1

I am stumped on a problem. Say you have the last ten years of wind generation data (let's pick August) in the form of every hour of every day. How would I go about creating a set (say of 1,000) wind generation profiles using this historical data?

The wind generation profile should be 'believable' in the sense that the curves are continuous. I.e. they shouldn't jump from one hour to the next (like 100MW at 2am to 1000MW at 3am of the same day).

I have gone over some ideas like ARIMA modelling, or predicative distributions but really I am just looking to output 1,000 'scenarios' of what wind could look like in August 2023 say!

A finger in the right direction would be appreciated (as long as as it's not your middle one pointed at me).

Cheers.

micoconnell
  • 59
  • 7
  • 1
    Great question, however, it's off topic here; try stats.stackexchange.com instead. That said, an easy way to create a sequence which has the same mean value as historical data is to generate independent random values with the appropriate mean, and then apply running average smoothing -- the longer the running average, the smoother it is. If you need for the generated data to have the same variance as historical, you might need a different approach. Anyway maybe try the simple approach first and go from there. – Robert Dodier Jul 06 '22 at 02:20
  • @RobertDodier thank you for the suggestions! I will try my luck over at stats.stack. Your time is much appreciated. – micoconnell Jul 06 '22 at 14:28

0 Answers0