In order to solve the problem that it is difficult to monitor the structural deformation of platform doors of operating lines in real time and accurately predict the structural deformation data of platform doors of high-speed railway when trains pass the station, this paper adopts artificial intelligent-based neural network method and selects 210 kinds of platform door structural deformation data with different train length, congestion ratio, installation distance and speed as network model training samples. By using two neural network models, CNN and GRNN based on K-Fold, the prediction models of platform door structure deformation under different working conditions of high-speed railway were established, and the comparison and verification were carried out with the remaining sample data. The research shows that both models can effectively predict the operation and maintenance data of railway platform door structure. In terms of prediction accuracy, the GRNN model based on K-Fold optimization is superior to the CNN model. The mean square error of prediction of the GRNN model based on K-Fold optimization can be maintained within 0.22, and the root mean square error is maintained within 0.27. At the leading level in the field of research. The GRNN model based on K-Fold optimization can better predict the structural shape variables of platform doors when trains pass the station, and provide data reference for the design and operation and maintenance of platform doors of high-speed railway.