Always on- real time?

Is the always-on watts that is shown on the bubble a real-time value? Or it something averaged over time? I have unplugged everything I can think of that would be in that category, but the bubble doesn’t report fewer watts?

Always On is a composite value based on the last 48 hours activity. Sense collects the minimum readings for the past 48 hours of Total Usage minus all identified devices and the Always On values for any smart plugs plus Hues. It then selects the 1% value (1% of the readings are lower than the 1% value) as Always On on a 1/2 second rolling basis. BTW - Dropouts are the same as having all the power turned off in your house.

So the Always On will persist for some extent in time even if you have turned EVERYTHING in the house off.

I did a little charting in another thread, and it turns out that the time things can be off before Always On goes to zero is related to the distribution of your power usage. The greater the standard deviation relative to the mean (coefficient of variation), the shorter the period of time before Sense shows Always On as zero…

Sense didn’t explain their 1% percentile method (there are many), but if it is like the quantile computation I did for my simulations, it’s a statistical calculation that is affected by distribution of your data. I did a a quick experiment with the 1% quantile function in R. I started with a 3 random normal distributions of power measurements for 48 hours (172,800 seconds), all with the same mean value (0.5kW) but different standard deviations (0.01, 0.1 and 0.2kW). Then I randomly replaced the normally distributed values with an increasing number of seconds of zeros representing dropout, from 1 second to 3600 seconds (1 hour). The results below show that if the standard deviation of your “Total Usage” or smartplug device usage is high relative to the mean value, far fewer dropouts are required to end up with an Always On of zero. That’s just the statistics of normal distributions. The blue curve shows what happens when the standard deviation of your power distribution is tight (0.01kW/10W), the orange middling (0.1kW) and red wide (0.2kW). These are all relative, so if you had a 50W mean and a 1W standard deviation, the curve would be similar to the one in blue. @serovner86 's smartplug power distribution is probably closer to the orange curve, so even a few zeros can take the Always On result down to 50% of the mean.

The blue and orange curves drop to zero at 28.8 minutes of zeros, precisely the point where 1% of all the data hits zeros. But the red curve goes to zero at 15min of zeros.

ps: This all assumes the quantile function. Plenty of other techniques can be used as well.

3 Likes

BTW welcome to the forums.

1 Like

thank you. Always-on clarified. … and way more detail than I can comprehend.

2 Likes

Every time I see @kevin1 responses I want to throw some sarcasm out there and ask why he’s always so vague.

2 Likes