Abstract:

Using the threshold regression model and the nonlinear sliding regression model, the nonlinear relationship between the daily electricity consumption and temperature in Tianjin from 2014 to 2019 was explored, and the threshold temperature under different response relationships was calculated. The results showed that the threshold temperature in Tianjin is significantly different under different response relationships. Under the "V"-shaped nonlinear response relationship, the threshold temperature is 18.8℃, while under the "U"-shaped response relationship, the comfort zone of the linear asymmetric model is 12.3-23.4℃, and the comfort zone range of the nonlinear model is 13.7-21.7℃. Comparing the prediction effects of different models, it is considered that the "U" model is better than the "V" model, and the nonlinear model is better than the linear model. The analysis on the influence factors of the threshold temperature showed that relative humidity has a greater impact on the threshold temperature of the comfort zone and the cooling zone. The threshold temperature is 2.2℃ higher when the relative humidity is 30%-50% than when the relative humidity is 50%-70%. In contrast, relative humidity has little effect on the threshold temperature of the comfort zone and heating zone. The threshold temperature in Tianjin will change significantly over time, and the range of the comfort zone in 2002-2005 is 1.4℃ larger than that in 2014-2019. In practical applications, the threshold temperature should be selected according to the needs of the model, and the influence of relative humidity and time changes should be fully considered, to improve the accuracy of electricity consumption prediction.

Key words: Threshold regression, Nonlinear sliding regression, "U"-shaped segmentation, "V"-shaped segmentation