Page 135 - Advances in Renewable Energies and Power Technologies

108 CHAPTER 3 Forecasting of Intermittent Solar Energy Resource

of NWP based on a postprocessing technique to reduce systematic forecast errors.
The MOS technique consists in using ground irradiance measurements to correct
localized errors from NWP models. The technique gives very good results and
should allow to generate a global and valuable method coupling machine learning
and NWP.
7.2.2 Nonhydrostatic Atmospheric Models
The goal of the NWP use is to produce a model of solar irradiation with an error
below 25% (nRMSE). Actually, the global models are not efficient but in the future,
local models based on nonhydrostatic equations will allow to obtain very good fore-
casts (for all the horizons). Note that, with this kind of models, the time and spatial
resolution can be increased and the global irradiation is available for an overall ter-
ritory with a resolution upper than 500 m.
7.2.3 High Accuracy Irradiance Measurement
For all the models tested it is important to validate the output with ground measure-
ment. The problem is obvious, but there is not enough installed sensors to validate all
the model through the world map. To overcome this problem, a database such as
Helioclim, for instance, should play an important role in the next few years. Indeed,
Helioclim-3 regroups time series of radiation components over a horizontal, fix-
tilted, and normal plane for the actual weather conditions as well for clear-sky con-
ditions. Geographical coverage corresponds to the Meteosat satellite field of view,
that is, it covers Europe, Africa, Atlantic Ocean, and Middle East. The spatial res-

olution is 3 km at Nadir, and approximately 4e5km at 45 latitude. Data are avail-
able with a time step ranging from 15 min to 1 month. The time coverage is from
February 2004 up to current day-2 for HC3v5 and up to day-1 for HC3v4 (http://
www.soda-pro.com/fr/home).

8. CONCLUSION
This chapter has demonstrated the fundamental importance of the production fore-
casting for intermittent and stochastics power plants. Without reliable predictions,
renewable energy will not reach a high level of integration as their management
would add considerable complexity to the electricity network.
Some information about the intermittence cost, increasing the energy production
cost, were introduced and a literature review is in total agreement with the fact that a
reliable forecasting of the renewable production at various temporal horizons can
decrease these costs and make the intermittent renewable energy more competitive.
The main forecasting methods for solar radiation have been presented and clas-
sified according to the temporal horizon. The accuracy of these methods depends on
the meteorological characteristics of the site where the solar plant is installed.
At last, new perspectives of the forecasting science were shown and prove that
significant progress remains to be made to reach the perfect forecasting model.