航空学报 > 2023, Vol. 44 Issue (13): 327860-327860   doi: 10.7527/S1000-6893.2022.27860

基于深度学习的高空风在线估计及预报方法

张荣升1, 吴燕生2(), 秦旭东1, 张普卓1   

  1. 1.北京宇航系统工程研究所,北京 100076
    2.中国航天科技集团有限公司,北京 100048
  • 收稿日期:2022-07-28 修回日期:2022-09-28 接受日期:2022-10-25 出版日期:2023-07-15 发布日期:2022-11-04
  • 通讯作者: 吴燕生 E-mail:WuYSh_CASC@163.com

A real⁃time in⁃flight wind estimation and prediction method based on deep learning

Rongsheng ZHANG1, Yansheng WU2(), Xudong QIN1, Puzhuo ZHANG1   

  1. 1.Beijing Institute of Astronautical Systems Engineering,Beijing 100076,China
    2.China Aerospace Science and Technology Corporation,Beijing 100048,China
  • Received:2022-07-28 Revised:2022-09-28 Accepted:2022-10-25 Online:2023-07-15 Published:2022-11-04
  • Contact: Yansheng WU E-mail:WuYSh_CASC@163.com

摘要:

针对运载火箭高空风数据预报偏差较大、飞行中难以获取的问题,提出了一种基于深度学习对运载火箭飞行中高空风进行在线估计及预报的方法。通过计算某地区近15年高空风实测数据的时空分布情况,获得该地区冬季风速明显偏高、风向集中在270°附近的规律,作为生成样本集和风场预报的依据;推导了高空风对火箭飞行状态的影响,根据推导结果设计了深度神经网络,由火箭飞行状态量估计高空风;提出了一种基于高空风时空分布的深度学习高空风预报方法,解释了通过深度学习方法预报高空风的合理性。在火箭飞行任务中进行算法的搭载飞行实验,验证了算法的准确性和实时性,算法具备工程实施的可行性。

关键词: 运载火箭, 高空风, 深度学习, 估计及预报, 高空风时空分布

Abstract:

The in-flight wind of launch vehicles is difficult to measure during flight, and the prediction error is great occasionally. In this paper, a real-time in-flight wind estimation and prediction method is proposed based on deep learning. By computing the spatial and temporal distribution of actual measurement data of in-flight wind in past 15 years, we acquire the rule that the wind velocity is higher and the wind direction is concentrated at 270° in winter, which is taken as the basis of sample set generation and wind prediction. Response of flight state to in-flight wind is obtained. Based on the obtained result, a deep neural network is designed to estimate the in-flight wind via the launch vehicle’s flight state. Next, a wind prediction method is proposed based on spatial and temporal distribution of in-flight wind. The rationality of wind prediction via deep learning is also given. We carry out a flight experiment on launch vehicle to verify the accuracy and real-timeliness of the proposed method. The proposed method is feasible for engineering implementation.

Key words: launch vehicle, in-flight wind, deep learning, estimation and prediction, spatial and temporal distribution of in-flight wind

中图分类号: