摘要
针对因卫星入境数据延迟,无法快速判断太阳能电池阵温度遥测数据是否发生异常问题,提出一种SE-TCN网络模型。首先借鉴SENet中的通道注意力机制,对时间卷积网络(TCN)进行改进,提高模型的特征提取能力;其次使用SE-TCN做为特征提取网络,训练出网络模型;最后对温度遥测数据做中长期预测(约4轨)。以某在轨卫星实际太阳能电池阵温度遥测数据作为实验数据。结果表明:本文提出的SE-TCN网络模型在评价指标上与传统TCN网络模型相比,平均绝对误差(MAE)降低了7.7%,均方根误差(RMSE)降低了5.2%,相关系数(R)提高了0.4%。当卫星入境时,该检测方法可根据预测值快速判断实时遥测数据是否发生异常。
An SE-TCN network model is proposed to solve the problem that it is impossible to quickly determine whether the temperature telemetry data of solar array is abnormal or not due to the delay of data for satellite passing territory.First,the squeeze-and-excitation block is introduced into the temporal convolutional network(TCN)so as to improve the feature extraction ability of the model.Second,the SE-TCN network is used to extract deep features of the telemetry data.Finally,a medium and long-term prediction(about 4 tracks)is made for the temperature telemetry data.Through the verification of the telemetry temperature data of an in-orbit satellite,it is shown that compared with the results obtained by the traditional TCN,the mean absolute error(MAE)is reduced by 7.7%,the root mean square error(RMSE)is reduced by 5.2%,and the correlation coefficient(R)is increased by 0.4%.When satellite passes territory,the proposed detection model can quickly judge whether the real-time telemetry data is abnormal or not according to the predicted value.
作者
何利健
张锐
陈文卿
HE Lijian;ZHANG Rui;CHEN Wenqing(Innovation Academy for Microsatellites,Chinese Aacdemy of Sciences,Shanghai 201203,China;University of Chinese Academy of Sciences,Beijing 100049,China)
出处
《上海航天（中英文）》
CSCD
2021年第5期8-16,共9页
Aerospace Shanghai（Chinese&English）
关键词
时间卷积网络
遥测数据
时序数据预测
异常检测
太阳能电池阵
time convolutional network(TCN)
telemetry data
time series data prediction
anomaly detection
solar array
