Multi-exposure adaptive threshold technique for cloud detection with sky imagers. (April 2015)

• Record Type: Journal Article
• Authors: Cazorla, A.; Husillos, C.; Antón, M.; Alados-Arboledas, L.
• Is Part Of: Solar energy. Volume 114(2015)
• Page Start: 268
• Page End: 277
• Publication Date: 2015-04
• Subjects: Sky imager -- High dynamic range -- Cloud detection -- Solar energy -- Forecasting; Solar energy -- Periodicals; Solar engines -- Periodicals; 621.47
• DOI: 10.1016/j.solener.2015.02.006
• Languages: English
• ISSNs: 0038-092X

Abstract

Highlights: Better cloud detection can increase accuracy in solar resource forecasting models. We developed a high dynamic range technique for cloud detection with sky imagers. An adaptive threshold based on sun-pixel distance is applied to the images. We increase the accuracy of cloud detection in the vicinity of the sun position. Abstract: Sky imagers have been used for cloud detection and classification in the last years, and one of the applications of these instruments is the use of cloud information in forecast algorithms for solar power technologies. These algorithms depend on an accurate classification of the complete sky dome cloud cover, but most system fail in the proximity of the sun due to saturation in the images. This work proposes a new method for cloud detection with sky imagers using images taken with different exposure times and applying an adaptive threshold to each one. The use of multiple exposure times avoids the saturation of the image in the vicinity of the sun position, while the adaptive threshold applied to the images helps in the accurate detection of cloud coverage, especially in the circumsolar area. The method is tested with a commercial sky imager, paying special attention to the detection of clouds close to the sun position. A case study is analyzed, showing an accurate detection of clouds in the vicinity of the sun. The method is also validated using statistical values for data recorded during almost one month which cover a great variety of cloudiness cases. For this purpose, the detection of clouds in the sun position is compared against the reduction of the direct normal irradiance (DNI) with respect to a modeled DNI.