5

What metrics are usually used to quantify forecasted power error for weather-dependent renewables such as wind or PV?

I guess that there are at least two kinds of errors: where the magnitude of the forecast is wrong, and where the timing of the forecast is wrong. Something like a typical root mean squared error (RMSE) would seem to penalise a timing error twice over.

So how is the skill of a forecasting method quantified?

Flyto
  • 3,701
  • 2
  • 21
  • 31
410 gone
  • 14,714
  • 6
  • 41
  • 95
  • 1
    You can't quantify errors; if you knew the error, you could simply correct it. You can, however, quantify uncertanties. – gerrit Oct 24 '16 at 13:10
  • For both PV and wind, the capacity factor per hour is predicted - how much of the primary energy source is available vs the maximum possible. So, this essentially includes both the magnitude and timing errors you describe. – LShaver Oct 24 '16 at 19:31
  • I just came across this article, which may help answer the question for wind power: http://www.nrel.gov/docs/fy11osti/50814.pdf – LShaver Jan 23 '17 at 04:03
  • 1
    You can quantify errors (quantify just means assign a quantity to; if you are measuring anything, you'll at least know after what your error was). You can also estimate the amount of error in an estimate with some probability if you have a data set to examine. – Jean-Paul Calderone Dec 09 '17 at 00:23

1 Answers1

1

Not really an answer, but an extended comment:

Power is dispatched on a very short time scale. If I know that I'm likely to have lots of wind power at 3 p.m. today, I'll start to lower the feed rate to the coal fired boilers at 2 p.m. picking up the slack with gas turbines, shutting down the gas as the wind picks up.

Forecasts a few hours ahead are quite accurate. (I routinely get rain forecasts that are accurate to within 15 minutes 6 hours ahead) I don't think short term forecasts are much of a problem.

Long term, you are working with climate data. Your non-renewables need to be able to cope with full renewable shutdown anyway, so any renewables you get just decrease your operating expenses.

PV is more predictable. You know the envelope -- no power at night. And even cloudy days produce reasonable amounts of power. (Germany with it's gloomy skies still generates about 800 kWh/year/installed kW while Arizona doesn't quite make double that.

Sherwood Botsford
  • 7,227
  • 17
  • 41