A virtual marshalling train tracking interval prediction and dynamic adjustment method and device

Abstract

The embodiment of the specification provides a virtual marshalling train tracking interval dynamic adjustment method and device, a vehicle-mounted controller and a train control center. The method comprises the following steps: acquiring historical data of influence factors of a virtual marshalling safety tracking interval; constructing an MLP neural network with the influence factors of the virtual marshalling safety tracking interval as input and the minimum safety interval as output; setting the influence factor data of the influence type label as input data, training the MLP neural network, and obtaining the trained MLP neural network; inputting the real-time acquired influence factor data of the virtual marshalling safety tracking interval into the trained MLP neural network to predict and obtain the minimum safety interval between the virtual marshalling trains. The present application not only guarantees the premise of safe operation of virtual marshalling, but also quickly predicts the virtual marshalling train tracking interval, efficiently and as short as possible to shorten the tracking interval and improve the train efficiency, and improve the flexibility and adaptability of operation.

Description

Technical Field

[0001] This invention relates to the field of train operation analysis and control technology, and in particular to a method, device, on-board controller and train control center for predicting and adjusting the tracking distance of virtual train formations. Background Technology

[0002] Virtual Coupling (VC) technology enables trains to operate in a coordinated manner with multiple trains at the same speed and with minimal intervals by communicating directly with each other. This is achieved through wireless communication, allowing subsequent trains to obtain the operating status of preceding trains and control their operation.

[0003] Virtual train formation technology helps signaling systems better adapt to different lines and operational needs, improving operational flexibility and adaptability. By dynamically adjusting train formations, trains can be flexibly configured according to actual conditions, enabling more efficient use of transportation resources. Traditional virtual train formation safety spacing settings are usually based on fixed rules or empirical formulas, which cannot adapt to changes in different road conditions and traffic flow. Furthermore, due to the complexity of virtual train formation systems and the safety risks associated with virtual couplers replacing physical couplers, how to efficiently improve train operation efficiency while ensuring safe operation of virtual train formations and minimizing tracking distances is a critical and challenging problem that urgently needs to be solved. Summary of the Invention

[0004] To overcome the problems existing in related technologies, this disclosure provides a method, device, on-board controller and train control center for dynamically adjusting the tracking distance of virtual train formations, so as to solve the technical problem in related technologies of how to improve train operation efficiency by efficiently and minimizing the tracking distance while ensuring the safe operation of virtual train formations.

[0005] This specification provides one or more embodiments of a method for dynamically adjusting the tracking spacing of virtual train formations, including the following steps:

[0006] Identify and acquire historical data on factors influencing the safety tracking spacing of virtual groupings;

[0007] An MLP neural network is constructed with the influencing factors of virtual group safety tracking distance as input and the minimum safety distance as output. The MLP neural network is trained with the influencing factor data with set influence type labels as input data to obtain a trained MLP neural network.

[0008] The data on factors affecting the safety tracking distance of the virtual train formation, acquired in real time, are input into a trained MLP neural network to predict and obtain the minimum safety distance between the virtual train formations.

[0009] Furthermore, the historical data for determining the influencing factors of the virtual grouping safety tracking spacing includes the following steps:

[0010] Identify the relevant factors affecting the safe distance for train tracking during train operation. These factors include those affecting the maximum deceleration capability, the process of establishing the maximum deceleration, and potential uncontrollable systemic risks.

[0011] The influencing factors of each related factor are determined based on the influence weight.

[0012] Furthermore, the influencing factors of the virtual grouping safety tracking distance and the influencing factors of each factor specifically include,

[0013] The motion status of the vehicles in front and behind, including speed, position, acceleration, load, traction / braking level, and operation plan;

[0014] Operating conditions include standard operating procedures, peak and off-peak hours, holidays, routes, passenger flow, and surge rate restrictions;

[0015] Vehicle performance parameters and health status, including traction characteristic curves, traction response time, and health of traction-related equipment that characterize the vehicle's traction capability; braking characteristic curves, braking response time, and health of braking-related equipment that characterize the vehicle's braking capability; ATP and ATO systems that characterize the onboard signaling system; and communication delay.

[0016] Vehicle operating environment conditions include weather, trackside equipment health, gradient, radius of curvature, speed limit, clearance intrusion, and track friction coefficient.

[0017] Furthermore, the construction of the MLP neural network model includes the following steps:

[0018] Construct a network model with one input layer, two hidden layers, and one output layer. Set bias vectors for the input layer and the two hidden layers.

[0019] The activation function of the model is set to the Sigmoid function;

[0020] The weight matrix and bias vector in the MLP neural network are randomly initialized, the batch gradient descent algorithm is used as the optimization algorithm for the network, and the learning rate of the optimization algorithm is initialized.

[0021] Mean squared error is set as the loss function of the model to supervise model training. The gradient of the loss function with respect to each parameter in the network is calculated by backpropagation algorithm. The network parameters are iteratively adjusted by batch gradient descent algorithm to optimize the MLP neural network.

[0022] Furthermore, it also includes steps,

[0023] Historical data on factors influencing the acquisition of virtual group safety tracking distances are preprocessed, including cleaning, deduplication, filling, and handling outliers of historical data, and then the data is normalized.

[0024] The preprocessed data is classified according to a preset ratio to obtain a training set, a validation set, and a test set. The data in the training set is labeled with influence type and used to train the MLP neural network model. The validation set is used to tune the model parameters. The test set is used to evaluate the trained model based on evaluation metrics.

[0025] This specification provides one or more embodiments of a virtual train formation tracking spacing dynamic adjustment device, including:

[0026] The data acquisition module is used to acquire historical data on factors affecting the safety tracking distance of virtual groupings;

[0027] The model building module is used to construct an MLP neural network that takes the influencing factors of the virtual grouping safety tracking distance as input and the minimum safety distance as output.

[0028] The model training module is used to train the MLP neural network model built by the model building module by setting the influencing factors of the influence type label as input, and obtain the trained MLP neural network.

[0029] The prediction module is used to input the data of the influencing factors to be predicted into the trained MLP neural network to obtain the predicted minimum safe distance of the train formation.

[0030] Furthermore, the MLP neural network is configured with an input layer, two hidden layers, and an output layer, with bias vectors set in both the input layer and the two hidden layers;

[0031] The activation function of the model is the Sigmoid function;

[0032] Initializing model parameters includes randomly initializing the weight matrix and bias vector in the MLP neural network, using batch gradient descent as the network optimization algorithm, and initializing the learning rate of the optimization algorithm.

[0033] The model's loss function is the mean squared error, and the gradient of the loss function with respect to each parameter in the network is calculated using the backpropagation algorithm. The network parameters are then iteratively adjusted using the batch gradient descent algorithm.

[0034] Furthermore, it also includes a data preprocessing module, which is used to preprocess historical data on factors affecting the acquisition of virtual grouping safety tracking spacing, including cleaning, deduplication, filling and processing of outliers, and standardizing the data.

[0035] This specification provides one or more embodiments of an on-board controller, including 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 virtual train formation tracking spacing dynamic adjustment method as described in any of the preceding embodiments. The on-board controller controls and adjusts the train formation spacing with the preceding train based on the predicted minimum safe distance.

[0036] This specification provides one or more embodiments of a train control center that is communicatively connected to the on-board controller. The center includes 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 virtual train tracking spacing dynamic adjustment method as described in any of the above embodiments, obtains prediction results, and sends them to the on-board controller.

[0037] The onboard controller adjusts the train spacing with the preceding train based on the predicted minimum safe distance.

[0038] The method, device, on-board controller, and train control center for dynamically adjusting the tracking distance of virtual train formations disclosed herein have the following advantages: They analyze the factors affecting the braking capacity of trains during the braking process, determine the factors affecting the safe tracking distance of virtual train formations during train operation, and, based on the constructed data-driven MLP model that predicts the minimum safe tracking distance of virtual train formations, continuously optimize the model through historical data and apply it to guide the dynamic adjustment of the tracking distance of virtual train formations. This not only ensures the safe operation of virtual train formations but also quickly predicts the tracking distance of virtual train formations, efficiently and minimizes the tracking distance to improve train operation efficiency. Furthermore, due to the wide range of application scenarios of neural network prediction, it can help the signaling system better adapt to different lines and operational requirements, improving operational flexibility and adaptability. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in one or more embodiments of this specification or in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 A flowchart illustrating a method for dynamically adjusting the tracking spacing of virtual train formations, provided for one or more embodiments of this specification;

[0041] Figure 2 A graph showing the three stages of a train emergency braking process as provided in one or more embodiments of this specification;

[0042] Figure 3 A diagram showing the relationship between influencing factors and influencing factors of safe tracking distance for virtual train formations provided in one or more embodiments of this specification;

[0043] Figure 4 MLP neural network structure diagram provided for one or more embodiments of this specification;

[0044] Figure 5 This is a schematic diagram of a virtual train formation tracking spacing dynamic adjustment device provided for one or more embodiments of this specification. Detailed Implementation

[0045] To enable those skilled in the art to 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 with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of this invention.

[0046] The present invention will now be described in detail with reference to specific embodiments and accompanying drawings.

[0047] Method Implementation Examples

[0048] According to embodiments of the present invention, a method for dynamically adjusting the tracking spacing of virtual train formations is provided, such as... Figure 1 The diagram shown is a flowchart of the virtual train tracking spacing dynamic adjustment method provided in this embodiment. The virtual train tracking spacing prediction and dynamic adjustment method according to this embodiment includes:

[0049] Step S1: Determine and obtain historical data on the factors influencing the virtual grouping safety tracking distance.

[0050] In this embodiment, by analyzing historical data, we can understand the operation and performance of the virtual grouping system under different conditions, provide a basis for future predictions, integrate data from various aspects, and select features related to the virtual grouping operation status.

[0051] This embodiment lists some factors that affect the safe tracking distance of virtual train formations based on historical experience. In the operation of the entire rail transit signaling system, there are a large number of factors that can affect the signaling equipment itself and the external environment. Therefore, calculating the minimum safe tracking distance of virtual train formations based on data-driven calculation requires a large amount of historical data on the operation of trains on the target line.

[0052] In this embodiment, there are many factors that affect the safe tracking distance of virtual grouping. For example, rainy or snowy weather or the health of equipment on and off the vehicle may cause changes in the vehicle's braking ability, or lead to higher system latency, or reduce the robustness of the system, which may further increase the vehicle's braking distance and bring about collision risks, etc.

[0053] In this embodiment, to further determine the influence of various factors on the minimum safe following distance of virtual train formations, a step is also included to analyze and determine the influence relationship between each influencing factor and the minimum safe distance, specifically:

[0054] S11, identify the relevant factors affecting the safe distance for train tracking during train operation, including factors affecting the maximum deceleration capability, the process of establishing the maximum deceleration, and potential uncontrollable systemic risks, as detailed below.

[0055] First, according to the kinematic formula, braking distance s = v 2 As shown in / 2a, the train's operating speed and acceleration capability are crucial factors in calculating the safe following distance for virtual train formations. The real-time speed and acceleration of a moving train can be measured using devices such as speedometers, accelerometers, and GPS. In actual train operation, the establishment of emergency braking is a slow process. Generally, the process of calculating the safe following distance is divided into three stages: traction cut-off, braking establishment, and maximum capacity braking. Figure 2 The figure shows the curves of the three stages of the emergency braking process of the train provided in this embodiment. At time t0, emergency braking is triggered; at time t1, traction is cut off; at time t2, braking is established; and at time t3, the train stops. L is the target minimum safe following distance to be calculated. Therefore, when calculating the minimum safe following distance, the factors that have the greatest impact on it are the maximum deceleration of the train and the process of establishing the maximum deceleration. Among the factors affecting the safe following distance of the virtual train formation, most factors directly or indirectly affect the maximum deceleration of the train and the process of establishing the maximum deceleration. Some factors may cause systemic risks due to equipment failure. Therefore, based on the influence relationship of these factors on the maximum deceleration and the process of establishing the maximum deceleration, they are re-correlated and classified into three categories of related factors: factors affecting the maximum deceleration capability, factors affecting the maximum deceleration establishment process, and factors that may bring uncontrollable systemic risks. Some factors may affect both the maximum deceleration capability and the process of establishing the maximum deceleration simultaneously.

[0056] S12, determine the influence factors and their influence relationships with each related factor based on their influence weights.

[0057] In this embodiment, to ensure both the accuracy and efficiency of the model's predictions, when determining the influencing factors that determine the relationships between related factors, we only consider the factors that have a significant impact on the related factors. Therefore, the influencing factors required for prediction in this embodiment can be determined through their influence weights, as detailed below:

[0058] Factors affecting maximum deceleration capacity include: load, gradient, radius of curvature, weather, track friction coefficient, vehicle clearance, and trackside equipment.

[0059] Factors affecting the establishment of maximum deceleration include: load, acceleration, traction / braking level, traction / braking characteristic curve, response delay of various equipment (operating cycle of signal equipment, communication delay, etc.), gradient, radius of curvature, weather, impact rate limit, etc.

[0060] Factors that contribute to uncontrollable systemic risks include: weather, the health of signal system equipment, and passenger flow.

[0061] The above lists some of the factors affecting the safe tracking distance of virtual train formations and their relationships with other influencing factors. In the operation of the entire rail transit signaling system, there are many other influencing factors, both from the signaling equipment itself and from external sources. Therefore, calculating the minimum safe tracking distance of virtual train formations based on data-driven methods requires a large amount of historical data on the target line's train operation. This data should include not only the data records related to the listed influencing factors but also the operating data of various equipment that may affect the train's traction and braking capabilities. Only in this way can the weight of each factor's influence on the calculation target be determined, and data overlooked by past experience be deeply explored.

[0062] Therefore, based on the magnitude of the influence weights of each influencing factor, this embodiment preferably categorizes the influencing factors of virtual grouping safety tracking distance into four main types, as referenced. Figure 3 As shown, the factors include four categories: the motion state of the vehicles ahead and behind, operating conditions, vehicle performance parameters and health status, and environmental conditions of vehicle operation. Each category contains multiple influencing factors. Because virtual train formation is based on a general signaling system, these four categories basically cover all aspects of the influencing factors of rail transit operation. Each category can include the following influencing factors:

[0063] ① The motion status of the vehicles in front and behind includes speed, position, acceleration, load, traction / braking level, and operation plan.

[0064] ②Operating conditions include standard operating procedures, peak and off-peak hours, holidays, routes, passenger flow, and impact rate restrictions.

[0065] ③ Vehicle performance parameters and health status include traction characteristic curves, traction response time, and health of traction-related equipment that characterize the vehicle's traction capability; braking characteristic curves, braking response time, and health of braking-related equipment that characterize the vehicle's braking capability; ATP (Automatic Train Protection) and ATO (Automatic Train Operation) of the onboard signaling system; and communication delays, etc.

[0066] ④ Vehicle operating environment conditions include weather (rain, snow, strong wind, fog, temperature and humidity), trackside equipment health (turnouts, signals, transponders), gradient, radius of curvature, speed limit, clearance intrusion, track friction coefficient, etc.

[0067] This embodiment should include not only the data records related to the listed influencing factors, but also the operating data of various equipment that may affect the traction and braking capabilities of the train, in order to determine the influence weight of each factor on the calculation target and to deeply mine data that has been overlooked in previous experience.

[0068] In this embodiment, due to the complexity and diversity of the acquired historical data, the collected influencing factor data may contain missing data, duplicate data, etc. If the acquired data is used directly for model training, it will produce a large training error. Therefore, this embodiment cleans and processes the data, including removing duplicate data, filling missing values, handling outliers, etc., and normalizes the data, selecting features in the data that are related to the virtual grouping safety tracking distance, so that the data set meets the requirements of modeling.

[0069] Step S2: Construct an MLP neural network with the influencing factors affecting the safe tracking distance of virtual grouping as input and the minimum safe distance as output. Use the influencing factor data with set influence type labels as input data to train the model and obtain the trained MLP neural network.

[0070] The MLP neural network model possesses nonlinear mapping and global optimization capabilities, generalizing from n-dimensional space to m-dimensional space, and is commonly used to solve classification and prediction problems. Furthermore, the MLP neural network is a generalization based on the single-layer perceptron, solving the problem that the single-layer perceptron cannot identify linearly inseparable data. It can fit any nonlinear functional relationship and has strong adaptive, self-learning, and fault-tolerant capabilities. This embodiment constructs an MLP neural network model to fit the nonlinear functional relationship between the factors affecting the safe tracking distance of virtual train formations and the minimum safe distance between two trains. The MLP neural network model can take multi-dimensional influencing factor data as input and map it to one-dimensional data as the minimum safe distance between two trains.

[0071] refer to Figure 4The diagram shown is a schematic of the MLP neural network structure provided in this embodiment. The MLP neural network model includes an input layer, two hidden layers and an output layer. Each circle represents a node of the MLP neural network. The left X1, X2, X3 and X4 represent the input feature vectors, the "+1" node is the bias vector of the neural network, and the rightmost node is the output node of the model. The calculation formula for each node is y = f(W*X + b), where W represents the weight matrix, b represents the bias vector, f(x) represents the activation function, and y represents the output value of the current node.

[0072] In this embodiment, it is also necessary to define the activation function of the MLP neural network model, initialize the parameters, and set the model training convergence conditions.

[0073] Using the Sigmoid function as the activation function and mapping the output to the range of 0-1 increases the model's expressive power and helps it better fit real-world data. The expression for the activation function is:

[0074] f(x) = 1 / (1+e -x ).

[0075] The model initialization process includes randomly initializing the weight matrix and bias vector in the MLP neural network, using batch gradient descent as the network optimization algorithm, and initializing the learning rate of the optimization algorithm.

[0076] In this embodiment, the MLP neural network model training includes backpropagation. During training, the model output value is compared with the actual value using mean squared error as the loss function, and the loss is calculated. Then, the gradient of the loss function with respect to each parameter in the network is calculated using the backpropagation algorithm. The network parameters are iteratively adjusted using the batch gradient descent algorithm to minimize the loss function, and finally, the optimization of the MLP neural network is achieved.

[0077] In this embodiment, the preprocessed data is classified according to a preset ratio to obtain a training set, a validation set, and a test set. The data in the training set has labels set with influence types, while the data in the validation and test sets are unlabeled. The training set samples are used as input to train the MLP neural network model. The validation set samples are used to evaluate the accuracy of the MLP neural network model and to tune the model's parameters, such as adjusting the learning rate and regularization parameters, to achieve better performance. The trained model is evaluated using the test set with accuracy, mean squared error, and model efficiency as evaluation metrics, thereby monitoring errors during the training process of the MLP neural network model.

[0078] Step S3: Input the data on the influencing factors of the virtual train formation safety tracking distance acquired in real time into the trained MLP neural network to predict and obtain the minimum safety distance between virtual train formations.

[0079] In this embodiment, the trained MLP neural network model is deployed into the train operation environment to collect on-site data of virtual train formation operation, predict the minimum safe distance of formation under the influence of multiple factors, and, taking into account the influence of operation routes and operation plans, dynamically adjust the spacing of virtual train formation according to on-site operation conditions and optimization objectives to adapt to real-time and efficient train operation requirements.

[0080] This embodiment analyzes the factors affecting train braking capability during the train braking process, determines the factors affecting the safe tracking distance of virtual train formations during train operation, and builds a data-driven MLP model that predicts the minimum safe tracking distance of virtual train formations. The model is continuously optimized through historical data and applied to guide the dynamic adjustment of the tracking distance of virtual train formations. This not only ensures the safe operation of virtual train formations but also quickly predicts the tracking distance of virtual train formations, efficiently and minimizes the tracking distance to improve train operation efficiency. Furthermore, due to the wide range of application scenarios of neural network prediction, it can help the signaling system better adapt to different lines and operational requirements, improving operational flexibility and adaptability.

[0081] Device Examples

[0082] According to an embodiment of the present invention, a virtual train formation tracking spacing dynamic adjustment device is provided. This device can be equipped on the onboard controllers of each train or on a train control center communicatively connected to the onboard controllers of each train, such as... Figure 5 The diagram shown is a block diagram of the virtual train tracking spacing dynamic adjustment device provided in this embodiment. According to an embodiment of the present invention, the virtual train tracking spacing dynamic adjustment device includes:

[0083] The data acquisition module 10 is used to acquire historical data on factors affecting the safety tracking distance of virtual grouping.

[0084] In this embodiment, the factors affecting the safe tracking distance of virtual train formations can be broadly categorized into four types: the motion state of vehicles before and after a specific location, operating conditions, vehicle performance parameters and health status, and environmental conditions of the vehicle's operating environment. Each category contains multiple influencing factors. Because virtual train formations are based on a general signaling system, these four categories basically cover all aspects of the influencing factors in rail transit operations. Each category can include the following influencing factors:

[0085] ① The motion status of the vehicles in front and behind includes speed, position, acceleration, load, traction / braking level, and operation plan.

[0086] ②Operating conditions include standard operating procedures, peak and off-peak hours, holidays, routes, passenger flow, and impact rate restrictions.

[0087] ③ Vehicle performance parameters and health status include traction characteristic curves, traction response time, and health of traction-related equipment that characterize the vehicle's traction capability; braking characteristic curves, braking response time, and health of braking-related equipment that characterize the vehicle's braking capability; ATP (Automatic Train Protection) and ATO (Automatic Train Operation) of the onboard signaling system; and communication delays, etc.

[0088] ④ Vehicle operating environment conditions include weather (rain, snow, strong wind, fog, temperature and humidity), trackside equipment health (turnouts, signals, transponders), gradient, radius of curvature, speed limit, clearance intrusion, track friction coefficient, etc.

[0089] In this embodiment, since the acquired historical data is complex and diverse, the collected influencing factor data may be missing or duplicated. If the acquired data is used directly for model training, it will produce a large training error. Therefore, this embodiment also sets up a data preprocessing module 101 to preprocess the data acquired by the data acquisition module 10.

[0090] Data preprocessing includes cleaning and processing the data, deduplicating data, filling missing values, handling outliers, and normalizing the data. It also involves selecting features in the data that are related to the virtual grouping safety tracking distance so that the dataset meets the requirements of modeling.

[0091] Model building module 20 is used to construct an MLP neural network with the influencing factors of the virtual grouping safety tracking distance as input and the minimum safety distance as output.

[0092] In this embodiment, the MLP neural network model includes an input layer, two hidden layers, and an output layer. The model contains nodes X1, X2, X3, and X4 on the left, representing input feature vector nodes, and nodes representing the bias vector "+1" of the neural network. The rightmost node is the output node of the model. The calculation formula for each node is y = f(W*X + b), where W represents the weight matrix, b represents the bias vector, f(x) represents the activation function, and y represents the output value of the current node.

[0093] In this embodiment, the Sigmoid function is set as the activation function of the model, and the output is mapped to the range of 0-1. This is used to increase the expressive power of the model and better fit the actual data. The expression of the activation function is:

[0094] f(x) = 1 / (1+e -x ).

[0095] The model initialization process includes randomly initializing the weight matrix and bias vector in the MLP neural network, using batch gradient descent as the network optimization algorithm, and initializing the learning rate of the optimization algorithm.

[0096] In this embodiment, the MLP neural network model training includes backpropagation. During training, the model output value is compared with the actual value using mean squared error as the loss function, and the loss is calculated. Then, the gradient of the loss function with respect to each parameter in the network is calculated using the backpropagation algorithm. The network parameters are iteratively adjusted using the batch gradient descent algorithm to minimize the loss function, and finally, the optimization of the MLP neural network is achieved.

[0097] The model training module 30 is used to train the MLP neural network model built by the model building module 20 by setting the influencing factors with influence type labels as input, and obtain the trained MLP neural network.

[0098] In this embodiment, according to the model training requirements, the preprocessed data is classified according to a preset ratio to obtain a training set, a validation set, and a test set. The data in the training set contains data with labeled influence types, while the data in the validation and test sets are unlabeled. The training set samples are used as input to train the MLP neural network model. The validation set samples are used to evaluate the accuracy of the MLP neural network model and to tune the model's parameters, such as adjusting the learning rate and regularization parameters, to achieve better performance. Accuracy, mean squared error, and model efficiency are used as evaluation metrics. The test set is used to evaluate the trained model, thereby monitoring errors during the MLP neural network model training process.

[0099] The prediction module 40 is used to input the data of the influencing factors to be predicted into the trained MLP neural network to obtain the predicted minimum safe distance between the trains. The on-board controller controls and adjusts the train spacing with the preceding train according to the predicted minimum safe distance and the current operation plan of the vehicle.

[0100] This embodiment of the device analyzes the factors affecting the braking capability of a train during the braking process, determines the factors affecting the safe tracking distance of virtual train formations during train operation, and, based on the constructed data-driven MLP model that predicts the minimum safe tracking distance of virtual train formations, continuously optimizes the model through historical data and applies it to guide the dynamic adjustment of the tracking distance of virtual train formations. This not only ensures the safe operation of virtual train formations but also quickly predicts the tracking distance of virtual train formations. Due to the wide range of application scenarios of neural network prediction, it can help the signaling system better adapt to different lines and operational requirements, improving operational flexibility and adaptability.

[0101] The embodiments of the present invention are device embodiments corresponding to the above method embodiments. The specific operations of each module processing step can be understood with reference to the description of the method embodiments, and will not be repeated here.

[0102] The present invention also provides an on-board controller, on which a computer program is stored. When the computer program is executed by a processor, it implements the method for dynamically adjusting the tracking distance of virtual train formations in the above embodiments, and the on-board controller controls and adjusts the train formation distance with the preceding train according to the predicted minimum safe distance.

[0103] The present invention also provides a train control center, which is communicatively connected to the on-board controller and stores a computer program thereon. When the computer program is executed by the processor, it implements the method for dynamically adjusting the tracking spacing of virtual train formations in the above embodiments, obtains prediction results, and sends them to the on-board controller.

[0104] The onboard controller adjusts the train spacing with the preceding train based on the predicted minimum safe distance.

[0105] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, 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), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0106] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for apparatus or system embodiments, since they are basically similar to method embodiments, the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. The apparatus and system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0107] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate 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 well known to those skilled in the art.

Claims

1. A method for dynamically adjusting the tracking spacing of virtual train formations, characterized in that, Including the following steps: Identify and acquire historical data on factors influencing the safety tracking spacing of virtual groupings; An MLP neural network is constructed with the influencing factors of virtual group safety tracking distance as input and the minimum safety distance as output. The MLP neural network is trained with the influencing factor data with set influence type labels as input data to obtain a trained MLP neural network. The real-time acquired data on factors influencing the safe tracking distance of virtual train formations is input into a trained MLP neural network to predict and obtain the minimum safe tracking distance between virtual train formations; the historical data for determining the influencing factors of the safe tracking distance of virtual train formations includes the following steps. Identify the relevant factors affecting the safe distance for train tracking during train operation. These factors include those affecting the maximum deceleration capability, the process of establishing the maximum deceleration, and potential uncontrollable systemic risks. The influencing factors of each related factor are determined based on the influence weight; the influencing factors of the virtual grouping safe tracking distance and the influencing factors of each influencing factor specifically include: The motion status of the vehicles in front and behind, including speed, position, acceleration, load, traction / braking level, and operation plan; Operating conditions include standard operating procedures, peak and off-peak hours, holidays, routes, passenger flow, and surge rate restrictions; Vehicle performance parameters and health status, including traction characteristic curves, traction response time, and health of traction-related equipment that characterize the vehicle's traction capability; braking characteristic curves, braking response time, and health of braking-related equipment that characterize the vehicle's braking capability; ATP and ATO systems that characterize the onboard signaling system; and communication delay. Vehicle operating environment conditions include weather, trackside equipment health, gradient, radius of curvature, speed limit, clearance intrusion, and track friction coefficient.

2. The method for dynamically adjusting the tracking spacing of virtual train formations as described in claim 1, characterized in that, The construction of the MLP neural network model includes the following steps: Construct a network model with one input layer, two hidden layers, and one output layer. Set bias vectors for the input layer and the two hidden layers. The activation function of the model is set to the Sigmoid function; The weight matrix and bias vector in the MLP neural network are randomly initialized, the batch gradient descent algorithm is used as the optimization algorithm for the network, and the learning rate of the optimization algorithm is initialized. Mean squared error is set as the loss function of the model to supervise model training. The gradient of the loss function with respect to each parameter in the network is calculated by backpropagation algorithm. The network parameters are iteratively adjusted by batch gradient descent algorithm to optimize the MLP neural network.

3. The method for dynamically adjusting the tracking spacing of virtual train formations as described in claim 1, characterized in that, It also includes steps, Historical data on factors influencing the acquisition of virtual group safety tracking distances are preprocessed, including cleaning, deduplication, filling, and handling outliers of historical data, and then the data is normalized. The preprocessed data is classified according to a preset ratio to obtain a training set, a validation set, and a test set. The data in the training set is labeled with influence type and used to train the MLP neural network model. The validation set is used to tune the model parameters. The test set is used to evaluate the trained model based on evaluation metrics.

4. A device for dynamically adjusting the tracking spacing of virtual train formations, characterized in that, include: The data acquisition module is used to acquire historical data on factors affecting the safe tracking distance of virtual train formations; specifically, it includes identifying various related factors that affect the safe tracking distance of trains during operation, including factors affecting the maximum deceleration capability, the process of establishing the maximum deceleration, and potential uncontrollable systemic risks. The influencing factors of each related factor are determined based on their influence weights. The model building module is used to construct an MLP neural network that takes the influencing factors of the virtual grouping safety tracking distance as input and the minimum safety distance as output. The model training module is used to train the MLP neural network model built by the model building module by setting the influencing factors of the influence type label as input, and obtain the trained MLP neural network. The prediction module is used to input the data of the influencing factors to be predicted into the trained MLP neural network to obtain the predicted minimum safe distance of the train formation. The influencing factors of the virtual grouping safe tracking distance and the influencing factors of each factor specifically include: The motion status of the vehicles in front and behind, including speed, position, acceleration, load, traction / braking level, and operation plan; Operating conditions include standard operating procedures, peak and off-peak hours, holidays, routes, passenger flow, and surge rate restrictions; Vehicle performance parameters and health status, including traction characteristic curves, traction response time, and health of traction-related equipment that characterize the vehicle's traction capability; braking characteristic curves, braking response time, and health of braking-related equipment that characterize the vehicle's braking capability; ATP and ATO systems that characterize the onboard signaling system; and communication delay. Vehicle operating environment conditions include weather, trackside equipment health, gradient, radius of curvature, speed limit, clearance intrusion, and track friction coefficient.

5. The virtual train formation tracking spacing dynamic adjustment device as described in claim 4, characterized in that, The MLP neural network is configured with one input layer, two hidden layers, and one output layer. The input layer and the two hidden layers are configured with bias vectors. The activation function of the model is the Sigmoid function; Initializing model parameters includes randomly initializing the weight matrix and bias vector in the MLP neural network, using batch gradient descent as the network optimization algorithm, and initializing the learning rate of the optimization algorithm. The model's loss function is the mean squared error, and the gradient of the loss function with respect to each parameter in the network is calculated using the backpropagation algorithm. The network parameters are then iteratively adjusted using the batch gradient descent algorithm.

6. The virtual train formation tracking spacing dynamic adjustment device as described in claim 4, characterized in that, It also includes a data preprocessing module, which is used to preprocess historical data on factors affecting the acquisition of virtual grouping safety tracking distance, including cleaning, deduplication, filling and processing of outliers of historical data, and normalizing the data.

7. An on-board controller, 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, it implements the method for dynamically adjusting the tracking distance of virtual train formations as described in any one of claims 1 to 3, wherein the on-board controller controls and adjusts the train formation distance with the preceding train based on the predicted minimum safe distance.

8. A train control center, communicatively connected to an onboard controller, 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, it implements the virtual train tracking spacing dynamic adjustment method as described in any one of claims 1 to 3, obtains the prediction result, and sends it to the on-board controller. The onboard controller adjusts the train spacing with the preceding train based on the predicted minimum safe distance.

Images