I was trying to do some calculations starting from an energy yearly balance of a city that we are going to analyse in a research project (I was using the yearly balance because the time series data didn't arrive to me, I will receive them in one month) and I put in Homer the average electric power, but if I add a noise in the hourly or daily data, it rescales the average (it is a little difference, but I can't understand from the explanation in the only help why it happens)

When it synthesizes the year of primary load data, HOMER first creates a one-year data set that contains no noise, then it multiplies the value in each time step by a random noise factor.  The noise factors come from a normal distribution with a mean of 1, so you wouldn’t expect them to affect the average.  If we were drawing an infinite number of them, the mean really would be exactly 1, but since we are only drawing 8760 of them, their mean is bound to be a little different from 1, so they will have a tiny effect on the average.  This effect is more noticeable that usual in your file because your average load is very high so the scaled average shows a lot of significant digits.  But it is still a very small effect.

You can rescale the average to an exact value by entering a number in the scaled average input box.