Employing sky images for ultra-short-term solar forecasts
Employing sky images for ultra-short-term solar forecasts lead image
Solar power is growing increasingly important for supplying the world’s electricity needs. Solar panels have many advantages, but the one major drawback is that they can only generate electricity when directly irradiated by the sun. Maximizing efficiency of solar panels therefore relies heavily on increasing the accuracy of irradiance predictions.
Sun et al. developed a method for ultra-short-term forecasting of solar irradiance using a combination of ground-based sky images with historical and meteorological data.
While many researchers have focused on long-term irradiance trends, short term fluctuations are also significant.
“Sudden changes can cause significant fluctuations in the power output of photovoltaic and solar thermal power plants,” said author Wenjie Zhang. “If not accurately predicted, they may lead to power shortages or surpluses, affecting the stability and reliability of the electrical grid.”
The authors collected cloud cover information using ground-based images of the skies above solar panels and coupled this information with meteorological forecasts and historical irradiance and meteorological data. They then fed this combined dataset into an ordinary differential equation network to generate forecasts.
In tests using public data from California and locally collected data from northern China, the researchers were able to significantly improve the short-term estimation of solar irradiance.
They aim to expand their testing and incorporate more sources of data in the future.
“We plan to use low-cost all-sky imagers to achieve irradiance predictions for more locations,” said Zhang. “We also aim to explore more advanced feature extraction techniques to comprehensively extract cloud dynamic information from sky images.”
Source: “Ultrashort-term prediction of solar irradiance with multiple exogenous variables by fusion of ground-based sky images,” by Xiaopeng Sun, Wenjie Zhang, Mifeng Ren, Zhujun Zhu, and Gaowei Yan, Journal of Renewable and Sustainable Energy (2025). The article can be accessed at https://doi.org/10.1063/5.0249194 .
