In order to further reduce the forecasting error of electric load data

a short-term power load forecasting method based on load decomposition and identification is proposed. First

for the electric power load data of each industry

the polynomial fitting error of temperature-sensitive load to the temperature series is taken as the objective function

and the load decomposition is transformed into a mathematical optimization problem

and the total load of each industry is decomposed into the weekly load based on load identification component and the temperature-sensitive load component. Second

the short-term load prediction is performed for the temperature-sensitive load component based on the long short-term memory network. Finally

the temperature-sensitive load prediction results are superimposed with the weekly load based on load identification component to obtain the complete load forecast results. The results show that the short-term load forecasting method based on load decomposition and identification proposed can effectively reduce the short-term load forecasting error.