Research On Key Technology For Railway Load Vehicle Flow Forecasting Based On The Railway Freight Big Data

In recent years,the industrial structure of China has been escalating.The freight logistics market of China has many changes.The technology concept has become more and more innovative.All the situation above promoted the revolution of China railway freight transportation.To improve the service levels and enhance the competitiveness of the railway freight transportation market,China Railway Corporation introduced the actual goods system reform and gradually increased the speed of the goods delivery.To realize the "quick,easy,fast" operation of railway freight express,it is inevitable to carry out the instant tracking management and dynamic forecasting for the railway vehicles during the entire process of transporting.The accurate and meticulous railway vehicle flow forecasting becomes the bases of the marketization,logistics,and intelligence of railway freight transportation.In addition,instant and accurate calculation of the railway vehicle flow is the very important guarantee for the dispatchers to make predictable railway vehicle flow adjustment plan,to ensure the balanced and unobstructed transportation of the whole railway network,to improve the vehicle flow organizational level,to short the delivery time of goods,to accelerate the utilization of freight cars and to increase the freight forwarding revenue.With the rapid development of big data technology,there has been a phenomenon of rapid growth of data explosion and fast development of artificial intelligence in all fields in the world.The field of railway transportation has also begun to use these new technologies.Railway informatization construction has accumulated huge amounts of vehicle flow data during the past 30 years,which has great value for the study of the railway vehicle flow.Therefore,applying big data,cloud computing and artificial intelligence technology,based on the accumulated historical data of the railway freight big data,it is a good opportunity for a new model and algorithm for calculating vehicle flow efficiently,instantly and accurately.Starting with the four stages of "data foundation,spatio-temporay distribution,dynamic coordination,system support",this paper takes the vehicle flow forecasting as the research object,focuses on the key technologies of load vehicle flow forecasting and takes the railway freight big data as the basis.Specific content includes:(1)The reconstruction and fusion method of basic data for serving the railway vehicle flow forecasting.Combining with the railway freight transportation production system and related data,the basic data model serving the railway vehicle flow forecasting was built.In view of the widespread railway production anomaly data,this paper designed the reconstruction and fusion method of the basic data for serving the railway vehicle flow forecasting from the aspects of anomaly analysis,anomaly identification,and abnormal remodeling.This method was to meet the basic data requirements of railway vehicle flow spatio-temporay forecasting.(2)Load vehicle flow routing trajectory pattern recognition and forecasting method.The traditional railway vehicle routing algorithm was described detailly.The reasons that the current routing algorithm has a low cashing ratio appeared in actual production situation were analyzed.An improved routing decision-making scheme that met the requirement of traffic flow estimation was proposed.Based on the railway freight big data,the spatial trajectory patterns of the vehicle routing between different ODs were tapped.The trajectory features were clustered by the reduced-dimensional similarity algorithm,and the railway vehicle routing trajectory recognition algorithm based on pattern-perception,which was recognizing and sensing from the mass vehicle flow data.To improve the accuracy of vehicle flow routing decision-making,a variable-length Markov model of railway vehicle flow routing based on the feature weighting was constructed.(3)Load vehicle on-station time mining and learning method.The railway vehicle flow period was an important guarantee to ensure forecast accurately.For vehicle on-station time,the traditional method adopted the statistical index as the calculation basis,which was too extensive and difficult to adapt to the different characteristics of the different station.According to the characteristics of the different station,based on the railway freight big data,this paper dug the characteristic indexes of vehicle on-station time in depth and constructed the vehicle on-station inference model.Based on the fluctuation of vehicle on-station time in the horizontal axis,the vehicle on-station time forecasting model was constructed,which was making the key nodes in the railway network more reliable to provide accurate forecasting service.(4)The whole vehicle flows dynamic collaborative forecasting method.Taking a single vehicle as the research object,a dynamic spatio-temporay vehicle flows forecasting model was constructed.The accuracy of the individual forecasting of the vehicle was guaranteed by the iterative calculation.At the macroscopic level of the whole railway network,a spatio-temporal forecasting model of arrival vehicle flows was constructed to ensure the accuracy of forecasting.A new method based on spatio-temporal iteration was proposed to reconstruct the railway vehicle flows.All the vehicles were divided according to spatio-temporal slices.With the change of vehicle state in the railway network,the spatio-temporal forecasting results and node states were updated and adjusted instantly.(5)Railway vehicle flow forecasting computing architecture and large-scale parallel forecasting task optimization.Considering the continuous and rapid growth of mass data flows,the paper built a railway vehicle flow forecasting computing framework for the whole vehicle flows,which fully coordinated the relationship between historical batch data and instant streaming data to ensure a healthy,rapid and stable forecasting operation.Combining with the characteristics of multi-node and large flow,a railway vehicle flow forecasting parallel assignment model based on the computing performance was constructed to make full use of cloud computing resources,which ensured the efficient operation of computing clusters.From the aspects of overall structural design and functional module design,the prototype system of load vehicle flow forecasting based on the railway freight big data was designed to enhance the practicality of the study and ensure its practical significance.