Affected by the pandemic coronavirus-19 (COVID-19)

significant changes have taken place in all aspects of social production and residents' lives

as well as in the energy supply and consumption characteristics of the power system. COVID-19 has brought unpredictable uncertainties to the power grid. These changes and uncertainties pose a challenge to conventional electric load forecasting. Therefore

aiming to load forecasting under the background of the pandemic

this paper proposes a power load segmented forecasting method based on the pandemic stage division method

attention mechanism

and bi-directional long and short-term memory artificial neural network quantile regression model (ESD-ABiLSTMQR). According to the development degree of the pandemic

considering characteristics of different development stages of the pandemic

the pandemic is divided into four stages by using the analytic hierarchy process method (AHP): initial stage

outbreak stage

control stage

and recovery stage. A segmented load forecasting model based on LSTM and attention mechanism is established to forecast load in different time series. Cases used data from the pandemic in Wuhan

China

for verification. Results show the segmented forecasting method can analyze load characteristics of each stage and can effectively improve the accuracy of load forecasting.