A train handling control method, device, equipment, medium and product

Abstract

The application discloses a train operation control method, device, equipment, medium and product, and relates to the field of heavy-haul train automatic driving. The method comprises the following steps: acquiring segmented vehicle parameters, segmented vehicle state information and line parameter information of a heavy-haul train; establishing a segmented longitudinal dynamics model to determine the coupler force of each vehicle segment connection in a prediction time domain; constructing a coupler force risk map; constructing a conditional tensor and inputting the conditional tensor into a condition denoising generation model to generate a plurality of candidate operation sequences; performing safety constraint screening on the plurality of candidate operation sequences, and selecting a sequence with the minimum comprehensive cost from the candidate operation sequences meeting preset safety constraints as an execution sequence; executing an operation instruction of the first time step, and after a rolling control cycle, reacquiring state information, reconstructing the coupler force risk map, and regenerating and screening the operation sequence to perform closed-loop rolling control on the heavy-haul train. The application realizes intelligent optimization and closed-loop accurate control of the heavy-haul train operation strategy.

Description

Technical Field

[0001] This application relates to the field of automatic driving of heavy-haul trains, and in particular to a train operation and control method, device, equipment, medium and product. Background Technology

[0002] When heavy-haul trains operate on long gradients and at gradient change points, the distribution of coupler forces is extremely uneven due to the length of the train formation and the different gradients of each car. Excessive pulling or pressing forces can lead to coupler breakage or derailment. Traditional control methods (PID speed regulation, rule-based segmented control) cannot predict the spatial distribution trend of coupler forces and are unable to proactively avoid high-risk areas. While optimization-based methods (such as MPC) can consider coupler force constraints, they involve large computational loads and are difficult to generate diverse solutions in real time. Therefore, there is an urgent need for a method that incorporates the spatial distribution of coupler force risk as a condition into the control sequence generation process, and rapidly generates multiple candidate control schemes that meet the constraints and proactively avoid coupler force risks. Summary of the Invention

[0003] The purpose of this application is to provide a train operation and control method, device, equipment, medium and product that can realize intelligent optimization and closed-loop precise control of heavy-load train operation strategies.

[0004] To achieve the above objectives, this application provides the following solution: In a first aspect, this application provides a train operation and control method, including: Obtain segment vehicle parameters, segment vehicle status information, and track parameter information for heavy-haul trains; Based on the segmented vehicle parameters, the segmented vehicle status information, and the line parameter information, a segmented longitudinal dynamic model is established, and the coupler force at the connection point of each vehicle segment in the prediction time domain is determined based on the segmented longitudinal dynamic model. Based on the coupler force at the connection points of each vehicle segment within the predicted time domain, a coupler force risk map is constructed. Based on the coupler force risk diagram and the line parameters, a conditional tensor is constructed. The conditional tensor is then input into a pre-trained conditional denoising generation model to generate multiple candidate manipulation sequences. Each candidate manipulation sequence includes manipulation instructions for each time step within the pre-test domain. The manipulation instructions include traction level, electric braking level, and air braking status. Multiple candidate manipulation sequences are screened for safety constraints, and candidate manipulation sequences that satisfy all preset safety constraints are retained. The candidate manipulation sequence with the lowest overall cost among the candidate manipulation sequences that satisfy the preset safety constraints is selected as the execution sequence. The operation command of the first time step in the execution sequence is executed. After one rolling control cycle, the segment vehicle status information is reacquired, the coupler force risk diagram is reconstructed, and the operation sequence is regenerated and filtered to perform closed-loop rolling control on the heavy-haul train.

[0005] Secondly, this application provides a train operation control device, including: The data acquisition module is used to acquire the segment vehicle parameters, segment vehicle status information, and track parameter information of heavy-haul trains. The dynamics prediction module is used to establish a segmented longitudinal dynamics model based on the segmented vehicle parameters, the segmented vehicle state information, and the line parameter information, and to determine the coupler force at the connection point of each vehicle segment in the prediction time domain based on the segmented longitudinal dynamics model. The risk map construction module is used to construct a coupler force risk map based on the coupler force at the connection points of each vehicle segment within the prediction time domain. The conditional diffusion generation module is used to construct a conditional tensor based on the coupler force risk diagram and the line parameters, input the conditional tensor into a pre-trained conditional denoising generation model, and generate multiple candidate manipulation sequences. Each candidate manipulation sequence includes manipulation instructions for each time step in the pre-test domain. The manipulation instructions include traction level, electric braking level, and air braking status. The safety constraint filtering module is used to filter multiple candidate manipulation sequences for safety constraints, retain candidate manipulation sequences that satisfy all preset safety constraints, and select the candidate manipulation sequence with the lowest overall cost from the candidate manipulation sequences that satisfy the preset safety constraints as the execution sequence. The rolling execution module is used to execute the operation command of the first time step in the execution sequence. After one rolling control cycle, it reacquires the segment vehicle status information, reconstructs the coupler force risk diagram, and regenerates and filters the operation sequence to perform closed-loop rolling control on the heavy-haul train.

[0006] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described train operation control method.

[0007] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described train operation control method.

[0008] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the above-described train operation control method.

[0009] According to the specific embodiments provided in this application, this application has the following technical effects: This application achieves intelligent optimization and closed-loop precise control of heavy-haul train operation strategies by integrating segmented dynamic modeling and conditional denoising generation models. First, a longitudinal dynamic model is established based on segmented parameters and track information, and a coupler force risk map is constructed. This allows for the early quantification and prediction of coupler force distribution in the time domain, effectively preventing coupler overload fracture or crush instability, and significantly improving the operational safety of long-formation trains. Second, the risk map and track parameters are constructed as conditional tensors and input into a pre-trained conditional denoising generation model. This efficiently generates diverse candidate operation sequences covering traction, electric braking, and air braking states, overcoming the shortcomings of traditional optimization algorithms that are prone to getting trapped in local optima or have excessive computational loads, thus significantly improving decision quality and real-time computation efficiency. Simultaneously, by selecting execution sequences through strict safety constraint screening and a comprehensive cost minimization criterion, not only are the operation commands satisfied with hard boundaries such as speed limits, signals, and coupler forces, but multiple economic objectives such as energy consumption and punctuality are also considered. Finally, by adopting a single-step execution and rolling time-domain update strategy, the risk map is reconstructed based on the real-time state and the sequence is regenerated, forming a closed-loop control architecture of "perception-decision-execution-correction". This architecture can effectively cope with line disturbances and time-varying parameters, greatly enhancing the system's adaptability and robustness, and providing reliable support for the smooth, safe and energy-saving operation of heavy-haul trains under complex conditions. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic flowchart of a train operation control method provided in an embodiment of this application.

[0012] Figure 2 This is a schematic diagram of the interface of the simulation software constructed in one embodiment of this application.

[0013] Figure 3 This is a schematic diagram of the functional modules of a train operation and control device provided in an embodiment of this application. Detailed Implementation

[0014] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0015] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0016] In one exemplary embodiment, such as Figure 1 As shown, a train operation control method is provided. This method is executed by a computer device, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method includes steps 101 to 106.

[0017] Step 101: Obtain the segment vehicle parameters, segment vehicle status information, and track parameter information for the heavy-haul train. Segment vehicle parameters include the mass of each vehicle segment. The segmented vehicle status information includes the speed of each vehicle segment. and location The route parameter information includes the route gradient. curve radius and line speed limits .in, Index the vehicle segments. The total number of vehicle segments. This represents the mileage location along the route.

[0018] Mass of each vehicle segment Sourced from train formation information and vehicle weight database (pre-entered); speed of each vehicle section Speed ​​from wheel axle sensors (pulse / Hall effect, accuracy) installed on locomotives and various vehicles km / h); Location of each vehicle segment Multi-source fusion positioning from satellite positioning receiver (GNSS, update rate 1Hz~10Hz) combined with wheel odometer; line gradient curve radius and line speed limits The route database (electronic map) is pre-stored in the vehicle's computer.

[0019] Segmented vehicle state information is represented as a state vector: ;in, For the first Each vehicle segment at time State vector ( The unit is meters. The unit is m / s, which is obtained by measuring speed sensors, accelerometers and train positioning systems installed on the locomotive.

[0020] Step 102: Based on the segmented vehicle parameters, the segmented vehicle status information, and the line parameter information, establish a segmented longitudinal dynamic model, and determine the coupler force at the connection point of each vehicle segment in the prediction time domain based on the segmented longitudinal dynamic model.

[0021] In a specific application example, step 102 includes steps 21 to 24.

[0022] Step 21: Determine the reference maneuver sequence in the prediction time domain based on the segment vehicle parameters, segment vehicle status information, and line parameter information at the current time.

[0023] Step 22: Based on the reference maneuvering sequence and the basic running resistance, additional slope resistance, additional curve resistance, and adjacent coupler force experienced by each vehicle segment, establish the longitudinal dynamic equations for each vehicle segment.

[0024] Specifically, based on the traction and braking forces corresponding to the reference control sequence, as well as the basic running resistance, additional slope resistance, additional curve resistance, and adjacent coupler forces experienced by each vehicle segment, the longitudinal dynamic equations for each vehicle segment are established.

[0025] Step 23: Based on the longitudinal dynamic equations of each vehicle segment, calculate the predicted speed and predicted position of each vehicle segment step by step in the prediction time domain.

[0026] For the The vehicle segment, in the future... Each prediction time step calculates the predicted acceleration based on the longitudinal force balance: ; in, For the first The vehicle segment is in the first Predicted acceleration at each prediction time step For the first The vehicle segment is in the first The traction of a predicted time step, For the first The vehicle segment is in the first The sum of the electric braking force and the air braking force at each predicted time step. For the first The vehicle segment is in the first The running resistance at each predicted time step, , For the first The vehicle segment is in the first The basic resistance at each predicted time step For the first The vehicle segment is in the first Additional slope resistance at each predicted time step , It is the acceleration due to gravity. The gradient angle of the track is a function of the track's location. Indicates the first The vehicle segment is in the first The line gradient angle corresponding to the predicted position at each predicted time step. For the first The vehicle segment is in the first The curve at each predicted time step adds resistance. , This is a function used to determine the additional resistance coefficient of a curve based on its radius. Let be the radius of the curve as a function of the line's position. Indicates the first The vehicle segment is in the first The radius of the line curve corresponding to the predicted position at each predicted time step. For the first The vehicle segment is in the first Predicted location at each prediction time step. For the first The vehicle segment is in the first The coupler force at the connection point at each predicted time step. For the first The vehicle segment is in the first The coupler force at the connection point at each predicted time step. For the first The quality of each vehicle segment.

[0027] After obtaining the predicted acceleration, discrete integration is used to calculate the velocity and position for the next predicted time step: ; ; in, For the first The vehicle segment is in the first Prediction speed per prediction time step For the first The vehicle segment is in the first Prediction speed per prediction time step For the first The vehicle segment is in the first Predicted location at each prediction time step. For the first The vehicle segment is in the first Predicted location at each prediction time step. To predict the time step. The initial condition is the state measured by the sensor at the current moment: ; ;in, This refers to the current moment.

[0028] Therefore, the future prediction time steps can be recursively obtained. and .

[0029] Step 24: Based on the predicted position difference and predicted speed difference between adjacent vehicle segments, the coupler force at the connection point of each vehicle segment in the prediction time domain is calculated using an equivalent spring-damping model.

[0030] No. The vehicle segment and the first At the time of the coupler connection between the vehicle segments The coupler force is calculated using a spring-damped model: ;in, For the first The vehicle segment and the first At the time of the coupler connection between the vehicle segments The coupler force (kN, positive values ​​represent the hook pulling force, negative values ​​represent the hook pressing force); For the first The vehicle segment and the first The equivalent stiffness coefficient (kN / m) of the coupler and buffer device at the coupler connection between vehicle sections is determined by the model of the coupler and buffer device and is a known parameter; For the first The vehicle segment and the first The equivalent damping coefficient (kN) of the coupler and buffer device at the coupler connection between vehicle sections. The value (s / m) is determined by the model of the hook and buffer device and is a known parameter. and Pre-stored in the vehicle's onboard computer; For the first The vehicle segment and the first The relative displacement at the coupler connection between each vehicle segment, minus the balance clearance. Effective displacement (m) after; For the first The vehicle segment and the first The speed difference (m / s) between the two ends of the coupler connection between the vehicle segments.

[0031] Specifically, the relative displacement and velocity difference are first calculated based on the predicted positions and velocities of adjacent vehicle segments: ; ; in, For the first The vehicle segment and the first The predicted position difference between vehicle segments For the first The vehicle segment and the first The predicted speed difference between vehicle segments For the first The vehicle segment is in the first Predicted location at each prediction time step. For the first The vehicle segment is in the first Prediction speed per prediction time step For the first The vehicle segment and the first The reference spacing between the couplers of each vehicle segment.

[0032] Considering the balance clearance of the hook and buffer device The effective relative displacement can be defined as: ; in, For the first The vehicle segment and the first The effective relative displacement between vehicle segments For the first The vehicle segment and the first The balance clearance of the coupler buffer device at the coupler connection between vehicle sections.

[0033] Then calculate the future coupling force: .in, For the first The vehicle segment and the first The coupling connection between the vehicle segments is at the first... The coupler force at each predicted time step.

[0034] In another specific application example, the segmented longitudinal dynamics model further includes differentiated modeling of the air brake wave propagation delay. After the air brake command is issued by the head brake control device, the brake pipe pressure change propagates longitudinally along the train at a finite speed. Different vehicle segments exhibit different air brake response delays due to their varying propagation distances from the head brake control device. The air brake response delay of each vehicle segment is determined based on its formation position, train formation length, and air brake wave propagation speed. ;in, For the first The time delay (s) from the issuance of the braking command to the start of brake cylinder pressure establishment for each vehicle segment. The inherent response delay (s) for the leading vehicle segment is determined by the brake valve characteristics; The total propagation delay (in seconds) of the braking wave along the train length from the front to the rear is determined by the train formation length and the air braking wave propagation speed. , This refers to the length of the train formation, in meters (m). The equivalent propagation speed of the air braking wave in the brake tube, in m / s, is determined by the brake tube structure, air pressure, and brake system characteristics. Therefore, the first The air brake response delay for each vehicle segment can also be expressed as: ;in, For the first The longitudinal distance between each vehicle segment and the head brake control device; when the vehicle segments are approximately uniformly distributed along the train length, we have: ; Corresponding to the head vehicle segment. Corresponding to the rear vehicle segment.

[0035] Step 103: Construct a coupler force risk map based on the coupler force at the connection points of each vehicle segment within the predicted time domain.

[0036] In a specific application example, based on the predicted coupler forces at each segment connection, the proximity of the hooking force and / or clamping force to the safety threshold in the future time period is normalized to form a coupler force risk map for subsequent operation. That is, the coupler force risk map is a normalized risk value matrix for each vehicle segment connection within the prediction time domain, expressed as: ; in, For the first The vehicle segment and the first The coupling connection between the vehicle segments is at the first... Coupler force risk value at each predicted time step (dimensionless, range of values) (A higher value indicates a higher risk). For the first The vehicle segment and the first The coupling connection between the vehicle segments is at the first... The coupler force (kN) at each predicted time step. The permissible coupler force safety threshold (kN) is determined by the vehicle design specifications and is a preset constant. For predicting the time step index, To predict the number of time-domain steps.

[0037] In another specific application example, the construction of the coupler force risk diagram further includes: constructing the hook risk diagram separately. Risk diagram of hook pressing : ; ; in, For the first The vehicle segment and the first The coupling connection between the vehicle segments is at the first... The hook risk value (dimensionless) for each prediction time step is non-zero only when the hook force is positive (under tension); For the first The vehicle segment and the first The coupling connection between the vehicle segments is at the first... The hook-over-load risk value (dimensionless) for each predicted time step is non-zero only when the coupler force is negative (under pressure).

[0038] Step 104: Construct a conditional tensor based on the coupler force risk diagram and the track parameters, and input the conditional tensor into a pre-trained conditional denoising generation model to generate multiple candidate manipulation sequences. Each candidate manipulation sequence includes manipulation commands at each time step within the pre-test domain; the manipulation commands include traction level, electric braking level, and air braking status.

[0039] In a specific application example, the coupler force risk diagram, track gradient, speed limit margin, air brake delay state, brake cylinder pressure prediction value, and distributed dynamic state are constructed into a conditional tensor. and condition tensor Input a pre-trained conditional denoising generative model.

[0040] Conditional denoising generation models use noisy manipulation sequences and conditional tensors Using the denoising step index as input, multiple candidate manipulation sequences are generated step by step. Each candidate manipulation sequence includes manipulation commands for each time step within the pre-test domain; the manipulation commands include traction level, electric braking level, and air braking status.

[0041] The conditional denoising generation model is described by a conditional inverse denoising process, for the first... The candidate manipulation sequence, from the _th ... Starting with the initial noisy manipulation sequence of each denoising step, the denoising is progressively reduced to step 0: For the The candidate manipulation sequence is defined in the nth... The noisy manipulation sequence for each denoising step is as follows: ;in, For the first In the candidate manipulation sequence, the first The control commands for each noise reduction step. ; For the first The candidate manipulation sequence in the ... The continuous encoded values ​​of the traction level bits at each prediction time step. For the first The candidate manipulation sequence in the ... The continuous encoded values ​​of the electric braking level bits for each prediction time step. For the first The candidate manipulation sequence in the ... The continuous encoded values ​​of the air braking state at each predicted time step. Therefore, For dimension Noisy manipulation sequences.

[0042] In the first rolling control cycle, the first The initial noisy manipulation sequence for each denoising step is randomly generated from standard Gaussian noise: ; .

[0043] Different candidate manipulation sequences use mutually independent random noise matrices This generates different candidate manipulation sequences: .in, For the first The candidate manipulation sequence in the ... A noisy manipulation sequence of denoising steps. For the first The candidate manipulation sequence in the ... A noisy manipulation sequence of denoising steps, with dimension . The three channels correspond to the continuous coded values ​​of the traction level, electric braking level, and air braking status, respectively. For the denoising step index, from Decrease to ; , and These are the noise scheduling parameters for the diffusion model; For a pre-trained denoising network, with the current noisy manipulation sequence Conditional tensor and denoising step index For input, output and Predictive noise of the same dimension; For standard Gaussian noise, when When introduced, Take zero at time; For candidate manipulation sequence index, This represents the total number of candidate manipulation sequences.

[0044] Conditional tensor For one 3D tensor: ;in, For the coupler force risk diagram channel (dimension) ); For the gradient channel of the line (dimensions) ), due to the gradient of the line The values ​​are obtained by interpolation along each segment position; For speed margin channel (dimension) ), ; For air braking delay state channel (dimension) This indicates whether the air brakes of each section have been fully established at the current moment, and is calculated in real time by the train braking control system based on the braking command issuance time and the delay model of each section. For brake cylinder pressure prediction channel (dimension) ), consisting of an air brake system pressure sensor (range 0kPa~1000kPa, accuracy ). The data was collected from the FS (Fuel Cell) and derived from the brake cylinder filling and exhaust model. For distributed dynamic state channels (dimensions) The system indicates whether each section is equipped with a main power unit, which is obtained by the train distributed power control system communication module. The traction / braking status of each locomotive is exchanged in real time through wireless communication (such as the Locotrol system or LTE-R). This indicates a splicing operation along the channel dimension; This represents the total number of channels.

[0045] Conditional tensor After being encoded by the conditional encoder, the signal is injected into the denoising network through feature splicing, feature-wise linear modulation (FiLM), or cross-attention mechanism. This allows the denoising network to progressively denoise the traction level channel, electric braking level channel, and air braking status channel based on the coupler force risk, gradient, speed limit margin, and braking status at each future time step, until a continuous candidate maneuver sequence is obtained. After performing inverse denoising, the candidate manipulation sequences in continuous form are then processed. By performing inverse normalization, rounding, and manipulation logic correction, the actual candidate manipulation sequence is obtained.

[0046] Let the maximum value of the traction level be The maximum value of the electric braking level is Then the traction level and electric braking level are converted according to the following formulas: ; ; When the air brake is in two states, application and release, it is switched according to the following formula: ; in, For the first The candidate manipulation sequence number The traction level of each predicted time step For the first The candidate manipulation sequence number The electric braking level position is predicted for each time step. For the first The candidate manipulation sequence number The air braking state at each predicted time step. Indicates that air braking is applied. This indicates that the air brake has been released. After discretization, conflicting traction and braking commands within the same time step are corrected to ensure that the traction level, electric braking level, and air brake status meet the train operation logic requirements.

[0047] Step 105: Perform security constraint screening on multiple candidate manipulation sequences, retain candidate manipulation sequences that satisfy all preset security constraints, and select the candidate manipulation sequence with the lowest overall cost from the candidate manipulation sequences that satisfy the preset security constraints as the execution sequence.

[0048] In a specific application example, multiple candidate maneuver sequences are successively substituted into the segmented longitudinal dynamics model for forward extrapolation. The predicted speed, predicted position, predicted coupler force at each coupler connection, and braking distance of each candidate maneuver sequence in the prediction time domain are calculated. Based on the prediction results, speed limit constraints, coupler force constraints, braking distance constraints, and maneuver rate of change constraints are verified respectively. Candidate maneuver sequences that satisfy all preset safety constraints are retained, and the sequence with the minimum comprehensive cost among the candidate maneuver sequences that satisfy the constraints is selected as the execution sequence.

[0049] No. A candidate manipulation sequence is defined as: Among them, the first The manipulation vectors for each prediction time step are: ;in, For the first The candidate manipulation sequence in the ... The traction level of each predicted time step For the first The candidate manipulation sequence in the ... The electric braking level position is predicted for each time step. For the first The candidate manipulation sequence in the ... The air braking state at each predicted time step.

[0050] For the Candidate manipulation sequences Based on the current train status The following results were obtained by forward extrapolating the line parameters and segmented longitudinal dynamic models: ;in, For the first The predicted state trajectory corresponding to each candidate manipulation sequence includes the speed and position of each vehicle segment at each predicted time step in the future; For each coupler connection point at each predicted time step in the future; This is the result of the braking distance calculation; For the first Total energy consumption corresponding to each candidate manipulation sequence; For a piecewise longitudinal dynamic model, For the set of line parameters, This is the set of parameters for the segmented longitudinal dynamics model.

[0051] The preset safety constraints are: ; ; ; ; in, For the first The first candidate manipulation sequence corresponding to the first The vehicle segment is in the first Prediction speed per prediction time step; For the first The first candidate manipulation sequence corresponding to the first The vehicle segment is in the first Predicted position at each prediction time step; Speed ​​limits will be imposed on the line corresponding to the predicted location; For the first The first candidate manipulation sequence corresponding to the first The vehicle segment and the first The coupling connection between the vehicle segments is at the first... Predicted coupler force at each predicted time step; In the first The expected braking distance calculated under each candidate manipulation sequence; The maximum permissible braking distance is determined by the braking system performance specifications and the current total weight of the train; For the first The candidate manipulation sequence in the ... Manipulation vectors for each predicted time step; The maximum permissible change in the control vector between adjacent prediction time steps is determined by the traction / braking system response characteristics and is a preset parameter.

[0052] Furthermore, when all candidate manipulation sequences If none of the preset security constraints are met, the rollback strategy will be executed according to the following priority: First priority: Increase the number of candidate manipulation sequences generated. to Candidate manipulation sequences are resampled from the conditional denoising generation model and constrained verification is performed. The number of expansions for the candidate manipulation sequence.

[0053] The second priority is to make a minimum change to the candidate maneuver sequence that is closest to satisfying the constraints: reduce the maneuver level that violates the speed limit constraint to the value that meets the speed limit, or adjust the traction / braking level that violates the coupler force constraint to the safe range of the coupler force.

[0054] The third priority is to switch to the preset conservative rule-based control scheme: determine the traction / braking level according to the current track gradient and speed limit based on the predefined safety control rules.

[0055] The overall cost function is: ; in, For the first The total cost of a candidate manipulation sequence (dimensionless). For the first Total energy consumption (kWh) corresponding to each candidate manipulation sequence; For the first The estimated arrival time error (s) of each candidate manipulation sequence; For the first The cumulative coupler force risk penalty for each candidate maneuver sequence; For the first The number of manipulation level changes in each candidate manipulation sequence; , , , To preset hyperparameters, The weighting factor representing total energy consumption. This represents the weighting coefficient for the estimated arrival time error. This represents the weighting coefficient of the cumulative coupler force risk penalty item. The weighting coefficient represents the number of times the manipulation level changes.

[0056] Total energy consumption Calculated based on the forward inference results corresponding to the candidate manipulation sequences, and expressed as follows: ;in, For the first The candidate manipulation sequence in the ... The traction power at each predicted time step. For the first The candidate manipulation sequence in the ... Auxiliary power for each prediction time step, For the first The candidate manipulation sequence in the ... The regenerative braking feedback power for each predicted time step; if the system does not consider auxiliary power or regenerative braking, the corresponding terms can be omitted.

[0057] Estimated arrival time error The deviation between the expected arrival time and the planned arrival time of the candidate manipulation sequence is expressed as: ;in, According to the first The estimated arrival time derived from the deduction of candidate manipulation sequences This refers to the planned arrival time.

[0058] The cumulative coupler force risk penalty item is represented as follows: .

[0059] Number of manipulation level changes The number of times the manipulation vector changes between adjacent time steps within the future prediction time domain is represented as: ;in, This is an indicator function that takes the value 1 when the condition is true and 0 otherwise.

[0060] Step 106: Execute the manipulation command of the first time step in the execution sequence. After one rolling control cycle, reacquire the segment vehicle status information, reconstruct the coupler force risk diagram, and regenerate and filter the manipulation sequence to perform closed-loop rolling control on the heavy-haul train.

[0061] Specifically, the manipulation instructions The output is sent to the Transmission Control Unit (TCU), the electric brake control unit, and the air brake control unit (via the brake control interface), and each actuator adjusts the traction / braking force according to the instructions; the updated train status is fed back to step 101 to start a new round of closed-loop control.

[0062] In a specific application example, the update rule executed in the rolling time domain is as follows: ;in, To execute the manipulation instructions for the first time step in the sequence (including traction level bits) Electric braking level and air braking state ); The rolling control period (s) has a value range of [value range missing]. s is adaptively determined by the line risk level and train operating status.

[0063] The rolling control cycle is adaptively adjusted according to the line risk level: ; in, For rolling control cycle, The rolling cycle (s) of medium- and high-risk areas. The rolling cycle (s) for medium-risk areas. The rolling period (s) in the stable region satisfies , The threshold for determining high risk. The threshold for determining low risk. This is the lower speed limit margin threshold (m / s). This is the high speed limit margin threshold (m / s). For the first The vehicle segment and the first The coupling connection between the vehicle segments is at the first... The risk value of the coupler force at each predicted time step This represents the maximum coupler force risk value in the coupler force risk diagram. For the first Rate limiting margin for each predicted time step, To predict the minimum speed margin (m / s) in the time domain, the speed limit constraint is used. definition, , For the first The vehicle segment is in the first The predicted mileage position along the route at each predicted time step.

[0064] When regenerating the manipulation sequence, the unexecuted portion of the manipulation sequence from the previous round is used as the initial prior trajectory for the backsampling process of the conditional denoising generation model in the current round: .in, This is the starting trajectory for denoising in the new generation process; The unexecuted portion from the second time step to the end of the previous manipulation sequence (i.e., retained generated but unexecuted prediction steps) is processed with noise and used as the prior for the new round of denoising. In the first rolling control cycle, samples obtained from a standard Gaussian distribution are used. A 3D random noise matrix is ​​used as the initial noisy manipulation sequence for the backsampling process; in non-first rolling control cycles, the unexecuted part of the previous round manipulation sequence is noise-added and used as the initial noisy manipulation sequence for the backsampling process.

[0065] In summary, this application achieves intelligent optimization and closed-loop precise control of heavy-haul train operation strategies by integrating segmented dynamic modeling and conditional denoising generation models. Specifically, firstly, a longitudinal dynamic model is established based on segmented parameters and track information, and a coupler force risk map is constructed. This allows for the early quantification and prediction of coupler force distribution in the time domain, effectively preventing coupler overload fracture or crush instability, and significantly improving the operational safety of long-formation trains. Secondly, the risk map and track parameters are constructed as conditional tensors and input into a pre-trained conditional denoising generation model. This efficiently generates diverse candidate operation sequences covering traction, electric braking, and air braking states, overcoming the shortcomings of traditional optimization algorithms that are prone to getting trapped in local optima or have excessive computational load, and significantly improving decision quality and real-time computation efficiency. Simultaneously, the execution sequence is selected through strict safety constraint screening and a comprehensive cost minimization criterion, ensuring that operation commands not only meet hard boundaries such as speed limits, signals, and coupler forces, but also taking into account multiple economic objectives such as energy consumption and punctuality. Finally, a single-step execution and rolling time-domain update strategy is adopted to reconstruct the risk map based on the real-time state and regenerate the sequence, forming a closed-loop control architecture of "perception-decision-execution-correction". This architecture can effectively cope with line disturbances and time-varying parameters, greatly enhance the system's adaptability and robustness, and provide reliable support for the smooth, safe and energy-saving operation of heavy-haul trains under complex conditions.

[0066] The technical solution of this application will be further explained below with reference to simulation experimental parameters. This embodiment uses a 20,000-ton heavy-haul combined train (approximately 102 C80 type open wagons, with a train length of approximately 1.5km) as the control object, simulating a long, continuous downhill slope and gradient change points ranging from -8‰ to -20‰. The conditional denoising generation model (based on the Transformer architecture) uses the coupler force risk map as the core condition input, and in the denoising step... Under the condition of generating M≥8 candidate manipulation sequences, the best sequence is selected through a multi-objective evaluation function that includes minimizing energy consumption, on-time rate, and coupler force risk. The core parameters of the specific embodiment are shown in Table 1, and the ablation experiment results are shown in Table 2.

[0067] Table 1 Core parameter settings for the embodiment

[0068] Table 2 Ablation Experiment Results

[0069] The simulation results above show that this application achieves a relatively balanced performance in terms of maximum coupler force, speed tracking error, and number of constraint violations. This is because the coupler force risk map provides feedforward risk information, and the candidate maneuvering sequence screening is further corrected through feedback corrections based on speed limits, coupler force, and maneuvering rate of change constraints; the combined effect of multiple weights allows the scheme to achieve a trade-off between safety and energy consumption.

[0070] Furthermore, this application also utilizes MATLAB's GUI to construct simulation software based on train operation and control methods, the interface of which is as follows: Figure 2 As shown.

[0071] Based on the same inventive concept, this application also provides a train control device for implementing the train control method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more train control device embodiments provided below can be found in the limitations of the train control method described above, and will not be repeated here.

[0072] In one exemplary embodiment, such as Figure 3 As shown, a train operation control device is provided, which includes the following multiple functional modules.

[0073] The data acquisition module 301 is used to acquire the segment vehicle parameters, segment vehicle status information and track parameter information of heavy-haul trains.

[0074] The dynamic prediction module 302 is used to establish a segmented longitudinal dynamic model based on the segmented vehicle parameters, the segmented vehicle state information and the line parameter information, and to determine the coupler force at the connection point of each vehicle segment in the prediction time domain based on the segmented longitudinal dynamic model.

[0075] The risk map construction module 303 is used to construct a coupler force risk map based on the coupler force at the connection points of each vehicle segment in the prediction time domain.

[0076] The conditional diffusion generation module 304 is used to construct a conditional tensor based on the coupler force risk diagram and the track parameters, and input the conditional tensor into a pre-trained conditional denoising generation model to generate multiple candidate manipulation sequences. Each candidate manipulation sequence includes manipulation commands at each time step within the pre-test domain; the manipulation commands include traction level, electric braking level, and air braking status.

[0077] The safety constraint filtering module 305 is used to perform safety constraint filtering on multiple candidate manipulation sequences, retain the candidate manipulation sequences that satisfy all preset safety constraints, and select the candidate manipulation sequence with the lowest overall cost from the candidate manipulation sequences that satisfy the preset safety constraints as the execution sequence.

[0078] The rolling execution module 306 is used to execute the operation command of the first time step in the execution sequence. After one rolling control cycle, it reacquires the segment vehicle status information, reconstructs the coupler force risk diagram, and regenerates and filters the operation sequence to perform closed-loop rolling control on the heavy-haul train.

[0079] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0080] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0081] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0082] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0083] Those skilled in the art will understand that all or part of the processes in 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. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0084] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, etc., and are not limited to these.

[0085] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0086] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A train operation and control method, characterized in that, include: Obtain segment vehicle parameters, segment vehicle status information, and track parameter information for heavy-haul trains; Based on the segmented vehicle parameters, the segmented vehicle status information, and the line parameter information, a segmented longitudinal dynamic model is established, and the coupler force at the connection point of each vehicle segment in the prediction time domain is determined based on the segmented longitudinal dynamic model. Based on the coupler force at the connection points of each vehicle segment within the predicted time domain, a coupler force risk map is constructed. Based on the coupler force risk diagram and the line parameters, a conditional tensor is constructed. The conditional tensor is then input into a pre-trained conditional denoising generation model to generate multiple candidate manipulation sequences. Each candidate manipulation sequence includes manipulation instructions for each time step within the pre-test domain. The manipulation instructions include traction level, electric braking level, and air braking status. Multiple candidate manipulation sequences are screened for safety constraints, and candidate manipulation sequences that satisfy all preset safety constraints are retained. The candidate manipulation sequence with the lowest overall cost among the candidate manipulation sequences that satisfy the preset safety constraints is selected as the execution sequence. The operation command of the first time step in the execution sequence is executed. After one rolling control cycle, the segment vehicle status information is reacquired, the coupler force risk diagram is reconstructed, and the operation sequence is regenerated and filtered to perform closed-loop rolling control on the heavy-haul train.

2. The train operation control method according to claim 1, characterized in that, Based on the segmented vehicle parameters, the segmented vehicle state information, and the track parameter information, a segmented longitudinal dynamic model is established, and the coupler force at the connection point of each vehicle segment in the prediction time domain is determined based on the segmented longitudinal dynamic model, including: Based on the segment vehicle parameters, segment vehicle status information, and line parameter information at the current moment, determine the reference maneuver sequence in the prediction time domain; Based on the reference control sequence and the basic operating resistance, additional slope resistance, additional curve resistance, and adjacent coupler force experienced by each vehicle segment, the longitudinal dynamic equations of each vehicle segment are established. Based on the longitudinal dynamic equations of each vehicle segment, the predicted speed and predicted position of each vehicle segment are calculated recursively step by step in the prediction time domain. Based on the predicted position and speed differences between adjacent vehicle segments, the coupler force at the connection point of each vehicle segment in the prediction time domain is calculated using an equivalent spring-damping model.

3. The train operation control method according to claim 2, characterized in that, The coupler force at each vehicle segment connection point in the prediction time domain is calculated using the following formula: ； ； ； ； in, For the first The vehicle segment and the first The coupling connection between the vehicle segments is at the first... The coupler force at a predicted time step. For the first The vehicle segment and the first The equivalent stiffness coefficient of the coupler buffer at the coupler connection between vehicle sections. For the first The vehicle segment and the first The equivalent damping coefficient of the coupler buffer at the coupler connection between vehicle sections. For the first The vehicle segment and the first The effective relative displacement between vehicle segments For the first The vehicle segment and the first The predicted position difference between vehicle segments For the first The vehicle segment and the first The predicted speed difference between vehicle segments For the first The vehicle segment and the first The balance clearance of the coupler buffer device at the coupler connection between vehicle sections. For the first The vehicle segment and the first The reference spacing between the vehicle segments at the coupler. For the first The vehicle segment is in the first Predicted location at each prediction time step. For the first The vehicle segment is in the first Predicted location at each prediction time step. For the first The vehicle segment is in the first Prediction speed per prediction time step For the first The vehicle segment is in the first Prediction speed per prediction time step.

4. The train operation control method according to claim 1, characterized in that, The coupler force risk map is a normalized risk value matrix at the connection points of each vehicle segment in the prediction time domain, expressed as: ； in, For the first The vehicle segment and the first The coupling connection between the vehicle segments is at the first... The risk value of the coupler force at each predicted time step For the first The vehicle segment and the first The coupling connection between the vehicle segments is at the first... The coupler force at a predicted time step. To allow for a safe threshold for coupler force.

5. The train operation control method according to claim 1, characterized in that, The conditional denoising generation model is described by a conditional inverse denoising process, for the first... The candidate manipulation sequence, from the first Starting with the initial noisy manipulation sequence of each denoising step, the denoising is progressively reduced to step 0: ； in, For the first The candidate manipulation sequence in the ... A noisy manipulation sequence of denoising steps. For the first The candidate manipulation sequence in the ... A noisy manipulation sequence of denoising steps. For the denoising step index, from Decrease to , , and These are the noise scheduling parameters for the diffusion model. For a pre-trained denoising network, For conditional tensors, It is standard Gaussian noise.

6. The train operation control method according to claim 1, characterized in that, The rolling control cycle is adaptively adjusted according to the line risk level: ； in, For rolling control cycle, The rolling cycle of medium- and high-risk areas For the rolling cycle of medium-risk areas, The rolling period in the stable region satisfies , The threshold for determining high risk. The threshold for determining low risk. This is the low speed limit margin threshold. This is the high speed limit margin threshold. For the first The vehicle segment and the first The coupling connection between the vehicle segments is at the first... The risk value of the coupler force at each predicted time step This represents the maximum coupler force risk value in the coupler force risk diagram. For the first Rate limiting margin for each predicted time step, This is for predicting the minimum speed margin in the time domain.

7. A train operation control device, applied to the train operation control method according to any one of claims 1-6, characterized in that, The device includes: The data acquisition module is used to acquire the segment vehicle parameters, segment vehicle status information, and track parameter information of heavy-haul trains. The dynamics prediction module is used to establish a segmented longitudinal dynamics model based on the segmented vehicle parameters, the segmented vehicle state information, and the line parameter information, and to determine the coupler force at the connection point of each vehicle segment in the prediction time domain based on the segmented longitudinal dynamics model. The risk map construction module is used to construct a coupler force risk map based on the coupler force at the connection points of each vehicle segment within the prediction time domain. The conditional diffusion generation module is used to construct a conditional tensor based on the coupler force risk diagram and the line parameters, input the conditional tensor into a pre-trained conditional denoising generation model, and generate multiple candidate manipulation sequences. Each candidate manipulation sequence includes manipulation instructions for each time step in the pre-test domain. The manipulation instructions include traction level, electric braking level, and air braking status. The safety constraint filtering module is used to filter multiple candidate manipulation sequences for safety constraints, retain candidate manipulation sequences that satisfy all preset safety constraints, and select the candidate manipulation sequence with the lowest overall cost from the candidate manipulation sequences that satisfy the preset safety constraints as the execution sequence. The rolling execution module is used to execute the operation command of the first time step in the execution sequence. After one rolling control cycle, it reacquires the segment vehicle status information, reconstructs the coupler force risk diagram, and regenerates and filters the operation sequence to perform closed-loop rolling control on the heavy-haul train.

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the train handling control method according to any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the train operation control method according to any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the train operation control method according to any one of claims 1-6.

Images