Include a longer-term average to smooth out short-term usage spikes

[kme77]

The integration works very well as far as I can see and I think a lot of that is due to the averaging of the previous 2 days usage in predicting the next days consumption.
If however, one day you use a lot of energy (e.g washing etc over the weekend) then predicted usage for Monday is going to be way too high and I'm wondering if a longer term slower reacting average could be built in to 'dampen' the effects of any short-term usage spikes on the prediction? (7-10-days to always include weekend usage?)

Like I said the integration works well and so far this hasn't presented as a problem (though my usage is VERY minimal) I just thought it might be an improvement to think about. 👍