Tropospheric Duct Characteristics And Its Inversion Method Based On Deep Learning

Tropospheric ducts is a natural phenomenon.It often occurs in the marine environment.It is an abnormal atmospheric state.The existence of this phenomenon will have a greater impact on the electromagnetic waves propagating in the troposphere,and they will be trapped in the duce layer,thus affecting the wireless communication equipment that requires electromagnetic wave to work.Therefore,it is necessary to conduct related studies on tropospheric ducts and its inversion method.Because the direct measurement of the tropospheric ducts structure is very difficult,a Refractivity From Clutter(RFC)technique using the easily measured radar clutter is proposed.This paper is different from the traditional method.Based on the currently more popular artificial intelligence field,the deep learning and RFC technology are combined to carry out the inversion of tropospheric ducts.The inversion process is introduced in detail,and the inversion results are analyzed.The feasibility of deep learning in the inversion of tropospheric ducts by sea clutter is verified,and a high-precision inversion model of different types of ducts is established.The main research contents are as follows:1.Several specific types of tropospheric ducts and their corresponding profile models are introduced,and the environmental conditions for the formation of tropospheric ducts are explained in detail.The parabolic equations used for studying the ducts propagation and its specific solutions are introduced.Several necessary conditions for the formation of tropospheric ducts propagation by electromagnetic waves are analyzed,which provides a basis for subsequent inversion studies.2.The basic characteristics and calculation basis of electromagnetic wave propagation loss in tropospheric ducts environment are discussed.Then,based on the deep learning LSTM network,a propagation loss prediction model in tropospheric ducts environment is established,and the trained model is verified by multiple sets of data.The results show that the LSTM prediction has the advantages of high accuracy and high efficiency,even at a far distance the accurate predicted results can also be obtained.3.According to the complex nonlinear relationship between tropospheric ducts profile parameters and sea clutter power,combined with Deep Neural Network(DNN),a model for inversion of tropospheric ducts profiles by using RFC technology is established.The factors that affect the accuracy of deep learning inversion are analyzed in many ways.According to the analysis results,we select the best parameters to train DNN,and finally get an optimal inversion model,and apply it to the problem of refraction index profile inversion.The results show that deep learning is more accurate in inversion of tropospheric ducts compared with previous algorithms.4.Because the tropospheric ducts also exhibits inhomogeneity on horizontal distance,the feature vectors of horizontal non-uniform parameters of the ducts' profile are extracted and modeled by PCA method.Then,the construction process of the inversion model is described in detail.Taking the evaporation duct and the surface duct as an example,the mapping relationship between the horizontal non-uniform ducts' profile parameters and the sea clutter power data is fitted by the deep neural network,and the horizontal heterogeneous tropospheric ducts inversion is realized.Compared with the traditional particle swarm optimization,it achieves better accuracy and saves a lot of time.