In view of the lack of meteorological visibility data in Anhui Province, this paper selects the meteorological data of four automatic meteorological stations (Mount Huangshan Station, Lingbi Station, Shannan Xigu Station, and Baize Lake Station) under different terrain conditions in Anhui Province from 2017 to 2019, and firstly selects the meteorological elements closely related to visibility by using the gray correlation analysis method. Then, a hybrid algorithm of Genetic Algorithm (GA) and Particle Swarm Optimization Algorithm (PSO) is constructed to optimize the prediction model of the Back Propagation neural network. The visibility of automatic weather stations under four different terrain conditions is predicted, and the prediction results are compared with the RF prediction model and XGBoost prediction model. The results indicate that the GA-PSO-BP neural network prediction model has smaller prediction error and higher model accuracy under any terrain condition.