Given that electricity usage here is monitored with a smart meter which periodically phones home to the electricity company over GPRS, this is the sort of information they get to see:
Consumption in blue, production in green. Since these are the final meter readings, those two data series will never overlap – ya can’t consume and produce at the same time!
There’s a lot of information to be gleaned from that. The recurring 2000+ W peaks are from a 7-liter kitchen boiler (3 min every 2..3 hours). Went out for dinner on Aug 31st, so no (inductive) home cooking, and yours truly burning lots of midnight oil into Sep 1st. Also, some heavy-duty cooking on the evening of the 1st (oven dish + stove).
During the day, it’s hard to tell if anyone is at home, but evenings and nights are fairly obvious (if only by looking at the lights lit in the house!). Here’s Sep 2nd in more detail:
This one may look a bit odd, but that double high-power blip is the dish washer with its characteristic two heating cycles (whoops, colours reversed: consumption is green now).
Note that whenever there is more sun, there would be fewer consumption cycles, and hence less information to glean from this single graph. But by matching this up with other households nearby, you’d still get the same sort of information out, i.e. from known solar power vs. returned power from this household. Cloudy patterns will still match up across a small area (you can even determine the direction of the clouds!).
I don’t think there’s a concern for (what little) privacy (we have left), but it’s quite intriguing how much can be deduced from this.
There is a slight lag in smart meter reporting (a value on the P1 port every 10s). This is not an issue of the smart meter though: the DyGraphs package is only able to plot step lines with values at the start of the step, even though these values pertain to the past 10 seconds.
Speaking of which – there was a problem with the way data got stored in Redis. This is no longer an issue in this latest version of HouseMon, because I’m switching over to LevelDB, a fascinating time- and space-efficient database engine.