The combination of increasing extreme weather and climate warming makes the disturbance of abnormal temperature fluctuations on regional electricity sales increasingly stronger. Based on the panel data constructed by regional electricity sales, temperature, humidity and economic variables, this paper uses abnormal temperature identification, temperature shock quantification, threshold effect test and panel smooth transformation model to empirically analyze the nonlinear response mechanism of electricity sales. The results show that there is a significant U-shaped relationship between regional electricity sales and temperature. The V-shaped single threshold is 18.8℃, the comfort zone identified by the U-shaped nonlinear model is 13.5 — 21.7℃, and its RMSE is 10.270, which is better than the V-shaped nonlinear and U-shaped linear models. In the model comparison, the RMSE, MAE and R² of the PSTR model reach 9.986, 7.516 and 0.796, respectively, and the comprehensive performance is the best. The increase of humidity reduced the threshold of high temperature side by about 2.2℃, and the width of comfort zone narrowed from 9.4℃ to 8.0℃ at different time periods. The research can provide more targeted quantitative basis for regional electricity sales forecasting, load scheduling and extreme weather response.
