Power load forecast using artificial neural network
Zhang Guo · Dianli zidonghua shebei · 2002
The method and steps of BP (Back Propagation) neural network for recognizing and forecasting power load in batch data processing of chronological sequence is presented. Batch training of network makes weight vectors and its partial derivative vectors proportionally follow the change of all training vectors. The application of additional momentum and adaptive learning rate overcomes the limitation effect of BP rule, accelerates the training speed and strengthens the generalization ability of network. The real power load of a district is forecasted based on it and the satisfied results are achieved.