遗传灰色神经网络的卫星钟差预报模型

Satellite clock error prediction model based on genetic grey neural network

针对目前导航卫星钟差预报精度低的问题,结合灰色模型与神经网络的特点,提出一种遗传算法优化灰色神经网络的钟差预报模型:采用遗传算法对网络权值与灰色参数进行动态调整,不仅能够有效避免权值选择的随机性,还能避免陷入局部最优的情况,从而有效提高预报精度;选取IGS提供的精密钟差数据进行实验,并与灰色模型、二次多项式模型、灰色神经网络与小波神经网络进行对比分析。实验结果表明,该模型平稳性强、预测精度高,能够满足实时钟差预报的要求。

Aiming at the problem of low accuracy in prediction of satellite clock error, combiningthe characteristics of the grey model and neural network, this paper proposed a model of clock error predition based on grey neural network optimized by genetic algorithm： genetic algorithm was used to dynamically adjust the network weights and grey parameters, in order to efficiently avoid not only the randomness of the weight selection, but also falling into the local optimal situation for improve the prediction accuracy. Finally, the precise clock error data provided by IGS were tested and comparatively analyzed with the grey model, the quadratic polynomial model, the grey neural network and the wavelet neural network. Result showed that the pro posed model could meet the requirements of real-time clock error prediction with high stability and accuracy.