基于QPSO-LSTM模型的电离层TEC预测  

Ionospheric TEC Prediction Based on QPSO-LSTM Model

作　　者：郭文韬 孙希延[1,2,3,4] 纪元法 贾茜子[1,2] GUO Wentao;SUN Xiyan;JI Yuanfa;JIA Qianzi(Guangxi Key Laboratory of Precision Navigation Technology and Application,Guilin University of Electronic Technology,Guilin 541004;School of Information and Communication,Guilin University of Electronic Technology,Guilin 541004;Satellite Navigation Positioning and Location Service,National&Local Joint Engineering Research Center,Guilin 541004;GUET-Nanning E-Tech Research Institute Co.,Ltd.,Nanning 530031)

机构地区：[1]桂林电子科技大学广西精密导航技术与应用重点实验室,桂林541004 [2]桂林电子科技大学信息与通信学院,桂林541004 [3]卫星导航定位与位置服务国家地方联合工程研究中心,桂林541004 [4]南宁桂电电子科技研究院有限公司,南宁530031

出　　处：《空间科学学报》2024年第5期772-781,共10页Chinese Journal of Space Science

基　　金：广西壮族自治区科学技术厅项目(桂科AB21196041,桂科AB22035074,桂科AD22080061);国家自然科学基金项目(U23A20280,62161007,62061010);广西高校中青年教师科研基础能力提升项目(2022KY0181)共同资助。

摘　　要：针对单一LSTM模型的电离层TEC短期预报存在参数调整和性能优化困难导致预测精度低的问题,结合量子粒子群算法(Quantum Particle Swarm Optimization,QPSO)和LSTM模型,通过量子粒子群算法自适应确定最优解,优化LSTM模型的参数配置,并利用该模型预测2014年和2018年共三个时段的低、中、高纬度提前5 d的电离层TEC,对地磁活动的平静期和扰动期的电离层TEC预测精度进行实验分析.结果表明,经过QPSO优化的LSTM模型对TEC进行连续5 d预测时,相对于单一LSTM模型,QPSO-LSTM模型在太阳活动低年均方根误差最多降低了0.34 TECU,而相对精度最多提高了2.68%,而在太阳活动高年,低纬度地区均方根误差最多下降了0.68 TECU,而相对精度在高纬度地区最多提高了2.36%.从不同的角度对比分析发现,QPSOLSTM模型的预测精度均优于单一LSTM模型.For the ionospheric TEC short-term prediction of a single LSTM model,there are difficulties in parameter adjustment and performance Optimization,resulting in low prediction accuracy.Quantum Particle Swarm Optimization(QPSO)and LSTM model are combined.The quantum particle swarm optimization algorithm was used to determine the optimal solution,optimize the parameter configuration of the LSTM model,and use the model to predict the ionospheric TEC of low,middle and high latitudes 5 d in advance for three periods in 2014 and 2018,and analyze the prediction accuracy of the ionospheric TEC during the quiet period and disturbance period of geomagnetic activity.The experimental results show that when the LSTM model optimized by QPSO is used to predict TEC for 5 consecutive days,compared with the single LSTM model,the root-mean-square error of the QPSO-LSTM model is reduced by 0.34 TECU at most in low solar activity years,and the relative accuracy is increased by 2.68%at most in high solar activity years.The RMS error decreases by up to 0.68 TECU at low latitudes,while the relative accuracy increases by up to 2.36%at high latitudes.From different analysis angles,it is found that the prediction accuracy of QPSO-LSTM model is better than that of single LSTM model.

关 键 词：LSTM 量子粒子群算法 地磁活动 预测精度 

分 类 号：P352[天文地球—空间物理学]