Solution & Technology
How solar power forecasting can limit the impact of the solar variability on the grid management ?
June 18th, 2015
In Germany, the last 17 and 18 of June 2015 were very different days in terms of solar photovoltaic production
On June 17th, the solar power exceeded 22GW in the middle of the day as compared to the following day during which it reached a ceiling at 10 GW only. These variations from one day to another may affect the grid and the cross-border exchanges, if not properly anticipated.
The following graphs show the German countrywide production (blue) vs. the forecast (orange) proposed by Steadysun the day before. The error level is very low.
Figure 1: Forecast vs Actual for Germany, on June 17th and 18th, 2015
Now, should one look at each TSO level, although the trend remain the same, one can notice less smooth production profiles. The accuracy level is a little degraded. It is particularly noticeable for the 50Hertz area of responsibility in the middle of the day.
At country level the forecast is benefiting from a statistical “cloaking” effect (1) which partly compensate the errors between the four areas.
Figure 2: Forecast vs Actual for TSO 50Hertz, on June 17th and 18th, 2015
Figure 3: Forecast vs Actual for TSO Amprion, on June 17th and 18th, 2015.
Figure 4: Forecast vs Actual for TSO TransnetBW, on June 17th and 18th, 2015.
Figure 5: Forecast vs Actual for TSO Tennet, on June 17th and 18th, 2015.
(1) which can be define as an obfuscation of errors due to statistical distribution.