CRTS II Track Slab Temperature Forecasting Method Based on CNN-LSTM Fusion Neural Network Network

Accurate prediction of the temperature in CRTS II track slab can give warning to the the occurrence of high temperature in the track slab in time, which can reduce the threat caused by the occurrence of the arch of slab to train operation in high temperature conditions. Considering the temporal lag effect of ambient temperature on the track slab temperature and the temperature change law of the track slab over time, a track slab temperature prediction method based on the convolutional neural network (CNN) and long short-term memory network (LSTM) fusion neural network was proposed. First, CNN was used to convolve the air temperature and rail temperature in the past period on the time-line to extract the influence of external meteorological conditions on the current track slab temperature. Then, the output of CNN at each time point was used as the input feature of LSTM to predict the track slab temperature by using the heat transfer law of track slab. The results show that the external meteorological conditions in the past 12 hours have a great impact on the current track slab temperature. Using the well-trained CNN-LSTM to predict the track slab temperature in the next 40 minutes, the mathematical expectation of the absolute value of prediction error is 0. 925 °C on the test set. © 2023 Science Press. All rights reserved.