Research On The Method Of Long-term And Short-term Forecast Of Severe Weather Along Railway

With the sustainable growth of passenger and freight traffic and operating mileage in China rail,the safety of the railway system is of paramount importance.During the train operation,it is susceptible to adverse weather such as strong wind,heavy rain and snowfall.The natural geological disasters are potential risks to train operation and likely cause driving accidents.Therefore,the real-time meteorological monitoring data along the railway is of great significance for weather prognosis and railway operation safety.In this thesis,the state of the art and national standard documents are analyzed first,and the prognostics methods of adverse weather along railway are studied.According to the actual application scenarios,the short-term and long-term prognostics models are introduced respectively.The performance of the two models is verified with the wind speed data.In the short-term prognostics method,a prognostics model based on Gated Recurrent Unit(GRU)is proposed to capture the varying information characteristics in the time series data and learn the data correlation.At the same time,a convolutional neural network(CNN)is used to extract the local deep features and scale-invariant features to optimize the previous GRU model,and a GRU-CNN prognostics model is thus constructed.Using the wind speed data with a time resolution of 30 seconds along the railway,the model is trained,and used to predict the wind speed ahead of half a minute,two minutes,and five minutes.The experimental results show the effectiveness of the short-term model.In the long-term prognostics method,a prognostics model based on Dual-Stage Attention-Based Recurrent Neural Network(DA-RNN)is proposed.Its temporal attention mechanism learns the target variable information from historical time sequence to predict its future state,while the input attention mechanism can adaptively select the sequence which has paramount influence on the current target variable as a complement to target variable prediction.Similarly,using the wind speed data with a time resolution of one hour,the long-term prediction of wind speed is conducted,and the experimental results show that the accuracy of wind speed prediction can be improved by considering other weather variables.In this thesis,the weather conditions along railway is predicted by the proposed prognostics models.With the wind speed data,the long-term and short-term prognostics models are verified.The outcome of this research work will be able to provide valuable reference for adverse weather monitoring and alarming.