Short-term Wind Energy Prediction Algorithm Based on SAGA-DBNs

Abstract

In order to overcome initial weights direction in the deep conviction network and improve the prediction accuracy of wind energy further, this paper proposes a new algorithm which combined the simulated annealing genetic algorithm and the deep belief network. Firstly, the main influencing factors of wind energy are selected. Then the superior global searching capability of the simulated annealing genetic algorithm is used to train the deep belief network and optimize the weights layer by layer. Finally, the deep belief networks weights were tuned by a BP neural network to make sure the networks are the best wind energy prediction system. The algorithm is verified by the experimental data from a wind farm in western China. The results show that the algorithm not only can overcome initial weights direction in the deep belief networks, but also further improve the prediction accuracy.