| Atmospheric aerosols have an important impact on the air environment and climate change,and lidar is an advanced instrument about detecting and monitoring atmospheric aerosol,through the research and calculation of lidar detection data,the necessary microphysical parameters can be obtained.Analyzing the microphysical parameter accurately can better understand the impact of atmospheric aerosols on the environment and climate.However,the traditional method of solving the lidar equation requires a lot of mathematical calculations and assumption,the process is complicated,and the neural network can solve the nonlinear mathematical relationship well.Therefore,in order to obtain the extinction coefficient better,this paper proposes a method of using neural networks to invert and predict the lidar equation.The main research contents are as follows:(1)Using the lidar system to detect atmospheric data and pre-processing it,establishing a BP neural network model.(2)BP neural network is easy to fall into local extremum,so GA and PSO are used to optimize the parameters of BP neural network to establish GA-BP and PSO-BP lidar equation inversion models.And comparing with the inversion results of the BP neural network,it is verified that the two optimized BP neural network model have more accurate than BP neural network model.(3)Using GA-BP and PSO-BP network models to invert the extinction coefficient,and compare the inversion with the inversion results obtained by the Klett method.Experiments show that the extinction coefficients obtained by the GA-BP and PSO-BP network models are both close to the Klett method,the overall trend remains the same,and the relative error is controlled below20%.This paper uses neural network method to invert atmospheric extinction coefficient,which eliminates the complicated calculation process of the traditional method and achieves a good result.Therefore,it provides a convenient and effective new idea for the inversion of atmospheric extinction coefficient. |
