Railway Train Conflicts Dynamic Prediction And Identification Strategies And Methods

The train dispatching is one of the cores of the railway transportation.As the brain and soul of the railway operation,its method and efficiency have a direct impact on the quality on railway.According to lots of reseach around the world,there is only about 4% of dispatcher’s working time spent on the process of selecting and making the conflicts resolution methods and strategies after train conflicts had been detected.About 16% of working time are spent on resolution methods’ s adjusting and implementation,compares to 80% of working time focused on identifying the forthcoming train conflicts happened or to happen.Although the railway technologies of china has great improvement,but in the field of train dispatching,especially on train conflicts prediction,conflicts identification and control still rely on human experience.The intelligence and automation are still in low level,which has become one of the important restrictive factors of rail development to achieve the requirement of safety,punctuality and high efficiency.It needs to be solved urgently.To improve the low level of intelligent and automation on train conflicts identification,prediction and measurement in dynamic environment,this paper focus on the railway train conflicts dynamic prediction and identification problem.Based on the novel ideas of“temporal and spatial evolution” combined with “predictive perception”,under the application of process mining,scene reconstruction,pattern matching,adaptive prediction and temporal and spatial evolution technologies,the dynamic train conflicts prediction and identification strategies and technology based on data-awareness in complex dynamic environment are studied.A dynamic train conflicts prediction and identification solution with high efficiency,intelligence and accuracy is proposed.Firstly,to solve the problem of traffic event data mining,this paper presents a process mining framework,mining algorithm and mining model based on signal state.Supported by it,a minimum traffic event scene reconstruction model is proposed.And a train event time extraction algorithm and conflict identification algorithm are proposed and verified by the historical data in the actual environment.Secondly,in the perspective of temporal prediction,a train event time prediction model based on data mining and statistical learning is constructed.Then,according to the dynamic characteristics of train events in the actual environment,a real-time prediction model based on train event graph in dynamic environment is proposed.An adaptive algorithm is used to correct the prediction result and the historical data of the dispatching command system is used to verify the model and algorithm.Thirdly,in the perspective of spatial prediction,According to the implied mapping relationship between signal state and train location,the relationship between the signal stateand train location and uncertainty of signal state are analyzed,then the mathmatic model of train position identificaiton based on signal state and real-time traffic plan is proposed based on the basic model of signal state.Also,a train position identification and prediction model based on hidden Markov model is proposed.Then a data oscillation filter algorithm is proposed to deal with the signal state oscillation and loss which may occur.Finally,the strategy and method of train conflict detection based on traffic event prediction are studied and the basic mathematic model of train conflict detection is constructed.Combined with the train event time and location prediction model and algorithm,the prediction framework of train conflict and the one-step scenario evolution method of train events are given,then it is extended to the generalized multi-step prediction of train events.Based on those research,temporal and spatial evolution prediction model and solution algorithm of train conflicts are proposed.