摘要
为提高极移的预报精度,本文建立了基于T-S模糊神经网络的极移预报模型。首先利用最小二乘对极移序列中的趋势项进行拟合外推,然后建立TSFNN预报模型对最小二乘拟合残差部分进行训练和预报,最后合并最小二乘外推值和TSFNN模型预报值得到最终的极移预报值。在试验中,本文基于T-S模糊神经网络模型对不同跨度的极移预报进行研究,并与BP神经网络的预报效果进行对比,结果表明,该预报模型能很好地克服神经网络收敛速度慢、易陷入局部极值、预报精度较低的缺点,可以有效地用于极移预报。
To improve the accuracy of pole motion prediction,this paper proposes the prediction model based on TSFNN(T-S fuzzy neural network).Firstly,the least square method is utilized to fit the trend term in the pole motion sequence.Then,the TSFNN prediction model is established to train and forecast the least square fitting residual.Finally,the least square extrapolation value and TSFNN model prediction value are combined to the final value of pole motion prediction.In the experiment,this paper studies different span of polar motion prediction based on T-S fuzzy neural network model,and compares with the effect of BP neural network prediction.The results show that the prediction model can well overcome the disadvantages of the neural network,such as slow convergence speed,easy to fall intolocal extremes and low prediction accuracy,and it can be used effectively in pole shift prediction.
作者
熊峰
李宗春
郭迎钢
付永健
汪文琪
XIONG Feng;Li Zongchun;GUO Yinggang;FU Yongjian;WANG Wenqi(Information Engineering University,Zhengzhou 450001,China)
出处
《测绘通报》
CSCD
北大核心
2020年第S01期205-209,共5页
Bulletin of Surveying and Mapping