A Transferable Real‐Time MUF Forecasting Model for HF Communication in China

Abstract High‐frequency (HF) communication is an important means of strategic and emergency communication, playing an irreplaceable role in both military and civil applications. The maximum usable frequency (MUF) is a key parameter controlling HF signal propagation, and its real‐time and accurate forecast has long been a major challenge in engineering. To address this issue, this study proposes a transferable short‐term MUF forecasting model based on mutual information for feature selection and the Neural Basis Expansion Analysis with Exogenous Variables architecture. The model integrates intelligent approaches such as deep learning and adopts a phased, progressive hyperparameter optimization strategy (random search and Tree‐structured Parzen Estimator) to enhance model performance. Experiments conducted on the Changchun–Jingyang circuit demonstrate that the proposed model achieves an RMSE of 1.66 MHz on the test set, representing a 27% improvement over baseline models. Further transfer experiments on the Chongqing–Hainan, Xinxiang–Suzhou, and Guangzhou–Hainan circuits demonstrate that the proposed model achieves an improvement of over 25% compared with the baseline models, confirming its strong generalization and adaptability. The experimental data we used span four seasons—spring, summer, autumn, and winter—and cover both high and low solar activity years, ensuring the reliability of the model. Overall, the proposed MUF forecasting model achieves high accuracy while exhibiting strong transferability, providing an effective technical solution for real‐time HF communication applications under complex environmental conditions.