I’ve recently purchased a TuYa power monitor clamp, connected to HA through Zigbee2Mqtt. Normally there would be two channels, one for each appliance, but thanks to funky wiring I have a two-plug faceplate with a single power input and two power outputs (washer and dryer).
I have successfully captured some metrics and I think I have an idea of which values to use, but I can’t figure out how to properly quantify it.
I’ve highlighted three areas. Pink and Gold are washer cycles, and Red is the single dryer cycle in the middle. I can see a really specific fingerprint of power usage for each, but I don’t know how to make a helper or automation (node red) that can recognise it.
The bottom graph is my attempt at using a derivative, and it almost looks like what I need, but it’s not consistent for some reason. There are large spikes in there, and some sections are missing where I’d expect to see them.
Can anyone help? I really think there is a simple solution to this, I’m just not experienced to do it.
How many cycles have you run since power monitoring?
Once you do you should be able to identify the peaks and valleys on the current power plug.
With them mixed tbh you’re going to have a really hard time differentiating when they run together…
Derivative and Reimann sum helpers are helpful for when you want to count the total amount of power used (convert power to energy for purposes of usage)
Here you’re probably looking at a series of threshold helpers (create stripes in your power use - from x to y is most likely the washer in spin) and probably having to tie them together with Bayseans.
Bayseans are useful if you have a ‘well it’s probably his but it could be that not sure…’ situation. They work off the statistical likelihood that something Is… For instance, I monitor power on the circuit that has a coffee maker and a microwave. If the mic comes on I pretty much know. It maxes the circuit. But if you turn on the coffee and one other device you get the same power draw +/- 5%. So I really don’t know.
But a Baysean looking at power as one of its conditions (power over x which I have as a template threshold sensor for the circuit denoting enough use for the kettle but not enough for the microwave ) also that I got out of bed within the last 30 minutes and the microwave threshold sensor has not otherwise been On within the last hour. With all of that AND the statistics on how much % of time my coffee maker is actually on… It’s close enough to work.
Do you have anything else going on in the house when you do laundry that you can key off of? (something like when I do laundry I always sort it on this table so I always turn on this lamp…)
Finally eventually this kind of ‘unwind my power use to the components’ will be the domain of AI but we haven’t trained the models for it… Yet.
