How confidence intervals can be used to properly assess short-term solar power production ? Case study for an island electrical system

18 Novembre 2015

The Gran Canary Island, located in the Canarias archipelago, is not interconnected to any continental grid. From an electrical viewpoint, it is an isolated system very sensitive to production hazards.

With over 35 MWp of photovoltaic installations deployed in the island, the variability of this energy source can affect the network stability.

Steadysun is updating its solar production forecasts all along the day thanks to satellite images. The forecasts include an information on the level of uncertainty with the use of probabilistic percentiles from 10% to 90%. These confidence intervals allow a proper assessment of the possible short term evolutions (typically one hour)

The day of November 18th , 2015 is a good example:

During the whole morning, the provided information shows a quite high production variability risk. Between 9.30 am and 10.00 am the actual is moving away from the most probable curve (P50%) to fluctuate between P10% and P40%. One can notice, then, a production increase by nearly 50% (from 11MW to 16 MW) within 15 minutes.

Afterwards, the uncertainty corridor is pretty narrow, which means a stable cloudy situation with a low probability of rapid fluctuation of the photovoltaic production.


1 hour ahead solar forecast Gran Canaria

Figure 1: Actual (in blue) vs Forecast (in red/orange for the percentiles) – 60 minutes ahead, time step: 15 minutes