Spacecraft Orbit Prediction Method Based on SSA-AM-LSTM

In order to improve the accuracy of spacecraft orbit prediction, a spacecraft orbit time series prediction method integrating Sparrow Search Algorithm (SSA), Attention Mechanism (AM) and Long Short-Term Memory (LSTM) network (denoted as SSA-AM-LSTM) is proposed. A nonlinear model between spacecraft orbit time series variables and spacecraft longitude, latitude and elevation is established using the LSTM network; the weights of the LSTM network are optimised based on the Attention Mechanism to solve the multi-value problem; and the AM-LSTM network is structurally optimised by incorporating the sparrow search algorithm to improve the model accuracy. Experiments are conducted using real orbital data from five in-orbit spacecraft. The experimental results show that the proposed model reduces the error evaluation indexes RMSE and MAE by 93.8% and 95.48% and 64.06% and 71.33% respectively compared to the LSTM network and the AM- LSTM model, and the MAPE indexes also decrease by 95.48% and 71.33%, which indicates that the accuracy of the proposed model is greatly improved compared to the traditional neural network.