The article uses the BP neural network method for short-term prediction of power load.The BP neural network method can not only eliminate some special data effects,but also have great advantages in processing non-linear data.The article first establishes the BP neural network model for training and learning,and extracts 4 sets of 24h power load historical data as a sample.Finally,Matlab is used for simulation prediction.The obtained predictive values and real values are similar,indicating that the method is feasible.