Atmospheric Properties Measurement Using Sky Images

Rémi Chauvin (1) , Stéphane Grieu (1) , Stéphane Thil (1) , Julien Nou (1)

1 PROMES - Procédés, Matériaux et Energie Solaire

Atmospheric properties play a key role in the evaluation of the solar power plant output. Indeed, properties as the cloud cover, its distribution or the atmospheric turbidity strongly influence the solar resource availability and variability. Consequently, it is recommended to integrate such information into the plant control strategy in order to avoid over- or underestimation of the electricity generation. This presentation deals with a novel methodology based on a sky-imaging system to measure the clear sky intensity distribution and compute the cloud cover map. According to the CIE clear sky standard model, the clear sky luminance distribution can be modeled by an equation which is a function of the Solar/Zenith Angle (SZA) and the Sun/Pixel Angle (SPA). From this equation, detailed models have been developed to simulate both sky luminance and sky radiance distribution for various sky conditions. A new equation, inspired by these works has been developed to fit with the clear sky images provided by our sky imager. The equation is a function of SZA, SPA and 7 coefficients determined by a regression analysis, minimizing the least square error. These coefficients are related to the atmospheric turbidity and give valuable information about the sunshape which impacts on the optical efficiency of solar concentrating systems. Once the equation is defined, it is also possible to generate a clear sky image. This image can be used as a reference to perform an optimized cloud detection. The cloud detection algorithm is based on a fixed thresholding technique applied on the Normalized Red/Blue Ratio (NRBR) of the image. The NRBR of the current sky image is compared to the NRBR of the associated clear sky image. Pixels for which the NRBR difference is above 5% are classified as cloud pixels. This cloud detection algorithm outperforms most of the standard algorithms, especially in the circumsolar area. From this pixel classification, it is possible to compute the cloud cover map which gives valuable information about the solar resource variability and helps the operator to optimize the plant control procedures.

Domaines: Automatique / Robotique

Rémi Chauvin, Stéphane Grieu, Stéphane Thil, Julien Nou. Atmospheric Properties Measurement Using Sky Images. 10th SFERA-SOLLAB Doctoral Colloquium on Solar Concentrating Technologies, Jun 2014, Odeillo, France. ⟨hal-01270004⟩