Solution & Technology
2016-02 / STEADYMET / trading
28 Febbraio 2016
Successive forecasts over a 6 day period of time for a 8MWp PV plant in France
The 5 graphs below are showing the forecasts produced in the week from February 21rst until 28th:
– Two days around 8 am
– Two days around 8 pm
– The day around 8 am
– The day around 8 pm
– The D day around 8 am
It is a week marked by a strong meteorological instability with important fluctuations.
For the same day, the forecasts produced every 12 hours (from J-2 to the D-day 8 am) are rather different. It is particularly the case for those concerning February 28th (the last day). We also notice that the forecast supplied on D -1 (8 pm) is more accurate than the one supplied on D day at 8 am. It may be a surprise!
In fact, the most recent forecast is not systematically the best, but it is it statistically over a long period of analysis. From an operational viewpoint, it means that it is necessary to take into account this aspect. A probabilistic approach of the production forecast is then useful, on the condition obviously to be able to use all the wealth of this information.
An evolution of the EMS (Energy Management System) or of the trading strategy is then suggested for this purpose.
Figure 1: Actual (in blue) vs Forecast (in orange) D-2, 8 am – Time step: 5 minutes
Figure 2: Actual (in blue) vs Forecast (in orange) D-2, 8 pm – Time step: 5 minutes
Figure 3: Actual (in blue) vs Forecast (in orange) D-1, 8 am – Time step: 5 minutes
Figure 4: Actual (in blue) vs Forecast (in orange) D-1, 8 pm – Time step: 5 minutes
Figure 5: Actual (in blue) vs Forecast (in orange) D-, 8 pm – Time step: 5 minutes