient computation. I also measured the constant UV/Ionizer consumptions.

does the kWh sensor match the actual wall readings?

Here's an hour of tracking versus an Ikea inspelning power socket, the estimation stays comfortably within half a Watt of measurements. Also on the energy tracking, both the socket and the estimate energy sensor agree on 0.004kWh for the period.

I don't have a spare plug though to check over e.g. a week. I'm not aiming for 100%, just a reasonable ballpark of usage.

I would expect some seconds to hour (60*60=3600) conversion?

(mainsplaining warning, apologies in advance)
The integration sensor just does the "Area under the curve" mathematical integration from a source sensor, which is exactly the total energy consumed within whatever monitored time period. This is why having a more detailed fan speed (thus power) graph results in better energy estimation.

The same 24kWh energy consumption can be 24_000W sustained over exactly one hour or 1000W sustained over 24h, or even 48kW for half an hour. Or in way more detail video form: Power is not energy

If you remove the quotes from the country numbers, and change the two uint8_t to uint32_t in components/miot/select/miot_select.h, does the country thing work then?

I'll check this over the weekend and get back to you.

[dhewg]

Alright, nice! 1s interval may clobber the serial line, does the rpm really fluctuate that much? In the end I don't mind either way, but 3s may be a less aggressive yet sufficient default?

[cristianchelu]

1s interval may clobber the serial line, does the rpm really fluctuate
that much?

I have the same setup in my Purifier 4 working without a hitch since I first submitted a PR for it aeons ago, so I would say it's safe.

I'm not interested as much in the 10rpm differences at steady state, just need to precisely detect the large changes (i.e. 100% or nearly - to min speed) as these will affect the "area under the graph" that's summed up.

but 3s may be a less aggressive yet sufficient default?

This is a question I'll leave to nerdier people with more free time to answer. 😄

This is also part of the reason I mentioned having these optional "addon" sensors in the wiki -- just that visibility suffers there.