Prediction of Satellite Solar Radiation Pressure Parameters Based on Recurrent Neural Network

The BeiDou Navigation Satellite System (BDS) has been put into operation. With the system's continuous operation, the data accumulated by the BDS has a growing trend. At the same time, the service is diversified and refined. Due to the huge system, operation and maintenance are increasingly difficult. To adapt to the high-quality and refined operation of the BDS, it is proposed to introduce artificial intelligence-related technologies to assist the BDS in achieving high-quality intelligent operation and maintenance. Based on the three network models of recurrent neural networks (RNN, LSTM, and GRU), this paper models and analyzes the BDS satellite's solar radiation pressure parameters. The optimal model and hyperparameters are obtained through data mining and analysis and model training, and the prediction model is used to verify the measured data. It is found that the prediction accuracy of the three recurrent neural network models is equivalent, the average accuracy is more than 90%, and the prediction accuracy from high to low is GRU, RNN, and LSTM. The prediction method based on the cyclic neural network adopted in this paper can be applied to the state prediction of time series data of the BDS and has certain reference significance for the construction of intelligent operation and maintenance of the BDS.