Three-layer BP artificial neural network（ANN） is used to set up a short-term load forecast model.Some major factors such as basic load

temperature difference

weather change and different type of day（working day and holiday） are used as data samples.Though the self-training and self-study of the network

BP method is introduced to revise the weight of ANN in the process of training and study.Thus the model of load forecast is improved.At the same time

the repairing of lines and equipments in power system is also considered

and the exact result is concluded due to this influencing factor.In practice

the above forecast method is proved to be prefect and its precision is very high too.