Coupling meteorological stations data and satellite data for prediction of global solar radiation with machine learning models. (October 2022)

• Record Type: Journal Article
• Authors: Zhao, Shuting; Wu, Lifeng; Xiang, Youzhen; Dong, Jianhua; Li, Zhen; Liu, Xiaoqiang; Tang, Zijun; Wang, Han; Wang, Xin; An, Jiaqi; Zhang, Fucang; Li, Zhijun
• Is Part Of: Renewable energy. Volume 198(2022)
• Page Start: 1049
• Page End: 1064
• Publication Date: 2022-10
• Subjects: Solar radiation -- Machine learning -- Input combination -- Meteorological factors; Renewable energy sources -- Periodicals; Power resources -- Periodicals; Énergies renouvelables -- Périodiques; Ressources énergétiques -- Périodiques; 333.794
• DOI: 10.1016/j.renene.2022.08.111
• Languages: English
• ISSNs: 0960-1481

Abstract

Abstract: The simulation of solar radiation is of great significance to the sustainable development of energy, engineering, and many other fields. The Himawari series of satellites has the characteristics of high temporal, spatial resolution, which helps to solve the problem of insufficient ground radiation observation in China. However, the accuracy of this data needs to be further improved. Thus, four machine learning models with 13 ground and satellite-based input combinations were used to simulate daily solar radiation. The results showed that the simulation accuracy of the model based on a combination of meteorological data from different sources was significantly improved compared with the model based on single-source data. The RMSE was 32.4% and 44.6% lower than those of the model based on the ground meteorological stations data and the model based on the satellite data, respectively. SVM13 model showed the optimal simulation performance compared with other models, and its RMSE and R 2 were 1.732 MJ m −2 day −1 and 0.939 in each climate region, respectively. Overall, we conclude that the SVM13 model is the most suitable model, and the model with a complex combination of more meteorological factors as input has higher simulation accuracy than the model with a relatively simple input combination.