Volume 20: Sustainable Energy Solutions for a Post-COVID Recovery towards a Better Future: Part III

Using Convolutional Neural Networks to Understand the Impact of COVID-19 on Electricity Demand in Texas Vincent Li, Xiaonan Wang



Economic consequences have been felt around the world as a result of COVID-19, among which have been changes in electricity demand. In this project, we use a convolutional neural network (CNN) to investigate whether there was a change in electricity demand in the state of Texas, located within the United States, during the pandemic, as compared to before it. Training the model on electricity demand and weather data, we were able to achieve a relative RMSE, relative MAE, and R2 of 0.049, 0.041, and 0.92, respectively, on a testing set that represented a normal, pre-pandemic year. The CNN showed better performance, as compared to a plain artificial neural network (ANN). Based on the predictions of the CNN and the actual demand in 2020–2021, we find that the hypothesis that demand decreased during the pandemic was partially supported. A larger decrease was present due to extreme weather events; therefore, we recommend that Texas fortify its electricity generation facilities against such events.

Keywords COVID-19, convolutional neural network, electricity demand, machine learning, lockdown policy

Copyright ©
Energy Proceedings