[1]谷吉海,姜兴渭,王晓锋,等.双向Elman神经网络在卫星电池阵功率预测中的应用研究[J].南京理工大学学报(自然科学版),2004,(04):404-408.
 GU Ji hai,JIANG Xing wei,WANG Xiao feng,et al.Application of Bi-directional Elman Neural Networks in Satellite Battery Array Power’s Prediction[J].Journal of Nanjing University of Science and Technology,2004,(04):404-408.
点击复制

双向Elman神经网络在卫星电池阵功率预测中的应用研究()
分享到:

《南京理工大学学报》(自然科学版)[ISSN:1005-9830/CN:32-1397/N]

卷:
期数:
2004年04期
页码:
404-408
栏目:
出版日期:
2004-08-30

文章信息/Info

Title:
Application of Bi-directional Elman Neural Networks in Satellite Battery Array Power’s Prediction
作者:
谷吉海1 姜兴渭1 王晓锋2 龙 兵1
1. 哈尔滨工业大学航天学院, 黑龙江哈尔滨150001;
2. 南京理工大学动力工程学院, 江苏南京210094
Author(s):
GU Ji hai 1JIANG Xing wei 1WANG Xiao feng 2LONG Bing 1
1.School of Astronautics, Harbin Institute of Technology, Harbin 150001, China;2.School of Power Engineering,NUST,Nanjing 210094,China
关键词:
Elman 神经网络 双向预测模型 卫星电池阵功率 预测
Keywords:
Elman neural network bi-directional predict ion model satellite bat tery array pow er predict ion
分类号:
V44
摘要:
针对神经网络预测电池阵功率存在的模型阶数难以确定及预测精度低下的问题 ,提出一种基于改进的Elman神经网络的双向预测模型。该模型利用关联层动态神经元的反馈连接 ,将未来预测网络和过去预测网络的信息进行融合 ,使网络对时间序列特征信息的记忆得到加强 ,从而提高预测精度。用该文提出的双向预测模型对电池阵功率进行预测 ,输入层仅需一个节点 ,不需事先对模型进行定阶。仿真预测表明 ,预测精度比单向模型明显提高 ,且网络具有较好的泛化能力。
Abstract:
In order to resolve the problem of determinat ion of the number of model order and predict ive accuracy being low in predict ing bat tery array power by ANN, the paper presents a bi-direct ional predict ive model based on improved Elman. Informat ion of future and past pre diction system is fused and memory of network for t ime-series feature’ s informat ion is strength ened by this model w hich depends on the feedback of connect ive lay er. T he method ef fect ively improves the predict ive accuracy. T he model presented by the paper needs not to determine the number of model order in predict ing bat tery array pow er. T he simulation results indicate that it s predict ive accuracy is greatly improved by comparison w ith unidirect ional model, and that the model has a bet ter generalization ability.

参考文献/References:

[ 1] Cholewo T , Zurada J M. Sequential networ k construction for time series prediction[ A] . Proceedings of the IEEE Internatio nal Joint Conference o n Neur al Netw orks[ C] . Houston: IEEE Computer Association, 1997. 2 034~ 2 039.
[ 2] 温广瑞, 屈梁生, 张西宁. 基于递归神经网络的多步预报方法[ J] . 西安交通大学学报, 2002, 36( 7) : 722~ 726.
[ 3] Aussem A. Dynamical recur rent neur al netw orks towards pr ediction and modeling of dynamical systems[ J] . Neurocomputing, 1999( 28) : 207~ 232.
[ 4] 阎平凡, 张长水. 人工神经网络与模拟进化计算[ M] . 北京: 清华大学出版社, 2000. 213~ 228.
[ 5] Wakuya H, Zurada J M . Bi-directional computing architecture for time ser ies pr ediction [ J] . Neural Net- wor ks, 2001, 1( 14) : 130 7~ 132 1.

备注/Memo

备注/Memo:
基金项目: 国家高技术研究发展计划( 863- 2002AA721063)
作者简介: 谷吉海( 1964- ) , 男, 吉林柳河人, 博士, 主要研究方向: 航天器故障诊断、预测技术, 人工智能等, E-mail: hgd-gjh@ 163. net。
更新日期/Last Update: 2013-03-11