Better in correlation than without the hour offset, but still not the max 1-2% difference you should be seeing.
3 thoughts:
- Try this form of an x-y plot, just comparing the two usage values at the same time. This type of x-y plot is going to be useful when you are comparing hundreds or thousands of readings and looking for patterns. Your comparison chart should look the one below once you get everything figured out.
-
double check the time base that your utility uses, to understand where this offset is coming from. I have seen some cases where the utility actually does 15min usage updates and the way the intervals are defined can lead to a 45min offset with respect to Sense. The error, even after the time shift, could still be consistent with a time base offset issue, though if you don’t get the 15 min data from your utility, there might not be any further you can go along these lines.
-
try to also figure out whether there is any way to tell when your utility is doing estimated reads of your smart meter. Together with other users, we have figured out how PG&E and NV Energy identify estimated reads in their exported data. Your time chart and associated error are consistent with estimated reads mixed in with actual reads like we saw in the NV Energy case. Example of a time comparison waveform showing many NV Energy estimated reads below - red is Sense, blue is utility reading. Dots are colored based on how many actual (vs. estimated) reads are done in an hour (4 reads of the meter per hour).
ps: Seems like you have very small solar component. In most cases with solar, I see several hours per day with negative net usage on Sense and utility.