Power signal prediction by neural network with a new fuzzy BP learning algorithm

In this paper, short-term power load signal forecasting based on neural network with a new fuzzy back-propagation (BP) learning algorithm is developed. This modified learning rule can effectively help the neural model escape from a local minimum while it is training. Consequently, the proposed neural forecaster has more accurate prediction in real forecasting operation. As a comparison, same experiments are also performed by using neural network with constant learning rate and momentum pairs of traditional BP learning algorithm.