1.China Electric Power Research Institute, Beijing 100192, China;2.Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Key Laboratory of Land Surface Pattern and Simulation, Beijing 100101, China

The climate change is intensifying, while the daily variation characteristics and relationships of wind, photovoltaic, and other renewable energy outputs and loads under regional-scale extreme temperature events remain unclear. Therefore, this paper uses independent models to simulate the impact of weather on load and wind, photovoltaic, and hydro power outputs. The joint probability and confidence intervals of outputs are calculated using Copula functions. Based on climate model data, the daily variation characteristics of power output and load as well as the power supply-demand relationship are estimated under heat waves and cold waves in typical provinces of North China and Southwest China by 2030, the target year for China’s carbon emission peak. A novel boosting ensemble learning model is proposed to predict the impact of weather on photovoltaic output, and calibrated with historical measured data. The validation shows that the model achieves a mean daily output error of 1.27% under actual extreme weather conditions, and the mean absolute error is significantly lower than other ensemble learning methods. The medium- and long-term forecast indicates that in 2030, typical provinces in North China and Southwest China are prone to experience power supply shortages during the evening hours of days under heat waves and cold waves, and indicators such as peak values and time of power supply and demand under extreme temperature events in the relevant regions in the future are given.