I’ve had my Sense running for about 4 days now, and I’m happy to report that it’s already identifying devices. The first one, a chest freezer, showed up about 30 hours after I hooked Sense up. That one made a lot of sense, as it was pretty frequent and had a very distinctive pattern, even to my human eyes looking at the graph.
Since then, it’s identified another device, which showed up on day 3 as an unknown heating element. I’m quite certain that it’s picking up my rice cooker, but I’m also thinking that Sense has it confused with another heating device, namely the coffee maker. If I had to guess, I think Sense first latched onto the coffee maker, but then switched its match to the rice cooker despite a significantly different power draw. As it is, either device seems capable of triggering a match to “Heat 1”, though I’ll have to make more coffee to be sure.
This, in and of itself, isn’t the problem. I expect to see some limitations in Sense’s ability to distinguish devices, especially early in the product’s life and early in its time in my home. What vexes me more, however, is the accounting for this heat source in the app data. I’ve uploaded a couple screenshots to show what I’m referring to.
The top chart is the overall chart from this morning that shows the blocky pattern of both devices. The first two spikes (about 1150 W) are the end of coffee maker’s warming plate cycling. I turned it off after the second spike. Then I started the rice cooker. As you can see, the spikes from that device (about every 30 sec) are on the order of 435 W.
The bottom chart is the data for the same time period, taken just from the “Heat 1” device history. It’s clear that we’re looking at the same pattern in time. But the size of the spikes makes no sense. Those 435-W stairsteps are now much higher, and all exactly 1146 W at the tail end. Whatever the Sense app is doing to assign wattage to the device, it’s clearly not just the delta from the surrounding “baseline”.
So what this means is that I’ve got a massive overreporting of this device’s power consumption. It’s obviously not using 1150 W every time it’s on, but that’s what the “Heat 1” device statistics say. This device (when it’s the rice cooker, anyway) is using well under half of that. That, to me, is an obvious problem when trying to identify your power consumers.
So now that Sense has learned this device and thinks it has it all figured out, will it continue to try to learn this same device better? Or does it say “This is a 1.1-kW heater whether it likes it or not, and I’m done messing with it”? I can handle some early wonkiness, but what I don’t know is Sense’s track record of refining already learned devices.
Thanks for reading all that, and for any insight you can provide.
