国产欧美精品一区二区,中文字幕专区在线亚洲,国产精品美女网站在线观看,艾秋果冻传媒2021精品,在线免费一区二区,久久久久久青草大香综合精品,日韩美aaa特级毛片,欧美成人精品午夜免费影视

When using grey incidence method to optimize the hidden layer of BP neural network structure, this method takes a short learning time to delete the hidden redundant nodes whose grey incidence degree is less than grey incidence threshold. But the selection of grey incidence threshold has certain subjectivity, which may result in deleting useful nodes by mistake or being unable to delete redundant nodes completely. When using sensitivity pruning method, only the node with the minimum sensitivity will be deleted each time, therefore it takes a long learning time. In view of this, this paper presents a BP structure optimization method based on the grey incidence and the sensitivity degree. At the early stage of the network learning, this paper uses grey incidence method to delete the redundant nodes rapidly, until the grey incidence degrees of the remaining hidden nodes are greater than dynamic grey incidence threshold value. Then in the later learning process of the network, it uses sensitivity pruning method to delete the hidden nodes precisely, until the learning error increases after the deletion of the node. Then keep the node and stop the learning. In this paper, the neural network with the optimized structure is applied to wind power prediction. The simulation results show that this method can meet the requests for forecasting error. And it can not only simplify the structure of neural network, but also solve the problem brought by grey incidence threshold’s precise determination of grey incidence method.