Regression analysis and prediction of monthly wind and solar power generation in China
Xueping Du, Zhikai Lang, Menglin Liu, Jiangtao Wu
Abstract
The development of clean energy power generation is imperative to improve the energy structure, protect the ecological environment and realize sustainable socio-economic advancements. The accurate prediction of monthly electricity generation from wind and solar power is essential for clean energy systems and power grids. This study aims to explore a novel combined forecasting model to offer theoretical references and data support for analyzing seasonal peaking and supply/demand balance in the power system. Multiple regression equations are established employing the monthly production data of industrial products in China as independent variables, and the corresponding monthly clean energy power generation as the dependent variable. Meanwhile, the monthly average temperature is incorporated as a correlating factor subsequent to grey correlation degree analysis. Furthermore, the monthly fluctuation pattern of China's clean energy power generation is subjected to quantitative analysis. This analysis determines the inhomogeneity coefficients and seasonal factors, which are crucial for addressing the short-term forecasting of monthly clean energy generation in China. Finally, the Surface Fitting-Seasonal Auto Regressive Integrated Moving Average (SF-SARIMA) model is established for China's monthly wind and solar power generation. The results indicate a significant correlation between the monthly production of ten non-ferrous metals and the monthly power generation from clean energy. The cumulative wind and solar power generation for the years 2025–26 is projected to be 1232.3 TW∙h and 450.9 TW∙h. The SF-SARIMA model is versatile and can be applied to both wind and solar power generation forecasts on a month-by-month basis, filling a gap in China's national medium- and long-term power planning for clean energy monthly load forecasting. It can also serve as a theoretical reference for long-term prediction of clean energy generation in other regions, as well as providing data insights for grid trading planning and stable operation.