Freight railway dispatching and train control coordination method and system

Abstract

Provided in the present invention is a freight railway dispatching and train control coordination method. The method includes the following steps: performing feature extraction on basic data, dispatching command data and train operation data of a freight railway line, and constructing a system model; acquiring various types of data in a train control production system, simulating a train operation process and a dispatching command policy in the constructed system model in view of an actual transportation production plan, and deducing, in real time, optimization solution policies of the system model in set scenarios; evaluating the optimization solution policies; and applying to the train control production system the optimization solution policy meeting requirements in the evaluation, so as to guide the actual transportation production. The present invention breaks down an information barrier between a dispatching system and a train control system, thereby effectively improving the transportation efficiency of freight railways.

Description

A method and system for coordinating freight railway dispatching and train control Technical Field

[0001] The present invention relates to the field of rail transit communications, and in particular to a method and system for coordinating freight railway dispatching and train control. Background Art

[0002] The daily operations of freight railways can be divided into three levels: planning, dispatching, and control. The planning system is responsible for compiling a planned operation diagram for a specific period and developing operational plans for train operations, locomotive operation, construction, and other operations. The dispatching system coordinates, configures, and operates stations, sections, trains, and locomotives within the system based on the planned operation diagram. The control system is responsible for managing routes, opening signals, and controlling train operations. The dispatching system and the train control system require real-time information exchange. Dispatchers send operation plans, dispatching instructions, train route information, temporary speed limit commands, and other information to the train control system through the dispatching system. The train control system then reports routes, signals, section status, train location, speed, and other information to the dispatching system, working together to complete transportation tasks.

[0003] The actual network train operation environment is complex and changeable. In addition to the general characteristics of general systems, it also has the following special features: (1) High complexity. The operating environment of freight railway trains is complex, and there are many conditions and internal interference factors that affect train operation. Dispatching command and train operation rely on the dispatcher and driver's control of environmental information and train performance. The handling and response capabilities of emergencies vary from person to person, resulting in poor stability and uniformity of train operation. After the operation control system developed from fixed block to mobile block, the train operation density is high, the interval is small, and the operation coupling between trains is strong. Poor train control effect will have a great impact on the efficient and punctual operation of the system. For example, it will cause unplanned stops of freight trains, which will not only increase the additional time for stopping, but also increase the energy consumption of locomotives. (2) High dynamics. Under the influence of internal and external factors, the operation process of freight trains often deviates from the existing plan, and the dispatching and train operation status are always in dynamic change. At present, dispatchers are unable to grasp fine-grained train status information (such as speed, position, train model, traction and braking characteristics), and the stage adjustment plans formulated based on this are not precise enough. At the same time, dispatchers have limited access to the internal status of the train control system, such as train communication failure, train degradation, etc., and cannot make effective decisions on the treatment methods in the first time. This is not conducive to the information coordination between dispatchers and drivers in different layers of the vehicle, line, and network, and reduces the speed of response to emergencies. (3) High coupling. The freight railway network is complex, with many interference factors. The train operation is highly coupled between vehicles, lines, and stations. The propagation process of train delays in the network is complex, making it difficult for dispatchers to make adjustments and decisions.

[0004] Passing trains in opposite directions is a crucial task for freight railway dispatching and train control systems. Only after the oncoming passing train arrives at the station and joins the track and stops can passing trains pass through the passing station. To improve throughput, the passing process is typically very compact.

[0005] Figure 2 shows a diagram of oncoming trains passing each other. In actual transport planning, this process requires coordination between personnel from multiple positions and the dispatching and train control systems. Dispatching personnel create a passing plan using the CTC (Centralized Control System) and issue it to the station's trackside train control system for routing. The train control system calculates a driving authorization for the train, and the onboard system calculates a speed profile based on the authorization and line data. The plan is also communicated to the driver, who then completes the passing operation under the protection of the onboard system. Dispatchers often need to repeatedly adjust the plan based on the train's progress. The train control system must repeatedly process routing and calculate authorizations, and the driver must adjust his driving strategy based on the plan and signals from ahead. Poor coordination can lead to problems such as unnecessary speed reduction or external stops on passing freight trains.

[0006] Therefore, it is urgent to study the coordinated optimization methods and technologies of dispatching command and train operation control, and form an efficient, safe and brand-new railway dispatching command and train operation control coordinated system. Currently, there is little research on the coordinated optimization of dispatching and train operation on freight railways. Technical issues

[0007] The present invention proposes a framework for coordinating the optimization of dispatching and train control systems in complex freight railway networks and time-varying environments, solving the modeling, decision-making, and optimization control issues of this problem. An improved train target speed curve optimization method using a dual-depth Q network is designed to solve the model. With the help of computational experiments, various operational scenarios and emergencies are "deduced and predicted" as a reference for actual dispatching command and train driving. By analyzing and utilizing the computational simulation results, coordinating the optimization of management and control is achieved. Technical Solutions

[0008] A method for coordinating freight railway dispatching and train control, comprising the following steps:

[0009] S1. Extract features of freight railway lines, including basic data, dispatching and command data, and train operation data, and build a system model;

[0010] S2. Acquire various data from the train control production system, combine it with the actual transportation production plan, simulate the train operation process and dispatching command strategy in the constructed system model, and deduce the optimization solution strategy of the system model in various set scenarios in real time;

[0011] S3. Evaluate the optimization solution strategy;

[0012] S4. Apply the optimization solution strategy that meets the requirements in the evaluation to the train control production system to guide actual transportation production.

[0013] Furthermore, the system model of step S1 includes: a basic data model, a scheduling system model, a train control system model and a train driving model.

[0014] Furthermore, the various types of data in step S2 include various emergencies during the operation of the train.

[0015] Furthermore, the optimization solution strategy in step S2 is a train speed curve and a scheduling adjustment plan suitable for a set scenario.

[0016] Furthermore, step S2 includes setting the operating environment of the passing train and the passing train using a Q network algorithm, simulating the passing train and the passing train process, and establishing a Markov decision process;

[0017] The operating environment includes line static data, the operating plans of passing trains and passing trains, the operating position, speed and running time of passing trains, and the position, speed and running time of passing trains.

[0018] Furthermore, the letting operation process is represented as the letting train running from station k-1 to station k and the passing train running from station k+1 to station k.

[0019] Furthermore, the interval between station k-1 and station k, and the interval between station k+1 and station k are divided into different sub-intervals according to the fixed speed limit, and the maximum traction speed and maximum braking speed that can allow the train or the passing train to pass are calculated from each sub-interval;

[0020] Based on the speed of the train or passing train in its current state, the running speed of the next state and the running speed of the previous state are calculated. Under the premise of not exceeding the speed limit of the corresponding sub-interval, the maximum traction speed is obtained according to the running speed of the next state, and the maximum braking speed is obtained according to the running speed of the previous state. The smaller value of the maximum traction speed and the maximum braking speed is taken as the speed value of the shortest running time.

[0021] Furthermore, the time it takes for the train to arrive at station k will be: , To make the train depart at station k-1, To make the train arrive at station k, is the length of the subinterval that the train will pass through, w is the number of subintervals that the train will pass through, The speed value that will allow the train to run in the shortest time;

[0022] The time it takes for the train to reach station k is: , is the departure time of the passing train at station k+1, is the arrival time of the passing train at station k, is the length of the subinterval through which the train passes, z is the number of subintervals passed by the train, It is the speed value of the shortest running time of a passing train.

[0023] Furthermore, the objective function of optimization is to always minimize the time: ,and , The time it takes to clear routes and open signals for passing trains.

[0024] Furthermore, if it is necessary to make the process run smoothly and reduce sudden acceleration and deceleration, the optimization objective function is to minimize the acceleration rate of change: ,in and are the length of the train subsection and the length of the passing train subsection, is the acceleration that will bring the train to the end of the sub-section, is the acceleration that will cause the train to reach the starting point of the sub-section, is the acceleration of the passing train at the end of the sub-section, is the acceleration of the train at the starting point of the sub-interval, It is the time for the train to run in the sub-section. It is the running time of the train in the sub-section.

[0025] Furthermore, the step S2 uses the Q network algorithm to optimize the train operation process, which includes the following steps:

[0026] S21. Initialize the approximation network parameters , let the target network parameters for ;

[0027] S22, will make the train running Depend on Sure, is the subinterval length, To speed up the train, is the distance from the passing train to the stopping point, is the speed of the passing train, is the weighted value of the objective function, The probability of randomly selecting the working condition action , with 1- The probability of choosing the current optimal state ;

[0028] S23, execute working condition action , which will allow the train operating environment to get the next state and reward value ;

[0029] S24, randomly select several from the experience playback area , update the network parameters according to the gradient descent algorithm ;

[0030] S25, update network parameters after the specified number of steps for , exit when the number of iterations is reached, otherwise return to S21.

[0031] The present invention also proposes a freight railway dispatching and train control coordination system for implementing the dispatching and train control coordination method, comprising the following steps:

[0032] The perception learning module is used to obtain basic data on freight railway lines, dispatching and command data, and train operation data features for extraction, and to build a computable and reconfigurable system model;

[0033] The collaborative optimization module is used to obtain various data from the production system, combine it with the actual transportation production plan, simulate the train operation process and dispatching command strategy in the production environment, and deduce the optimization solution strategy of the system model in real time;

[0034] Strategy evaluation module, which tests and evaluates the optimization solution strategy generated by the collaborative optimization module;

[0035] If the strategy evaluation module pushes the optimization solution strategy that meets the predefined goals to the production system for execution.

[0036] Furthermore, the system model includes: a basic data model, a scheduling system model, a train control system model and a train driving model.

[0037] Furthermore, the various types of data come from various types of random emergencies.

[0038] Furthermore, the optimization solution strategy includes: train speed curve and scheduling adjustment plan. Beneficial effects

[0039] Break down the information barriers between the dispatching system and the train control system, and effectively improve the efficiency of freight railway transportation.

[0040] A freight railway train passing model was developed to describe the operation and planned arrival times of passing and passing trains in detail. This model accurately describes the coordination between passing and passing trains as they arrive at passing stations, laying the foundation for optimizing passing operations.

[0041] Considering the speed and position of the opposite train, the calculation method of the double-depth Q network is improved. The advantage is that when calculating the train curve and output working conditions, the speed of the passing train and the distance to the passing point are considered. From the perspective of scheduling, the operation process of the two trains can be coordinated and optimized at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] FIG1 is a scheduling and train control collaborative optimization framework of the present invention;

[0043] Figure 2 is a schematic diagram of a passing train and a giving way train;

[0044] Figure 3 shows the train scheduling and control process;

[0045] Figure 4 shows the process of reinforcement learning optimization by the train;

[0046] FIG5 is a schematic diagram of a speed curve taking into account the position and speed of the train. Best Mode for Carrying Out the Invention

[0047] The scheduling and train control coordination method and system proposed by the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. The advantages and features of the present invention will become more apparent from the following description.

[0048] The present invention provides a coordinated freight railway dispatching and train control system consisting of a perception learning module, a collaborative optimization module, a strategy evaluation module, and a strategy generation module. Its core is to utilize perception learning and artificial intelligence technologies to break down the information barriers between the dispatching and train control systems, achieving coordinated optimization of dispatching and train operation control. The generated strategies are then fed back to the production dispatching and train control systems in real time and dynamically, guiding the train control systems to ensure safe, punctual, and smooth operation of trains. This system also assists dispatchers in understanding the status of operational resources such as trains, locomotives, freight cars, station routes, and sections, enabling unified and optimized dispatching and command, thereby improving transportation efficiency.

[0049] As shown in Figure 1, the perception learning module is responsible for extracting features from freight railway line basic data, dispatching and command data, and train operation data using knowledge engineering methods to build a computable and reconfigurable system model. The system model includes a basic data model, a dispatching system model, a train control system model, and a train driving model.

[0050] The collaborative optimization module is responsible for simulating train operations and dispatching strategies within the production environment using various data captured by the production system, combined with actual transportation production plans. This module then develops, in real time, an optimization solution for the system model constructed by the perception learning module. This optimization solution involves designing daily freight railway operation scenarios, manipulating trains within the simulation according to a given dispatch plan, and randomly generating various unexpected events. This automatically generates train speed curves and dispatching strategies to achieve a driving strategy appropriate for the scenario.

[0051] The strategy evaluation module is responsible for testing and evaluating the scheduling strategies and speed curves generated by the collaborative optimization module around specific operating scenarios and goals.

[0052] If the strategy evaluation module finds that a strategy meets the predefined goals, the strategy generation module will push it to the dispatchers and drivers in the production system to guide them in transportation production.

[0053] The present invention constructs a freight railway dispatching and train control collaborative system, and with the help of simulation methods, designs dispatching adjustment and train control experiments for complex operating scenarios, extracts state parameters under different scenarios from the actual production system, uses the input real data to train the model, and generates a large amount of simulation data based on this, so that the system can continuously learn from massive data, improve the system's scenario recognition and problem-solving capabilities, and effectively evaluate and estimate the decision-making effects of dispatchers and drivers. The learning results are not just a "simulation" of the actual situation, but provide an effective solution for actual production.

[0054] Train passing is a critical task for freight railway dispatching and train control systems. Figure 2 shows a diagram of a passing train m passing through train n at station K (the horizontal axis represents time, the vertical axis represents stations, and the slash represents the train's movement between stations). Only after passing train n arrives at the station and stops on the track can passing train m pass through the passing station. To improve throughput, the passing process is typically very compact.

[0055] Currently, this type of train passing requires coordination between personnel in multiple positions and the dispatching and train control systems: dispatchers create a passing plan through the centralized dispatching system and issue it to the station's trackside train control system for routing. The train control system calculates driving authorization for the train, and the onboard system calculates the speed curve based on the authorization and line data. The plan is also communicated to the driver, who then drives the train to complete the passing under the protection of the onboard system. Dispatchers often need to repeatedly adjust the plan based on the train's operation, the train control system must repeatedly handle routing and calculate authorization, and the driver must adjust his driving strategy based on the plan and signals from ahead. If the connection is not good, passing freight trains can suffer from problems such as unnecessary speed reduction or external parking.

[0056] In order to achieve accurate passing of two trains, the dispatching and train control collaborative system proposed in the present invention constructs a unified system model for the passing process. The collaborative control will accelerate the train to enter the station and stop, and the passing train will appropriately slow down.

[0057] Based on the above-mentioned dispatching and train control coordination system, the present invention provides a freight railway dispatching and train control coordination method, comprising the following steps:

[0058] S1. Extract features of freight railway lines, including basic data, dispatching and command data, and train operation data, and build a system model;

[0059] S2. Acquire various data from the train control production system, combine it with the actual transportation production plan, simulate the train operation process and dispatching command strategy in the constructed system model, and deduce the optimization solution strategy of the system model in various set scenarios in real time;

[0060] S3. Evaluate the optimization solution strategy;

[0061] S4. Apply the optimization solution strategy that meets the requirements in the evaluation to the train control production system to guide actual transportation production.

[0062] Furthermore, step S2 includes using a Q network algorithm to set the operating environment of passing trains and passing trains, simulating the passing process of passing trains and passing trains, and establishing a Markov decision process; the operating environment includes line static data, the operating plans of passing trains and passing trains, the operating position, speed and running time of passing trains, and the position, speed and running time of passing trains.

[0063] The letting operation process is represented as the running process of letting the train from station k-1 to station k and the running process of passing train from station k+1 to station k.

[0064] Let train m run in the section as an example. represents the departure time of the train at station k-1, = represents the arrival time of the train at station k. The line from station k-1 to station k is divided into different sub-intervals according to the fixed speed limit. Each sub-interval has a starting position , and the end position , and the speed limit of the sub-interval From the starting point of each sub-interval The maximum traction speed of the train is calculated at , calculate the running speed of the next state .like > , then the running speed of the next state is the speed limit value of the corresponding sub-interval , otherwise the running speed of the calculation , and finally get the maximum traction speed From the starting point of each subinterval Calculate the maximum braking speed of the train forward. According to the speed of the train at the current state Calculate the running speed of the previous state .like > , then the running speed of the previous state is the speed limit value of the corresponding sub-interval , and finally get the maximum braking speed .Compare and , choose the smaller value from the two, that is , get the shortest running time speed value .

[0065] The time it takes for the train to arrive at station k is: , is the number of sub-intervals. The train is at the end of the sub-interval The maximum pulling speed is: , when calculating the braking speed The net force acting on the train is: For traction, is the traction coefficient, is the braking force coefficient, For braking force, is the basic resistance, is the additional resistance, and the train acceleration is: , The mass of the train will be. The speed of the train will be in the range: .

[0066] Similarly, a process model of the train n passing through the station K+1 to the station k can be established. The difference between the two is that the train must stop at the station as planned, so its final speed is , passing the final speed of the train .

[0067] Usually the departure time of train m at station k-1 and the departure time of train n at station k+1 are determined, and the objective function to be optimized is to always minimize the time: . It indicates the time it takes for a passing train to arrive at station k. However, this time must satisfy the time it takes for the passing train to enter the route and open the signal. ,and .

[0068] It is hoped that the train will run smoothly with less sudden acceleration and deceleration, that is, the acceleration change rate should be minimized: ,in and are the length of the m-subinterval that will allow the train to pass, and the length of the n-subinterval that will allow the train to pass. Is to make train m at the end of its sub-section The acceleration of Is to make train n at the starting point of its sub-interval The acceleration of is the terminal position of train n at its sub-interval The acceleration of The train is at the starting point of the sub-section The acceleration of is the time for train m to run in its subinterval, It is the running time of the train in its subsection.

[0069] Furthermore, the Q network reinforcement learning environment can be set to the actual train operation environment, including line static data, train operation plan, the running position, speed and running time of the passing train, and the position, speed and running time of the opposite train. ,in, The values ​​for the train's operating state, state transition probability, and reward are respectively represented. The reward value is determined by the total let-off time and driving smoothness. The coordination process is illustrated using the let-off train as an example.

[0070] The shortest value is the weighted average of the two objective functions of minimum time and acceleration rate of change, namely: , are the coefficients of the two objectives. The advantage of including the oncoming train in the train operation status is that when calculating the train curve and output conditions, the speed of the passing train and the distance to the passing point are taken into account. This allows for coordinated control optimization of the operation of both trains from the perspective of scheduling. The optimization method for passing train m in the interval is as follows:

[0071] S21. Initialize the approximation network parameters , let the target network parameters for ;

[0072] S22, will make the train running Depend on Sure, is the subinterval length, To speed up the train, is the distance from the passing train to the stopping point, is the speed of the passing train, is the weighted value of the objective function, The probability of randomly selecting the working condition action , with 1- The probability of choosing the current optimal state ;

[0073] S23, execute working condition action , which will allow the train operating environment to get the next state and reward value ;

[0074] S24, randomly select several from the experience playback area , update the network parameters according to the gradient descent algorithm ;

[0075] S25, update network parameters after the specified number of steps for , exit when the number of iterations is reached, otherwise return to S21.

[0076] The onboard intelligent agents of the train and passing trains will continuously interact with the reinforcement learning environment according to the above optimization process. The onboard intelligent agents will calculate the train operating conditions under the current state based on the train position, speed and running time, so that the train and passing trains will tend to run in the shortest time and smoothly.

[0077] The train operating environment allows the train and the onboard agent to calculate the train's acceleration sequence and the train's arrival time at station k based on their operating conditions. This generates a new train operating state, reward value, and action-value function. The two agents select the maximum action-value function and feed it back to the reinforcement learning environment. Using this closed-loop structure, the agents continuously update the operating conditions, ultimately achieving the optimal operating condition-action sequence. The speed curve corresponding to this operating condition-action sequence is the optimal speed curve, as shown in Figure 4.

[0078] Taking the data from a local railway line as an example, the section from JK to MP is 6.6 km long and takes 15 minutes to run. The section from MP to BF is 10.8 km long and takes 20 minutes to run. The JK-MP-BF direction is a loaded train, while the BF-MP-JK direction is an empty train. The locomotive is a DF7G, pulling a 26-carriage train with a gross train mass of 2,500 tons and a maximum tractive force of 270 kN. The train's starting speed is 0 km / h and its ending speed is 70 km / h. Passing MP Station on the mainline results in a starting and ending speed of 0 km / h, with the train finally stopping on the MP siding. The test line data is shown in Tables 1 and 2.

[0079] Table 1 Line conditions in test section 1

[0080] Section number 1234567891011 Length (m) 406390860510260430270310450760800 Slope (‰) -0.402.009.50-10.30-5.80-3.600.507.003.700.203.30 Speed ​​limit (km / h) 3030505050505050507070

[0081] Table 2 Line conditions in test section 2

[0082] Section number 12345678910111213 Length (m) 8003402101100260720460166018409806501700800 Slope (‰) -4.00-7.303.6011.105.903.001.005.203.605.303.500.803.30 Speed ​​limit (km / h) 505050505050505050505050

[0083] The target network has a fixed update step count of 300, an experience replay data area capacity of 10,000, a single experience replay data size of 64, and a deep learning rate of , discount factor , initial greed rate , the final greed rate , each state transition .

[0084] With the goal of minimizing travel time and ensuring smooth operation, the coordinated optimization of train group scheduling and operational control is achieved. Train operation information is collected and combined with the actual operation diagram, basic section conditions, and basic train parameters to generate a feasible optimized speed and position curve. The ultimate goal of the coordinated optimization of train running time and speed curve is to minimize the weighted sum of train delay time and smoothness. The core strategy is to continuously adjust the train running time in each section and obtain the optimal speed curve for that running time. This running time and acceleration set is then fed back to the train running time adjustment model, ultimately resulting in an optimal running time adjustment strategy.

[0085] Typically, when not considering oncoming trains yielding, the passing train generates a speed curve with maximum speed as the target. During the simulation experiment, considering the possibility that the oncoming train might not reach the target speed during station entry for some reason, resulting in delayed entry, the passing train continuously adjusted its speed in each section to ensure that its arrival time exactly matched the yielding interval, avoiding unexpected stops. Table 3 describes the corresponding position and speed of the passing train during the optimization process, based on the yielding train's position and speed information.

[0086] Table 3 Speed ​​strategy of passing trains considering oncoming trains

[0087] No. Remaining distance to opposite train (m) Speed ​​of opposite train (km / h) Remaining time for opposite train to arrive (s) Remaining distance to own train (m) Maximum speed of own train (km / h) 1155050239.71600482155040295.51600383155030388.51600294155020574.51600205137050213.781370466137040263.11370377137030345.31370288137020509.7137019960030160.5600261060020232.560018

[0088] During the simulation, the speed curve generated for the passing train considering the position and speed of the oncoming train is shown in Figure 5. The passing train slows down in advance based on the position of the passing train about two kilometers from the station, and then increases its speed after the passing train enters the station, passing through the station quickly.

[0089] Therefore, by adopting the above method, when the passing train approaches the MP station, the passing train can be automatically controlled to slow down in advance according to the position and speed of the two cars, which will allow the train to accelerate into the station and improve the passing efficiency of the two cars. On the one hand, it can effectively avoid the passing train from stopping outside the machine due to waiting for the passing train to enter the station. On the other hand, it can reduce the frequent communication costs between the dispatcher and the drivers of the two cars during the passing.

[0090] Compared with the prior art, the present invention has the following beneficial effects:

[0091] 1. Break the information barriers between the dispatching system and the train control system, and effectively improve the efficiency of freight railway transportation.

[0092] 2. A freight train passing model was established, detailing the operation and planned arrival times of passing and passing trains. This model accurately describes the coordination between passing and passing trains as they arrive at passing stations, laying the foundation for optimizing passing operations.

[0093] 3. Considering the speed and position of the oncoming trains, the calculation method of the dual-depth Q network is improved. The advantage is that when calculating the train curve and output working conditions, the speed of the passing train and the distance to the passing point are taken into account. From the perspective of scheduling, the operation process of the two trains can be coordinated and optimized at the same time.

[0094] Although the present invention has been described in detail through the above preferred embodiments, it should be understood that the above description is not intended to limit the present invention. After reading the above description, various modifications and substitutions of the present invention will become apparent to those skilled in the art. Therefore, the scope of protection of the present invention should be defined by the appended claims.

Claims

1. A method for coordinated freight railway dispatching and train control, characterized in that: Include: S1. Extract features of basic data, dispatching command data and train operation data of freight railway lines and build a system model; S2. Acquire various data in the train control production system, simulate the train operation process and dispatching command strategy in the constructed system model in combination with the actual transportation production plan, and deduce the optimization solution strategy of the system model in each set scenario in real time; S3, evaluating the optimization solution strategy; S4. Apply the optimization solution strategies that meet the requirements in the evaluation to the train control production system to guide the actual transportation production.

2. The dispatching and train control coordination method according to claim 1, characterized in that: The system model of step S1 includes: a basic data model, a dispatching system model, a train control system model and a train driving model.

3. The dispatching and train control coordination method according to claim 1, characterized in that: The various types of data in step S2 include various emergencies during the operation of the train.

4. The dispatching and train control coordination method according to claim 1, characterized in that: The optimization solution strategy in step S2 is a train speed curve and a scheduling adjustment plan suitable for a set scenario.

5. The dispatching and train control coordination method according to claim 1, characterized in that: The step S2 includes using a Q network algorithm to set the operating environment of the passing train and the passing train, simulating the passing process of the passing train and the passing train, and establishing a Markov decision process; The operating environment includes static data of the line, the operating plans of passing trains and passing trains, the operating positions, speeds and operating times of passing trains, and the positions, speeds and operating times of passing trains.

6. The dispatching and train control coordination method according to claim 5, characterized in that: The operation process is represented as the operation process of the train from station k-1 to station k and the operation process of the passing train from station k+1 to station k.

7. The dispatching and train control coordination method according to claim 6, characterized in that: Divide the section from station k-1 to station k and the section from station k+1 to station k into different subsections according to the fixed speed limit, and calculate the maximum traction speed and maximum braking speed that will allow the train or the passing train to move in each subsection; According to the speed that will allow the train or the passing train to enter the current state, the running speed of the next state and the running speed of the previous state are calculated. Under the premise of not exceeding the speed limit of the corresponding sub-interval, the maximum traction speed is obtained according to the running speed of the next state, and the maximum braking speed is obtained according to the running speed of the previous state. The smaller value of the maximum traction speed and the maximum braking speed is taken as the shortest running time speed value.

8. The dispatching and train control coordination method according to claim 7, characterized in that: The time it takes for the train to arrive at station k is: , To make the train depart at station k-1, To make the train arrive at station k, is the length of the subinterval that the train will pass through, w is the number of subintervals that the train will pass through, is the speed value that will allow the train to run in the shortest time; The time it takes for the train to reach station k is: ， is the departure time of the passing train at station k+1, is the arrival time of the passing train at station k, is the length of the sub-interval through which the train passes, z is the number of sub-intervals through which the train passes, It is the speed value of the shortest running time of a passing train.

9. The dispatching and train control coordination method according to claim 8, characterized in that: The objective function of optimization is to always minimize the time: ,and , The time it takes to clear the route and open the signal for passing trains.

10. The dispatching and train control coordination method according to claim 8, characterized in that: If it is necessary to make the process run smoothly and reduce sudden acceleration and deceleration, the optimization objective function is to minimize the acceleration change rate: ,in and are the length of the train subsection and the length of the train subsection, is the acceleration that will bring the train to the end of the subsection, is the acceleration that will cause the train to reach the starting point of the subsection. is the acceleration of the passing train at the end of the subsection, is the acceleration of the train at the starting point of the sub-section, It is to let the train run in the sub-section for a certain period of time. It is the running time of the train in the sub-section.

11. The dispatching and train control coordination method according to claim 1, characterized in that: The step S2 uses the Q network algorithm to optimize the train operation process, which includes the following steps: S21, Initialize the approximation network parameters , let the target network parameters for ； S22, will make the train running Depend on Sure, is the subinterval length, To speed up the train, is the distance from the passing train to the stopping point, is the speed of the passing train, is the weighted value of the objective function, The probability of randomly selecting the working condition action , with 1- The probability of choosing the current optimal state ； S23, Execute working condition action , according to which the train running environment will get the next state and reward value ； S24, randomly select a number of , update the network parameters according to the gradient descent algorithm ； S25, update network parameters after a specified number of steps for , exit when the number of iterations is reached, otherwise return to S21.

12. A freight railway dispatching and train control coordination system, used to implement the dispatching and train control coordination method according to any one of claims 1 to 11, characterized in that: Include: The perception learning module is used to obtain basic data of freight railway lines, dispatching and command data, and train operation data features for extraction, and to build a computable and reconfigurable system model; Collaborative optimization module, which is used to obtain various data in the production system, simulate the train operation process and dispatching command strategy in the production environment in combination with the actual transportation production plan, and deduce the optimization solution strategy of the system model in real time; The strategy evaluation module tests and evaluates the optimization solution strategy generated by the collaborative optimization module; If the strategy evaluation module pushes the optimization solution strategy that meets the predefined goals to the production system for execution.

13. The dispatching and train control coordination system according to claim 12, characterized in that: The system model includes: a basic data model, a dispatching system model, a train control system model and a train driving model.

14. The dispatching and train control coordination system according to claim 12, characterized in that: The various types of data are derived from various types of emergencies that are randomly generated.

15. The dispatching and train control coordination system according to claim 12, characterized in that: The optimization solution strategy includes: train speed curve and scheduling adjustment plan.

Images