Design of an intelligent interference prediction model for high-speed frequency hopping satellite communication systems

This study presents an intelligent interference prediction method to enhance the performance of high -speed frequency hopping communication systems operating in "blind hopping" mode under random comb-shaped interference, ensuring reliable communication in complex electromagnetic environments. First, a random comb-shaped interference data set, incorporating dynamic comb-shaped characteristics that span the frequency hopping channel, is constructed. Next, a frequency-specific loss function is introduced to address the label imbalance between the interference and non-interference categories. Finally, a novel temporal interference prediction model based on an enhanced Transformer architecture is proposed, integrated with convolutional networks and attention mechanisms. The model is designed to predict the interference conditions in the next slot of high-speed frequency hopping communication systems by leveraging historical spectral interference data. Simulation results show that the model achieves an interference prediction accuracy of 99.54% and a false alarm rate of 9.77%. In conclusion, the proposed method, through accurate interference prediction, provides a reliable solution for adaptive frequency hopping pattern reconstruction in space-based systems for next-generation applications.