A Prediction Model Of Evaporation Duct Characteristics Based On Machine Learning

The atmospheric duct is a horizontal layer that appears under specific weather conditions in the lower atmosphere.It is an essential part of low-altitude wireless channels and has a high probability of appearing in marine and coastal environments.An evaporation duct is an essential form of maritime atmospheric duct,and its height is generally 0-40 meters.The occurrence probability of the evaporation duct in the ocean and its coastal areas can reach more than 80%.It is of great significance for researching and predicting electromagnetic wave transmission characteristics and exploring new communication means.In order to improve the accuracy of predicting the characteristics of the evaporation duct in the marine environment,to support the planning and design of electronic information systems,command and control,and to occupy the advantages of electromagnetic environment and spectrum warfare,this thesis introduces the method of machine learning to address the limitations of the traditional multi-parameter evaporation duct prediction model in terms of performance,a reconstruction method of the evaporation waveguide model is proposed.The main ideas of this method are:(1)Choose multi-parameter model as the object of machine learning,and introduce data such as waveguide thickness,mixing layer slope,critical height,and atmospheric refractive index for model reconstruction;(2)Determine the mean square error Minimal as a machine learning strategy;(3)Use multiple linear regression method to train the model,based on the measured data of the atmospheric refractive index profile,inversion to obtain the optimal parameters of the thickness of the evaporation waveguide,the refractive index of the mixed layer,and the critical height.At the same time,the above method is used to reconstruct the evaporation duct prediction model based on the global surface meteorological parameter assimilation database of the National Center for Atmospheric Research(NCAR).The comparations results show that the deviation between the model built in this thesis and the measured profile is 0.45M-unit,the relative error is reduced to 0.14%,the root-mean-square error is 1.20 m,and the accuracy improvement is 3.46% compared to the Gerstoft model.The waveguide predictive analysis is of great significance,and it also verifies the effectiveness of the improved method in this thesis.The proposed method has great potential in improving the accuracy of evaporation duct height prediction in China and even globally.