联合SSA与BiLSTM的北斗卫星钟差预报算法

BeiDou satellite clock bias prediction algorithm by integrating of SSA and BiLSTM

针对现有的卫星钟差预报模型难以捕捉其非线性特性的问题,提出了一种联合麻雀搜索算法(SSA)与双向长短期记忆神经网络(BiLSTM)的北斗卫星钟差预报算法。将BiLSTM应用于钟差预报中,并引入SSA进行网络超参数选择,能够更好地捕捉钟差数据中的特征关系,提高模型预报的准确性。利用德国地球科学研究中心提供的北斗三号精密卫星钟差数据,进行了1 h、3 h、6 h、12 h、24 h和48 h的钟差预报实验;与常用模型从卫星轨道类型和模型普适性方面,进行了单天与多天的预报对比分析。结果表明,相对于多项式模型、小波神经网络、长短期记忆神经网络模型和BiLSTM模型,所提算法的钟差预报平均精度分别提升了75.12%、67.44%、75.18%和48.65%。

In response to the challenge of existing satellite clock bias prediction models in capturing its nonlinear characteristics,a Beidou satellite clock bias prediction algorithm by integrating sparrow search algorithm(SSA)and bidirectional long short-term memory network(BiLSTM)is proposed.BiLSTM is employed for forecasting clock bias,and SSA is introduced for network hyperparameter selection,which can better capture the characteristics in sequence data and improve the accuracy of model prediction.Experimental validations are conducted using precise BDS-3 satellite clock bias data provided by the German Research Centre for Geosciences,encompassing clock bias predictions for 1 h,3 h,6 h,12 h,24 h,and 48 h intervals.In terms of satellite orbit types and model universality,single-day forecast and multi-day forecast are compared with common models.The results show that compared with the polynomial model,wavelet neural network,long short-term memory model,and BiLSTM model,the average accuracy of clock bias prediction of the proposed algorithm is improved by 75.12%%,67.44%,75.18%and 48.65%,respectively.