Improved Active Disturbance Rejection Control Method for Virtual Marshalling of High-Speed ​​Trains Based on GRO Algorithm

Abstract

The present invention relates to the field of high-speed train control, and in particular to an improved virtual marshaling anti-disturbance control method for high-speed trains based on a GRO algorithm, including: establishing a dynamic model of a virtual marshaling high-speed train; establishing an ADRC controller, which includes: a tracking differentiator, a nonlinear state error feedback control law, and an extended state observer; and using the fastest comprehensive function in the tracking differentiator to establish a fastest feedback system; after completing population initialization and setting a fitness function, using the GRO algorithm to implement self-tuning of five parameters to be tuned of the ADRC controller to obtain a GRO-ADRC control model; setting line parameters and the initial state of a virtual marshaling high-speed train, obtaining the actual vehicle distance during the operation of the dynamic model of the virtual marshaling high-speed train, and inputting the actual vehicle distance into the GRO-ADRC control model to output a control torque for control. The method can ensure the stable running of virtual marshaling high-speed trains.

Description

Technical Field

[0001] The invention relates to the field of high-speed train control, and in particular to an improved GRO algorithm-based high-speed train virtual marshaling anti-disturbance control method. Background Art

[0002] With the growing demand for rail transit, how to improve the utilization rate of existing lines and vehicles within limited costs, resources and environment is an urgent problem to be solved. Virtual marshaling technology can achieve precise adjustment of transportation capacity by optimizing the use of vehicle resources to cope with the uneven distribution of intercity passenger flow, which not only improves the quality of operating services but also reduces operating costs. Ensuring the stable operation of virtual marshaling trains is the foundation for realizing the application of virtual coupling technology, and this field urgently needs in-depth research and exploration. Due to the different braking performance of the power unit and the uncertainty of external line conditions, it is difficult for virtual marshaling trains to maintain an ideal state when encountering interference, which limits the actual effectiveness of the technology.

[0003] In the face of complex interference, there is an urgent need for an improved GRO algorithm-based high-speed train virtual marshaling anti-disturbance control method to ensure the stable operation of virtual marshaling high-speed trains, thereby improving the utilization efficiency of track and vehicle resources. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide an improved high-speed train virtual marshaling anti-disturbance control method based on the GRO algorithm.

[0005] To solve the above problems, the technical solution adopted by the present invention is:

[0006] An improved GRO algorithm-based high-speed train virtual marshaling anti-disturbance control method, the method comprising:

[0007] Step 1: Establish a dynamic model of a virtual marshaling high-speed train, which includes a dynamic model of a single power unit of a 4-section virtual marshaling train;

[0008] Step 2: Establish an ADRC controller, which includes: a tracking differentiator, a nonlinear state error feedback control law and an extended state observer; and the tracking differentiator uses the fastest synthesis function fhan( x 1 ,x 2 ,r 0 ,h 0) Establish the fastest feedback system;

[0009] Step 3: After completing population initialization and fitness function setting, the GRO algorithm is used to self-tune the parameters to be adjusted of the ADRC controller to obtain the GRO-ADRC short-distance coupling tracking cruise control model, wherein the parameters to be adjusted include 5, which are the system gain in the nonlinear state error feedback control law β 1 and β 2, and the system gain parameter in the extended state observer β 01 , β 02 , β 03 ;

[0010] Step 4: Set the line parameters and the initial state of the virtual marshaling high-speed train, obtain the actual vehicle distance during the operation of the virtual marshaling high-speed train, input the actual vehicle distance into the GRO-ADRC short-spacing coupling tracking cruise control model to output the control torque for control.

[0011] As an implementation mode of the present invention, in step 1, a dynamic model of a virtual marshaling single-section power unit is established in the following manner:

[0012] Step S101, establishing a topological diagram of a dynamic model of a single power unit, and determining suspension parameters and structural parameters of the dynamic model; wherein the topological diagram includes: 1 vehicle body, 2 frames, 4 wheel sets and 8 axle boxes, and the vehicle body, each frame, and each wheel set have 6 degrees of freedom, namely, lateral, vertical, longitudinal, roll, nod, and shake, and each axle box has only 1 degree of freedom, namely, nod;

[0013] Step S102, according to the multi-rigid body dynamics theory, using the suspension parameters and structural parameters, assembling each structure into a power unit bogie subsystem; wherein the frame and the axle box are connected through a swing arm node;

[0014] Step S103: For the vertical damper, lateral damper and anti-snaking damper in the power unit bogie subsystem, nonlinear modeling is performed according to the relationship curve between their displacement velocity and force; for other mechanical elements, they are directly simplified into linear force elements;

[0015] Step S104, in combination with the body parameters of the virtual marshaling high-speed train, the front bogie on the outside of the power unit bogie subsystem is adjusted to construct the rear bogie; then, the front bogie and the rear bogie are connected to the body model through the secondary suspension system, and the assembly of the whole vehicle is completed to obtain the dynamic model of the virtual marshaling single-section power unit.

[0016] As an implementation mode of the present invention, the fastest comprehensive function fhan(x 1 ,x 2 ,r 0 ,h 0) is as follows:

[0017] ;

[0018] in, x 1. x 2 are the target workshop distances x The transition and differential values ​​of r 0 、h 0 is the function control parameter, r The value range of 0 is [10,100]; d , a 0,y, a 1. a 2. s y , a , s a is the function fhan( x 1 ,x 2 ,r 0 ,h 0) Derived quantity;

[0019] The expression of the fastest feedback system is as follows:

[0020] ;

[0021] in, h is the sampling step length.

[0022] As an implementation mode of the present invention, the expression of the nonlinear state error feedback control law is as follows:

[0023] ;

[0024] in, e 1 and e 2 is the difference between the transition value and the observed value; u 0 is the initial control amount; z 1 is the observed displacement, z 2 is the observed speed; β 1 and β 2 is the system gain; α 1The value range is [0,1], α 2>1;fal( e,α,δ ) is a nonlinear saturation function, and its expression is as follows:

[0025] ;

[0026] in,e is the error signal; α is a controller parameter with a value range of [0,1], used to reflect the fal( e,α,δ ) Whether the function takes non-smooth feedback; δ is the width of the nonlinear interval;

[0027] The disturbance compensation forms the control quantity, which is the control torque input to the train. u The expression is as follows:

[0028] ;

[0029] in, z 3 is the observed disturbance, b 0 is the compensation factor.

[0030] As an implementation mode of the present invention, the expression of the extended state observer is as follows:

[0031] ;

[0032] in, β 01 , β 02 , β 03 is the system gain parameter; h is the sampling step length; α 3 and α The value range of 4 is [0,1].

[0033] As an implementation mode of the present invention, in step 3, the population initialization is completed by using the following steps:

[0034] A sufficient set of parameters is randomly generated according to the parameters of the ADRC controller as an initial population, which is represented by an M×N matrix as shown below:

[0035] ;

[0036] Wherein, M is the size of the population particles; N is the dimension of the particles. Since the parameters to be adjusted include 5, N=5.

[0037] As an implementation mode of the present invention, in step 3, the fitness function is expressed as follows:

[0038] ;

[0039] in, e ( t ) is the system error; μ ( t ) is the output of the system; a (t ) is the acceleration of a single power unit of the virtual marshaling; ω 1. ω 2. ω 3 is the weight coefficient.

[0040] As an implementation mode of the present invention, a method for self-tuning the parameters to be tuned of the ADRC controller using the GRO algorithm includes:

[0041] (1) Initialize the population particles and determine the size of the population particles M and the maximum number of iterations T max ;

[0042] (2) Calculate the fitness value of the current population particles at the new position, update the position of the current population particles according to the equation in the GRO algorithm, and update the optimal position of the particles at the same time;

[0043] The equation in the GRO algorithm is: ;

[0044] in, X i For gold prospectors i location; t is the current iteration number; Xnew i For gold prospectors i The latest location of f (.) is the fitness function;

[0045] (3) Update according to the following formula l 1 and l 2:

[0046] ;

[0047] ;

[0048] in, l 1 and l 2 are convergent components;

[0049] (4) Calculate the next position of the population particle according to one of the following three formulas:

[0050] ; ;

[0051] ;

[0052] in, X r For randomly selected current surveyors r location;r 1 is a random vector with a value range of [0,1];

[0053] D 1 , A 1 , D 2 , A 2 , D 3 It can be expressed by the following formula:

[0054] ; ;

[0055] ; ;

[0056] ;

[0057] in C 1=2 r 2, r 2 is a random vector with a value range of [0,1]; X g1 , X g2 are two randomly selected gold prospectors; D3 is the collaboration vector; X * is the best gold mine location;

[0058] (5) Update the fitness value and the best population particle position. If the termination condition is met, output the result; if not, return to step (2) to continue the iterative search.

[0059] The beneficial effects of adopting the above technical solution are:

[0060] The invention provides an improved GRO algorithm-based high-speed train virtual marshaling anti-disturbance control method. After establishing a virtual marshaling high-speed train dynamics model including a dynamics model of a single power unit of a 4-unit virtual marshaling, an ADRC controller is established. The ADRC controller includes: a tracking differentiator, a nonlinear state error feedback control law and an extended state observer; and the tracking differentiator uses the fastest comprehensive function fhan( x 1 ,x 2 ,r 0 ,h0) establish the fastest feedback system; then, after completing the population initialization and the fitness function setting, the GRO algorithm is used to realize the self-tuning of the parameters to be tuned of the ADRC controller to obtain the GRO-ADRC short-distance coupling tracking cruise control model, wherein the parameters to be tuned include 5, which are the system gain in the nonlinear state error feedback control law β 1 and β 2, and the system gain parameter in the extended state observer β 01 , β 02 , β 03 ; Finally, the line parameters and the initial state of the virtual marshaling high-speed train are set, the actual vehicle distance during the operation of the virtual marshaling high-speed train is obtained, and the actual vehicle distance is input into the GRO-ADRC short-spacing coupling tracking cruise control model to output the control torque for control.

[0061] The improved GRO algorithm-based high-speed train virtual marshaling anti-disturbance control method provided by the present invention is based on the theory of vehicle motion equations, combined with vehicle dynamics and structural parameters, and a dynamic model of a virtual marshaling high-speed train is established, which provides a research basis for the subsequent research on the GRO-ADRC short-spacing coupling tracking cruise control model; in addition, the ADRC controller is improved by the GRO algorithm, and the operation of a single power unit of a four-section virtual marshaling under straight and curved line conditions is simulated, and the control effect is evaluated from four dimensions of speed deviation, overshoot, acceleration and its rate of change, which further confirms that the GRO-ADRC short-spacing coupling tracking cruise control model can meet the control requirements of the high-speed train virtual marshaling system and achieve fast and accurate tracking of the target spacing curve. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is a flow chart of a high-speed train virtual marshaling self-disturbance rejection control method based on an improved GRO algorithm provided in an embodiment of the present invention.

[0063] Figure 2 It is a topological diagram of a dynamic model of a single-section power unit provided in an embodiment of the present invention.

[0064] Figure 3 It is a structural schematic diagram of a bogie subsystem provided in an embodiment of the present invention.

[0065] Figure 4 It is a damping force and speed characteristic curve diagram provided by an embodiment of the present invention.

[0066] Figure 5 It is a lateral stop force and displacement characteristic curve diagram provided by an embodiment of the present invention.

[0067] Figure 6 It is a schematic diagram of a dynamic model of a virtual marshaling single-section power unit provided in an embodiment of the present invention.

[0068] Figure 7 It is a structural block diagram of an ADRC parameter self-tuning structure based on a GRO algorithm provided in an embodiment of the present invention.

[0069] Figure 8 This is the first straight-line coupling tracking operation result of a following power unit provided by an embodiment of the present invention; wherein, (a) the vehicle distance-time curve, (b) the speed-time curve, (c) the acceleration-time curve, and (d) the acceleration time-varying rate-time curve.

[0070] Fig. 9 It is the third straight-line coupling tracking operation result of the following power unit provided by an embodiment of the present invention; (a) vehicle distance-time curve, (b) speed-time curve, (c) acceleration-time curve, (d) acceleration time-varying rate-time curve.

[0071] Fig.10 It is the first power unit curve-linked tracking operation result provided by an embodiment of the present invention; (a) vehicle distance-time curve, (b) speed-time curve, (c) acceleration-time curve, (d) acceleration time-varying rate-time curve.

[0072] Fig.11 It is the third power unit curve-linked tracking operation result provided by an embodiment of the present invention; (a) vehicle distance-time curve, (b) speed-time curve, (c) acceleration-time curve, (d) acceleration time-varying rate-time curve.

[0073] Among them: 100-bogie subsystem, 1-air spring, 2-frame, 3-anti-snaking shock absorber, 4-primary vertical shock absorber, 5-secondary lateral shock absorber. DETAILED DESCRIPTION

[0074] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention is clearly and completely described below in conjunction with specific embodiments.

[0075] The embodiment of the present invention provides a high-speed train virtual marshaling self-disturbance rejection control method based on an improved GRO algorithm, such as Figure 1 As shown, the method includes:

[0076] Step 1: Establish a dynamic model of a virtual marshaling high-speed train, which includes a dynamic model of a single power unit of a 4-section virtual marshaling train;

[0077] In this step, the dynamic model of the virtual marshaling single-section power unit is established in the following way:

[0078] The present invention establishes a power unit with a high-speed train as a power parameter. For each component in the vehicle multi-rigid body model, assuming that the component has n degrees of freedom, the expression of its dynamic calculation differential equation is as follows:

[0079] ;

[0080] in, M ij is the mass matrix; C ij is the damping matrix; K ij is the stiffness matrix; x n is the state vector of the system; F n ( t ) is the external excitation force vector.

[0081] Step S101, establish a topological diagram of the dynamic model of a single power unit, and determine the suspension parameters and structural parameters of the dynamic model; wherein the topological diagram includes: 1 car body, 2 frames (frames I and II), 4 wheel sets (wheel sets I, II, III, IV) and 8 axle boxes (axle box I-axle box VIII), and the car body, each frame, and each wheel set have 6 degrees of freedom, namely, lateral, vertical, longitudinal, roll, nod, and shake, and each axle box has only 1 degree of freedom, namely, nod. The topological diagram is as follows: Figure 2 As shown;

[0082] The suspension parameters and structural parameters are shown in Table 1 and Table 2 respectively.

[0083] Table 1 Suspension parameters of the dynamic model of a single power unit

[0084]

[0085] Table 2 Structural parameters of the dynamic model of a single power unit

[0086]

[0087] Step S102: according to the multi-rigid body dynamics theory, using the suspension parameters and structural parameters, assemble each structure into a power unit bogie subsystem 100; wherein the frame and the axle box are connected through a swing arm node to achieve a positioning auxiliary effect;

[0088] The structure can be obtained from Table 1 and Table 2. The structure of the bogie subsystem 100 finally obtained is as follows: Figure 3As shown, it includes: an empty spring 1, a frame 2, an anti-snake shock absorber 3, a first-order vertical shock absorber 4, and a second-order lateral shock absorber 5.

[0089] Step S103: For the vertical damper, lateral damper and anti-snaking damper in the power unit bogie subsystem 100, nonlinear modeling is performed according to the relationship curve between their displacement velocity and force; for other mechanical elements, they are directly simplified into linear force elements;

[0090] The above displacement velocity and force relationship curve can be obtained through experiments, such as the damping force and velocity characteristic curve as shown in Figure 4 As shown in the figure, the lateral stop force and displacement characteristic curve is as follows Figure 5 shown.

[0091] Step S104, in combination with the body parameters of the virtual high-speed train, the front bogie on the outside of the power unit bogie subsystem 100 is adjusted to construct the rear bogie; then, the front bogie and the rear bogie are connected to the body model through the secondary suspension system, and the assembly of the whole vehicle is completed to obtain the dynamic model of the virtual single-section power unit.

[0092] The body parameters of the high-speed train can be obtained based on the model of the high-speed train, and the dynamic model of the virtual marshaling single-section power unit is obtained as follows: Figure 6 shown.

[0093] Afterwards, a dynamic model of a virtual marshaling high-speed train is established according to the dynamic model of a single power unit of a four-section virtual marshaling. The train can be a leading-car-following topology structure (which includes a leading power unit and three following power units). The dynamic models of any two adjacent single power units of the virtual marshaling are communicated and connected, and are separated by a certain distance (i.e., the "vehicle distance" described later).

[0094] Step 2: Establish an ADRC controller, which includes: a tracking differentiator, a nonlinear state error feedback control law and an extended state observer; and the tracking differentiator uses the fastest synthesis function fhan( x 1 ,x 2 ,r 0 ,h 0) Establish the fastest feedback system.

[0095] The ADRC controller can be established in combination with the virtual marshaling high-speed train dynamics model. The tracking differentiator, the nonlinear state error feedback control law and the extended state observer are respectively described below.

[0096] 1. Tracking Differentiator

[0097] The tracking-differentiator is also called the fastest tracking differentiator, which aims to balance the conflict between the overshoot and response speed of the system. The error between the target vehicle distance and the actual vehicle distance and its error differential are set as signal input. y ( t )for:

[0098] ;

[0099] in, S i ( t )and S i-1 ( t ) are respectively the displacements of the current virtual marshaling single-section power unit and the virtual marshaling single-section power unit in front of it;

[0100] Using the fastest synthesis function fhan( x 1 ,x 2 ,r 0 ,h 0), its expression is as follows:

[0101] ;

[0102] in, x 1. x 2 are the target workshop distances x The transition and differential values ​​of r 0 、h 0 is the function control parameter, r 0 value range is [10,100], h 0 is consistent with the simulation step size; d , a 0,y, a 1. a 2. s y , a , s a is the function fhan( x 1 ,x 2 ,r 0 ,h 0) Derived quantity;

[0103] The expression of the fastest feedback system is as follows:

[0104] ;

[0105] in, h is the sampling step length.

[0106] 2. Nonlinear state error feedback control law

[0107] A nonlinear combination strategy in the form of PD is used to combine various control links. The expression of the nonlinear state error feedback control law is as follows:

[0108] ;

[0109] in, e 1 and e 2 is the difference between the transition value and the observed value; u 0 is the initial control amount; z 1 is the observed displacement, z 2 is the observed speed; β 1 and β 2 is the system gain; α 1The value range is [0,1], α 2>1;

[0110] fal( e,α,δ ) is a nonlinear saturation function, and its expression is as follows:

[0111] ;

[0112] in, e is the error signal; α is a controller parameter with a value range of [0,1], used to reflect the fal( e,α,δ ) function whether to take non-smooth feedback, δ is the width of the nonlinear interval;

[0113] The disturbance compensation forms the control quantity, which is the control torque input to the train. u The expression is as follows:

[0114] ;

[0115] in, z 3 is the observed disturbance, b 0 is the compensation factor.

[0116] 3. Extended State Observer

[0117] The nonlinear extended state observer, the core of ADRC, mainly focuses on the impact of sudden changes in system parameters and external disturbances on the system, and regards these impacts as the total disturbance of the system and as new state variables. Then the input and output of the controlled object are accepted as the input of the extended observer, and all state variables of the control system are estimated through the state observer.

[0118] In the present invention, the expression of the extended state observer is as follows:

[0119] ;

[0120] in, β 01 , β 02 , β 03 is the system gain parameter; h is the sampling step length; α 3 and α The value range of 4 is [0,1]. For example, α 3=0.5, α 4=0.25.

[0121] Step 3: After completing population initialization and setting the fitness function, the GRO algorithm is used to self-tune the parameters to be tuned of the ADRC controller to obtain the GRO-ADRC short-distance coupling tracking cruise control model.

[0122] In the nonlinear state error feedback control law, β 1 mainly affects the speed of system regulation, that is, it determines the rapidity of system response. β If 1 is set too large, it may cause oscillation and overshoot in the system, thus affecting stability. β 2 plays a damping role, which helps to suppress the oscillation of the system and improve stability. These two parameters usually need to be repeatedly tested and adjusted to achieve the best control performance. The parameters that need to be carefully adjusted in the extended state observer are β 01 , β 02 , β 03 The setting of these parameters needs to be continuously tested and optimized according to the characteristics of system errors and disturbances. Appropriate parameter values ​​can ensure that the extended state observer accurately estimates the system state and unknown disturbances, thereby improving the overall performance of the control system.

[0123] Therefore, the parameters to be adjusted include 5, which are the system gain in the nonlinear state error feedback control law β 1 and β 2, and the system gain parameter in the extended state observer β 01 , β 02 , β 03 .

[0124] The flowchart of this step is as follows Figure 7 As shown, the population initialization is completed using the following steps:

[0125] A sufficient set of parameters is randomly generated according to the parameters of the ADRC controller as an initial population, which is represented by an M×N matrix as shown below:

[0126] ;

[0127] Wherein, M is the size of the population particles; N is the dimension of the particles. Since the parameters to be adjusted include 5, N=5.

[0128] In the present invention, the output torque of the controller, i.e., the traction and braking torque of the power unit, is μ ( t ), and the acceleration of the power unit are introduced into the fitness function, and the expression of the fitness function is as follows:

[0129] ;

[0130] in, e ( t ) is the system error; μ ( t ) is the output of the system; a ( t ) is the acceleration of a single power unit of the virtual marshaling; ω 1. ω 2. ω 3 is the weight coefficient.

[0131] The process of self-tuning the parameters to be tuned of the ADRC controller using the GRO algorithm includes:

[0132] (1) Initialize the population particles and determine the number of population sizes M and the maximum number of iterations T max ;

[0133] (2) Calculate the fitness value of the current population particles at the new position, update the position of the current population particles according to the equation in the GRO algorithm, and update the optimal position of the particles at the same time;

[0134] The equation in the GRO algorithm is: ;

[0135] in, X i For gold prospectors i location; t is the current iteration number; Xnew i For gold prospectors i The latest location of f (.) is the fitness function;

[0136] (3) Update according to the following formula l 1 and l 2:

[0137] ;

[0138] ;

[0139] in, l 1 and l 2 are convergent components;

[0140] (4) Calculate the next position of the population particle according to one of the following three formulas:

[0141] ; ;

[0142] ;

[0143] in, X r For randomly selected current surveyors r location; r 1 is a random vector with a value range of [0,1];

[0144] D 1 , A 1 , D 2 , A 2 , D 3 It can be expressed by the following formula:

[0145] ; ;

[0146] ; ;

[0147] ;

[0148] in C 1=2 r 2, r 2 is a random vector with a value range of [0,1]; X g1 , X g2 are two randomly selected gold prospectors; D3 is the collaboration vector; X * is the best gold mine location;

[0149] (5) Update the fitness value and the best population particle position. If the termination condition is met, output the result; if not, return to step (2) to continue the iterative search.

[0150] Therefore, the virtual marshaling train system can be controlled according to the set target vehicle distance curve, and the error between the actual output and the expected output fed back by the virtual marshaling train system can be approached to zero through ADRC control, ensuring that the power unit can accurately complete the coupling marshaling.

[0151] Step 4: Set the line parameters and the initial state of the virtual marshaling high-speed train, obtain the actual vehicle distance during the operation of the virtual marshaling high-speed train dynamic model, input the actual vehicle distance into the GRO-ADRC short-spacing coupling tracking cruise control model to output the control torque for control.

[0152] Line parameters and initial states can be preset in advance as needed. Line parameters include line type and the length of each line type; initial states include initial distance, initial speed and disturbance type.

[0153] In a possible implementation, the selected line has a straight section, a curve section and a long ramp, the slope of the long ramp is 6%, the curve section parameters of the virtual train coupling tracking are shown in Table 3, and the initial state is shown in Table 4 and Table 5.

[0154] Table 3 Curved line parameters

[0155]

[0156] Table 4 Initial state settings for linear working conditions

[0157]

[0158] Table 5 Initial state settings of curve conditions

[0159]

[0160] The present invention also uses the method to conduct simulation experiments based on the above line parameters and initial state, and compares the simulation results with those of ordinary ADRC controllers to illustrate the excellent performance of the method provided by the present invention. The simulation results are as follows: Figure 8-Figure 11 , and as shown in Table 6 and Table 7 (in the present invention, counting starts with the leading power unit, the one adjacent to it and located behind it is the first following power unit; the following power unit located at the end is the third following power unit).

[0161] 1. Four-car straight-line tracking performance

[0162] Combined with simulation results Figure 8 and Fig. 9It can be seen that the ADRC controller and the GRO-ADRC short-spacing coupling tracking cruise control model are not much different in terms of the control effect of the vehicle distance. However, in the entire coupling tracking process, the ADRC controller still has large errors in speed and acceleration tracking, resulting in large energy consumption in the virtual marshaling process of high-speed trains, and the constant adjustment of acceleration increases the amplitude of the acceleration time-varying rate of change, reducing the comfort of passengers. The GRO-ADRC short-spacing coupling tracking cruise control model shows good overall performance, can timely perceive the subtle changes in the real-time vehicle distance, and adjust the control strategy accordingly to ensure that the power unit quickly and accurately tracks the target vehicle distance and realizes efficient coupling tracking operation; at the same time, the maximum speed and maximum acceleration of the power unit can be effectively reduced, indirectly making the energy consumption and passenger comfort of high-speed train marshaling in the global coupling tracking process effectively improved.

[0163] In addition, through simulation analysis, it is found that as the initial vehicle distance increases, the time required for the power unit to connect and hang in the process remains almost unchanged, but the maximum speed and acceleration of the power unit increase with the increase of the connection and hanging distance.

[0164] Table 6 Comparison of performance evaluation indicators of two control strategies

[0165]

[0166] It can be seen from Table 6 that under time-varying disturbances, the GRO-ADRC short-spacing coupling tracking cruise control model shows better performance than the traditional ADRC controller in terms of adjustment time, overshoot, and RMS value of acceleration change. This optimization performance further confirms that the short-spacing coupling tracking cruise controller designed based on GRO-ADRC can effectively meet the strict requirements of the high-speed train virtual marshaling control system and provide excellent control performance and stable cruising effect in actual operation.

[0167] 2. Four-car curve tracking performance

[0168] Combined with simulation results Fig.10 and Fig.11It can be seen that there is little difference between the ADRC controller and the GRO-ADRC short-spacing coupling tracking cruise control model in terms of the control effect of the vehicle distance. However, in the entire coupling tracking process, although the ADRC controller performs well in some aspects, its tracking accuracy of speed and acceleration needs to be improved, and there are large errors, which leads to large energy consumption in the virtual marshaling process of high-speed trains. The constant adjustment of acceleration increases the amplitude of the acceleration time-varying rate, which reduces the comfort of passengers. The GRO-ADRC short-spacing coupling tracking cruise control model has good comprehensive performance, can respond to the changes in real-time spacing, and control the power unit to accurately and quickly couple and track the target vehicle distance; at the same time, the maximum speed and maximum acceleration of the power unit can be effectively reduced, which indirectly improves the energy consumption and passenger comfort of high-speed train marshaling in the global coupling tracking process.

[0169] Table 7 Comparison of performance evaluation indicators of two control strategies

[0170]

[0171] It can be seen from Table 7 that under time-varying disturbances, the GRO-ADRC short-spacing coupling tracking cruise control model shows better performance than the traditional ADRC controller in terms of adjustment time, overshoot, and RMS value of acceleration change. It proves that the short-spacing coupling tracking cruise controller designed by GRO-ADRC can effectively meet the strict requirements of the high-speed train virtual marshaling control system, and provides excellent control performance and stable cruising effect in actual operation.

[0172] Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art may still modify the technical solutions described in the aforementioned embodiments, or make equivalent substitutions for some of the technical features therein; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An improved GRO algorithm based high-speed train virtual marshaling anti-disturbance control method, characterized in that: The method comprises: Step 1: Establish a dynamic model of a virtual marshaling high-speed train, which includes a dynamic model of a single power unit of four virtual marshaling trains; the dynamic model of the virtual marshaling high-speed train is a leading train-following topology structure, which includes a leading power unit and three following power units, and the dynamic models of any two adjacent single power units of the virtual marshaling train are connected in communication and are separated by a certain distance; Step 2: Establish an ADRC controller, which includes: a tracking differentiator, a nonlinear state error feedback control law and an extended state observer; and the tracking differentiator uses the fastest synthesis function fhan( x 1 ,x 2 ,r 0 ,h 0) Establish the fastest feedback system; in the tracking differentiator, the error between the target vehicle distance and the actual vehicle distance and its error differential are set as signal input, and the actual vehicle distance y (t) is: ; in, S i (t ) and Si -1 (t) are the displacements of the current virtual marshaling single-section power unit and the virtual marshaling single-section power unit in front of it; Step 3: After completing population initialization and fitness function setting, the GRO algorithm is used to self-tune the parameters to be adjusted of the ADRC controller to obtain the GRO-ADRC short-distance coupling tracking cruise control model, wherein the parameters to be adjusted include 5, which are the system gain in the nonlinear state error feedback control law β 1 and β 2, and the system gain parameter in the extended state observer β 01 , β 02 , β 03 ; Step 4: setting the line parameters and the initial state of the virtual marshaling high-speed train, obtaining the actual vehicle distance during the operation of the virtual marshaling high-speed train, inputting the actual vehicle distance into the GRO-ADRC short-distance coupling tracking cruise control model to output the control torque for control; In step 3, the population is initialized using the following steps: A sufficient set of parameters is randomly generated according to the parameters of the ADRC controller as the initial population, which is represented by an M×N matrix as shown below: ； Among them, M is the size of the population particles; N is 5; In step three, the output torque of the controller, i.e., the traction and braking torque of the power unit, and the acceleration of the power unit are introduced into the fitness function, and the expression of the fitness function is as follows: The expression of the fitness function is as follows: ； in, e ( t ) is the system error; μ ( t ) is the output torque of the controller, i.e. the traction and braking torque of the power unit; a ( t ) is the acceleration of a single power unit of the virtual marshaling; ω 1. ω 2. ω 3 is the weight coefficient.

2. According to the improved GRO algorithm based high-speed train virtual marshaling anti-disturbance control method of claim 1, it is characterized in that: In step 1, the dynamic model of a single-section power unit of a virtual marshaling is established in the following manner: Step S101, establishing a topological diagram of a dynamic model of a single power unit, and determining suspension parameters and structural parameters of the dynamic model; wherein the topological diagram includes: 1 vehicle body, 2 frames, 4 wheel sets and 8 axle boxes, and the vehicle body, each frame, and each wheel set have 6 degrees of freedom, namely, lateral, vertical, longitudinal, roll, nod, and shake, and each axle box has only 1 degree of freedom, namely, nod; Step S102, according to the multi-rigid body dynamics theory, using the suspension parameters and structural parameters, assembling each structure into a power unit bogie subsystem; Step S103: For the vertical damper, lateral damper and anti-snaking damper in the power unit bogie subsystem, nonlinear modeling is performed according to the relationship curve between their displacement velocity and force; for other mechanical elements, they are directly simplified into linear force elements; Step S104, in combination with the body parameters of the virtual high-speed train, the front bogie on the outside of the power unit bogie subsystem is adjusted to construct the rear bogie; then, the front bogie and the rear bogie are connected to the body model through the secondary suspension system, and the assembly of the whole vehicle is completed to obtain the dynamic model of the virtual single-section power unit.

3. The improved GRO algorithm-based high-speed train virtual marshaling anti-disturbance control method according to claim 1 is characterized in that: In step 2, the fastest comprehensive function fhan( x 1 ,x 2 ,r 0 ,h 0) is as follows: ； in, x 1. x 2 are the transition value and differential value of the target vehicle-to-vehicle distance respectively; r 0 、h 0 is the function control parameter, r The value range of 0 is [10,100]; d , a 0,y, a 1. a 2. s y , a , s a is the function fhan( x 1 ,x 2 ,r 0 ,h 0) Derived quantity; The expression of the fastest feedback system is as follows: ； in, h is the sampling step length.

4. The improved GRO algorithm-based high-speed train virtual marshaling auto-disturbance rejection control method according to claim 3 is characterized in that: The expression of the nonlinear state error feedback control law is as follows: ； in, e 1 and e 2 is the difference between the transition value and the observed value; u 0 is the initial control amount; z 1 is the observed displacement, z 2 is the observed speed; β 1 and β 2 is the system gain; α 1The value range is [0,1], α 2>1;fal( e,α,δ ) is a nonlinear saturation function, and its expression is as follows: ； in, e is the error signal; α is the controller parameter, and its value range is [0,1]; δ is the width of the nonlinear interval; The disturbance compensation forms the control quantity, which is the control torque input to the train. u The expression is as follows: ； in, z 3 is the observed disturbance, b 0 is the compensation factor.

5. The improved GRO algorithm-based high-speed train virtual marshaling anti-disturbance control method according to claim 4 is characterized in that: The expression of the extended state observer is as follows: ； in, α 3 and α The value range of 4 is [0,1].

6. The improved GRO algorithm-based high-speed train virtual marshaling auto-disturbance rejection control method according to claim 1 is characterized in that: The method for implementing self-tuning of the parameters to be tuned of the ADRC controller using the GRO algorithm includes: (1) Initialize the population particles and determine the size of the population particles M and the maximum number of iterations T max ; (2) Calculate the fitness value of the current population particles at the new position, update the position of the current population particles according to the equation in the GRO algorithm, and update the optimal position of the particles at the same time; The equation in the GRO algorithm is: ; in, X i For gold prospectors i location; t is the current iteration number; Xnew i For gold prospectors i The latest location of f (.) is the fitness function; (3) Update according to the following formula l 1 and l 2: ； ； in, l 1 and l 2 are convergent components; (4) Calculate the next position of the population particle according to one of the following three formulas: ； ； ； in, X r For randomly selected current surveyors r location; r 1 is a random vector with a value range of [0,1]; D 1 , A 1 , D 2 , A 2 , D 3 It can be expressed by the following formula: ； ； ； ； ； in C 1=2 r 2, r 2 is a random vector with a value range of [0,1]; X g1 , X g2 are two randomly selected gold prospectors; D3 is the collaboration vector; X * is the best gold mine location; (5) Update the fitness value and the best population particle position. If the termination condition is met, output the result; if not, return to step (2) to continue the iterative search.

Images