Page 123 - Advances in Renewable Energies and Power Technologies

96 CHAPTER 3 Forecasting of Intermittent Solar Energy Resource

The influence of ambient temperature via the cell temperature into the PV power
model is low and that of the wind speed even more. For the ambient temperature,
often that measured at the previous temporal horizon can be used.
Thus, the accuracy of the PV production forecasting depends mainly on the solar
irradiation forecasting.

6.2 VERY SHORT-TERM FORECASTING IN A TEMPORAL RANGE
FROM 0 TO 6 H
In the time horizon, three methods are available:

• Total sky imagery (TSI)
• Satellite Cloud Motion Vector Approach
• Stochastic learning methods or Time seriesebased method

6.2.1 Sky Imagery
TSI is used or real-time forecasting called now-time, up to 10e30 min ahead; it con-
sists in applying image processing and cloud tracking techniques to sky photos
(Fig. 3.11). The method generally assumes a persistence of the opacity, direction,
and velocity of cloud movements [62].
The cameras have a fish-eye lens that gives a 360-degree vision of the sky above
the plant (Fig. 3.11). Clouds are detected and segmented on successive images and
their past displacements are estimated so as to anticipate their next moves. The direct
sun onto the camera is responsible for one difficulty in such applications. Indeed, it
may burn the captors and lead to color saturation for several regions of the sky. Most
often, it is solved by applying a mask that follows the path of the sun (but hide some
parts of the sky). Several devices have been developed including the whole-sky
imager [63], TSI [64] and the infrared sky imager [65], which considers the infrared
rays. The Steadysun company has also developed his own camera (Steadysun) [66].
Some works were undertaken to estimate the movement of clouds on an image of
the sky [62] or the segmentation and classification of clouds [67,68] (Fig. 3.12).
Additional steps were taken for the prediction of global horizontal irradiance
[62,69e71] (Fig. 3.12) or to predict the direct normal irradiance [72].
6.2.2 Satellite Cloud Image
Clouds can be detected by satellites using visible and/or infrared images; the cloud-
iness has a strong impact on surface solar irradiance. A semiempirical method al-
lows to estimate solar irradiance from satellite using a set of cloud index images
according to the Heliosat method [73]. To calculate the cloud index image in a
first-step motion, vector fields are derived from two consecutive images [4]. The pre-
dicted image then is determined by applying the calculated motion vector field to the
actual image. Finally, solar surface irradiance is derived from the predicted cloud
index images via the Heliosat method. Fig. 3.13 gives an overview of these steps
to derive the irradiance forecast.