1

This plot shows hourly time series data of a households power usage. The house is only occupied for short periods.

What simple alg. or technique can I use to find the start of these irregularities?

The amplituede of these irregularities will mostly by above the mean of the previous hours, but the duration can vary. Also, the "human noise" will be present in these periodes.

enter image description here

NorwegianClassic
  • 145
  • 1
    https://stats.stackexchange.com/questions/1142 has many constructive suggestions that could be applied to this series. https://stats.stackexchange.com/questions/20612 looks like a duplicate question: indeed, it concerns an even more challenging version of your series (with a great many such unusual episodes to detect). https://stats.stackexchange.com/questions/261115 is also related and gives an approach based on the increase in variability in the high signal (the "human noise"). – whuber Mar 16 '22 at 13:22
  • 1
    This is great, thanks. Didn't find these while googling. – NorwegianClassic Mar 16 '22 at 13:28

0 Answers0