Path Loss Prediction in Evaporation Ducts Based on Deep Neural Network
Bingwei Shu, Wensheng Zhang, Yunfei Chen, Jian Sun, Cheng‐Xiang Wang
Abstract
The evaporation duct effect is a critical issue in maritime wireless communications. This letter presents a novel scheme based on the deep neural network (DNN) for accurately predicting path loss in evaporation ducts. The environment and antenna parameters are employed as inputs for the lightweight model, enhancing its applicability to natural scenarios. The proposed scheme can achieve superior prediction performance than the k-nearest neighbor model, random forests model, and linear regression model at varying frequencies. In addition, this letter studies the impact of the frequency, receiver height, and transmission distance on the prediction accuracy of DNN. Simulation results show that DNN exhibits high prediction accuracy at low frequencies but experiences a slight accuracy reduction at higher frequencies due to the presence of complex peak regions. The impact of receiver height and transmission distance on the prediction accuracy is not significant.