1

I have a database with time series data of different solar power plants: how strong was the sun and how much power that plant created / harvested. This data is in 15 min increments.

I would like to use data mining to get new insights and to then visualize the findings to the users.

I know this falls into the domain of data mining, but my problem is maybe more specific (dealing with time series data). So what can I extract from this kind of data or where can I read about this?

duality_
  • 17,738
  • 23
  • 77
  • 95
  • Well, the first thing you need to answer is: **what have you already got**? You won't want to find things again and again that you already know. Such as: there is more sun and more solar power output during the day than during the night. That is a really strong pattern! – Has QUIT--Anony-Mousse Oct 29 '12 at 22:40

1 Answers1

1

Time Series Analysis is a whole field in itself. That said, you can always start with a few basics and keep adding more to your analysis.

Here are a few things to try for starters from your solar power data:

  1. First, profile your solar power data. That is, calculate Min, Max, daily averages, hourly peaks and lows etc. to get a feel for the data. Plotting with x-axis as time will give you visual information.
  2. Time Series data can be decomposed into "Trend" & "Seasonality" (can be for any repeating time interval)
  3. Look for outliers, abnormalities in your data stream. Missing values, repeats etc.

If you want to learn more about time-series, (and if know R) then the forecast package is a good way to get started. (Especially this free e-book)

Any search on Time Series will take you to Prof. Hyndman's pages, and I have found the free chapters of his forecasting book very useful.

Hope that helps you get started.

Ram Narasimhan
  • 22,341
  • 5
  • 49
  • 55