基于CNN-BiLSTM的短波通信频率预测研究

受电离层变化影响,短波通信频率如何实现优选一直是影响短波通信效果的关键。针对目前短波频率预测方法在远程通信中出现的预测精度不高、不能较好满足通信需要的现状,提出一种基于历史通信数据的卷积神经网络(CNN)和双向长短期记忆神经网络(BiLSTM)相结合的预测模型实现对短波通信频率值进行预测,并与单特征、多特征输入长短期记忆神经网络(LSTM)预测模型进行对比。仿真结果表明,该模型能够实现短波通信频率预测且相对于单特征输入LSTM预测用时更短、多特征输入LSTM预测精度更高,具有一定的可行性。

Frequency Prediction of HF Communication based on CNN-BiLSTM

Influenced by the ionosphere change,how to optimize the frequency of HF communication is always the key to the effect of HF communication.In view of the current situation that the prediction accuracy of shortwave frequency prediction method in long-distance communication is not high enough to meet the needs of communication,a prediction model based on the combination of CNN(Convolutional Neural Network)and BiLSTM(Long Short-Term Memory)using historical communication data is proposed to predict the frequency value of HF communication,and its comparison with the single feature and multi-feature input LSTM(Long Short-Term Memory)prediction model is also done.The simulation results indicate that this model can predict the frequency of HF communication,and compared with the single feature input LSTM,the prediction time is shorter and the prediction accuracy of multi-feature input LSTM is higher,so it has certain feasibility.