基于CNN-GRU的对流层散射通信短期传输损耗预测

Short-term Transmission Loss Prediction of Troposcatter Communication Based on CNN-GRU

针对对流层散射通信短期传输损耗预测这一难题,提出了一种基于卷积神经网络联合门控循环单元(Convolutional Neural Networks-Gated Recurrent Neural Network, CNN-GRU)的混合神经网络模型,利用CNN提取气象数据的特征,再利用GRU提取数据的时间相关性特征,最后通过全连接层将特征映射到信道传输损耗,从而利用气象数据实现对散射通信短期传输损耗的预测。通过实测数据进行了验证,结果显示该模型相对现有的短期预测算法精确度提升10.98%,最大误差下降8.57 dB,能更准确地预测对流层散射通信短期传输损耗。

To address the challenging task of predicting short-term transmission losses in troposcatter communication,a hybrid neural network model based on convolutional neural networks combined with gated recurrent units(CNN-GRU)is proposed.This model utilizes CNN to extract features from meteorological data and GRU to capture the temporal correlations in the data.The features are mapped to channel transmission losses through fully connected layers,enabling the use of meteorological data for predicting short-term transmission losses in troposcatter communication.Validation using actual measurement data shows that this model improves accuracy by 10.98%compared with existing short-term forecasting algorithms,with a decrease of 8.57 dB in maximum error.It can predict short-term transmission losses in troposcatter communication more accurately.