A multi-step length parameter self-adjusting MPC train control method

Abstract

The application discloses a kind of multi-step length parameter self-adjusting MPC train control method, it is related to rail transit control technical field, including: with tracking train speed as target 1, with smooth control as target 2, establish the MPC quadratic programming problem with different step length, objective function and constraint condition;Solving MPC quadratic programming problem, obtain the control output of two targets respectively;Select appropriate control output as the final MPC control output;Test and record the model parameters that influence MPC accuracy, according to train operation data, real-time self-adjusting MPC model parameters by table lookup method.The application has good speed tracking control performance;At the same time, the control is smooth, there is no back and forth switching or big and small traction, braking level of traction, braking;It also benefits from the design of parameter self-adjusting, with excellent model accuracy.

Description

Technical Field

[0001] This invention relates to the field of rail transit control technology, specifically to a multi-step parameter self-adjustment MPC train control method. Background Technology

[0002] PB-TACS (Point Based Train Autonomous Control System) is a fifth-generation urban rail transit train control system that enhances the level of autonomous train operation control through reliable environmental perception technology, building upon the fourth-generation FAO (Fully Automatic Operation) urban rail transit train operation control system. With the increased autonomy of the train control system, the importance of automatic driving in operation has become more prominent. At the same time, the generational advancement of the train control system also places higher demands on the performance of automatic driving.

[0003] The core technologies of autonomous driving are the calculation of target speed and the tracking and control algorithms for target speed.

[0004] Traditionally, target speed tracking control algorithms rely on PID (Proportional Integral Derivative) control. This method is simple to implement, requires little computing power, and was the best engineering choice when equipment had limited computing power. However, the PID algorithm still has some problems when performing train target speed tracking control due to its inherent limitations. First, the control effect of the PID algorithm depends on the selection of PID parameters, and in engineering practice, continuous debugging is often required to obtain the best control effect. Second, the PID algorithm inevitably produces output oscillations, affecting control stability. In addition, the PID algorithm cannot solve the problem of energy-saving control. Due to the limitations of the PID algorithm, in order to achieve good control effects for multiple objectives, it is often necessary to superimpose a lot of special processing on the PID algorithm to meet performance requirements such as energy saving and stability, making the software logic complex and disordered, and with poor adaptability.

[0005] Model Predictive Control (MPC) is a feedback control method based on rolling optimization, offering superior control performance compared to the PID algorithm. MPC also has applications in train control, with existing research showing it provides better speed tracking performance than PID. However, MPC still faces some challenges in the target speed tracking problem for trains. First, MPC performance largely depends on the accuracy of the model; overly accurate models can cause computational complexity to exceed equipment performance limits, while directly abstracting parts of the model into fixed parameters leads to poor adaptability. Second, when performing speed tracking, MPC typically considers only a few seconds, while achieving smooth control requires a longer timeframe. Summary of the Invention

[0006] To address the aforementioned shortcomings in existing technologies, this invention provides a multi-step parameter self-adjustment MPC train control method that solves the problems of existing MPC train control technologies being unable to simultaneously achieve real-time speed tracking and stable control, and having high computational complexity in accurate model calculations.

[0007] To achieve the above-mentioned objectives, the technical solution adopted by this invention is as follows:

[0008] A multi-step parameter self-adjusting MPC train control method includes the following steps:

[0009] S1. Taking the train speed as objective 1 and smooth control as objective 2, establish an MPC quadratic programming problem with different step sizes, objective functions and constraints.

[0010] S2. Solve the MPC quadratic programming problem to obtain the control outputs of the two objectives;

[0011] S3. Select a suitable control output as the final MPC control output;

[0012] S4. Test and record the model parameters that affect the accuracy of MPC, and adjust the model parameters of MPC in real time by using a lookup table method based on train operation data.

[0013] Further, step S1 includes the following sub-steps:

[0014] S11. Define state variables and output variables, and establish the original state-space equations describing the motion state of the train.

[0015] S12. Perform zero-order sampling and preservation on the original state-space equations to obtain the sampled state-space equations of the vehicle system.

[0016] S13. Introduce correction feedback into the state-space equation after sampling the vehicle system to obtain the corrected state-space equation;

[0017] S14. Based on the corrected state-space equation, establish the prediction equation to obtain the train's state sequence and output sequence in the prediction time domain.

[0018] S15. For objective 1 and objective 2, set different step sizes and different constraints, and construct their respective objective functions.

[0019] Furthermore, in S11, the state variables are: ,

[0020] in, For state variables, For the train's position, For train speed, For train acceleration, superscript This is a transpose operation;

[0021] The output variable is: ,

[0022] in, For output variables;

[0023] The original state-space equations are: ,

[0024] in, for The differential, Let be the first parameter matrix. For the second parameter vector, For the third parameter vector, To control the output;

[0025] ,

[0026] in, Input inertial element coefficients;

[0027] ,

[0028] in, To output the percentage of the traction level;

[0029] ,

[0030] in, For additional acceleration.

[0031] Furthermore, in S12, the state-space equation of the vehicle system after sampling is: ,

[0032] ,

[0033] in, For the periodic number, for Periodic state variables, for Periodic state variables, for Periodic output variables, This is the first sampling parameter matrix. This is the second sampling parameter vector. For the third sampling parameter vector, for Periodic control output, This is the fourth sampling parameter matrix;

[0034] ,

[0035] ,

[0036] ,

[0037] ,

[0038] in, It is a natural constant. For time, Step size;

[0039] In S13, the corrected state-space equation is: ,

[0040] ,

[0041] in, For correction Periodic state variables, Periodic state variables, periodic state variables, periodic state variables Periodic output variables, for The periodic error correction vector, This is the error correction coefficient matrix;

[0042] ,

[0043] in, The difference between the current actual test location and the predicted location from the previous period. The difference between the current actual test speed and the speed predicted in the previous cycle. This is the difference between the actual acceleration measured at present and the acceleration predicted in the previous cycle.

[0044] ,

[0045] in, This is the position error correction coefficient. This is the speed error correction coefficient. This is the acceleration error correction coefficient.

[0046] Furthermore, in S14, the predicted state sequence of the train in the time domain is as follows: ,

[0047] in, To predict the state sequence of trains in the time domain, For the first parameter sequence, For the second parameter sequence, For the third parameter sequence, To control the output sequence;

[0048] The matrix form is ,in to For correction to Periodic state variables, The matrix form is ,in to for to Periodic control output, Predict step size To control the number of steps;

[0049] ,

[0050] ,

[0051] ,

[0052] In S14, the predicted train output sequence in the time domain is: ,

[0053] in, To predict the train output sequence in the time domain, For the fourth parameter sequence, For the fifth parameter sequence, The sixth parameter sequence;

[0054] The matrix form is , to For correction to Periodic output variables;

[0055] ,

[0056] ,

[0057] .

[0058] Furthermore, in S15, for objective 1, the constraints include input level bit range constraints and input level bit change constraints; for objective 2, only input level bit range constraints are included.

[0059] For objective 1, the step size is set to the calculation cycle of the ATO (Automatic Train Operation) software used; for objective 2, the step size is set to 50 times or an equivalent order of magnitude of the calculation cycle of the ATO software used.

[0060] The objective function for objective 1 is: ,

[0061] The objective function for objective 2 is: ,

[0062] in, Let the objective function be objective function 1. Let the objective function be 2. For the target weight matrix, For the target vector, for OK The identity matrix of columns, To avoid oscillations in target weights;

[0063] , ,

[0064] in, to For the first To the The target position of the cycle, to For the first For the first The target speed of the cycle, to The weights are for positions 1 to p. to The velocity weights are from the 1st to the pth.

[0065] Furthermore, the method in S3 is as follows:

[0066] When the control output of target 1 is negative, if the control output of target 2 is even more negative, then the control output of target 2 is selected as the final MPC control output; otherwise, the control output of target 1 is selected as the final MPC control output.

[0067] When the control output of target 1 is positive, if the control output of target 2 is less than the control output of target 1, then the control output of target 2 is selected as the final MPC control output; otherwise, the control output of target 1 is selected as the final MPC control output.

[0068] Furthermore, the model parameter affecting the accuracy of MPC in S4 is the additional acceleration.

[0069] Furthermore, step S4 includes the following sub-steps:

[0070] S41. Divide the line into sections according to the characteristics of the line the train travels on;

[0071] S42. Conduct a train traction test and obtain the speed curve of the train from zero speed to maximum speed under full traction on a flat slope;

[0072] S43. Construct and initialize the additional acceleration information table:

[0073] The initial value of the additional acceleration information table is set to 0, and it has two query dimensions: velocity range and position range.

[0074] The speed ranges correspond to different speed ranges in the S42 speed curve;

[0075] The location interval corresponds to the line segment of S41;

[0076] S44. Before the train is put into operation, the additional acceleration information table is updated based on the speed curve described in S42 through a test run.

[0077] S45. During train operation, if a deterioration in the MPC control effect is detected, the additional acceleration information table is queried based on the train's current speed range and position range, and the additional acceleration of the MPC is corrected.

[0078] Furthermore, S44 includes the following sub-steps:

[0079] S441. Differentiate the current speed of the train to obtain the current longitudinal acceleration of the train;

[0080] S442. Calculate the current control acceleration of the train according to the following formula: ,

[0081] in, The current control acceleration of the train, To output the percentage of the traction level, The current speed of the train. The traction acceleration with velocity as the independent variable in the S42 velocity curve;

[0082] S443. Based on the train's current speed and track segmentation, fill in the acceleration values ​​in the corresponding speed and position intervals in the additional acceleration information table. ,in This is the longitudinal acceleration.

[0083] The beneficial effects of this invention are as follows: This invention has good speed tracking control performance; at the same time, the control is smooth, without back-and-forth switching of traction and braking or sudden changes in traction and braking levels; it also benefits from the parameter self-adjustment design, and has excellent model accuracy. Attached Figure Description

[0084] Figure 1 This is a flowchart of a multi-step parameter self-adjusting MPC train control method according to an embodiment of the present invention. Detailed Implementation

[0085] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0086] like Figure 1 As shown, a multi-step parameter self-adjusting MPC train control method includes the following steps:

[0087] S1. Taking the train speed as objective 1 and smooth control as objective 2, establish an MPC quadratic programming problem with different step sizes, objective functions and constraints.

[0088] S1 includes the following steps:

[0089] S11. Define state variables and output variables, and establish the original state-space equations describing the motion state of the train.

[0090] The equation of motion for the train is:

[0091] ,

[0092] in, For the train's position, For train speed, To accelerate the train, for The derivative, for The derivative, for Derivative; This represents the acceleration corresponding to external forces other than traction and braking (i.e., additional acceleration). In the prediction time domain, the train's speed and gradient do not change significantly, therefore the change in additional acceleration is also small. To reduce computational complexity, let... It is a constant; Assuming the input inertial element coefficients are used, the input-to-response model is considered to be a first-order inertial element model, meaning the input needs to pass through... The response takes time to fully materialize, and the rate of change of the response remains constant during the process. To control the output.

[0093] In this embodiment, the state variable is defined as: ,

[0094] in, State variables, superscript This is for the transpose operation.

[0095] It is worth noting that in this article, all superscripts... All of these are transpose operations.

[0096] Since the control objective of the train is to track the target position-velocity curve, the output variable is defined as: ,

[0097] in, For output variables.

[0098] The original state-space equations are as follows: ,

[0099] in, for The differential, Let be the first parameter matrix. The second parameter vector, The third parameter vector, To control the output;

[0100] ,

[0101] in, Input the inertial element coefficient;

[0102] ,

[0103] in, To output the percentage of the traction level;

[0104] .

[0105] S12. Perform zero-order sampling and preservation on the original state-space equations to obtain the sampled state-space equations of the vehicle system.

[0106] The sampling of the train state by the onboard system is equivalent to zero-order hold sampling. In this embodiment, the state-space equation of the onboard system after sampling is as follows:

[0107] ,

[0108] ,

[0109] in, For the periodic number, for Periodic state variables, for Periodic state variables, for Periodic output variables, This is the first sampling parameter matrix. This is the second sampling parameter vector. For the third sampling parameter vector, for Periodic control output, This is the fourth sampling parameter matrix;

[0110] ,

[0111] ,

[0112] ,

[0113] ,

[0114] in, It is a natural constant. For time, The step size.

[0115] S13. Introduce correction feedback into the state-space equation after sampling the vehicle system to obtain the corrected state-space equation.

[0116] Due to model and measurement errors, there is a discrepancy between the train state predicted by the state-space equations and the actual state. To reduce this discrepancy, correction feedback is introduced to revise the state-space equations. The corrected state-space equations are as follows:

[0117] ,

[0118] ,

[0119] in, For correction Periodic state variables, Periodic state variables, periodic state variables, periodic state variables Periodic output variables, for The periodic error correction vector, This is the error correction coefficient matrix;

[0120] ,

[0121] in, The difference between the current actual test location and the predicted location from the previous period. The difference between the current actual test speed and the speed predicted in the previous cycle. This is the difference between the actual acceleration measured at present and the acceleration predicted in the previous cycle.

[0122] ,

[0123] in, This is the position error correction coefficient. This is the speed error correction coefficient. This is the acceleration error correction coefficient.

[0124] S14. Based on the corrected state-space equation, establish the prediction equation to obtain the train's state sequence and output sequence in the prediction time domain.

[0125] Let the prediction step size be Control step size is , All texts in the entire text , and or Expressions with addition and subtraction operations indicated by the superscript are all exponentiation operations.

[0126] The predicted state sequence of the train in the time domain is as follows: ,

[0127] in, To predict the state sequence of trains in the time domain, For the first parameter sequence, For the second parameter sequence, For the third parameter sequence, To control the output sequence;

[0128] The matrix form is ,in to For correction to Periodic state variables, The matrix form is ,in to for to Periodic control output, Predict step size To control the number of steps;

[0129] ,

[0130] ,

[0131] ,

[0132] The predicted train output sequence in the time domain is as follows: ,

[0133] in, To predict the train output sequence in the time domain, For the fourth parameter sequence, For the fifth parameter sequence, It is the sixth parameter sequence;

[0134] The matrix form is , to For correction to Periodic output variables;

[0135] ,

[0136] ,

[0137] .

[0138] For objective 1 and objective 2, different step sizes and different constraints are set, and their respective objective functions are constructed.

[0139] Train speed control must take into account multiple objectives:

[0140] Objective 1: Track the train speed, that is, control the actual speed of the train to match the target speed;

[0141] Objective 2: Smooth control, meaning there should be no back-and-forth switching of traction and braking or sudden changes in traction and braking levels.

[0142] Objective 1 prioritizes real-time performance, requiring the use of the minimum step size for computation, i.e., the step size. The computation cycle of the ATO software is taken into account; Objective 2 requires computation within a larger time domain, necessitating the use of a larger step size. Use an ATO computation cycle that is 50 times or equivalent in length. For targets with different cycle lengths, establish and solve the objective functions separately, and then synthesize the results.

[0143] The objective function for objective 1 is defined as: ,in For the target weight matrix, The target vector.

[0144] , ,

[0145] in, to For the first To the The target position of the cycle, to For the first For the first The target speed of the cycle, to The weights are for positions 1 to p. to The velocity weights are from the 1st to the pth.

[0146] Expand ,get , To accumulate the sequence numbers, substituting them into the prediction equation described in S14, and omitting terms irrelevant to the input, we obtain: ,

[0147] Right now Let be the objective function for objective 1.

[0148] Objective 2 requires controlling the train speed to closely approach the target speed while avoiding input oscillations, i.e. It should be as small as possible, among which This represents the average value of the input oscillations, which is set to 0 for model simplification. Adding this objective to the objective function of objective 1 yields the objective function of objective 2: ,

[0149] Right now The objective function for objective 2; for OK The identity matrix, i.e. , Target weights to avoid oscillations.

[0150] In this embodiment, for objective 1, the constraints include input level range constraints and input level change constraints; for objective 2, only input level range constraints are included.

[0151] When using ATO (Automatic Train Control) to control the vehicle, the following restrictions apply to the input level:

[0152] Constraint 1: The input level bit (MPC control output) cannot exceed a certain range, i.e. , To control the minimum output value, To control the maximum limit of the output;

[0153] Constraint 2: The input level bit (MPC control output) change cannot exceed a certain value, i.e. and , Let be any meaningful positive integer index. For the limit of variation, This represents the control output value calculated by the ATO software in the previous cycle.

[0154] Transform the above constraints into matrix form.

[0155] Since objective 1 considers both constraint 1 and constraint 2, the constraint expression of objective 1 is in matrix form as follows: ,in , , , , , and These are all intermediate matrices and vectors involved in the calculations. In this embodiment, they will not be given specific names. The expressions for these quantities are defined as follows:

[0156] ,

[0157] ,

[0158] ,

[0159] ,

[0160] ,

[0161] ,

[0162]

[0163] Considering only constraint 1 for objective 2, the constraint expression for objective 2 is in matrix form as follows:

[0164] , and The defined expression is:

[0165] ,

[0166] .

[0167] S2. Solve the MPC quadratic programming problem to obtain the control outputs of the two objectives.

[0168] In this embodiment, the quadratic programming solver (QuadProg++) is used to solve the following MPC quadratic programming problem.

[0169] The solution to objective 1 can be expressed as: , st That is, in In the bundle, minimize the objective function of objective 2.

[0170] S3. Select a suitable control output as the final MPC control output.

[0171] When the control output of target 1 is negative, that is, when braking is required to ensure that the speed can track the target speed, if the control output of target 2 is even more negative, then the control output of target 2 is selected as the final MPC control output; otherwise, the control output of target 1 is selected as the final MPC control output.

[0172] When the control output of target 1 is positive, meaning that traction is required to ensure that the speed can track the target speed, if the control output of target 2 is less than the control output of target 1, then the control output of target 2 is selected as the final MPC control output; otherwise, the control output of target 1 is selected as the final MPC control output.

[0173] S4. Test and record the model parameters that affect the accuracy of MPC, and adjust the model parameters of MPC in real time by using a lookup table method based on train operation data.

[0174] In S4, the model parameter affecting the accuracy of MPC is the additional acceleration, i.e., the aforementioned... .

[0175] Influence Many factors influence the speed of a vehicle, including wind resistance, gradient, and vehicle mechanical structure, making it impossible to calculate using a simple linear formula. To address this problem, this invention employs a lookup table method to directly query pre-stored... It is updated in real time based on train running data.

[0176] S4 includes the following steps:

[0177] S41. Based on the characteristics of the train's route, such as the gradient, curves, and whether it is a tunnel, the line is divided into sections to affect the train's acceleration.

[0178] S42. Conduct a train traction test to obtain the speed curve of the train being pulled from zero speed to maximum speed on a flat slope.

[0179] S43. Construct and initialize an additional acceleration information table in the ferroelectric memory of the vehicle system;

[0180] The initial value of the additional acceleration information table is set to 0, and it has two query dimensions: velocity range and position range.

[0181] The speed ranges correspond to different speed ranges in the S42 speed curve;

[0182] The location interval corresponds to the line segment of S41.

[0183] S44. Before the train is put into operation, the additional acceleration information table is updated based on the speed curve described in S42 through a test run.

[0184] S44 includes the following steps:

[0185] S441. By differentiating the current speed of the train using ATO software, the current longitudinal acceleration of the train is obtained.

[0186] S442. Calculate the current control acceleration of the train according to the following formula: ,in, The current control acceleration of the train, To output the percentage of the traction level, The current speed of the train. The traction acceleration with velocity as the independent variable in the S42 velocity curve;

[0187] S443. Based on the train's current speed and track segmentation, fill in the acceleration values ​​in the corresponding speed and position intervals in the additional acceleration information table. ,in This is the longitudinal acceleration.

[0188] S45. During train operation, if a deterioration in the MPC control effect is detected, the additional acceleration information table is queried based on the train's current speed range and position range, and the additional acceleration of the MPC is corrected.

[0189] In summary, this invention has good speed tracking control performance; at the same time, the control is smooth, without back-and-forth switching of traction and braking or sudden changes in traction and braking levels; it also benefits from the parameter self-adjustment design, resulting in excellent model accuracy.

[0190] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A multi-step parameter self-adjusting MPC train control method, characterized in that, Includes the following steps: S1. Taking the train speed as objective 1 and smooth control as objective 2, establish an MPC quadratic programming problem with different step sizes, objective functions and constraints. S2. Solve the MPC quadratic programming problem to obtain the control outputs of the two objectives; S3. Select the control output as the final MPC control output; S4. Test and record the model parameters that affect the accuracy of MPC, and adjust the model parameters of MPC in real time by using a lookup table method based on train operation data. S1 includes the following steps: S11. Define state variables and output variables, and establish the original state-space equations describing the motion state of the train. S12. Perform zero-order sampling and preservation on the original state-space equations to obtain the sampled state-space equations of the vehicle system. S13. Introduce correction feedback into the state-space equation after sampling the vehicle system to obtain the corrected state-space equation; S14. Based on the corrected state-space equation, establish the prediction equation to obtain the train's state sequence and output sequence in the prediction time domain. S15. For objective 1 and objective 2, set different step sizes and different constraints, and construct their respective objective functions; In S14, the predicted state sequence of the train in the time domain is as follows: ， in, To predict the state sequence of trains in the time domain, For the first parameter sequence, For the second parameter sequence, For the third parameter sequence, To control the output sequence, for Periodic state variables; The matrix form is ,in to For correction to Periodic state variables, The matrix form is ,in to for to Periodic control output, To predict the step size, To control the number of steps; In S14, the predicted train output sequence in the time domain is: ， in, To predict the train output sequence in the time domain, For the fourth parameter sequence, For the fifth parameter sequence, The sixth parameter sequence; The matrix form is , to For correction to Periodic output variables; In S15, for objective 1, the constraints include input level bit range constraints and input level bit change constraints; for objective 2, only input level bit range constraints are included. For objective 1, the step size is set to the calculation cycle of the ATO software used; for objective 2, the step size is set to 50 times or the same order of magnitude as the calculation cycle of the ATO software used. The objective function for objective 1 is: ， The objective function for objective 2 is: ， in, Let the objective function be objective function 1. Let the objective function be 2. For the target weight matrix, For the target vector, for OK The identity matrix of columns, To avoid oscillations in the target weight, the superscript... This is for the transpose operation.

2. The MPC train control method with multi-step parameter self-adjustment according to claim 1, characterized in that, In S11, the state variables are: ， in, For state variables, For the train's position, For train speed, For train acceleration, superscript This is a transpose operation; The output variable is: ， in, For output variables; The original state-space equations are: ， in, for The differential, Let be the first parameter matrix. For the second parameter vector, For the third parameter vector, To control the output; ， in, Input inertial element coefficients; ， in, This represents the percentage of the output traction level. ， in, For additional acceleration.

3. The MPC train control method with multi-step parameter self-adjustment according to claim 2, characterized in that, In S12, the state-space equation of the vehicle system after sampling is: ， ， in, For the periodic number, for Periodic state variables, for Periodic state variables, for Periodic output variables, This is the first sampling parameter matrix. This is the second sampling parameter vector. For the third sampling parameter vector, for Periodic control output, This is the fourth sampling parameter matrix; ， ， ， ， in, It is a natural constant. For time, Step size; In S13, the corrected state-space equation is: ， ， in, For correction Periodic state variables, For correction Periodic output variables, for The periodic error correction vector, This is the error correction coefficient matrix; ， in, This is the difference between the current actual test location and the predicted location from the previous period. The difference between the current actual test speed and the speed predicted in the previous cycle. This is the difference between the actual acceleration measured at present and the acceleration predicted in the previous cycle. ， in, This is the position error correction coefficient. This is the speed error correction coefficient. This is the acceleration error correction coefficient.

4. The MPC train control method with multi-step parameter self-adjustment according to claim 1, characterized in that, The method described in S3 is as follows: When the control output of target 1 is negative, if the control output of target 2 is even more negative, then the control output of target 2 is selected as the final MPC control output; otherwise, the control output of target 1 is selected as the final MPC control output. When the control output of target 1 is positive, if the control output of target 2 is less than the control output of target 1, then the control output of target 2 is selected as the final MPC control output; otherwise, the control output of target 1 is selected as the final MPC control output.

5. The MPC train control method with multi-step parameter self-adjustment according to claim 1, characterized in that, The model parameter in S4 that affects the accuracy of MPC is the additional acceleration.

6. The MPC train control method with multi-step parameter self-adjustment according to claim 5, characterized in that, S4 includes the following sub-steps: S41. Divide the line into sections according to the characteristics of the line the train travels on; S42. Conduct a train traction test and obtain the speed curve of the train from zero speed to maximum speed under full traction on a flat slope; S43. Construct and initialize the additional acceleration information table: The initial value of the additional acceleration information table is set to 0, and it has two query dimensions: velocity range and position range. The speed ranges correspond to different speed ranges in the S42 speed curve; The location interval corresponds to the line segment of S41; S44. Before the train is put into operation, the additional acceleration information table is updated based on the speed curve described in S42 through a test run. S45. During train operation, if a deterioration in the MPC control effect is detected, the additional acceleration information table is queried based on the train's current speed range and position range, and the additional acceleration of the MPC is corrected.

7. The MPC train control method with multi-step parameter self-adjustment according to claim 6, characterized in that, S44 includes the following sub-steps: S441. Differentiate the current speed of the train to obtain the current longitudinal acceleration of the train; S442. Calculate the current control acceleration of the train according to the following formula: ， in, The current control acceleration of the train, To output the percentage of the traction level, The current speed of the train. The traction acceleration with velocity as the independent variable in the S42 velocity curve; S443. Based on the train's current speed and track segmentation, fill in the acceleration values ​​in the corresponding speed and position intervals in the additional acceleration information table. ,in This is the longitudinal acceleration.

Images