Though, if someone has the time and technical ability to implement this kind of surveillance then there are probably easier ways to get better info than inferring activity from energy readings.
I don't know, the original mapping probably isn't the most difficult thing in the world, and then you can just passively (programmatically) wait for oddities in consumption patterns against an entire neighbourhood.
You can have "small hands" doing the initial survey (e.g. triangulate signal strengths to meter locations), automate the surveillance/pattern checking, and call in more professional break-in teams when you have high confidence.