CN113911172A - High-speed train optimal operation control method based on self-adaptive dynamic planning - Google Patents

High-speed train optimal operation control method based on self-adaptive dynamic planning Download PDF

Info

Publication number
CN113911172A
CN113911172A CN202111187459.3A CN202111187459A CN113911172A CN 113911172 A CN113911172 A CN 113911172A CN 202111187459 A CN202111187459 A CN 202111187459A CN 113911172 A CN113911172 A CN 113911172A
Authority
CN
China
Prior art keywords
train
representing
vehicle
model
function
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202111187459.3A
Other languages
Chinese (zh)
Inventor
金柏
姜正
杜海宾
王瑞
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
CRRC Dalian Institute Co Ltd
Original Assignee
CRRC Dalian Institute Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by CRRC Dalian Institute Co Ltd filed Critical CRRC Dalian Institute Co Ltd
Priority to CN202111187459.3A priority Critical patent/CN113911172A/en
Publication of CN113911172A publication Critical patent/CN113911172A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B61—RAILWAYS
    • B61L—GUIDING RAILWAY TRAFFIC; ENSURING THE SAFETY OF RAILWAY TRAFFIC
    • B61L23/00—Control, warning, or like safety means along the route or between vehicles or vehicle trains
    • B61L23/08—Control, warning, or like safety means along the route or between vehicles or vehicle trains for controlling traffic in one direction only
    • B61L23/14—Control, warning, or like safety means along the route or between vehicles or vehicle trains for controlling traffic in one direction only automatically operated

Abstract

The invention discloses a high-speed train optimal operation control method based on self-adaptive dynamic programming, which establishes a multi-mass-point coupling motion model of a high-speed train, mainly defines the rule of influence of coupling coupler force on the operation of adjacent vehicles, meanwhile, the running resistance in the model and the dynamic parameters of the coupler buffer device are identified in real time by utilizing a particle swarm algorithm with global optimization capability, according to the dispersive characteristic of the power or the braking force of the established train model, the tracking precision, the operating punctuality, the operating stability and the operating energy saving of the train form a constraint variable, the weight of the decision variable is adaptively adjusted according to the operating state and the operating target of the train, under the constraint of an optimization objective function, the generalized tracking error value is input into the designed dispersed robust self-adaptive controller again to obtain control force output so as to realize the optimized and efficient operation of the high-speed train.

Description

High-speed train optimal operation control method based on self-adaptive dynamic planning
Technical Field
The invention relates to the technical field of high-speed train optimized operation control and automatic driving, in particular to a high-speed train optimized operation control method based on self-adaptive dynamic planning.
Background
At present, as one of the main factors restricting the development of the high-speed train, most of the operation control systems of the high-speed train still depend on import, and the core algorithm of the operation control systems cannot be completely made into a home, in addition, the operation of the high-speed train at present mainly depends on the operation of a driver, the condition that the driving efficiency is influenced due to different driving experiences of the driver can occur, and the actual labor intensity is also increased for the driver. Therefore, the algorithm for detecting the automatic driving and optimizing the operation control of the high-speed train can effectively reduce the working amount of a driver, effectively ensure the operation consistency of a locomotive interval and further improve the operation advantage of the high-speed railway; on the other hand, safe, intelligent, efficient and green travel of people can be realized, and the increasing material requirements of people are met; more importantly, the development of train control systems and the like with high safety and strong intelligence and independent intellectual property rights has important practical significance.
The high-speed train operation process is a complex dynamic system with large time lag, large inertia, high nonlinearity and coupling characteristics, and the design of the high-precision modeling and control strategy is two core works for realizing automatic driving of the train operation control system. At the present stage, the single-element point modeling scheme has the defects that the front and rear stress of train vehicles is generally different in a multi-marshalling train or a long-distance complex road section, and the defects are obviously amplified; the action rules of the multi-power unit model and the multi-mass point single displacement model for describing the car coupler force are not clear; although the accuracy of the virtual structure modeling scheme is high, the scheme depends on the designed model structure too much, has certain limitations, is difficult to improve the pertinence of the structure or parts, and is not enough to deal with the sudden factors such as railway signal change and the like. The control algorithm is also deficient to a certain extent, for example, PID control has a slow response speed to a high-speed train with large mass, and is a test for the smooth running performance of the train; the methods of fuzzy control and self-adaptive fuzzy control excessively depend on advanced experience, and have long time period and poor applicability; the predictive control has large calculation amount and long running time, and is matched with a computer with high performance.
Disclosure of Invention
The invention provides a high-speed train optimal operation control method based on self-adaptive dynamic programming, which aims to overcome the defects that the existing control method has long time period and poor applicability; the predictive control has large calculation amount and long running time.
In order to achieve the purpose, the technical scheme of the invention is as follows:
a high-speed train optimal operation control method based on self-adaptive dynamic planning comprises the following steps:
step 1, constructing the destination function γ of the train operation performancek；
Step 2, constructing an index weight value change rule;
step 3, utilizing the target function gammakAnd calculating a train robust adaptive control function F according to the index weight value change rule.
Further, step 1, constructing an objective function gamma of the train operation performancekThe method specifically comprises the following steps:
Figure BDA0003299853000000021
f1(e)＝X-Xd＝e
f2(ΔF)＝(ΔF)/κ1
Figure BDA0003299853000000022
wherein f is1(e) Performance function representing guaranteed train tracking accuracy and punctuation, f2(deltaf) represents the guaranteed train operation stability function,
Figure BDA0003299853000000023
representing the energy function of the train, k1,κ2To adjust the parameters, W1～W3Weights of key indexes and decision variables respectively, X represents the real-time feedback position of the train, and XdIndicating the period of the trainA desired position, e a position error, Delta F a change in control force of the train,
Figure BDA0003299853000000024
Representing the energy function of the train.
Further, the step 2 of constructing an index weight value change rule specifically includes:
Figure BDA0003299853000000031
Figure BDA0003299853000000032
χ＝∫eidt
wherein χ represents a set positive point performance deviation threshold,
Figure BDA0003299853000000033
a variation function showing the performance of the error of the running position of the train,
Figure BDA0003299853000000034
Shows the variation function of the train operation stability,
Figure BDA0003299853000000035
Represents a function of variation, mu, affecting the performance of the energy of operation of the train1Correction coefficient mu representing error performance of train running position2Correction coefficient f representing train running stability1(e) A train position deviation function is shown, and e represents a position error.
Further, the step 3 of constructing the train robust adaptive control function F specifically includes: f ═ U1+U2
Figure BDA0003299853000000036
U2＝-Qz-∑[ηhtanh(hz/ε)]
Ρ＝[m,ρT]
Figure BDA0003299853000000037
Figure BDA0003299853000000038
F represents the control force of the input train, i.e. the robust adaptive control function of the train, U1Representing adaptive terms, U2Representing a robust term, wherein p is a matrix formed by parameters in a multi-texture point coupling motion model, phi and z are intermediate variables for calculation, and the matrix has no practical physical significance; eta > 0, epsilon > 0, zeta > 0, Q > 0 are design parameters,
Figure BDA0003299853000000039
and
Figure BDA00032998530000000310
the design parameters A representing the parameter matrixes K and C in the multi-prime coupled motion model represent a matrix (actually determined by the train) consisting of parameters in the multi-prime coupled motion model, X represents the running position of each vehicle,
Figure BDA00032998530000000311
which represents an estimate of a design parameter,
Figure BDA00032998530000000312
the running speed of each vehicle is shown, z represents the running position error of each vehicle,
Figure BDA00032998530000000313
indicating the desired operating acceleration of each section of the vehicle,
Figure BDA00032998530000000314
indicating the desired operating speed of each section of the vehicle,
Figure BDA00032998530000000315
representing running speed error of each vehicle, e'XRepresenting the running position error, p, of each section of the vehicleTH and
Figure BDA00032998530000000316
and the design parameters for ensuring the stability of the system have no corresponding physical significance.
Further, the method for constructing the multi-prime point coupling motion model specifically comprises the following steps:
3.1, constructing a mutual coupling force model between vehicles;
and 3.2, constructing a multi-quality-point coupling motion model by utilizing the inter-vehicle mutual coupling force model.
Further, the specific formula of the inter-vehicle mutual coupling force model is as follows:
Figure BDA0003299853000000041
wherein, F(i-1)iRepresenting the coupling force during the operation of two vehicles
Figure BDA0003299853000000042
Delay and inertia links, k, representing coupler forcesi(xi-1-xi) The force representing the spring rate of the coupler is,
Figure BDA0003299853000000043
indicating the hook damping force, deltamaxThe travel capacity of the coupler buffer.
Further, the specific formula of the multi-texture-point coupling motion model is as follows:
Figure BDA0003299853000000044
Figure BDA0003299853000000045
Figure BDA0003299853000000046
m＝diag(m1,m2,…,mi,…,mn)
Figure BDA0003299853000000047
F＝[F1,…,Fi,…,Fn]T
ρ＝diag([αi,βi,γi]T)
Figure BDA0003299853000000051
wherein m represents the mass of each section of the vehicle;
Figure BDA0003299853000000052
representing the acceleration of each section of the vehicle; f represents the power or braking force input of each section of the vehicle; rho represents an operation resistance coefficient in the multi-prime point coupling motion model; rhoTRepresenting the transpose of the running resistance coefficient for the purpose of matching model calculations;
Figure BDA0003299853000000053
representing a state variable matrix composed of state variables in a multi-prime coupled motion model.
Further, the method comprises an online updating method for the running resistance coefficient and the car coupler model in the multi-quality-point coupling motion model, and the online updating method for the running resistance coefficient and the car coupler model in the multi-quality-point coupling motion model specifically comprises the following steps:
step 3.2.1, set 5 particles representing unknown parameters from the off-line data, each being αi、βi、 γi、ciAnd kiSimultaneously determining the initial position and velocity of the particles;
step 3.2.2, constructing a matrix w consisting of the resistance coefficient and the coupler model coefficientiThe matrix wiSpecifically is wi＝[ρivi]＝[αi βi γi ci ki]Where ρ isi＝[αi βi γi]，vi＝[ci ki]And i represents a vehicle number;
3.2.3, calculating a fitness function value min | J | of the parameter particles in each multi-quality point coupling motion model by the initial particles;
step 3.2.4, at the t-th running time of the train, comparing the fitness function value min | | J | | of the parameter particle in each multi-quality point coupling motion model in a preset area with the fitness function value of the best position which the parameter particle has been subjected to, and if the fitness function value is smaller than a set error, taking the particle parameter of the position as the current global best position to output wi(t), the current global best position output is specifically denoted as wi(t)＝[αi(t) βi(t) γi(t) ci(t) ki(t)]Executing the step six; if the error is greater than the set error, go to step 3.2.5;
step 3.2.5, respectively updating the speed and the position of the parameter particles in the multi-mass-point coupling motion model, executing step 3.2.4, and executing step 3.2.6 if a termination condition is met or the maximum iteration number is reached;
and 3.2.6, comparing the fitness function value min | | J | | | of the parameter particle in each multi-quality point coupling motion model with the fitness function value of the best position which is experienced in the whole situation, outputting the state of the particle with the minimum fitness function as the current best position in the whole situation, and executing the step 3.2.7.
3.2.7, recording and storing the state of the particles at the moment, namely the optimal solution of the fitness function;
step 3.2.8: at the t +1 moment of the operation of the first train, repeating the steps from 3.2.4 to 3.2.7, and solving the current global best position wi(t) optimal solution | w under fitness function constraintiIf w | |iRecording and outputting when | | is more than or equal to δ, otherwise outputting the optimal solution at the previous moment, wherein δ is an updating threshold;
step 3.2.9: until the train operation is finished, outputting the optimal solution wiAnd | | l corresponds to the model parameter value of the time series.
Further, the specific formula for obtaining the fitness function value min | | J | | | is as follows:
Figure BDA0003299853000000061
has the advantages that:
(1) according to the running characteristics of the high-speed train and the action mechanism of the coupler, a coupled motion model of the high-speed train is established, the physical significance of the model is clear, and the calculation process is simple. Compared with a simple substance point model and a virtual structure model, the calculation efficiency is obviously improved; compared with a virtual structure model, the method has visual physical significance, and can be used for carrying out targeted structure improvement and process improvement;
(2) when the multi-prime-point coupling motion model is constructed, dynamic parameters of the high-speed train coupling motion model and a real-time updating strategy of an added threshold are identified by utilizing a particle swarm algorithm, so that the model precision can be effectively improved;
(3) according to the actual running state of the high-speed train, a weight adjustment strategy of the running indexes of the high-speed train is designed in the step 2, namely, the change rule of the weight values of the indexes is stronger in flexibility and accords with the actual running of the train;
(4) and the fitness function result solved by the constraint index is input into the controller again, so that the train operation is more optimized.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a schematic diagram of coupled motion of a high-speed train;
FIG. 2 is a model time varying parameter update strategy;
FIG. 3 is a flow chart of a high-speed train operation optimization control method based on adaptive dynamic programming;
FIG. 4 is a weight adjustment rule for the optimization index;
FIG. 5 is a diagram of the vehicle corresponding coefficient adaptation model absolute error;
FIG. 6 is a second section vehicle tracking speed profile;
FIG. 7 is an entire train speed tracking curve;
FIG. 8 is a second section vehicle acceleration profile;
FIG. 9 is a graph of weight variation of constrained variables.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In the embodiment, starting from the longitudinal stress condition of a high-speed train in the running process, the whole train is described as a non-rigid-link multi-mass-point vibration system, the influence of the coupler force on adjacent trains is emphasized, a multi-mass-point coupling motion model of the high-speed train is established, and dynamic parameters in the model are identified in real time by utilizing a particle swarm algorithm with global optimization capability. On the basis, the influence rule of the optimization adjustment of the constraint variable on the result is researched, the target function related to the safety, stability and energy-saving indexes of train operation is designed, and the target function and the robust adaptive control method form a high-speed train optimization operation control scheme together. The running performance function of the train running process mainly comprises four indexes of accurate tracking, punctuality, stability and energy conservation, and the invention is designedThe optimization control algorithm takes the running position of the train as input, namely e is X-XdWhere X represents the output of the train strong coupling model, XdIf the condition is met, two index functions of stability and energy conservation are set simultaneously, and a comprehensive operation performance function of the train operation process can be constructed; for the train, if the condition is not met, two indexes of stability and energy conservation are not needed to be considered. As shown in fig. 1-9, comprising the following steps:
step 1, constructing the destination function γ of the train operation performancek；
Step 2, constructing an index weight value change rule;
step 3, utilizing the target function gammakAnd calculating a train robust adaptive control function F according to the index weight value change rule.
In a specific embodiment, step 1 constructs the destination function y of the train operation performancekThe method specifically comprises the following steps:
Figure BDA0003299853000000081
f1(e)＝X-Xd＝e
f2(ΔF)＝(ΔF)/κ1
Figure BDA0003299853000000082
wherein, γkThe smaller the absolute value of the target function established for considering the accurate punctuality, stability and energy conservation of the train is, the higher the comprehensive operation performance of the train is; f. of1(e) Representing a performance function for ensuring the tracking precision and the positive point of the train; f. of2(Δ F) represents a guaranteed train operation stability function;
Figure BDA0003299853000000086
energy index k representing train1,κ2For adjusting the parameters, the purpose is to make f1,f2,f3E is as good as theta, theta is f1,f2,f3Set of a priori intervals, W, to which data values are adapted1～W3Weights of key indexes and decision variables respectively, X represents the real-time feedback position of the train, and XdA desired position of the train, a position error, a control force change of the train, and a delta F,
Figure BDA0003299853000000087
And expressing an energy function of the train, wherein the key indexes and decision variables refer to indexes corresponding to accurate punctuality, stability and energy conservation of the train.
The invention utilizes the thought of the fuzzy theory to continuously provide an adaptive optimization weight algorithm related to the train operation punctuality index, realizes the performance index function of online adjustment, is used for exciting the result of target function decision to be more objective and reasonable, and further improves the operation efficiency and the operation performance of the high-speed train. Due to W1～W3Weights corresponding to the key index and the decision variable respectively should satisfy
Figure BDA0003299853000000083
Calculating the weight values of all indexes of the train running state by using the fuzzified idea, and constructing an index weight value change rule in the step 2 specifically comprises the following steps:
Figure BDA0003299853000000084
Figure BDA0003299853000000085
χ＝∫eidt
wherein, χ represents the set positive point performance deviation threshold, if the positive point performance deviation threshold is greater than the threshold, only the running tracking precision is considered, the train is ensured to run at the preset speed, and the stability index and the energy index are not considered; if the weight is within the threshold value range, the weight is calculated according to a weight adjustment algorithm, and W is worth mentioning1～W3The value should be changed according to the change rule set by the prior experience and the actual requirement as long as the set parameters are met
Figure BDA0003299853000000091
That is, take in the text
Figure BDA0003299853000000092
Figure BDA0003299853000000093
μ1＝0.1，μ2Simulation experiment is carried out when W is 0.91＝0.8,W2＝0.1,W3And when the positive point index proportion is calculated and the stability index and the energy-saving index proportion are calculated at the same time, the absolute value of the fitness function is the minimum.
Figure BDA0003299853000000094
A variation function showing the performance of the error of the running position of the train,
Figure BDA0003299853000000095
Representing a variation function affecting the running stability of the train,
Figure BDA0003299853000000096
Representing a function of variation, mu, affecting the running energy performance of the train1Correction coefficient mu representing error performance of train running position2Correction coefficient f representing train operation stability performance1(e) Represents a train position deviation function, and e represents a position error. At this time, it can be shown in fig. 4, specifically explained that, when the tracking error value is within the threshold range, the train operation stability and the required energy index can be considered, and when the tracking error value is outside the threshold range, only the operation tracking precision is considered, the operation safety is ensured, and the sum of the weights of the objective functions can be ensured to be 1 at any time, where W1, W2, and W3 are the weights of the key index and the decision variable, respectively. The method is simple and easy to understand, accords with the reality, has small calculated amount and has stronger practical value.
Supplementary explanation, the objective function gamma of the train operation performance is constructedkMainly utilizes the control quantity of the upper moment and the train running state in the running process to plan the tracking error of the lower moment, namely in a formula
Figure BDA0003299853000000097
Only e '(t) is an unknown variable, and the calculated value is input into the robust adaptive controller (1.. i.. n) again as e' (t +1), so that the optimal calculation of the running objective function can be realized.
In a specific embodiment, the step 3 of constructing the train robust adaptive control function F specifically includes:
F＝U1+U2
Figure BDA0003299853000000098
U2＝-Qz-∑[ηhtanh(hz/ε)]
Ρ＝[m,ρT]
Figure BDA0003299853000000099
Figure BDA00032998530000000910
f represents the control force of the input train, i.e. the robust adaptive control function of the train, U1Representing adaptive terms, U2Representing a robust term, wherein p is a matrix formed by parameters in a multi-texture point coupling motion model, phi and z are intermediate variables for calculation, and the matrix has no practical physical significance; eta > 0, epsilon > 0, zeta > 0, Q > 0 are design parameters,
Figure BDA00032998530000000911
and
Figure BDA00032998530000000912
representing parameters in coupled motion model for multiple prime pointsThe design parameters of the matrixes K and C have no practical physical significance, necessary satisfying conditions are given, so that the system tends to be gradually stable, A represents a matrix (actually determined by the train) formed by the parameters in the multi-quality-point coupling motion model, X represents the running position of each section of vehicle,
Figure BDA0003299853000000101
which represents an estimate of a design parameter,
Figure BDA0003299853000000102
the running speed of each vehicle is shown, z represents the running position error of each vehicle,
Figure BDA0003299853000000103
indicating the desired degree of running acceleration of each section of the vehicle,
Figure BDA0003299853000000104
indicating the desired operating speed of each section of the vehicle,
Figure BDA0003299853000000105
representing running speed error of each vehicle, e'XIndicating the running position error of each section of the vehicle.
The method comprehensively considers the running state and the running target of the train, namely three indexes of the right running performance, the running stability performance and the required energy performance of the train. The concrete expression is as follows: the method comprises the steps of respectively obtaining a punctual index and a tracking precision index according to the running position and the tracking error state of a train, obtaining a running stability performance index according to the variable quantity of control force, and calculating required energy according to the running control force and the real-time speed of the train to obtain a running energy index. And obtaining an optimized objective function through a set weight regulation rule related to the running state of the train, inputting a tracking error value output by the objective function into a design robust adaptive controller, wherein the tracking error value comprises the functional characteristics of specific vehicles of the high-speed train, the coupler link state of each vehicle and the coupler force determined by the position change information and the speed change information of each vehicle, and the coupling motion mechanism of the high-speed train is described together. Because each section of vehicle is taken as a mass point for research, and the parameter time-varying characteristics of the running resistance and the elastic structure of the car coupler are considered, the acting rule of the car coupler force is determined, the method has the advantage of simulating the actual running process, and a set of optimized control running scheme of a high-speed train coupling model is formed.
In a specific embodiment, the constructing the multi-texture-point coupling motion model specifically includes:
3.1, constructing a mutual coupling force model between vehicles;
and 3.2, constructing a multi-quality-point coupling motion model by utilizing the inter-vehicle mutual coupling force model.
In a specific embodiment, a calculation method for determining coupler force is required, a gap exists between a coupler and a coupler under normal conditions, however, in a movement process, a distance between the coupler and the coupler is changed due to the effect of coupling force of a vehicle, when a vehicle behind the running direction displaces more than a vehicle ahead, the coupler is in a compressed state, and represents power for the vehicle ahead and resistance for the vehicle behind. Similarly, when the front vehicle is displaced more than the rear vehicle, the coupler is in a tensile state, and the coupler represents tensile force to the front vehicle and power to the rear vehicle. At the moment, the mutual coupling force between vehicles can be calculated through the physical characteristics of the coupler device, and the specific formula of the mutual coupling force model between vehicles is as follows:
Figure BDA0003299853000000106
wherein, F(i-1)iThe coupling acting force in the running process of the two vehicles is represented, and the coupling acting force can be simply calculated in real time;
Figure BDA0003299853000000111
the delay link and the inertia link of the coupler force are represented, the delay link and the inertia link change within a certain range in the actual running of the train, the coefficient of the delay link and the inertia link is determined by the buffer structure, and the delay link and the inertia link can be measured through experiments. k is a radical ofi(xi-1-xi) The force representing the spring rate of the coupler is,
Figure BDA0003299853000000112
indicating coupler damping force, deltamaxThe travel capacity of the coupler buffer.
In a specific embodiment, combining newton's second law, a specific formula for obtaining the multi-prime point coupling motion model is:
Figure BDA0003299853000000113
Figure BDA0003299853000000114
Figure BDA0003299853000000115
m＝diag(m1,m2,…,mi,…,mn)
Figure BDA0003299853000000116
F＝[F1,…,Fi,…,Fn]T
ρ＝diag([αi,βi,γi]T)
Figure BDA0003299853000000117
wherein m represents the mass of each section of the vehicle;
Figure BDA0003299853000000118
representing the acceleration of each section of the vehicle; f represents the power or braking force input of each section of the vehicle; rho represents an operation resistance coefficient in the multi-prime point coupling motion model; rhoTRepresenting the transpose of the running resistance coefficient for the purpose of matching model calculations;
Figure BDA0003299853000000121
representing a state variable matrix composed of state variables in a multi-particle coupled motion model.
The multi-quality coupling motion model distinguishes trailers and vehicles of the trains and considers coupling force between the vehicles and the vehicles, and although the complex coupling force is generated by the whole train together, the actual action is only completed by the action of the adjacent vehicles and reacts to the adjacent vehicles. In the actual running process, the trailer vehicle has no power control input and only has the control input of the braking force; with respect to trailers, power vehicles can provide both power control inputs and braking force inputs for high speed trains. The multi-mass-point coupling motion model describes the complex change rule of the car coupler force on the running condition of each section of car, better accords with the actual condition than a multi-power unit model and a multi-mass-point single-displacement model in a general form, and has visual physical significance compared with a virtual structure model.
In the specific embodiment, the particle swarm optimization algorithm is a heuristic optimization algorithm simulating a natural law based on population evolution, has the advantage of directly searching without depending on information such as gradient, curvature and the like, and is very suitable for searching the optimal solution of unknown parameters in a certain range. Therefore, the high-speed train coupled motion model can be obtained by single-step cyclic optimization under the constraint condition by utilizing a particle swarm algorithm with global optimization capability, and the parameter to be identified is set to be a matrix formed by an unknown running resistance coefficient and an unknown coupler model coefficient. In order to make the model change discontinuously, an update threshold is added and the model is updated. The method specifically comprises an online updating method for the running resistance coefficient and the car coupler model in the multi-quality-point coupling motion model, and the online updating method for the running resistance coefficient and the car coupler model in the multi-quality-point coupling motion model specifically comprises the following steps:
step 3.2.1, set 5 particles representing unknown parameters from the off-line data, each being αi、βi、 γi、ciAnd kiSimultaneously determining the initial position and velocity of the particles;
step 3.2.2, constructing the resistorMatrix w formed by force coefficients and coupler model coefficientsiThe matrix wiSpecifically is wi＝[ρivi]＝[αi βi γi ci ki]Where ρ isi＝[αi βi γi]，vi＝[ci ki]And i represents a vehicle number;
3.2.3, calculating a fitness function value min | | J | | | of parameter particles in each multi-quality point coupling motion model by the initial particles, namely calculating a coupling model error function;
step 3.2.4, at the t-th running time of the train, comparing the fitness function value min | | J | | of the parameter particle in each multi-quality point coupling motion model in a preset area with the fitness function value of the best position which the parameter particle has been subjected to, and if the fitness function value is smaller than a set error, taking the particle parameter of the position as the current global best position to output wi(t), the current global best position output is specifically denoted as wi(t)＝[αi(t) βi(t) γi(t) ci(t) ki(t)]Executing the step six; if so, go to step 3.2.5, where the best position represents the most appropriate value, so that the min | | J | | function is the smallest and the value that is most appropriate for the adjacent optimization;
step 3.2.5, respectively updating the speed and the position of the parameter particles in the multi-mass-point coupling motion model, executing step 3.2.4, and executing step 3.2.6 if a termination condition is met or the maximum iteration number is reached;
and 3.2.6, comparing the fitness function value min | | J | | | of the parameter particle in each multi-quality point coupling motion model with the fitness function value of the best position which is experienced in the whole situation, outputting the state of the particle with the minimum fitness function as the current best position in the whole situation, and executing the step 3.2.7.
3.2.7, recording and storing the state of the particles at the moment, namely the optimal solution of the fitness function;
step 3.2.8: repeating the steps 3.2.4 to 3.2.7 at the t +1 moment of the operation of the first train to solve the current global maximumGood position wi(t) optimal solution | w under fitness function constraintiIf w | |iWhen | | | is more than or equal to δ, recording and outputting, otherwise, outputting the optimal solution at the previous moment, wherein δ is an updating threshold value, and the purpose is to add the updating threshold value in order to ensure that the model is discontinuously transformed;
step 3.2.9: until the train operation is finished, outputting the optimal solution wiAnd | | l corresponds to the model parameter value of the time series.
Will | wiEach parameter in | is set as a reasonable initial value determined by high-speed train off-line data, that is, each possible solution is expressed as a particle in the group, each particle has its own position vector and velocity vector, and a fitness function determined by a coupled motion model, all the particles move at a certain speed in a range space, and an optimal solution is searched through iteration.
In a specific embodiment, the specific formula for obtaining the fitness function value min | | J | | | is as follows:
Figure BDA0003299853000000131
regarding the problem of uncertain parameters in the model, it can be known from prior experience that although the model parameters are uncertain, the variation range is bounded, and it is assumed that each model parameter varies in a known range, that is, each model parameter has a corresponding prior interval, and can be written as the following form, ΩK＝[Kmin,Kmax]、ΩC＝[Cmin,Cmax]To control the quantity U1The variable constraint in (1) represents a matrix formed by the maximum value and the minimum value of an elastic coefficient matrix and a damping coefficient matrix of the coupler and omegaΡ＝[Ρmin,Ρmax]To control the quantity U1The intermediate variables of (2) are constrained, and have no actual physical significance. And satisfy the inequality relation
Figure BDA0003299853000000132
Defining a projection calculator
Figure BDA0003299853000000133
Figure BDA0003299853000000141
The parameter adaptation law can be given directly as follows:
Figure BDA0003299853000000142
in the formula, gamma1,γ2Γ is a positive adaptive gain matrix,
Figure BDA0003299853000000143
the estimates of the parameter matrices K, C and the design parameter matrix p in the model are represented separately.
A six-action two-towing CRH380A type high-speed motor train unit is selected as a simulation verification object, consists of 8 vehicle units, and has the same dynamic mechanism as that described in the figure 1. The acquired actual operation data of 1500 groups of speed traction force of the high-speed train in a certain section of the model are used as modeling samples and tracking experiment data, the modeling scheme and the control scheme provided by the invention are applied, the obtained experiment results are shown in figures 5-9, and the mean square deviation of the required energy and speed curves is shown in table 1.
TABLE 1 comparison of data collected with data values after optimization (e +09 denotes multiplication by the power of nine of ten)
Type (B) Required energy (J) Mean square error of velocity curve
Originally collected data 2.1987e+09 5.2555e+09
Operation scheme of the invention 4.3174e+07 8.0763e+07
Fig. 5 shows modeling errors, where the absolute modeling error is large in the initial stage of operation, but relative errors are within an acceptable range with respect to traction data of hundreds of kn and speed data of tens of meters per second, and furthermore, the modeling errors gradually stabilize as the operation time increases. Finally, when the train decelerates, the error has space for further improvement, but the overall error tends to be flat.
Fig. 6 is a graph of the operating position and speed of the second vehicle of the high-speed train of this type, where W4 gives a speed curve, and W5 is a tracking curve of the second vehicle, and it can be seen from the graph that for a given speed curve, the ability of implementing preprocessing for a given speed is firstly provided, then the high-speed train performs tracking operation according to the optimized curve, in the area where the operating speed curve changes greatly, the actual operation tends to be smooth, and finally, both can reach the destination at the same time, and reach the on-time target.
Fig. 7 shows the running speed and running time curves of various vehicles of the train with the model under the action of the coupling coupler force, and the good running speed law is basically maintained. Wherein W4 gives the speed curve, and W5 is 1-8 vehicle tracking curve, because 1-8 vehicle tracking curve has coupler force, actually have little difference, for easy understanding, so merge into a curve, see from figure 7 that actual operation has optimized the original speed curve, make the given speed curve more smooth operation, make the locomotive more steady energy-conserving. FIG. 8 is a graph illustrating the change of acceleration at 0.25m/s2The passengers can ride comfortablyThe degree is particularly good. Fig. 9 is a weight change curve, which illustrates that the influence of the tracking accuracy can be reduced appropriately when the tracking accuracy is high, and the effectiveness of the present solution is described to some extent.
Finally, the optimal control operation scheme adopted by the coupling motion model established by the scheme can ensure the safe, stable, correct and energy-saving operation of the high-speed train in the operation process.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and such modifications and substitutions do not depart from the spirit and scope of the present invention as defined by the appended claims.

Claims (9)

1. A high-speed train optimal operation control method based on self-adaptive dynamic planning is characterized by comprising the following steps:
step 1, constructing the destination function γ of the train operation performancek；
Step 2, constructing an index weight value change rule;
step 3, utilizing the target function gammakAnd calculating a train robust adaptive control function F according to the index weight value change rule.
2. The method for controlling the optimized operation of the high-speed train based on the adaptive dynamic programming as claimed in claim 1, wherein the step 1 of constructing the objective function γ of the train operation performancekThe method specifically comprises the following steps:
Figure FDA0003299852990000011
f1(e)＝X-Xd＝e
f2(ΔF)＝(ΔF)/κ1
Figure FDA0003299852990000012
wherein f is1(e) Performance function representing guaranteed train tracking accuracy and punctuation, f2(deltaf) represents a guaranteed train operation stability function,
Figure FDA0003299852990000013
representing the energy function of the train, k1,κ2To adjust the parameters, W1～W3Weights of key indexes and decision variables respectively, X represents the real-time feedback position of the train, and XdA desired position of the train, e a position error, Delta F a control force change of the train,
Figure FDA0003299852990000014
Representing the energy function of the train.
3. The method for controlling the optimized operation of the high-speed train based on the adaptive dynamic programming as claimed in claim 2, wherein the step 1 of constructing the index weight value change rule specifically comprises the following steps:
Figure FDA0003299852990000015
Figure FDA0003299852990000016
χ＝∫eidt
wherein χ represents a set positive point performance deviation threshold,
Figure FDA0003299852990000021
a variation function showing the performance of the error of the running position of the train,
Figure FDA0003299852990000022
Shows the variation function of the train operation stability,
Figure FDA0003299852990000023
Representing a function of variation, mu, affecting the running energy performance of the train1Correction coefficient mu representing error performance of train running position2Correction coefficient f representing train operation stability performance1(e) Represents a train position deviation function, and e represents a position error.
4. The high-speed train optimal operation control method based on the adaptive dynamic programming as claimed in claim 1, characterized in that: step 3, constructing a robust adaptive control function F of the train specifically comprises the following steps:
F＝U1+U2
Figure FDA0003299852990000024
U2＝-Qz-∑[ηhtanh(hz/ε)]
Ρ＝[m,ρT]
Figure FDA0003299852990000025
Figure FDA0003299852990000026
wherein F represents the control force of the input train, i.e. the robust adaptive control function of the train, U1Representing adaptive terms, U2Representing a robust term, wherein p is a matrix formed by parameters in a multi-texture point coupling motion model, phi and z are intermediate variables for calculation, and the matrix has no actual physical significance; eta > 0, epsilon > 0, zeta > 0, Q > 0 are design parameters,
Figure FDA0003299852990000027
and
Figure FDA0003299852990000028
the design parameters A representing parameter matrixes K and C in the multi-prime point coupling motion model represent matrixes formed by parameters in the multi-prime point coupling motion model, X represents the running position of each section of vehicle,
Figure FDA0003299852990000029
which represents an estimate of a design parameter,
Figure FDA00032998529900000210
represents the running speed of each vehicle, z represents the running position error of each vehicle,
Figure FDA00032998529900000211
indicating the desired operating acceleration of each section of the vehicle,
Figure FDA00032998529900000212
indicating the desired operating speed of each section of the vehicle,
Figure FDA00032998529900000213
representing running speed error of each vehicle, e'XRepresents the running position error, p, of each section of the vehicleTH and
Figure FDA00032998529900000214
and the design parameters for ensuring the stability of the system have no corresponding physical significance.
5. The method for controlling the optimized operation of the high-speed train based on the adaptive dynamic programming as claimed in claim 4, wherein the method for constructing the multi-prime-point coupling motion model specifically comprises the following steps:
3.1, constructing a mutual coupling force model between vehicles;
and 3.2, constructing a multi-quality-point coupling motion model by utilizing the inter-vehicle mutual coupling force model.
6. The method for controlling the optimal operation of the high-speed train based on the adaptive dynamic programming as claimed in claim 5, wherein the specific formula of the inter-vehicle mutual coupling force model is as follows:
Figure FDA0003299852990000031
wherein, F(i-1)iRepresenting the coupling force during the operation of two vehicles
Figure FDA0003299852990000032
Delay and inertia links, k, representing coupler forcesi(xi-1-xi) The force representing the spring rate of the coupler is,
Figure FDA0003299852990000033
indicating coupler damping force, deltamaxThe travel capacity of the coupler buffer.
7. The method for controlling the optimal operation of the high-speed train based on the adaptive dynamic programming as claimed in claim 5, wherein the specific formula of the multi-prime-point coupling motion model is as follows:
Figure FDA0003299852990000034
Figure FDA0003299852990000035
Figure FDA0003299852990000036
Figure FDA0003299852990000037
m＝diag(m1,m2,…,mi,…,mn)
Figure FDA0003299852990000038
F＝[F1,…,Fi,…,Fn]T
ρ＝diag([αi,βi,γi]T)
Figure FDA0003299852990000039
wherein m represents the mass of each section of the vehicle;
Figure FDA0003299852990000041
representing the acceleration of each section of the vehicle; f represents the power or braking force input of each section of the vehicle; rho represents an operation resistance coefficient in the multi-prime point coupling motion model; rhoTRepresenting the transpose of the running resistance coefficient;
Figure FDA0003299852990000042
representing a state variable matrix composed of state variables in the multi-prime coupled motion model.
8. The method for controlling the optimized operation of the high-speed train based on the adaptive dynamic programming as claimed in claim 5, further comprising an online updating method for the operation resistance coefficient and the coupler model in the multi-prime-point coupling motion model, wherein the online updating method for the operation resistance coefficient and the coupler model in the multi-prime-point coupling motion model specifically comprises the following steps:
step 3.2.1, setting 5 unknown parameters for representation according to off-line dataNumber of particles, each being alphai、βi、γi、ciAnd kiSimultaneously determining the initial position and velocity of the particles;
step 3.2.2, constructing a matrix w consisting of the resistance coefficient and the coupler model coefficientiThe matrix wiIn particular wi＝[ρi vi]＝[αi βi γi ci ki]Where ρ isi＝[αi βi γi]，vi＝[ci ki]And i represents a vehicle number;
3.2.3, calculating a fitness function value min | J | of the parameter particles in each multi-quality point coupling motion model by the initial particles;
step 3.2.4, at the t-th running time of the train, comparing the fitness function value min | | J | | of the parameter particle in each multi-quality point coupling motion model in a preset area with the fitness function value of the best position which the parameter particle has been subjected to, and if the fitness function value is smaller than a set error, taking the particle parameter of the position as the current global best position to output wi(t), the current global best position output is specifically denoted as wi(t)＝[αi(t) βi(t) γi(t) ci(t) ki(t)]Executing the step six; if the error is greater than the set error, go to step 3.2.5;
3.2.5, respectively updating the speed and the position of the parameter particles in the multi-mass-point coupling motion model, executing step 3.2.4, and executing step 3.2.6 if the termination condition is met or the maximum iteration number is reached;
and 3.2.6, comparing the fitness function value min | | J | | | of the parameter particles in each multi-quality point coupling motion model with the fitness function value of the best global position, outputting the state of the particles with the minimum fitness function as the current best global position, and executing the step 3.2.7.
3.2.7, recording and storing the state of the particles at the moment, namely the optimal solution of the fitness function;
step 3.2.8: and repeating the step 3 at the t +1 moment of the operation of the first train.2.4 to 3.2.7, solving for the current global best position wi(t) optimal solution | w under fitness function constraintiIf w | |iRecording and outputting when | | is more than or equal to δ, otherwise outputting the optimal solution at the previous moment, wherein δ is an updating threshold;
step 3.2.9: until the train operation is finished, outputting the optimal solution wiAnd | | l corresponds to the model parameter value of the time series.
9. The method for controlling the optimized operation of the high-speed train based on the adaptive dynamic programming as claimed in claim 8, wherein the specific formula for obtaining the fitness function value min J is as follows:
Figure FDA0003299852990000051
CN202111187459.3A 2021-10-12 2021-10-12 High-speed train optimal operation control method based on self-adaptive dynamic planning Pending CN113911172A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111187459.3A CN113911172A (en) 2021-10-12 2021-10-12 High-speed train optimal operation control method based on self-adaptive dynamic planning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111187459.3A CN113911172A (en) 2021-10-12 2021-10-12 High-speed train optimal operation control method based on self-adaptive dynamic planning

Publications (1)

Publication Number Publication Date
CN113911172A true CN113911172A (en) 2022-01-11

Family

ID=79239601

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111187459.3A Pending CN113911172A (en) 2021-10-12 2021-10-12 High-speed train optimal operation control method based on self-adaptive dynamic planning

Country Status (1)

Country Link
CN (1) CN113911172A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114275012A (en) * 2022-03-04 2022-04-05 西南交通大学 Self-adaptive control method for train control level
CN115128957A (en) * 2022-07-20 2022-09-30 西南交通大学 Heavy-duty train operation control method based on iterative learning
CN115130701A (en) * 2022-08-30 2022-09-30 长沙润伟机电科技有限责任公司 Locomotive coupler force monitoring method
CN117389158A (en) * 2023-12-11 2024-01-12 华东交通大学 High-speed train tracking control method and system based on hook buffer constraint condition

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140005877A1 (en) * 2012-06-29 2014-01-02 Mitsubishi Electric Research Laboratories Method for Determining Run-Curves for Vehicles in Real-Time Subject to Dynamic Travel Time and Speed Limit Constraint
CN104731096A (en) * 2015-02-04 2015-06-24 北京交通大学 Training section running track planning method based on rolling optimization
CN110428093A (en) * 2019-07-19 2019-11-08 北京工商大学 City railway train running optimizatin method based on improved cuckoo algorithm
CN110497943A (en) * 2019-09-03 2019-11-26 西南交通大学 A kind of municipal rail train energy-saving run strategy method for on-line optimization based on intensified learning
CN111391887A (en) * 2019-12-05 2020-07-10 中车工业研究院有限公司 High-speed train control method and design method of robust controller thereof

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140005877A1 (en) * 2012-06-29 2014-01-02 Mitsubishi Electric Research Laboratories Method for Determining Run-Curves for Vehicles in Real-Time Subject to Dynamic Travel Time and Speed Limit Constraint
CN104731096A (en) * 2015-02-04 2015-06-24 北京交通大学 Training section running track planning method based on rolling optimization
CN110428093A (en) * 2019-07-19 2019-11-08 北京工商大学 City railway train running optimizatin method based on improved cuckoo algorithm
CN110497943A (en) * 2019-09-03 2019-11-26 西南交通大学 A kind of municipal rail train energy-saving run strategy method for on-line optimization based on intensified learning
CN111391887A (en) * 2019-12-05 2020-07-10 中车工业研究院有限公司 High-speed train control method and design method of robust controller thereof

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
金柏: "基于动车组强耦合模型的速度跟踪控制方法" *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114275012A (en) * 2022-03-04 2022-04-05 西南交通大学 Self-adaptive control method for train control level
CN115128957A (en) * 2022-07-20 2022-09-30 西南交通大学 Heavy-duty train operation control method based on iterative learning
CN115130701A (en) * 2022-08-30 2022-09-30 长沙润伟机电科技有限责任公司 Locomotive coupler force monitoring method
CN117389158A (en) * 2023-12-11 2024-01-12 华东交通大学 High-speed train tracking control method and system based on hook buffer constraint condition
CN117389158B (en) * 2023-12-11 2024-03-08 华东交通大学 High-speed train tracking control method and system based on hook buffer constraint condition

Similar Documents

Publication Publication Date Title
CN113911172A (en) High-speed train optimal operation control method based on self-adaptive dynamic planning
Zhou et al. Smart train operation algorithms based on expert knowledge and reinforcement learning
Novi et al. Real-time control for at-limit handling driving on a predefined path
CN107791773B (en) Whole vehicle active suspension system vibration control method based on specified performance function
CN109815523B (en) Multi-target differential evolution algorithm for train operation based on decomposition
CN103324085A (en) Optimal control method based on supervised reinforcement learning
CN103879414B (en) A kind of railway locomotive optimized handling method based on self adaptation A-Star algorithm
CN104656452B (en) A kind of subway train optimal control method discrete based on matrix and device
CN106056238B (en) Planning method for train interval running track
CN110758413B (en) Train speed self-adaptive control method based on system parameter identification
CN108897928A (en) A kind of intelligent vehicle slope road energy conservation speed optimization method based on the search of nested Monte Carlo tree
CN114379583A (en) Automatic driving vehicle trajectory tracking system and method based on neural network dynamics model
CN112507464A (en) Freight train operation curve optimization method based on improved multi-target wolf algorithm
CN106708044A (en) Full-hovering hovercraft course control method based on grey prediction hybrid genetic algorithm-PID
CN114852105A (en) Method and system for planning track change of automatic driving vehicle
CN114253274A (en) Data-driven-based online hybrid vehicle formation rolling optimization control method
CN111591324B (en) Heavy-load train energy consumption optimization method based on gray wolf optimization algorithm
CN106842924A (en) EMUs optimal control method based on multi-state ANFIS models
Zhang et al. A flexible and robust train operation model based on expert knowledge and online adjustment
CN111598311B (en) Novel intelligent optimization method for train running speed curve
CN112947056B (en) Magnetic-levitation train displacement speed tracking control method based on IGWO-BP-PID
Zhao et al. Integrated longitudinal and lateral control system design and case study on an electric vehicle
Wu et al. An optimal longitudinal control strategy of platoons using improved particle swarm optimization
CN113759701A (en) High-speed train speed control method and system
CN113033902A (en) Automatic driving track-changing planning method based on improved deep learning

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination