I'm making a simple correlation study (Person's correlation) between different activities electricity consumption and the load of the overall power grid.

While it is no problem to create a correlation number for the entire year (e.g. with 8760 values if using hourly resolution), I'm stuck in how to make some kind of correlation between the highest numbers.

If I take, lets say the 150 largest values from the overall power grid and their indexes (i.e. when the grid is strained the most), how could I compare them in a smart way with the activity's electricity usage? That is, to examine to what extent the activity has its highest loads compared to the grids?

Just "cherry picking" the corresponding values from the activity using the indexes from the overall power grid and make a correlation analysis does not give any meaningful answer since it disregards the overall behavior of the both time series data.

  • 1
    $\begingroup$ Could you tell us what property of the two series this "correlation number" should reflect? $\endgroup$
    – whuber
    Jun 22, 2021 at 12:52

1 Answer 1


This is a question belonging to the field known as disaggregation. Plain correlation analysis may not yield to a reliable estimate of usage peaks and grid peaks. For example Disaggregation of Electric Appliance's Consumption Using Collected Data by Smart Metering System, uses sparse coding.

If you insist using Pearson correlation, you may want to do ranking in the peak times of the grid for initial understanding.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.