Multiscale Decomposition Prediction of Propagation Loss for EM Waves in Marine Evaporation Duct Using Deep Learning

A tropospheric duct (TD) is an anomalous atmospheric refraction structure in marine environments that seriously interferes with the propagation path and range of electromagnetic (EM) waves, resulting serious influence on normal operation radar. Since loss (PL) can reflect characteristics EM waves inside layer, it important to obtain accurate cognition PL TDs. However, strongly non−linear due trapped effect which makes prediction more difficult. To resolve this problem, a novel multiscale decomposition method (VMD−PSO−LSTM) based long short−term memory (LSTM) network, variational mode (VMD) particle swarm optimization (PSO) algorithm proposed study. Firstly, VMD used decompose into several smooth subsequences different frequency scales. Then, LSTM−based model for each subsequence built predict corresponding subsequence. In addition, PSO optimize hyperparameters LSTM model. Finally, predicted are reconstructed final results. The performance VMD−PSO−LSTM verified by combining measured PL. minimum RMSE MAE indicators VMD−PSO−PSTM 0.368 0.276, respectively. percentage improvement compared other methods reach at most 72.46 77.61% MAE, respectively, showing has advantages high accuracy outperforms comparison methods.