基于SSA-PSO-LSTM模型的电离层TEC预报OA

Due to various factors,the time series of total electron content(TEC)in the ionosphere has nonlinear and nonstationary characteristics.To improve the accuracy of the long short-term memory(LSTM)neural network model in predicting TEC in the ionosphere,this paper introduced singular spectrum analysis(SSA)and particle swarm optimization(PSO)algorithms into the neural network model and constructed a new SSA-PSO-LSTM model.On the one hand,SSA was used for data preprocessing of the time series of TEC,and on the other hand,the PSO algorithm was used to improve the parameters of the LSTM neural network model.The experiment was conducted using the time series data of TEC in the ionosphere in grid points provided by the European Reference Organisation for Quality Assured Breast Screening and Diagnostic Services(EUREF).The experimental results show that during the magnetic calm period and the magnetic storm period,the root mean square errors of the TEC prediction results of the combined model are 0.28 TECu and 0.83 TECu,respectively,with an average relative accuracy of 96.35%and 91.33%,which are superior to those of the comparative model,verifying the effectiveness and superiority of the proposed combined prediction model in this paper.