摘要
传统的轨道预报方法,主要分为数值法和解析法,数值法计算精度高但计算速度慢,星载计算机的算力要求很高;解析法计算精度低但计算速度快,对星载计算机的算力要求较低。随着卫星技术的发展,卫星的自主定轨和智能化变得越来越重要,而传统的轨道预报方法不能很好的满足这一要求,即兼顾计算精度和计算速度。针对这一问题,提出一种基于LSTM神经网络的轨道预报算法,该方法通过使用LSTM神经网络对解析法的预测值与卫星的真实位置的误差进行训练,从而来预测解析法在未来时刻的预报误差,通过对误差的修正来提高预测的准确性,且具有较快的计算速度。
The traditional orbit prediction methods are mainly divided into numerical method and analytical method.The numerical method has high accuracy but is slow in calculation and requires high computation power of space-borne computer.The analytical method is low in accuracy but fast in computation,and requires less computation power for the space-borne computer.With the development of satellite technology,the independent orbit determination and intelligence of satellite become more and more important,but the traditional orbit prediction methods can not well meet the requirements,that is,give consideration to the calculation accuracy and calculation speed.To solve this problem,an orbit prediction algorithm based on LSTM neural network is proposed.The method uses LSTM neural network to train the error between the predicted value of the analytical method and the real position of the satellite,so as to predict the prediction error of the analytical method at the future time,and improve the accuracy of prediction by correcting the error,and it has a fast calculation speed.
作者
韩雨恒
Han Yuheng(Central South University,School of Aeronautics and Astronautics,Changsha 410083,China)
出处
《科学技术创新》
2022年第21期88-91,共4页
Scientific and Technological Innovation
作者简介
韩雨恒(1997-),男,汉族,四川成都人,硕士研究生,主要研究领域:应用深度学习,轨道预报。
