Advanced cloud motion tracking increases solar forecasting accuracy
DOI: 10.1063/1.5109127
Advanced cloud motion tracking increases solar forecasting accuracy lead image
By predicting solar irradiance and solar power generation, solar forecasting helps engineers estimate fluctuations in sunlight availability for an area over time and better integrate the intermittency of solar power production as part of the power grid.
Current solar forecasting methods generally rely on analyzing static images of the sky by examining features such as cloud position and shape. Pedro et al. introduce a new solar forecasting model that could improve the current accuracy by also considering dynamic features in sky images.
The new method couples a block-matching algorithm that identifies the bulk motion of clouds relative to the position of the sun in the sky using an extreme gradient boost algorithm. This approach limits feature extraction to image sections that contain clouds moving toward the sun, and ignores clouds moving away from the sun or in a line that does not intercept with the sun’s trajectory. In doing so, it can extract adaptive image features that change with the computed average cloud direction in a specific region.
The researchers tested their method with sky images taken by a low-cost fisheye camera, and found that it has improved the predictability of the magnitude and direction of solar irradiance ramps, which are step changes in the measured irradiance time series.
Carlos Coimbra, one of the authors, said that this simple but robust model bypasses some of the common steps of solar forecasting using sky images, including the computation of cloud height and speed, thus reducing error accumulation during these steps.
This paper is part of a collection called “Best Practices in Renewable Energy Resourcing and Integration” assembled by the Journal of Renewable and Sustainable Energy.
Source: “Adaptive image features for intra-hour solar forecasts,” by Hugo T. C. Pedro, Carlos F. M. Coimbra, and Philippe Lauret, Journal of Renewable and Sustainable Energy (2019). The article can be accessed at https://doi.org/10.1063/1.5091952 .
