Optimization And Application Of LSTM Model In Multi-point Deformation Combination Prediction Of Operating High-speed Railway

Operating high-speed railway will deform due to the combined effects of various influencing factors.Reasonable deformation prediction analysis can provide important guarantee for safe operation of line.Traditional deformation prediction methods have problems such as insufficient accuracy and low efficiency when processing non-stationary and nonlinear deformation data,which can no longer meet actual needs.In order to improve the accuracy and efficiency of deformation prediction,strengthen the prevention and control of operational risks,and promote the intelligent early warning of high-speed railway,this paper carries out research on the deformation prediction and analysis of operating high-speed railway.The main contents are as follows:1.The LSTM prediction model optimized by swarm intelligence algorithm is established.Two kinds of swarm intelligence algorithms,GWO and WOA,are used to optimize the initial learning rate and the number of neural units in the hidden layer of LSTM globally,which overcomes the problem of slow manual tuning of hyperparameters and easy convergence to local optimal solutions.The corresponding prediction models are applied to the deformation prediction of some missing points in the periodic settlement monitoring of key areas.The results show that the two prediction models can automatically adjust the hyperparameters,extract the deformation information from the spatial correlation of adjacent points effectively,and show good prediction performance.2.The WOA-LSTM multi-point deformation prediction model based on multi-task learning is established.Based on the spatiotemporal correlation of adjacent points,the multitask learning method is used to complete the future short-term deformation prediction tasks of multiple points in the same frame.Through experimental verification,the results show that compared with the single-point prediction model,the proposed model can not only realize the multi-point parallel prediction of future short-term deformation,save the time of deformation prediction,but also fully excavate the deformation law,show strong generalization ability.3.The multi-point deformation combined prediction model of operating high-speed railway is established.The WOA-SVR and WOA-LSTM multi-point deformation prediction models are integrated by using several weight allocation methods,and the useful information of each single prediction model is comprehensively used for combined prediction.Through experimental verification,the results show that the combined prediction model with reasonable weight distribution takes more time than a single prediction model,but it has higher prediction accuracy and higher reliability..