利用灰色关联极限学习机预报日长变化

发布时间：2018-01-15 04:07

本文关键词：利用灰色关联极限学习机预报日长变化　出处：《中国科学院大学学报》2015年05期 　论文类型：期刊论文

【摘要】：针对日长变化难以用精确模型进行预报的问题,将一种新型人工神经网络——极限学习机(extreme learning machine,ELM)用于日长变化预报中.首先针对时间序列预测问题中存在的嵌入维数选取和网络结构设计问题,提出一种基于灰色关联分析(grey relational analysis,GRA)的ELM算法(GRA-ELM),该算法将灰色关联分析输入节点选取嵌入到ELM网络的训练过程中,同时完成嵌入维数和隐层节点规模的确定.然后根据日长变化数据的特点对其进行预处理,建立一种能够高精度、近实时预报日长变化的GRA-ELM预报模型.最后将GRA-ELM模型的预报结果同标准ELM、反向传播神经网络、广义回归神经网络和地球定向参数预报比较竞赛的结果进行比较.结果表明,通过本方法得到的日长变化较其他方法在精度上有较大改善.
[Abstract]:Aiming at the problem that it is difficult to predict the diurnal variation with accurate model, a new type of artificial neural network-extreme learning machine (LLM-extreme learning machine) is proposed. Elm) is used in the prediction of diurnal variation. Firstly, the embedded dimension selection and network structure design problems in time series prediction are discussed. A ELM algorithm based on grey relational analysis (GRA-ELM) is proposed. In this algorithm, the gray correlation analysis input node selection is embedded in the training process of ELM network. At the same time, the embedding dimension and the size of hidden layer nodes are determined. Then, according to the characteristics of day length change data, we preprocess them to establish a high accuracy. Finally, the prediction results of GRA-ELM model are compared with standard Elm and back propagation neural network. The results of the generalized regression neural network and the Earth orientation parameter prediction competition are compared. The results show that the diurnal variation obtained by this method is better than that of other methods.
【作者单位】：中国科学院国家授时中心;中国科学院时间频率基准重点实验室;中国科学院大学;
【基金】：中国科学院西部之光联合学者项目(201491)资助
【分类号】：P126
【正文快照】： Lei Y,Cai H B,Zhao D N.Prediction of length-of-day variation using grey relational analysis and extreme learningmachine[J].Journal of University of Chinese Academy of Sciences,2015,32(5):588-593,604.generalization regression neural networks and Earth ori

【相似文献】