Assessing the Impact of the COVID-19 Pandemic on Electricity Consumption: A Machine Learning Approach

This study aims to quantitatively assess the impact of the COVID-19 pandemic on societal electricity consumption behavior, particularly considering the interference of special events. Currently, there is insufficient quantitative analysis of the extent to which the COVID-19 pandemic affects electricity consumption behavior. Therefore, we endeavor to introduce new theories and methods to comprehensively understand the potential impact of the pandemic on electricity usage in this field. Our proposed machine learning methods demonstrate significant results and advantages in two key aspects. Firstly, we innovatively introduce an abnormality detection algorithm based on the lunar calendar, thereby establishing a detection system capable of accurately identifying abnormal fluctuations in electricity consumption, to precisely determine the onset of the pandemic. Secondly, we design an evaluation system that integrates CNN-LSTM predictive models and controlled variable strategies, enabling us to reconstruct electricity usage patterns less affected by the pandemic. By comparing actual electricity consumption with reconstructed patterns, we deeply evaluate the impact of the pandemic on different regions and industries and introduce the pandemic impact percentage as a quantification method to precisely assess the extent of the impact. The findings indicate a significant impact of the COVID-19 pandemic on commercial and industrial electricity consumption. Furthermore, our methods are not limited to pandemic research but can also be applied to investigate the effects of other events such as holidays, typhoons, and special production schedules on electricity consumption. In summary, by delving into the complex associations underlying electricity consumption, this study provides a solid theoretical and methodological foundation for future similar research and offers crucial decision support for the electricity industry and emergency management.