Identifying baseline consumption

Question

I have data of intraday electricity consumptions (by half hours - 48 a day) over a year of 4000 households. Task is to establish baseline consumption of each of these households - possibly also differentiated on seasonality.

One way how to do this would be just taking the mean of the consumption signals. What would be more sophisticated method for this?

I would be very grateful for pointing out to methods I could look into.

Answer 1

You can try and model your data as having trend/seasonal/cyclic components if you want something a little more sophisticated. Here is an intro reference to get you started.

Answer 2

I don't think there is need for very complex methods here, I would take a look at several granularity levels and take means for these levels. One important thing is if all the households are in the same area/timezone. After establishing this I would first take the general mean and see per household how they compare to this, then do this for every halfhour segment of the day to get rid of time-of-day seasonality, do this for day of the week, month of the year etc. It really depends on what you want to know what to do with this data.