Prediction of Low-Visibility Events on Expressways Based on the Backpropagation Neural Network (BPNN)

Low visibility events, including heavy fog and mass fog, are critical threats to travel safety on expressways. However, it is relatively difficult to make predictions of some types of low visibility events in meteorological ways, such as mass fog. Therefore, mathematical modeling can be utilized as a supplement to meteorological ways to predict low visibility events on expressways. Under this context, this study innovatively applies the backpropagation neural network (BPNN) to predict visibility in 30 min using data collected from meteorological monitoring stations on the expressway of Shandong Province, China. A classification for five warning levels is then conducted based on the predicted visibilities according to the related National Standard of China. The results suggest that the total prediction accuracy of the warning levels reaches 86.63%, which is leading in relevant research. Accurate predictions of low visibility events can prompt operators of the expressways to take countermeasures in advance to improve the safety of road travel during foggy days.