Research On Delay Prediction And Operation Adjustment Of High-Speed Railway Trains Under Rainfall Conditions

During the operation of high-speed railway trains,the climatic conditions are changeable.In order to ensure the safety and efficiency of high-speed railway operation,it is necessary to pay attention to the influence of external environmental conditions.Rainfall,as the most common meteorological factor in train delays,often causes trains to run at a speed limit,which leads to train delays.Therefore,predicting train delays under rainfall conditions can help dispatchers to predict the train status in advance,so as to better predict train delays.Control the spread of late.While predicting the delay time,it is of great research significance to take the exact operation adjustment strategy in time to make the efficient recovery of the delay time.Benefiting from the rapid development of big data technology,data such as actual train operation performance and operating external environmental factors are preserved,which provides conditions for using machine learning to study train delays under rainy weather conditions.Based on the records of the actual operation process of high-speed railway trains,this paper excavates the law of the evolution of delays during train operation,and uses machine learning methods to establish a high-speed train delay prediction model under different typical rainy weather.Finally,the method of association rules is used to extract the scheduling policies under various operating scenarios.The main research contents are as follows:(1)Introduced the basic theories.Classified the causes of high-speed railway delays and summarized the process of train delays.The effect of rainfall on train delays and the relevant concepts of train operation adjustment were shown in the form of event causal diagrams,and the emergency response work under rainfall conditions was summarized.(2)Basic data samples were processed.The work of data processing has been done on the original data.Firstly,the source and main content of the data were introduced,and then the train operation data and weather data were combined based on time to construct the weather-delay data.After the merger process,multiple data were obtained,including meteorological and train operation indicators and train number information.Finally,data preprocessing was completed,and the cleaned data results were used as samples for subsequent research.(3)Statistics and analysis of the data were completed.Based on the actual operation data of trains,statistics and visual analysis of delay indicators,mainly including: late arrivals and departures,additional delays at stations and sections,and redundant time at stops and sections.Combined with the visualization results,the main research objects of train delay law under rainfall conditions were sorted out,which provided support for the selection of eigenvectors in railway delay prediction.(4)Selected typical rainy weather.Firstly,the binary logistic method was used to analyze the characteristics of the weather data samples,and the meteorological factors that affect the delay were screened out.Then,the K-Means algorithm was used to cluster the selected meteorological factors to form different weather categories.Five typical rainy weathers were selected from all the categories as research samples for late prediction.(5)Build a delay prediction model under typical rainy weather.Two kinds of delay prediction models for departure and arrival were constructed respectively.Firstly,the attribute features of the training set of the model were selected from the railway level,and then combined with the weather indicators selected by binary logistic regression in the previous section,they were synthesized as candidate features.Then,the three methods of variance selection,recursive elimination of features,and XGBoost tree-based feature selection were used to screen the feature indicators of the weather and railway levels together,and the screening results were used as the input variables of the machine learning model.Three machine learning models were used to compare the effects.The results show that the random forest model has high prediction accuracy in most typical rainy weather.(6)Extracted the association rules.Using the Light GBM model,firstly,from a macro perspective,delay recovery models of the compression stop time strategy and the compression interval running time strategy were constructed.Then,from a microscopic perspective,based on the Apriori algorithm,the association rule mining model of the compressed stop time strategy and the compressed interval running time strategy was constructed.Finally,the results of the association rules were sorted out,and the rules between the train operation indicators,rainfall conditions and operation adjustment were analyzed.It was found that under the premise of sufficient redundancy time,When operating across the Wuhan-Guangzhou highspeed railway line,there is a great possibility that a delayed train will become punctual,and when the train encounters heavy rainfall,the short delay can be recovered,and most trains choose a small range Compressed stop or interval running time.