2
$\begingroup$

I am wondering what is the best way to extract knowledge about periodic time series data?

In my case, I am trying to analyse historical hourly electricity prices to get information about its volatility, when the peaks occur, what is its distribution among the hours of the day or something similar to be able to make good decisions on when to throttle devices to save money and stay energy efficient.

My original idea was to simply aggregate the time series by calculating mean values for each hour over a period of several months and then sort them to pick the top M% hours and say that they are the most expensive on the average and should be avoided.

This is a very crude approach, so I'm wondering whether I'm making some obvious mistake, disregarding some information? Is there some better approach? It seems to me that the information on how high the current prices are in comparison to the average should also be used to sometimes exclude, say, 3 hours from a day and on other times only 1 hour.

$\endgroup$
  • $\begingroup$ How well-versed are you with time-series analysis? I'm not an expert on ts, but looking for regular periods (often called seasonality) is fairly standard. $\endgroup$ – gung Sep 6 '12 at 16:21
  • $\begingroup$ You could say I'm a beginner. I have seen the basic moving average method to detect seasonal impacts before and I'm aware that something similar (maybe a more advanced method) could be used to get the info I'm looking for, but I was hoping to get some input from somebody experienced about the state-of-the-art methods, not to wander in the wrong direction. $\endgroup$ – metakermit Sep 9 '12 at 16:50
  • $\begingroup$ I'm not an expert on time-series; I know a little bit about it & I'm under the impression that it's more involved / difficult than (say) standard regression methods. You will have to learn a good bit to do it well; I doubt you could just pick up a tip on CV & then go knock out a good model. You may want to work w/ a statistical consultant. This question: books-for-self-studying-time-series-analysis may help you get started w learning about ts. $\endgroup$ – gung Sep 9 '12 at 16:59
  • $\begingroup$ OK, thanks a lot, gung! I guess for now I will use a simple solution and slowly study the field some more to get a better grasp of it. $\endgroup$ – metakermit Sep 10 '12 at 13:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.