Two-dimensional Parabolic Equation Correction Model Based on DNN for Radio Wave Propagation
To overcome the defect that Two-dimensional Parabolic Equation(2DPE)can only model two-dimensional communication links without considering lateral diffraction and backward reflection of electromagnetic waves near buildings in urban environment,a modified 2DPE model based on Deep Neural Network(DNN)is proposed.The three-dimensional model of urban buildings is built through digital elevation map,from which seven characteristics such as propagation distance,propagation angle,and building coverage are extracted to characterize the distribution of buildings on the propagation path and the deployment of transceiver antennas.Then combined with the measured data,the dataset for correcting 2DPE is constructed;through the training of DNN,the modified model of 2DPE is constructed to make it suitable for the prediction of radio wave propagation in a complicated three-dimensional environment.The simulation results show that compared to linear regression,support vector regression,and decision tree model,the calculation accuracy of the 2DPE correction model based on DNN is high in the three-dimensional propagation environment,and the prediction error on the test set is reduced by a maximum of 46.8%.
radio wave propagationurban environment2DPEmachine learning
国家科技期刊平台
NSTL
万方数据
维普
