Highway icing early warning method based on meteorological characteristic modeling

Abstract

The invention discloses an expressway icing early warning method based on meteorological characteristic modeling, and the method comprises the steps: obtaining historical meteorological data of each meteorological station, including regional station meteorological data and station meteorological data along an expressway, and constructing a time sequence window sample; inputting a time sequence window sample of a regional site, and pre-training the model; extracting real meteorological data of a target day corresponding to the historical meteorological data from the historical meteorological data, substituting the real meteorological data into an icing logistic regression formula to calculate an icing probability, and taking the icing probability as an icing soft label; constructing an icing early warning model on the basis of the preprocessing model, and training the model by using time sequence window samples and icing soft labels of stations along the highway; and taking the real-time meteorological data as input, predicting an icing risk probability at a future moment, and performing graded early warning based on the icing risk probability. According to the method, meteorological physical rule constraint and data-driven modeling capability are considered, and the method is suitable for predicting the icing risk of the expressway under complex climate conditions.

Images