Train Time Delay Prediction for High-Speed Train Dispatching Based on Spatio-Temporal Graph Convolutional Network

Abstract

Train delay prediction can improve the quality of train dispatching, which helps the dispatcher to estimate the running state of the train more accurately and make reasonable dispatching decision. The delay of one train is affected by many factors, such as passenger flow, fault, extreme weather, dispatching strategy. The departure time of one train is generally determined by dispatchers, which is limited by their strategy and knowledge. The existing train delay prediction methods cannot comprehensively consider the temporal and spatial dependence between the multiple trains and routes. In this paper, we don't try to predict the specific delay time of one train, but predict the collective cumulative effect of train delay over a certain period, which is represented by the total number of arrival delays in one station. We propose a deep learning framework, train spatio-temporal graph convolutional network (TSTGCN), to predict the collective cumulative effect of train delay in one station for train dispatching and emergency plans. The proposed model is mainly composed of the recent, daily and weekly components. Each component contains two parts: spatio-temporal attention mechanism and spatio-temporal convolution, which can effectively capture spatio-temporal characteristics. The weighted fusion of the three components produces the final prediction result. The experiments on the train operation data from China Railway Passenger Ticket System demonstrate that TSTGCN clearly outperforms the existing advanced baselines in train delay prediction.

A high-speed railway network dataset from train operation records and weather data

“…The high-speed railway network dataset can be processed as the materials for effective methods to issue the problems in large-scale complex network, complex dynamical system, intelligent transportation, deep learning, data mining and other fields, including but not limited to complex network modeling 10 – 12 , complex dynamic system pattern mining 5 , 13 – 15 , travel demand analysis 16 , community detection and discovery 17 – 19 , urban accessibility research 20 , 21 , train delay analysis 6 , 7 , 22 – 24 , task mining on multi-scale and dynamic graphs 25 – 27 . In addition, it can be used to optimize the actual railway operation and management, such as (a) train operation scheme and schedule adjustment, (b) passenger service network improvement, (c) train speed, punctuality, capacity, and energy consumption prediction, (d) real-time dispatching, (e) intelligent driving assistance, (f) fault or accident detection and (g) maintenance plans making.…”

DCGCN: Dynamic community graph convolutional network for traffic forecasting

“…(5) TSTGCN [12]: Spatiotemporal convolutional train delay forecasting model considering station distance.…”

“…The innovation and use of GNNs has enabled the modeling of non-Euclidean properties and directionality of traffic data [12], but previous studies, such as Zhao et al [13], Dai et al [14], and Lee et al [15], mainly focused on modeling spatial dependence using distance only, and this approach often does not represent the traffic state propagation patterns between traffic monitoring stations well, while GNNs are extremely dependent on graph construction. Therefore, many studies have now attempted to capture more possible correlations between stations during the training process.…”

Stance Detection Based on User Feature Fusion

“…Based on the deep forest, a fusion-enhanced cascade model (FECM) is proposed, which fuses the surface features from the multigranularity module and combines the cascaded gradient descent tree (GBDT) and random forest models to fuse the features [ 17 ]. Zhang et al [ 18 ] proposed a train spatiotemporal graph convolutional network (TSTGCN) that includes two parts, spatiotemporal attention mechanism and spatiotemporal convolution. This method is used to capture spatiotemporal characteristics in three levels; therefore, the final result is predicted by the weighted fusion of the three components.…”