Train delay time prediction method, device, equipment and medium

Abstract

The invention provides a train delay time prediction method and device, equipment and a medium. The method comprises the steps that a delay feature data set based on cause events and a train operation data set based on a working diagram are constructed; performing feature engineering processing on the preprocessed late feature data set to obtain a first feature data set; a train delay time prediction model is constructed, a convolutional neural network is trained with the first feature data set as input and the delay train delay time as output, and the output of the convolutional neural network, the planned running diagram parameters and the actual running diagram parameters as input; the late time of each station after the late train arrives at the S station is used as output to train the long short term memory network; and receiving train delay information and running diagram information of the fault train, inputting the train delay information and the running diagram information into the convolutional neural network to obtain delay time, and predicting and determining the delay time of the fault train arriving at each station after the S station by combining the running diagram information long-short memory network.

Description

Technical Field

[0001] The present invention relates to the field of data analysis technology, and in particular to a method, device, equipment and medium for predicting train delay time. Background Art

[0002] With the development of communication technology, modern rail transit systems have adopted more advanced automatic train monitoring technologies to ensure safe, reliable, and efficient train operation. Leveraging advanced computer and communication technologies, subway monitoring systems can collect and process various train operation data in real time, such as train location, speed, and signal status. Based on this information, they automatically adjust train operation plans to ensure safe and punctual operation. Furthermore, they can provide real-time and historical train operation information to support decision-making by operations management departments.

[0003] However, when train delays occur during current station operations, station attendants' ability to access information and coordinate with passengers remains limited. For one thing, when a train is delayed due to various faults, only a rough estimate of the delay time in the current direction of travel is provided at the specific location of the delay, without considering the specific delay time at subsequent stations caused by the train in question. Furthermore, coordination between stations cannot be synchronized in a timely manner. With existing traditional handheld radios, station attendants cannot quickly understand train delays and related train operation information. If this information is relayed through the station's integrated control room, communication efficiency is further reduced, making it difficult to adapt to the growing needs of passengers. Summary of the Invention

[0004] In order to overcome the problems existing in the related art, the present disclosure provides a train delay time prediction method, device, equipment and medium to solve the technical problem in the related art that when faced with emergencies and emergency events causing train delays, station attendants can understand the situation in a timely and accurate manner, and solve the technical problem of being unable to respond to passengers' demands in a timely manner, which easily leads to passenger complaints.

[0005] One or more embodiments of this specification provide a method for predicting train delay time, including the following steps:

[0006] 1) Obtain historical train operation data and construct a delay feature dataset based on causal events and a train operation dataset based on a train operation diagram. The historical train operation data includes train delay information based on causal events and corresponding train operation diagram information. The train delay information includes basic information, operation information, and delay time information of the train causing the delay. The basic information includes the operation parameters of the train causing the delay. The operation information includes the location of the train accident, the location of the next station, and the human factors, machine factors, and environmental factors that cause the train delay. The delay time information includes the fault start time, fault end time, and delay time. The train operation diagram information includes the planned train operation diagram parameters and the actual train operation diagram parameters based on the causal event.

[0007] 2) Preprocessing the delay feature data set and the train operation data set, and performing feature engineering processing on the preprocessed delay feature data set train delay information to obtain a first feature data set;

[0008] 3) Construct a train delay time prediction model composed of a convolutional neural network and a long short-term memory network, wherein the first feature data set is used as input, and the delayed time of the delayed train arriving at station S / departing from station S is used as output to train the convolutional neural network, and the delayed time of the delayed train arriving at station S / departing from station S, the planned operation diagram parameters, and the actual operation diagram parameters are used as input, and the delayed time of each station after the delayed train arrives at station S is used as output to train the long short-term memory network, thereby obtaining a trained train delay time prediction model;

[0009] 4) Receive the train delay information and timetable information of the faulty train sent by the ATS system and input them into the train delay time prediction model. The convolutional neural network predicts the delay time of the faulty train at station S / departure time from station S, and combines the timetable information with the long-short memory network to predict the delay time of the faulty train at each station after station S. The model then feeds back the basic information and operation information of the faulty train to the operation management system of the corresponding station.

[0010] One or more embodiments of this specification provide a train delay time prediction device, including:

[0011] A data construction module is used to obtain historical train operation data and construct a delay feature dataset based on causal events and a train operation dataset based on a train operation diagram. The historical train operation data includes train delay information based on causal events and corresponding train operation diagram information. The train delay information includes basic information, operation information, and delay time information of the train causing the delay. The basic information includes the operation parameters of the train causing the delay. The operation information includes the location of the train accident, the location of the next station, and the human factors, machine factors, and environmental factors that cause the train delay. The delay time information includes the fault start time, fault end time, and delay time. The train operation diagram information includes the planned train operation diagram parameters and the actual train operation diagram parameters based on the causal event.

[0012] A data processing module is used to preprocess the delay feature data set and the train operation data set, and perform feature engineering processing on the train delay information in the preprocessed delay feature data set to obtain a first feature data set;

[0013] A model construction and training module is used to construct a train delay time prediction model composed of a convolutional neural network and a long short-term memory network, wherein the first feature data set is used as input, and the delay time of the delayed train arriving at station S / departing from station S is used as output to train the convolutional neural network, and the delay time of the delayed train arriving at station S / departing from station S, the planned operation diagram parameters, and the actual operation diagram parameters are used as input, and the delay time of each station after the delayed train arrives at station S is used as output to train the long short-term memory network, thereby obtaining a trained train delay time prediction model;

[0014] A prediction module is used to receive the train delay information and timetable information of the faulty train sent by the ATS system and input them into the train delay time prediction model. The convolutional neural network predicts the faulty train's delay time at station S / departure time from station S, and combines the timetable information with the long-short memory network to predict the delay time of the faulty train at each station after station S; and

[0015] The information exchange module is configured to transmit the basic information and operational information of the faulty train, as well as the delay times of the faulty train at each station after arriving at station S, as predicted by the prediction module, to the corresponding station's operations management system. One or more embodiments of this specification provide a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the train delay prediction method described above.

[0016] One or more embodiments of this specification provide a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the method for predicting train delay time as described above is implemented.

[0017] The present invention provides a train delay time prediction method, device, equipment and medium, which have the advantages of extracting historical train operation data, analyzing the basic information, operation information and delay information data of the train based on the causal event to determine the characteristic data set composed of the delay influencing factors corresponding to each delayed train, and then using the train delay time prediction model to learn the multivariate data about the train delay type and estimate the delay time, so as to obtain the delay time that can be used to predict the arrival at the remaining stations under the corresponding failure situation of the train, and through the operation management system such as the collaborative train ATS system and the station service individual system for communication, it can realize the explanation of the two situations of train delay caused by station equipment failure and train delay caused by the operation process, so as to improve the station staff's ability to understand the situation in time when facing sudden incidents and emergency events causing train delays, and comprehensively improve the station operation service level, thereby realizing a comprehensive solution mainly based on station service individuals that can be applied to subway delays. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate one or more embodiments of this specification or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0019] Figure 1 A flow chart of a method for predicting train delay time provided in one or more embodiments of this specification;

[0020] Figure 2 A framework diagram of the model training process provided for one or more embodiments of this specification;

[0021] Figure 3 A diagram of the train delay prediction process using a neural network model based on ensemble learning provided in one or more embodiments of this specification;

[0022] Figure 4 A block diagram of a train delay time prediction device provided in one or more embodiments of this specification;

[0023] as well as

[0024] Figure 5 A schematic diagram of the structure of a computer device provided in one or more embodiments of this specification. DETAILED DESCRIPTION

[0025] In order to help those skilled in the art better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below in conjunction with the drawings in one or more embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this invention.

[0026] The present invention will be described in detail below with reference to specific implementation methods and the accompanying drawings.

[0027] Method Example

[0028] According to an embodiment of the present invention, a method for predicting train delay time is provided. Figure 1-2 As shown, Figure 1 The flow chart of the train delay time prediction method provided in this embodiment is as follows: Figure 2 The model training process framework diagram provided in this embodiment, according to the train delay time prediction method of the embodiment of the present invention, includes the following steps:

[0029] Step S1: Acquire historical train operation data, wherein the historical train operation data includes train delay information based on the causal event and corresponding operation diagram information; wherein the train delay information includes basic information, operation information, and delay time information of the train causing the delay;

[0030] Basic information is the operating parameters of the train causing the delay, which may include the line, train number, sequence number, destination number, and trainset number;

[0031] Operational information includes the location of train accidents, the location of the next station, and the causes of train delays. The causes of train delays include human factors, machine factors, and environmental factors. Each factor contains n features, and each feature has multiple influencing feature subsets. For example, machine factors include features such as vehicle failure, operation failure, line failure, and power supply failure. Vehicle failures include feature subsets such as door failure and brake system failure.

[0032] Operational failures include feature subsets such as signal failures and ATP failures;

[0033] Line faults include feature subsets such as track damage, foreign object intrusion, and turnout failure;

[0034] Power supply faults include feature subsets such as catenary faults and substation faults;

[0035] Environmental factors include a subset of characteristics such as rainfall and snowfall intensity, wind speed, temperature, and visibility;

[0036] Human factors include feature subsets such as passenger density and passenger events.

[0037] It should be noted that under each subset category, if there is a more detailed division, for example, the door fault category is A1, if this subset can be further drilled down, the next subset is represented as a 1n , i=1,2,3…n.

[0038] The delay time information includes the fault start time, fault end time, and delay time. The delay time is equal to the sum of the fault repair time and the operation preparation time after the fault repair, which can be understood as the end time minus the start time.

[0039] The operation diagram information includes the planned operation diagram parameters and the actual operation diagram parameters based on the causal event. The planned operation diagram parameters include the redundant time of the train staying at station S, the planned operation time of the train between stations S-1 and S, and the redundant time of the train running between stations S-1 and S.

[0040] The actual operation diagram parameters include the actual stay time of the train at S station, the deviation between the planned arrival and actual arrival of the train at S station, the deviation between the planned departure and actual departure of the train at S station, and the actual running time of the train between S-1 station and S station.

[0041] Step S2: Preprocess the delay feature dataset and the train operation dataset, and perform feature engineering on the preprocessed train delay feature dataset to obtain a first feature dataset. Preprocessing the historical train operation data may include data cleaning to remove noise, fill in missing values, and handle outliers.

[0042] Step S3, constructing a train delay time prediction model consisting of a convolutional neural network and a long short-term memory network, wherein the first feature data set is used as input, and the delay time of the delayed train arriving at station S or the delay time of departing from station S is used as output to train the convolutional neural network, and the delay time of the delayed train arriving at station S or the delay time of departing from station S, the planned operation diagram parameters and the actual operation diagram parameters are used as input, and the delay time of each station after the delayed train arrives at station S is used as output to train the long short-term memory network, thereby obtaining a trained train delay time prediction model.

[0043] Step S4: Receive the train delay information and operation diagram information of the faulty train sent by the ATS system and input them into the train delay time prediction model. The convolutional neural network predicts the delay time of the faulty train arriving at station S / departing from station S, and combines the operation diagram information with the long-short memory network to predict and determine the delay time of the faulty train at each station after station S. The delay time is then fed back to the operation management system of the corresponding station together with the basic information and operation information of the faulty train.

[0044] The train delay time prediction method provided in this embodiment extracts historical train operation data of trains, analyzes basic train information, operation information, and delay information data based on causal events, determines a feature data set composed of delay influencing factors corresponding to each delayed train, and then uses a train delay time prediction model constructed by a neural network to learn multivariate data about train delay information and operation diagram information and estimate delay time, thereby obtaining a delay time that can be used to predict the arrival time of the train at the remaining stations under the corresponding failure situation. The method also communicates with the operation management system such as the collaborative train ATS system and / or the station service individual system to explain and explain train delays caused by station equipment failures and train delays caused during operation. This improves the station staff's ability to understand the situation in a timely manner when facing sudden incidents and emergency events that cause train delays, and comprehensively improves the station operation service level, thereby realizing a comprehensive solution based on station service individuals that can be applied to subway delays.

[0045] In one embodiment, step S2 specifically includes the following implementation steps:

[0046] Step S21: Regularize the acquired train delay information to obtain fault feature data, then determine the Pearson correlation coefficient between the fault feature and the delayed time of the train arriving at station S / delayed time of the train departing from station S, and filter the train delay data based on the Pearson correlation coefficient.

[0047] In one embodiment, since fault or delay events are essentially unstructured text data stored in the form of documents, train delay information is processed using regular expressions to extract fault feature data and perform structural unification. Based on the fault feature data, the Pearson correlation coefficient between the fault feature and the delayed arrival time / departure time of the train causing the delay at station S is determined. Train delay data is filtered based on the Pearson correlation coefficient. The Pearson correlation coefficient formula is as follows:

[0048]

[0049] Among them, ρ xy is the Pearson correlation coefficient, x i and y i are the observed values ​​of the two variables, x′ and y′ are the means of the two variables, and M is the number of observations.

[0050] Step S22: Process the filtered train delay data through feature engineering to obtain a first feature data set.

[0051] In this embodiment, in step S22, obtaining a first feature data set by performing feature engineering on the filtered train delay data specifically includes the following steps:

[0052] Step A1: Feature analysis;

[0053] The cleaned historical train operation data is analyzed to extract factors related to train delays, including six main features: vehicle failure, operation failure, line failure, power supply failure, weather factors, and passenger factors. Each main feature also includes n feature subsets, as shown in Table 1 below.

[0054] Table 1. Main features and feature subsets

[0055]

[0056] Step A2: Data dimensionality reduction;

[0057] Since there are many factors related to train delays and the amount of data is large, the K-means mean classification method can be used to extract and classify core features based on the correlation between each factor and the corresponding train delay, and obtain feature subsets of each main feature.

[0058] Step A3: feature parameterization;

[0059] After determining all the main features and feature subsets, in order to reduce the impact of dimensional differences, the data after dimensionality reduction needs to be normalized. The mapping values ​​of the feature subsets refer to the normalization principle. For the data set X = {x1, x2…x n}, x n The mapping value is:

[0060]

[0061] Taking the main feature A1 as an example, if the main feature A1 includes 3 sub-factors, it is recorded as {0, 1, 2}, and after normalization it should be {0, 0.5, 1}.

[0062] This embodiment can achieve data mining and expansion based on a small amount of data through feature engineering, extracting information useful for result prediction from historical train operation failure data, allowing the model to learn and capture the relationship between factors causing train delays, thereby improving the model's prediction accuracy.

[0063] In this embodiment, reference Figure 3 As shown, this is a diagram of the train delay time prediction process based on the ensemble learning neural network model provided in this embodiment. The convolutional neural network in step S3 can use models such as CNN and RCNN. This embodiment uses CNN as an example to illustrate the specific structure of the convolutional neural network.

[0064] CNN consists of five parts: input layer, convolution layer, pooling layer, fully connected layer and output layer: the convolution layer is used to extract feature data. The convolution layer is activated by the Relu function to improve nonlinear relationships and enhance expression capabilities; the pooling layer reduces the dimension of the convolution data to improve operating efficiency and uses maximum pooling; the fully connected layer classifies the extracted features and outputs them in combination with the activation function.

[0065] The CNN network structure determines the number of convolutional layers, the size and number of convolution kernels, the size of the maximum pooling layer, and inputs prediction parameter data for feature extraction and prediction analysis. The model optimization method is Adam, and the mean square error is selected as the loss function. The loss function is as follows:

[0066]

[0067] Where, E mse is the mean square error, n is the number of samples; y i is the true value; is the predicted value.

[0068] In this embodiment, during the prediction process, the convolutional neural network predicts the arrival / departure time of the delayed train at station S, which is used as input data for the long-short-term memory network along with the planned and actual timetable parameters. Regardless of the reason for a train's arrival or departure delay, the train's timetable includes a buffer time. Short delays can be absorbed by the preset redundancy time, but longer delays cannot be promptly recovered and will affect the arrival times of the next several stations in the train's direction. The delayed train's arrival time at the next several stations is predicted until the delay is absorbed. Therefore, the prediction process takes into account the redundant time the train spends at station S, the planned travel time between stations S-1 and S, the redundant time the train spends between stations S-1 and S, the actual time the train spends at station S, the deviation between the planned and actual arrival time at station S, the deviation between the planned and actual departure time at station S, and the actual travel time between stations S-1 and S, resulting in more accurate prediction results.

[0069] In this embodiment, since the planned operation diagram parameters and the actual operation diagram parameters are time-series variables, the RNN model has the problem of gradient vanishing in the long sequence training mode. Therefore, LSTM (Long Short-Term Memory) is used to learn time series data. In order to more comprehensively understand the performance of the prediction model in different aspects and ensure the objectivity of the evaluation system, the mean absolute error (MAE), root mean square error (RMSE), and mean percentage error (MAPE) are selected as the evaluation indicators of the prediction effect:

[0070]

[0071] Where N is the sample size, is the observed value, y i is the simulated fitted value.

[0072] The model of this embodiment calculates the error between the model fitting value and the true value, and uses the error back propagation algorithm to back propagate the error from the output layer to the input layer to update the neuron weights to reduce the model loss until the convergence condition is reached, thereby obtaining a train delay time prediction model.

[0073] In this embodiment, during the model training process, the training data is divided into a training sample data set, a validation sample data set, and a test sample data set in a ratio of 70%, 15%, and 15%; when data is scarce, smaller validation sets and test sets can be used to ensure data training.

[0074] After training is completed, the performance of the train delay prediction model is evaluated on the test set, and relevant indicators such as loss value, accuracy, F1 value, etc. are calculated.

[0075] In this embodiment, in order to improve information coordination between stations, to enable station staff to promptly understand and convey the situation to passengers when unexpected events or emergency situations cause train delays, and to comprehensively improve the station operation service level, step S4 specifically includes the following execution steps:

[0076] Step S31: When a train encounters a fault, the ATS system obtains the train delay information and timetable information of the faulty train through monitoring equipment and sends it to the central processing system. The central processing system inputs the train delay information and timetable information into the train delay time prediction model to predict the delay time of the faulty train at each station after station S.

[0077] In step S32, the central processing system pushes the predicted train delay information for each station to the operation management system, including the corresponding station integrated control room, station staff, and the maintenance personnel closest to the current fault location, so that the station staff at the corresponding station can obtain the predicted train delay time in a timely manner;

[0078] In step S33, relevant personnel use the system interface service to promptly determine the various operating indicators of the delayed train. The various operating indicators include but are not limited to the train number, the current position of the train, the reason for the delay, the time of the delay, and the delay so that the station staff can perceive the content of the delay in multiple dimensions.

[0079] Station staff obtain delay information according to the above steps and convey the delay situation to passengers, realizing the coordination of multiple information between train and station, and between stations when the train is delayed. By establishing an information sharing platform, station staff can obtain and share status related to station operations in real time, allowing station staff to understand delay information in a timely manner and adjust travel plans in time, and synchronize it with passengers, thereby improving passengers' travel experience as a whole, and improving the overall work level and ability to respond to emergencies.

[0080] Device embodiment

[0081] According to an embodiment of the present invention, a train delay time prediction device is provided. Figure 4 FIG. 1 is a block diagram of a train delay time prediction device provided in this embodiment. The train delay time prediction device according to an embodiment of the present invention includes:

[0082] The data construction module 10 is used to obtain historical train operation data and construct a delay feature dataset based on the causal event and a train operation dataset based on the train operation diagram. The historical train operation data includes train delay information based on the causal event and corresponding train operation diagram information. The train delay information includes basic information, operation information, and delay time information of the train causing the delay. The basic information includes the operation parameters of the train causing the delay. The operation information includes the location of the train accident, the location of the next station, and the human factors, machine factors, and environmental factors that cause the train delay. The delay time information includes the fault start time, fault end time, and delay time. The train operation diagram information includes the planned train operation diagram parameters and the actual train operation diagram parameters based on the causal event.

[0083] The data processing module 20 is used to pre-process the delay feature data set and the train operation data set, and perform feature engineering processing on the pre-processed train delay information in the delay feature data set to obtain a first feature data set;

[0084] The model construction and training module 30 is used to construct a train delay time prediction model composed of a convolutional neural network and a long short-term memory network, wherein the convolutional neural network is trained using the first feature data set as input and the delayed time of the delayed train arriving at station S / departing from station S as output, and the long short-term memory network is trained using the delayed time of the delayed train arriving at station S / departing from station S, the planned operation diagram parameters, and the actual operation diagram parameters as input and the delayed time of each station after the delayed train arrives at station S as output, thereby obtaining a trained train delay time prediction model;

[0085] Prediction module 40 is configured to receive train delay information and timetable information of the faulty train sent by the ATS system and input the information into the train delay prediction model. The convolutional neural network predicts the faulty train's arrival / departure delay times at station S and, combined with the timetable information, the long-short memory network predicts the faulty train's arrival delay times at each station after station S.

[0086] The information interaction module 50 is used to feed back the basic information of the faulty train, the operation information, and the delay time of each station after the faulty train arrives at station S as predicted by the prediction module to the operation management system of the corresponding station.

[0087] The train delay time prediction device provided in this embodiment has a data processing module 20 that extracts historical train operation data, analyzes the basic train information, operation information, and delay information data to determine a feature data set composed of delay influencing factors corresponding to each delayed train. The prediction module 40 learns multivariate data on train delay types and estimates delay times through a train delay time prediction model, thereby obtaining delay times that can be used to predict the arrival time of the train at the remaining stations under corresponding failure conditions. The information interaction module 50 communicates with other operation management systems such as the coordinated train ATS system and the station service individual system to explain and explain both train delays caused by station equipment failures and train delays caused during operation. This improves the station staff's ability to understand the situation in a timely manner when faced with sudden incidents or emergency events that cause train delays, thereby comprehensively improving the station's operational service level, thereby realizing a comprehensive solution based on station service individuals that can be applied to subway delays.

[0088] In this embodiment, in order to ensure data quality and accuracy, a data preprocessing module 50 is further provided for performing regularization processing on the acquired train delay information to obtain fault characteristic data, and then determining the Pearson correlation coefficient between the fault characteristic and the delayed time of the train arriving at station S / delayed time of the train departing from station S that caused the delay, and filtering the train delay data based on the Pearson correlation coefficient.

[0089] In this embodiment, a feature engineering processing module 60 is further provided for obtaining a first feature data set by performing feature engineering processing on the filtered train delay data. The feature engineering processing module 60 is configured to perform the following steps.

[0090] Step B1: Feature analysis;

[0091] The cleaned historical train operation data is analyzed to extract factors related to train delays, including six main features: vehicle failure, operation failure, line failure, power supply failure, weather factors and passenger factors. Each main feature also includes n feature subsets.

[0092] In this embodiment, the feature subsets of vehicle faults include door faults, brake system faults, etc.

[0093] Operational failures such as signal failures, ATP failures, and other feature subsets;

[0094] Line fault feature subsets such as track damage, foreign object intrusion, and turnout failure;

[0095] Power supply faults include feature subsets such as catenary faults and substation faults;

[0096] Weather factors include a subset of characteristics such as rainfall and snowfall intensity, wind speed, temperature, and visibility;

[0097] Passenger factors include feature subsets such as passenger density and passenger events.

[0098] Step B2: Data dimensionality reduction;

[0099] Since there are many factors related to train delays and the amount of data is large, the K-means mean classification method can be used to extract and classify core features based on the correlation between each factor and the corresponding train delay, and obtain feature subsets of each main feature.

[0100] Step BA3: Feature parameterization;

[0101] After determining all the main features and feature subsets, in order to reduce the impact of dimensional differences, the data after dimensionality reduction needs to be normalized. The mapping values ​​of the feature subsets refer to the normalization principle. For the data set X = {x1, x2…x n}, x n The mapping value is:

[0102]

[0103] The embodiment of the present invention is an apparatus embodiment corresponding to the above-mentioned method embodiment. The specific operations of the processing steps of each module can be understood by referring to the description of the method embodiment, and will not be repeated here.

[0104] like Figure 5 As shown, the present invention further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the train delay time prediction method in the above embodiment is implemented. Alternatively, when the computer program is executed by a processor, the train delay time prediction method in the above embodiment is implemented. When the computer program is executed by the processor, the following method steps are implemented:

[0105] Step S1, obtain historical train operation data, and construct a delay feature dataset based on the causal event and a train operation dataset based on the operation diagram, wherein the historical train operation data includes train delay information based on the causal event and the corresponding operation diagram information; wherein the train delay information includes the basic information, operation information and delay time information of the train causing the delay; the basic information is the operation parameters of the train causing the delay, the operation information includes the train accident location, the next station location and the train delay cause information human factors, machine factors and environmental factors, the delay time information includes the fault start time, fault end time and delay time; the operation diagram information includes the planned operation diagram parameters and the actual operation diagram parameters based on the causal event.

[0106] Step S2: pre-processing the delay feature data set and the train operation data set, and performing feature engineering processing on the pre-processed train delay information of the delay feature data set to obtain a first feature data set.

[0107] Step S3, constructing a train delay time prediction model consisting of a convolutional neural network and a long short-term memory network, wherein the first feature data set is used as input, and the delay time of the delayed train arriving at station S / delayed time of departure from station S is used as output to train the convolutional neural network, and the delay time of the delayed train arriving at station S / delayed time of departure from station S, the planned operation diagram parameters and the actual operation diagram parameters are used as input, and the delay time of each station after the delayed train arrives at station S is used as output to train the long short-term memory network, thereby obtaining a trained train delay time prediction model.

[0108] Step S4: Receive the train delay information and operation diagram information of the faulty train sent by the ATS system and input them into the train delay time prediction model. The convolutional neural network predicts the delay time of the faulty train arriving at station S / departing from station S, and combines the operation diagram information with the long-short memory network to predict and determine the delay time of the faulty train at each station after station S. The delay time is then fed back to the operation management system of the corresponding station together with the basic information and operation information of the faulty train.

[0109] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0110] Each embodiment in this specification is described in a progressive manner. The same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the device or system embodiments, since they are basically similar to the method embodiments, the description is relatively simple. For the relevant parts, refer to the partial description of the method embodiments. The device and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the scheme of this embodiment. A person of ordinary skill in the art can understand and implement it without making any creative efforts.

[0111] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and the contents not described in detail in the specification of the present invention are common knowledge to those skilled in the art.

Claims

1. A method for predicting train delay time, characterized in that The following steps are involved: 1) Obtain historical train operation data, and construct a delay feature dataset based on causal events and a train operation dataset based on a running diagram, wherein the historical train operation data includes train delay information based on causal events and corresponding running diagram information; wherein the train delay information includes basic information, operation information and delay time information of the train causing the delay; the basic information is the operation parameters of the train causing the delay, the operation information includes the location of the train accident, the location of the next station, and the human factors, machine factors and environmental factors of the train delay cause information, the delay time information includes the fault start time, fault end time and delay time; the running diagram information includes the planned running diagram parameters and the actual running diagram parameters based on the causal event; 2) Preprocessing the delay feature data set and the train operation data set, and performing feature engineering processing on the preprocessed delay feature data set to obtain a first feature data set; 3) Construct a train delay time prediction model composed of a convolutional neural network and a long short-term memory network, wherein the first feature data set is used as input, and the delay time of the delayed train arriving at station S / delayed time from station S is used as output to train the convolutional neural network, and the delay time of the delayed train arriving at station S / delayed time from station S, the planned operation diagram parameters and the actual operation diagram parameters are used as input, and the delay time of each station after the delayed train arrives at station S is used as output to train the long short-term memory network, thereby obtaining a trained train delay time prediction model; 4) Receive the train delay information and operation diagram information of the faulty train sent by the ATS system and input them into the train delay time prediction model. The convolutional neural network predicts the delay time of the faulty train at station S / departure time from station S, and combines the operation diagram information with the long short-term memory network to predict the delay time of the faulty train at each station after station S, and feeds it back to the operation management system of the corresponding station together with the basic information and operation information of the faulty train.

2. The train delay time prediction method according to claim 1, characterized in that: The planned operation diagram parameters include the redundant time for the train to stay at the S station, the planned operation time for the train between the S-1 station and the S station, and the redundant time for the train to run between the S-1 station and the S station; The actual operation diagram parameters include the actual stay time of the train at station S, the deviation between the planned arrival and actual arrival of the train at station S, the deviation between the planned departure and actual departure of the train at station S, and the actual running time of the train between station S-1 and station S.

3. The train delay time prediction method according to claim 1, characterized in that: The train delay information is preprocessed, and feature engineering is performed on the preprocessed train delay information, specifically including the following steps: The acquired train delay information is processed by regularization to obtain fault feature data, and then the Pearson correlation coefficient between the fault feature and the delayed time of the train arriving at station S / delayed time of the train departing from station S is determined, and the train delay data is screened according to the Pearson correlation coefficient; The filtered train delay data is processed through feature engineering to obtain the first feature data set.

4. The train delay time prediction method according to claim 3, characterized in that: The obtaining of the first feature data set by feature engineering processing specifically includes: Train delay data is processed through feature engineering to obtain main features and feature subsets, where the main features include vehicle failure, operation failure, line failure, power supply failure, weather factors and passenger factors; Operational failures include signal failures and ATP failure feature subsets; Line faults include track damage, foreign object intrusion, and turnout fault feature subsets; Power supply faults include the characteristic subsets of overhead line faults and substation faults; Weather factors include a subset of rainfall intensity, snowfall intensity, wind speed, temperature, and visibility characteristics; Passenger factors include passenger density and passenger event feature subsets.

5. The train delay time prediction method according to claim 3, characterized in that: The convolutional neural network and the long short-term memory network select mean square error as the loss function, and the long short-term memory network selects mean absolute error, root mean square error and average percentage error as evaluation indicators of the network prediction effect.

6. The train delay time prediction method according to claim 1, characterized in that: The step 4) specifically comprises the following steps: When a train encounters a fault, the ATS system obtains the train delay information and operation diagram information of the faulty train through monitoring equipment and sends it to the central processing system. The central processing system inputs the train delay information and operation diagram information into the train delay time prediction model to predict the delay time of the faulty train at each station after station S; The central processing system pushes the predicted delay information of each station to the corresponding operation management system, so that the station staff of the corresponding station can check the predicted delay time through the station staff; and Relevant personnel use the system interface service to promptly determine the various operating indicators of delayed trains.

7. A train delay time prediction device, characterized in that include: A data construction module is used to obtain historical train operation data and construct a delay feature data set based on the causal event and a train operation data set based on the operation diagram, wherein the historical train operation data includes train delay information based on the causal event and the corresponding operation diagram information; wherein the train delay information is the basic information, operation information and delay time information of the causal delay train; the basic information is the operation parameters of the causal delay train, the operation information includes the train accident location, the next station location and the train delay cause information human factors, machine factors and environmental factors, the delay time information includes the fault start time, the fault end time and the delay time; the operation diagram information includes the planned operation diagram parameters and the actual operation diagram parameters based on the causal event; a data processing module is used to preprocess the delay feature data set and the train operation data set, and perform feature engineering processing on the preprocessed delay feature data set train delay information to obtain a first feature data set; The model building and training module is used to build a train delay time prediction model composed of a convolutional neural network and a long short-term memory network, wherein the first feature data set is used as input, and the delay time of the delayed train arriving at station S / delayed time from station S is used as output to train the convolutional neural network, and the delay time of the delayed train arriving at station S / delayed time from station S, the planned operation diagram parameters and the actual operation diagram parameters are used as input, and the delay time of each station after the delayed train arrives at station S is used as output to train the long short-term memory network, thereby obtaining a trained train delay time prediction model; The prediction module is used to receive the train delay information and operation diagram information of the faulty train sent by the ATS system and input them into the train delay time prediction model. The convolutional neural network predicts the delay time of the faulty train at station S / delay time from station S, and combines the operation diagram information with the long short-term memory network to predict and determine the delay time of the faulty train at each station after station S; and the information interaction module is used to feed back the basic information of the faulty train, operation information, and the delay time of the faulty train at each station after station S obtained by the prediction module to the operation management system of the corresponding station.

8. The train delay time prediction device according to claim 7, characterized in that: It also includes data preprocessing module and feature engineering processing module; The data preprocessing module is used to perform regularization processing on the acquired train delay information to obtain fault feature data, and then determine the Pearson correlation coefficient between the fault feature and the delayed time of the train arriving at station S / delayed time of the train departing from station S, and filter the train delay data according to the Pearson correlation coefficient; The feature engineering processing module is used to obtain a first feature data set by performing feature engineering processing on the screened train delay data.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the train delay time prediction method as described in any one of claims 1 to 6 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the train delay time prediction method as described in any one of claims 1 to 6 is implemented.

Images