基于PSO-Elman神经网络BDS导航卫星钟差预报

BDS navigation satellite clock difference prediction based on PSO-Elman neural network

卫星钟差是导航定位系统定位精度的重要影响因子之一.针对北斗卫星导航系统(BDS)精密钟差预报性能寻优问题,提出一种基于粒子群优化(PSO)算法优化Elman神经网络的钟差预报模型方法(PSO-Elman模型),以解决Elman神经网络局部最优问题对钟差预报结果的影响.首先对钟差产品进行预处理;然后通过PSO算法迭代寻优确定Elman神经网络权值、阈值的初始值,并将进行预处理之后的序列数据进行训练建模;再采用武汉大学国际GNSS服务(IGS)数据分析中心(WHU)提供的BDS精密钟差产品数据进行钟差预测;最后将预测结果还原为预报钟差.结果表明:对比于二次多项式(QP)模型、附加周期项多项式(SA)模型、灰色(GM)模型,PSO-Elman模型精度分别提高90.7%、84.2%、81.6%,稳定度提高85.3%、76.3%、36.1%.实验表明:PSO-Elman模型在1~12 h短期预报模拟结果的预报精度和稳定性有显著提高,验证了提出方法的可行性.

Satellite clock error is one of the important factors affecting the positioning accuracy of navigation and positioning system.Aiming at the problem of optimizing the precision clock error prediction performance of the BeiDou Navigation Satellite System(BDS),a method of optimizing the Elman neural network clock error prediction model based on particle swarm optimization(PSO)is proposed to solve the influence of the local optimal problem of Elman neural network on the clock error prediction results.Firstly,the clock error product is preprocessed.The initial weights and thresholds of Elman neural network are determined by iterative optimization of PSO algorithm,and the preprocessed sequence data are used for training modeling.The BDS precision clock error product data provided by IGS Data Analysis Center(WHU)of Wuhan University are used to predict the clock error,and then the prediction results are restored to predict the clock error.The results show that compared with the quadratic polynomial(QP)model,the polynomial(SA)model with additional period term,and the grey(GM)model,the accuracy is improved by 90.7%,84.2%,81.6%,and the stability is improved by 85.3%,76.3%,36.1%,respectively.The experimental results show that the prediction accuracy and stability of PSO-Elman model are significantly improved in 1−12 h short term forecast simulation,which verifies the feasibility of the proposed method.