建立一种基于改进型BP神经网络的卫星钟差高精度预报方法.使用PSO算法对BP神经网络结构参数和连接权值阈值进行优化;引入自适应变异因子,以一定概率初始化部分变量改进PSO算法.通过实验验证本文提出的改进BP神经网络算法对于解决BP神经网络容易陷入局部最小值以及训练收敛速率低等问题,以及常规PSO算法早熟收敛等问题具有较好的效果.选用取自IGS网站提供的4颗GPS卫星钟差数据进行288次连续5 min、24次连续1h以及连续7次1d的预报研究.结果表明,研究预报方法的预报精度和稳定性要明显优于使用常规BP神经网络和LSSVM算法建立的模型.%The conventional clock difference data are from IGS data center.Although final ephemeris of satellite clock error precision is available 0.1ns,there exists certain delay.Consequently,it cannot meet the requirement of real time single point positioning.This paper presents a method based on the improved BP neural network to raise prediction accuracy satellite clock.The BP neural network parameters and connection weights are optimized by PSO algorithm,and the adaptive mutation factor is introduced to improve the PSO algorithm with a certain probability.Experiments show that the improved BP neural network algorithm can solve the problems that the BP neural network is easy to fall into local minimum and the training convergence rate is low,and the conventional PSO algorithm may be premature convergence.The forecasting accuracy and stability are tested under the conditions of 288 GPS satellite clock error data from 4 consecutive 5 min,24 continuous 1 hours and 1 time 7 days.The results show that the prediction accuracy and stability of this research are better than thaose of the conventional BP neural network and LSSVM algorithm.
展开▼
