Abstract:

Based on the daily social electricity consumption and meteorological data of Shijiazhuang in summer from 2017 to 2021, we calculated temperature and humidity index, hotness index, and comfort index. We employed multiple linear regression analysis and the BP neural network algorithm to explore the correlation between daily social electricity consumption and these meteorological indices in the region. We then developed models for electricity consumption, i.e.a multiple linear regression model and a neural network model. The results indicated that the spatial and temporal distribution of daily electricity consumption and the number of uncomfortable days of living environments during summer are largely similar. Moreover, there is a notably positive correlation between the summer daily meteorological indices and social electricity consumption. Among the indices, the correlation between daily electricity use and the number of environmentally uncomfortable days is the most significant. A model parameterized by the comfort index for predicting daily summer electricity consumption is found to be particularly applicable. The study demonstrates that both the multiple linear regression analysis and the BP neural network algorithm can effectively capture the general trend of daily social electricity consumption, although the latter exhibits a higher degree of error. The accuracy of the social daily electricity consumption forecast model can be enhanced by focusing on the month of June, which significantly contributes to the error in the multiple linear regression analysis model. Additionally, establishing thresholds for meteorological factors and comfort indices that define the onset of peak and trough periods can further refine the model′s predictive capabilities.

Key words: Multiple linear regression, BP neural network, Comfort index