基于MSW-LSTM的卫星功率负荷预测方法  

Prediction Method of Satellite Power Load Based onMSW-LSTM

作　　者：姜垚先 蒋硕[1] 李键 张文芳[1] 王利然[1] 刘鹏[1] JIANG Yaoxian;JIANG Shuo;LI Jian;ZHANG Wenfang;WANG Liran;LIU Peng(Institute of Telecommunication and Navigation Satellites,China Academy of Space Technology,Beijing 100094,China)

机构地区：[1]中国空间技术研究院通信与导航卫星总体部,北京100094

出　　处：《航天器工程》2024年第5期43-49,共7页Spacecraft Engineering

摘　　要：文章提出一种基于多尺度小波变换的长短期记忆神经网络(Multi-scale Wavelet-Long Short-term Memory,MSW-LSTM)卫星功率负荷预测方法,利用多尺度小波变换的时频特性将卫星功率负荷数据进行平稳化处理,结合LSTM算法针对多尺度小波分解并降噪后的各尺度功率负荷序列进行预测,并对预测得到的各序列数据进行重构得到时域功率负荷数据,从而提高卫星功率负荷预测的精度。通过某在轨卫星真实功率负荷数据进行分析,仿真结果表明本文提出的算法可以有效地提高功率负荷预测的精度,且针对不同轨道的卫星功率负荷预测有较高的普适性和鲁棒性。This paper proposes a prediction method of satellite power load using multi-scale wavelet-long short-term memory(MSW-LSTM)neural networkbased on the multi-scale wavelet transform.The time-frequency characteristics of multi-scale wavelet transform are used to flatten the satellite power load data,the LSTM algorithm is combined to predict the power load sequences of various scales after multiscale wavelet decomposition and denoising,and the predicted sequence data is reconstructed to obtain time-domain load power data,thereby improving the accuracy of satellite power load prediction.By analyzing the load power data of a certain satellite in orbit,the simulation results show that the method proposed in this paper can effectively improve the accuracy of load power prediction,and own high universality and robustness for satellites in different orbits.

关 键 词：卫星 小波变换 长短期记忆 功率负荷预测 

分 类 号：V43[航空宇航科学与技术—航空宇航推进理论与工程]