To address the issue of insufficient consideration of the spatial heterogeneity of wind speed in existing methods for fog visibility prediction, which leads to low accuracy and stability, this paper constructs a Long Short-Term Memory Neural Network with Attention Mechanism (LSTM-AM) model for fog visibility prediction that takes into account the spatial heterogeneity of wind speed. The model quantifies the variation characteristics of wind speed at different spatial locations using a semi-variogram, integrating the spatial distribution of neighboring points and differences in wind speed. It employs wind direction angles and variation values to weight the features of wind speed spatial heterogeneity, effectively extracting these characteristics. Additionally, the Attention Mechanism (AM) enhances the LSTM method by improving its focus on key information, enabling the model to effectively capture and reflect the impact of critical meteorological factors on fog visibility. This enhances the model"s ability to pay attention to important temporal information and improves the accuracy of predictions under conditions of wind speed spatial heterogeneity. The results indicate that the proposed model improves R2 by 10%-20%, reduces RMSE by 25%-40%, and decreases MAE by 26.3%-39.1%, demonstrating high accuracy and stability in fog visibility prediction.