基于LSTM网络的卫星寿命预测研究  

Prediction of Satellite Lifetime Based on Short and Long Time Memory Networks

作　　者：张伟[1,2] 王宪勇[1,2,3] 崔秀艳 何旭 ZHANG Wei;WANG Xian-yong;Cui Xiu-yan;HE Xu(State Key Laboratory of Robotics,Shenyang Institute of Automation,Chinese Academy of Sciences,Shenyang Liaoning 110016,China;Institutes for Robotics and Intelligent Manufacturing,Chinese Academy of Sciences,Shenyang Liaoning 110169,China;Shenyang Ligong University,School of automation and electrical engineering,Shenyang Liaoning 110159,China;Hbei Softeare Institute,Baoding Hebei 071000,China)

机构地区：[1]中国科学院沈阳自动化研究所机器人学国家重点实验室,辽宁沈阳110016 [2]中国科学院机器人与智能制造创新研究院,辽宁沈阳110169 [3]沈阳理工大学自动化与电气工程学院,辽宁沈阳110159 [4]河北软件职业技术学院,河北保定071000

出　　处：《计算机仿真》2021年第9期80-83,128,共5页Computer Simulation

摘　　要：针对卫星寿命预测的需求和以往方法退化失效的不足,提出了基于长短时记忆网络(Long Short-Term Memory)进行卫星寿命预测的方法。针对卫星系统中锂电池充放电次数增加,导致电池性能衰退的特性。对电池进行衰退分析,选取等压降时间序列数据作为数据集,对电池每次充放电的使用时间进行特征提取,并对数据进行规范化处理,在不同细胞长度及层数的长短时记忆网络模型中进行训练和测试,选取最合适的网络结构,实验表明使用长短时记忆网络模型计算复杂度小,预测准确率高。Aiming at the demand of satellite life prediction and the shortage of degradation and failure of previous methods, a method of satellite life prediction based on long-term and short-term memory networks was proposed.In view of the characteristics that the increase of charge and discharge times of lithium batteries in satellite systems leads to the decline of battery performance, the battery decay analysis was carried out, the equal voltage drop time series data were selected as the data set, the service time of each battery charge and discharge was extract, and the data were standardized.The long-term and short-term memory network models with different cell lengths and layers were trained and tested, and the most appropriate network structure was selected.The experiments show that the long-term and short-term memory network model has low computational complexity and high prediction accuracy.

关 键 词：卫星 寿命预测 数据规范化 锂离子电池 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]